Дисертації з теми "Lungs Cancer Imaging"

Introdução: A definição do volume-alvo macroscópico, principalmente referente a casos de câncer de pulmão, exige o maior número de informações possíveis no que diz respeito à localização, extensão e mobilidade tumoral. A literatura demonstra um importante avanço quando utilizada imagens metabólicas como é o caso da tomografia por emissão de pósitron/tomografia computadorizada (PET/CT), porém a sua aplicação nos planejamentos radioterápicos ainda é muito discutida devido ao seu grau de complexidade. Objetivos: Avaliar o impacto da PET/CT no delineamento tumoral em casos de câncer de pulmão não-pequenas células (CPNPC) e linfonodos regionais. Materiais e Métodos: Foram selecionados retrospectivamente estudos de PET/CT de 26 casos de câncer de pulmão. Todos foram confirmados por biópsia, sendo em sua totalidade CPNPC. Todos os estudos foram realizados em um equipamento de PET/CT dedicado com parâmetros de aquisição idênticos. A interpretação das imagens e posterior delineamento foram realizados por dois médicos experientes, um radioterapeuta e um nuclear/radiologista. Os parâmetros ótimos de visualização foram pré-definidos, sendo mandatórios para os delineamentos. Os delineamentos foram realizados em duas etapas principais. A primeira relacionada ao desenho tumoral somente pela CT e a segunda, após no mínimo duas semanas de descanso visual, referindo-se ao desenho tumoral pela PET/CT. Somente o volume tumoral macroscópico (GTV) e os linfonodos regionais aumentados ou PET positivos foram delineados. Índices de conformidades (IC) foram calculados, tanto interobservadores (11 casos), quanto intra-observador (26 casos). Para a comparação entre observadores e entre delineamentos em relação ao volume, foi considerado o teste não-paramétrico de Wilcoxon. As comparações em relação ao IC foram feitas usando-se o teste t de Student para amostras pareadas. Em todos os testes, valores de p <0,05 indicaram significância estatística. Os dados foram analisados com o programa computacional SPSS® Statistics 17.0 (EUA). Resultados: A análise dos dados demonstrou diferença significativa entre os volumes médios delineados na CT e na PET/CT (p = 0,02), com evidente redução volumétrica no delineamnto por PET/CT. Houve diferença significativa entre os volumes CT delineados pelos dois observadores (p = 0,03) e uma tendência a apresentar diferença significativa entre volumes PET/CT (p = 0,05). A avaliação volumétrica intraobservador foi significativa (p < 0,01) apenas para o médico nuclear/radiologista, com redução de até 51% do volume CT e uma relação entre modalidades de 2,11 ± 0,22. Na análise dos IC, não houveram diferenças significativas entre as duas modalidades de imagem (p = 0,598). A análise dos IC intra-observadores demonstrou que para o radioterapeuta a PET/CT apresenta um impacto de 46% (IC médio = 0,54 ± 0,06), já para o nuclear/radiologista, o impacto foi de 65% (IC médio = 0,35 ± 0,06), representando uma diferença significativa (p = 0,03) em relação ao IC entre o médicos observadores. Para a análise linfonodal a PET/CT apresentou importante diferença na visualização de linfonodos, alterando 10 dos 26 casos, sendo 9 para a positividade apenas na fusão. Conclusão: A PET/CT apresentou significativo impacto no desenho do GTV e linfonodos regionais para casos de CPNPC.
Introduction: The definition of gross target volume, especially concerning cases of lung cancer, requires the greatest amount of information possible with regard to location, tumor size and tumor mobility. The literature demonstrates an important advancement using metabolic images such as PET/CT, however, its application in radiotherapy planning is still controversial due to its complexity. Objectives: To assess the impact of PET/CT in tumor delineation in cases of non-small cell lung cancer and regional lymph nodes as additional findings. Materials and Methods: Retrospectively studies of PET/CT of 26 lung cancer cases were selected. All were confirmed by biopsy, in its entirety NSCLC. All studies were performed on a PET/CT with dedicated acquisition identical parameters. Image interpretation and subsequent delineation were performed by two experienced physicians, one radiotherapist and the another nuclear/radiologist. The optimal parameters display were pre-defined, being mandatory for the designs. Each case received an identification of three random letters to access the medical images to be analyzed. The delineation was made in two main steps. The first reference to the drawing only in tumor CT and the second, after two weeks of visual rest, referring to the drawing on tumor PET/CT. Only the gross tumor volume (GTV) and regional lymph nodes were enlarged or PET + outlined. Conformity index (CI) were calculated both interobserver (11 cases), and intra-observer (26 cases). For comparison between observers and between designs in relation to the volume, was considered the nonparametric Wilcoxon test. Comparisons regarding the conformity index were made using the Student t test for paired samples. To assess the degree of agreement regarding positive lymph nodes were estimated with kappa coefficients of agreement. In all tests, p values <0.05 were considered statistically significant. Data were analyzed with the software SPSS Statistics 17.0 (USA). Results: Data analysis showed significant difference between the average volumes delineated on CT and PET/CT (p = 0.02), with obvious volume reduction. Significant difference between the volumes delineated by CT observars medical distinct classes (p = 0.03) and a tendency to present significant difference between volumes PET / CT (p = 0.05). The intraobserver volumetric evaluation was significant (p <0.001) only for observer 2, being the nuclear medicine physician / radiologist, reducing up to 51% of the volume CT and a relationship between methods of 2.11 ± 0.22. In the analysis of CI, there were no significant differences between the two imaging modalities (p = 0.598).CI analysis showed that intra-observer to observer 1 PET / CT has an impact of 46% (average CI = 0.54 ± 0.06). The viewer 2, the impact was greater, 46% (average IC = 0.39 ± 0.03), representing a difference of opinion regarding the CI (p = 0.03) between the medical classes. To regional lymph nodes with PET/CT revealed an important difference in the visualization of lymph nodes, changing 10 of the 26 cases, 9 to positivity only in the image fusion.Conclusion: PET/CT has a significant impact on the design of the GTV and regional lymph nodes in cases of NSCLC.

Quantitative imaging is a rapidly growing area of interest within the field of bioinformatics and biomarker discovery. Due to the routine nature of medical imaging, there is an abundance of high-quality imaging linked to clinical and genetic data. This data is particularly relevant for cancer patients who receive routine CT imaging for staging and treatment purposes. However, current analysis of tumor imaging is generally limited to two-dimensional diameter measurements and assessment of anatomic disease spread. This conventional tumor-node-metastasis (TNM) staging system stratifies patients to treatment protocols including decisions regarding adjuvant therapy. Recently there have been several studies suggesting that these images contain additional unique information regarding tumor phenotype that can further aid clinical decision-making. In this study I aimed to develop the predictive capability of medical imaging. I employed the principles of quantitative imaging and applied them to patients with non-small cell lung cancer (NSCLC). Quantitative imaging, also termed radiomics, seeks to extract thousands of imaging data points related to tumor shape, size and texture. These data points can potentially be consolidated to develop a tumor signature in the same way that a tumor might contain a genetic signature corresponding to mutational burden. To accomplish this I applied radiomics analyses to patients with early and late stage NSCLC and tested these for correlation with both histopathological data as well as clinical outcomes. Patients with both early and late stage NSCLC were assessed. For locally advanced NSCLC (LA-NSCLC), I analyzed patients treated with preoperative chemoradiation followed by surgical resection. To assess early stage NSCLC, I analyzed patients treated with stereotactic body radiation therapy (SBRT). Quantitative imaging features were extracted from CT imaging obtained prior to chemoradiation and post-chemoradiation prior to surgical resection. For patients who underwent SBRT, quantitative features were extracted from cone-beam CTs (CBCT) at multiple time points during therapy. Univariate and multivariate logistic regression were used to determine association with pathologic response. Concordance-index and Kaplan-Meier analyses were applied to time dependent endpoints of overall survival, locoregional recurrence-free and distant metastasis. In this study, 127 LA-NSCLC patients were identified and treated with preoperative chemoradiation and surgical resection. 99 SBRT patients were identified in a separate aim of this study. Reduction of CT-defined tumor volume (OR 1.06 [1.02-1.09], p=0.002) as continuous variables per percentage point was associated with pathologic complete response (pCR) and locoregional recurrence (LRR). Conventional response assessment determined by diameter (p=0.213) was not associated with pCR or any survival endpoints. Seven texture features on pre-treatment tumor imaging were associated with worse pathologic outcome (AUC 0.61-0.66). Quantitative assessment of lymph node burden demonstrated that pre-treatment and post-treatment volumes are significantly associated with both OS and LRR (CI 0.62-0.72). Textural analyses of these lymph nodes further identified 3 unique pre-treatment and 7 unique post-treatment features significantly associated with either LRR, DM or OS. Finally early volume change showed associated with overall survival in CBCT scans of early NSCLC. Quantitative assessment of NSCLC is thus strongly associated with pathologic response and survival endpoints. In contrast, conventional imaging response assessment was not predictive of pathologic response or survival endpoints. This study demonstrates the novel application of radiomics to lymph node texture, CBCT volume and patients undergoing neoadjuvant therapy for NSCLC. These examples highlight the potential within the rapidly growing field of quantitative imaging to better describe tumor phenotype. These results provide evidence to the growing radioimics literature that there is significant association between imaging, pathology and clinical outcomes. Further exploration will allow for more complete models describing tumor imaging phoentype with clinical outcomes.

Dans ce travail, les s´séquences Imagerie par Résonance Magnétique (IRM) radiales à temps d’écho ultra-court (UTE) sont analysées pour évaluer leur potentiel dans l’étude non-invasive de différents modèles expérimentaux de maladies pulmonaires chez la souris. Chez le petit animal, les séquences radiales UTE peuvent efficacement limiter l’impact négatif sur la qualité de l’image dû au déphasage rapide des spins causé par les nombreuses interfaces air/tissu. En plus, les séquences radiales UTE sont moins sensibles aux artefacts de mouvement par rapport aux séquences Cartésiennes classiques. En conséquence, chez le petit animal, les séquences radiales UTE peuvent permettre d’obtenir des images du poumon avec une résolution bien inférieure au millimètre avec des rapports signal/bruit importants dans le parenchyme pulmonaire, tout en travaillant en conditions physiologiques (animaux en respiration spontanée). Dans cette thèse, il sera démontré que les séquences d’IRM protonique UTE sont outils efficaces dans l’étude quantitative et non-invasive de différents marqueurs distinctifs de certaines pathologies pulmonaires d’intérêt général. Les protocoles développés serontsimples, rapides et non-invasifs, faciles à implémenter, avec une interférence minimale sur la pathologie pulmonaire étudiée et, en définitive, potentiellement applicables chez l’homme. Il sera ainsi démontré que l’emploi des agents de contraste, administrés via les voies aériennes, permet d’augmenter la sensibilité des protocoles développés. Parallèlement, dans cette thèse des protocoles suffisamment flexibles seront implémentés afin de permettre l’étude d’un agent de contraste paramagnétique générique pour des applications aux poumons
In this work, ultra-short echo time (UTE) Magnetic Resonance Imaging (MRI) sequences are investigated as flexible tools for the noninvasive study of experimental models of lung diseases in mice. In small animals radial UTE sequences can indeed efficiently limit the negative impact on lung image quality due to the fast spin dephasing caused by the multiple air/tissue interfaces. In addition, radial UTE sequences are less sensitive to motion artifacts compared to standard Cartesian acquisitions. As a result, radial UTE acquisitions can provide lung images in small animals at sub-millimetric resolution with significant signal to noise ratio in the lung parenchyma, while working with physiological conditions (freely-breathing animals). In this thesis, UTE proton MRI sequences were shown to be efficient instruments to quantitatively investigate a number of hallmarks in longitudinal models of relevant lung diseases with minimal interference with the lung pathophysiology, employing easilyimplementable fast protocols. The synergic use of positive contrast agents, along with anadvantageous administration modality, was shown to be a valuable help in the increase of sensitivity of UTE MRI. At the same time, UTE MRI was shown to be an extremely useful and efficacious sequence for studying positive contrast agents in lungs

Background To combat one of the leading causes of death worldwide, lung cancer treatment techniques and modalities are advancing, and the treatment options are becoming increasingly individualized. Modern cancer treatment includes the option for the patient to be treated with proton therapy, which can in some cases spare healthy tissue from excessive dose better than conventional photon radiotherapy. However, to assess the benefit of proton therapy compared to photon therapy, it is necessary to make both treatment plans to get information about the Tumour Control Probability (TCP) and the Normal Tissue Complication Probability (NTCP). This requires excessive treatment planning time and increases the workload for planners.  Aim This project aims to investigate the possibility for automated prediction of the treatment dose distribution using a deep learning network for lung cancer patients treated with photon radiotherapy. This is an initial step towards decreasing the overall planning time and would allow for efficient estimation of the NTCP for each treatment plan and lower the workload of treatment planning technicians. The purpose of the current work was also to understand which features of the input data and training specifics were essential for producing accurate predictions.  Methods Three different deep learning networks were developed to assess the difference in performance based on the complexity of the input for the network. The deep learning models were applied for predictions of the dose distribution of lung cancer treatment and used data from 95 patient treatments. The networks were trained with a U-net architecture using input data from the planning Computed Tomography (CT) and volume contours to produce an output of the dose distribution of the same image size. The network performance was evaluated based on the error of the predicted mean dose to Organs At Risk (OAR) as well as the shape of the predicted Dose-Volume Histogram (DVH) and individual dose distributions.  Results  The optimal input combination was the CT scan and lung, mediastinum envelope and Planning Target Volume (PTV) contours. The model predictions showed a homogenous dose distribution over the PTV with a steep fall-off seen in the DVH. However, the dose distributions had a blurred appearance and the predictions of the doses to the OARs were therefore not as accurate as of the doses to the PTV compared to the manual treatment plans. The performance of the network trained with the Houndsfield Unit input of the CT scan had similar performance as the network trained without it.  Conclusions As one of the novel attempts to assess the potential for a deep learning-based prediction model for the dose distribution based on minimal input, this study shows promising results. To develop this kind of model further a larger data set would be needed and the training method could be expanded as a generative adversarial network or as a more developed U-net network.

Respiratory gating in lung PET imaging to compensate for respiratory motion artifacts is a current research issue with broad potential impact on quantitation, diagnosis and clinical management of lung tumors. However, PET images collected at discrete bins can be significantly affected by noise as there are lower activity counts in each gated bin unless the total PET acquisition time is prolonged, so that gating methods should be combined with imaging-based motion correction and registration methods. The aim of this study was to develop and validate a fast and practical solution to the problem of respiratory motion for the detection and accurate quantitation of lung tumors in PET images. This included: (1) developing a computer-assisted algorithm for PET/CT images that automatically segments lung regions in CT images, identifies and localizes lung tumors of PET images; (2) developing and comparing different registration algorithms which processes all the information within the entire respiratory cycle and integrate all the tumor in different gated bins into a single reference bin. Four registration/integration algorithms: Centroid Based, Intensity Based, Rigid Body and Optical Flow registration were compared as well as two registration schemes: Direct Scheme and Successive Scheme. Validation was demonstrated by conducting experiments with the computerized 4D NCAT phantom and with a dynamic lung-chest phantom imaged using a GE PET/CT System. Iterations were conducted on different size simulated tumors and different noise levels. Static tumors without respiratory motion were used as gold standard; quantitative results were compared with respect to tumor activity concentration, cross-correlation coefficient, relative noise level and computation time. Comparing the results of the tumors before and after correction, the tumor activity values and tumor volumes were closer to the static tumors (gold standard). Higher correlation values and lower noise were also achieved after applying the correction algorithms. With this method the compromise between short PET scan time and reduced image noise can be achieved, while quantification and clinical analysis become fast and precise.

Improved radiotherapy (RT) outcomes may be facilitated through monitoring of physiological processes implicated in radio-resistance such as proliferation. To this end, we studied 16 patients with non-small cell lung cancer with dynamic 3'-deoxy-3'-fluorothymidine (FLT) PET-CT before and after a week of radical RT. In absence of changes in primary tumour volume manually delineated on CT, RT induced a significant, moderately variable decrease in maximum and mean standard uptake values (SUVmax and SUVmean) of the order of 25%. Metastatic nodes showed a larger relative decrease in uptake approximating 40% associated with volumetric regression and only partially accountable by partial volume effect. Implementation of different segmentation approaches including manual delineation by a second operator and PET-based semi-automatic algorithms [two fixed thresholds, 2/3-cluster Fuzzy C-means (FCM-2, FCM-3) and 2/3-cluster fuzzy locally adaptive Bayesian algorithm (FLAB-2, FLAB-3)] yielded substantially different volumes and SUVs but consistent SUV responses. Reproducibility comparison favoured manual delineation, while thresholding delivered poor volumetric robustness and no apparent SUV reproducibility advantage over SUVmax or SUVpeak. FCM-2/FLAB-2 demonstrated intermediate reproducibility. In contrast to anatomical volumes, metabolic volumes exhibited significant increases with treatment, which for FLAB-2 correlated with changes of intratumoural uptake heterogeneity quantified by the coefficient of variation. Normal tissue analysis revealed an anterior-posterior gradient of lung uptake and an association of baseline marrow SUV with type/timing of neo-adjuvant chemotherapy. RT induced a dramatic (≈-76%), sharply demarcated marrow SUV decline in response to a minimum of 5Gy and a small (≈-20%), consistent decline in normal lung SUV. Kinetic analysis revealed a significant increase in the tumour delivery constant K1 (+32%) and a decrease in Ki/K1, larger (-36%) and more variable than the Ki (-26%) and SUV responses. Furthermore, despite baseline independence, we found a strong negative correlation between Ki/K1 and K1 at the response level. Kinetic analysis of the most uptake-avid tumour cluster extracted with FCM-3 yielded similar results with attenuated changes in delivery and retention. Overall, we found that RT induces early measurable changes in lung tumour FLT uptake. Spatial analysis indicated a variable dissociation of anatomical and metabolic volumes, while temporal analysis showed a variable antagonistic effect on delivery and phosphorylation, indicating that SUV analysis may misrepresent the magnitude and variability of RT anti-proliferative effect.

Medical imaging is a powerful tool for clinical practice allowing in-vivo insight into a patient’s disease state. Many modalities exist, allowing for the collection of diverse information about the underlying tissue structure and/or function. Traditionally, medical professionals use visual assessment of scans to search for disease, assess relevant disease predictors and propose clinical intervention steps. However, the imaging data contain potentially useful information beyond visual assessment by trained professional. To better use the full depth of information contained in the image sets, quantitative imaging characteristics (QICs), can be extracted using mathematical and statistical operations on regions or volumes of interests. The process of using QICs is a pipeline typically involving image acquisition, segmentation, feature extraction, set qualification and analysis of informatics. These descriptors can be integrated into classification methods focused on differentiating between disease states. Lung cancer, a leading cause of death worldwide, is a clear application for advanced in-vivo imaging based classification methods. We hypothesize that QICs extracted from spatially-linked and size-standardized regions of surrounding lung tissue can improve risk assessment quality over features extracted from only the lung tumor, or nodule, regions. We require a robust and flexible pipeline for the extraction and selection of disease QICs in computed tomography (CT). This includes creating an optimized method for feature extraction, reduction, selection, and predictive analysis which could be applied to a multitude of disease imaging problems. This thesis expanded a developmental pipeline for machine learning using a large multicenter controlled CT dataset of lung nodules to extract CT QICs from the nodule, surrounding parenchyma, and greater lung volume and explore CT feature interconnectivity. Furthermore, it created a validated pipeline that is more computationally and time efficient and with stability of performance. The modularity of the optimized pipeline facilitates broader application of the tool for applications beyond CT identified pulmonary nodules. We have developed a flexible and robust pipeline for the extraction and selection of Quantitative Imaging Characteristics for Risk Assessment from the Tumor and its Environment (QIC-RATE). The results presented in this thesis support our hypothesis, showing that classification of lung and breast tumors is improved through inclusion of peritumoral signal. Optimal performance in the lung application achieved with the QIC-RATE tool incorporating 75% of the nodule diameter equivalent in perinodular parenchyma with a development performance of 100% accuracy. The stability of performance was reflected in the maintained high accuracy (98%) in the independent validation dataset of 100 CT from a separate institution. In the breast QIC-RATE application, optimal performance was achieved using 25% of the tumor diameter in breast tissue with 90% accuracy in development, 82% in validation. We address the need for more complex assessments of medically imaged tumors through the QIC-RATE pipeline; a modular, scalable, transferrable pipeline for extracting, reducing and selecting, and training a classification tool based on QICs. Altogether, this research has resulted in a risk assessment methodology that is validated, stable, high performing, adaptable, and transparent.

ADVANCED IMAGING ANALYSIS FOR PREDICTING TUMOR RESPONSE AND IMPROVING CONTOUR DELINEATION UNCERTAINTY By Rebecca Nichole Mahon, MS A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy at Virginia Commonwealth University. Virginia Commonwealth University, 2018 Major Director: Dr. Elisabeth Weiss, Professor, Department of Radiation Oncology Radiomics, an advanced form of imaging analysis, is a growing field of interest in medicine. Radiomics seeks to extract quantitative information from images through use of computer vision techniques to assist in improving treatment. Early prediction of treatment response is one way of improving overall patient care. This work seeks to explore the feasibility of building predictive models from radiomic texture features extracted from magnetic resonance (MR) and computed tomography (CT) images of lung cancer patients. First, repeatable primary tumor texture features from each imaging modality were identified to ensure a sufficient number of repeatable features existed for model development. Then a workflow was developed to build models to predict overall survival and local control using single modality and multi-modality radiomics features. The workflow was also applied to normal tissue contours as a control study. Multiple significant models were identified for the single modality MR- and CT-based models, while the multi-modality models were promising indicating exploration with a larger cohort is warranted. Another way advances in imaging analysis can be leveraged is in improving accuracy of contours. Unfortunately, the tumor can be close in appearance to normal tissue on medical images creating high uncertainty in the tumor boundary. As the entire defined target is treated, providing physicians with additional information when delineating the target volume can improve the accuracy of the contour and potentially reduce the amount of normal tissue incorporated into the contour. Convolution neural networks were developed and trained to identify the tumor interface with normal tissue and for one network to identify the tumor location. A mock tool was presented using the output of the network to provide the physician with the uncertainty in prediction of the interface type and the probability of the contour delineation uncertainty exceeding 5mm for the top three predictions.

Positron Emission Tomography (PET) is an imaging modality that has become increasingly beneficial in Radiotherapy by improving treatment planning (1). PET reveals tumor volumes that are not well visualized on computed tomography CT or MRI, recognizes metastatic disease, and assesses radiotherapy treatment (1). It also reveals areas of the tumor that are more radiosensitive allowing for dose painting - a non-homogenous dose treatment across the tumor (1). However, PET is not without limitations. The quantitative unit of PET images, the Standardized Uptake Value (SUV), is affected by many factors such as reconstruction algorithm, patient weight, and tracer uptake time (2). In fact, PET is so sensitive that a patient imaged twice in a single day on the same machine and same protocol will produce different SUV values. The objective of this research was to increase the capabilities of PET by exploring other quantitative PET/CT measures for Radiotherapy treatment applications. The technique of quantitative image feature analysis, nowadays known as radiomics, was applied to PET and CT images. Image features were then extracted from PET/CT images and how the features differed between conventional and respiratory-gated PET/CT images in lung cancer was analyzed. The influence of noise on image features was analyzed by applying uncorrelated, Gaussian noise to PET/CT images and measuring how significantly noise affected features. Quantitative PET/CT measures outside of image feature analysis were also investigated. The correlation of esophageal metabolic tumor volumes (tumor volume demonstrating high metabolic uptake) and endoscopically implanted fiducial markers was studied. It was found that certain image features differed greatly between conventional and respiratory-gated PET/CT. The differences were mainly due to the effect of respiratory motion including affine motion, rotational motion and tumor deformation. Also, certain feature groups were more affected by noise than others. For instance, contour-dependent shape features exhibited the least change with noise. Comparatively, GLSZM features exhibited the greatest change with added noise. Discordance was discovered between the inferior and superior tumor fiducial markers and metabolic tumor volume (MTV). This demonstrated a need for both fiducial markers and MTV to provide a comprehensive view of a tumor. These studies called attention to the differences in features caused by factors such as motion, acquisition parameters, and noise, etc. Investigators should be aware of these effects. PET/CT radiomic features are indeed highly affected by noise and motion. For accurate clinical use, these effects must be account by investigators and future clinical users. Further investigation is warranted towards the standardization of PET/CT radiomic feature acquisition and clinical application.

Abstract Purpose: The purpose of this paper is to quantify if using an average intensity projection (AIP) scan or a 30% phase (mid-ventilation surrogate, MidV) scan as the reference image for patient position verification affects reproducibility of lung cancer patient alignment under free-breathing cone beam computed tomography (CBCT) image guidance and to analyze the effects of common clinical issues on registration variability. Methods: AIPs were retrospectively created for 16 lung patients (14 SBRT, 2 conventional treatments) originally planned/treated using the 30% phase MidV surrogate scan as reference. The study included 3-5 CBCTs from each patient. Registrations were performed between the AIP-CBCT and between the MidV-CBCT by 5 individuals (student, medical physics resident, medical resident, medical physicist, and attending physician) using MIM 6.2 image registration platform (Beachwood, OH). The images were rigidly registered, internal tumor volume (ITV) contours were displayed, and no rotational adjustments were allowed to reflect real treatment conditions. Additionally, the registrations for AIP-CBCT and MidV-CBCT were repeated 3 times by one individual for intra-observer variability assessment. Patient setup rotations, tumor volume, tumor motion, and breathing variability were estimated for correlation with registration variability. Results: The magnitude of the average intra-observer standard deviations from the lateral (LAT), anterior-posterior (AP), and superior-inferior (SI) directions for the AIP/CBCT and MidV/CBCT registrations were 0.9 mm and 1.2 mm, respectively. The magnitude of the average inter-observer standard deviations for the AIP/CBCT and MidV/CBCT were 1.7 mm and 1.8 mm, respectively. Average discrepancies over the whole population were found to be small; however, some individual patients presented high variability. Patient-specific cases with high variability were analyzed and observations on its potential causes are discussed. Conclusion: The differences in alignment using AIP versus MidV as the reference images are, when averaged over the population studied, very small and clinically irrelevant for PTV margins > 5mm; however, individual patients may be impacted in a clinically relevant manner if smaller margins, 3 mm and below, are used instead.

Responsable d'environ 30000 décès/an en France, le cancer du poumon est un problème de santé publique majeur. Un des enjeux actuels est d'adapter le traitement du cancer du poumon pour proposer des thérapeutiques ciblées plus efficaces et moins agressives. Les inhibiteurs de l'activité tyrosine kinase du récepteur de l'EGF (EGFR-TKI, gefitinib et erlotinib) constituent un réel progrès pour le traitement des cancers du poumon. Cependant, des mécanismes de résistance ont été décrits et des traitements combinés de thérapies ciblées avec des EGFR-TKI pourraient permettre de surmonter les résistances dans les cancers du poumon.Dans ce contexte, nous avons étudié les mécanismes impliqués dans la résistance à ces traitements. Nous montrons que l'activation de la voie de signalisation PI3K/AKT joue un rôle majeur dans la résistance aux EGFR-TKI, en inhibant l'apoptose par des mécanismes dépendant de l'acétylation. Les histones déacétylases (HDACs) et les sirtuines interviennent dans ces mécanismes de résistance, en modulant l'activation de la voie PI3K/AKT et l'apoptose. Ainsi l'utilisation d'inhibiteurs des HDACs (HDACi) et des sirtuines permettent de restaurer la sensibilité aux EGFR-TKI. Ces résultats confirment l'intérêt thérapeutique de l'association EGFR-TKI/HDACi et montrent le potentiel thérapeutique d'associer des inhibiteurs de l'EGFR et de la voie PI3K/AKT pour contourner la résistance aux EGFR-TKI.Les molécules thérapeutiques doivent atteindre spécifiquement le site tumoral, nécessitant parfois de les protéger contre leur dégradation, de réduire leurs effets indésirables, et de contrôler leur libération dans le temps et l'espace, à l'aide de transporteurs. Ainsi dans la deuxième partie de cette thèse, nous avons évalué les capacités de ciblage des tumeurs pulmonaires de nanoparticules à base de copolymère amphiphile, contenant une partie polysaccharidique hydrophile (le hyaluronane) et une partie polypeptidique hydrophobe (le poly(γ‐benzyl L‐glutamate, PBLG). Nos travaux mettent en évidence la capacité de ciblage tumoral de ces nanoparticules injectées par voie intraveineuse, ouvrant ainsi de nouvelles perspectives thérapeutiques. Notre objectif est de charger les combinaisons de traitements EGFR-TKI/HDACi que nous avons identifiées dans ces vecteurs, afin de traiter les tumeurs pulmonaires résistantes aux EGFR-TKI
Responsible of 30000 deaths each year in France, lung cancer is a major public health problem. One of the current challenges is to adapt the treatment of lung cancer to offer more effective and less aggressive targeted therapies. EGFR tyrosine kinase inhibitors (EGFR-TKI, gefitinib and erlotinib) represent a real progress in lung cancer therapy. However resistance mechanisms have been described and combination of targeted therapy with EGFR-TKI could overcome resistance in lung cancer.In this context, we studied mechanisms involved in resistance to EGFR-TKI. We show that PI3K/AKT activation is a major pathway leading to EGFR-TKI resistance leading to apoptosis inhibition through acetylation-dependent mechanisms. Histone deacetylase (HADCs) and sirtuin are involved in these mechanisms and modulate PI3K/AKT activation and apoptosis. The use of HDACs inhibitors (HDACi) and sirtuins inhibitors thus restores the sensitivity to EGFR-TKI. Altogether these results confirm the therapeutic effect of the EGFR-TKI/HDACi combination and show the therapeutic potential of the association of EGFR and PI3K/AKT inhibitors to overcome EGFR-TKI resistance.Therapeutic molecules must specifically reach the tumor site, sometimes requiring to protect them against degradation, to reduce their side effects, and to control their release in time and space, using transporters. In the second part of this thesis, we have thus evaluated the lung tumors targeting capabilities of amphiphilic copolymer-based nanoparticles, containing an hydrophilic polysaccharidic block (hyaluronan) and an hydrophobic polypeptidic block (the poly(γ‐benzyl L‐glutamate PBLG). Our work highlights the tumor targeting capability of these nanoparticles injected intravenously, offering new lung cancer therapy perspectives. Our aim is to load the drugs combination (EGFR-TKI/HDACi) in these vectors, to treat the lung tumors resistant to EGFR-TKI

Le troisième Plan Cancer lancé en 2013 désigne la précocité du diagnostic comme l'un des enjeux majeurs pour l'amélioration de la prise en charge des patients. Malgré l’essor des modalités et des performances de l’imagerie médicale, il reste des défis à relever pour l’aide au diagnostic et optimiser le recourt à la biopsie.L’imagerie photonique et spécialement la fluorescence résolue spectralement a déjà été éprouvée pour la caractérisation ex vivo des tumeurs mammaires et pulmonaires, sans agent de contraste ou traitement des échantillons. Notre objectif est de caractériser les capacités d'un dispositif médical innovant, développé au laboratoire, utilisant une aiguille fibrée de faible calibre pour l’analyse spectrale de la fluorescence endogène de ces lésions in situ. Nos premiers travaux dans le cadre d’études précliniques et cliniques ont montré des différences significatives de signatures spectrales entre tumeurs bénignes et malignes ex vivo et in vivo. Nos résultats ont également mis en évidence les limites d’utilisation du dispositif, en termes de spécificité, pour certains types de lésions.Une étude secondaire a été entreprise sur des tumeurs mammaires afin d'identifier les entités tissulaires majeures à l'origine des signatures spectrales obtenues avec notre dispositif fibré. L'imagerie spectrale en microscopie confocale et seconde harmonique (SHG), en multiphoton, ont été mises en œuvre afin d’établir une cartographie de biomarqueurs endogènes des tissus mammaires. Nous avons confronter ses résultats aux données obtenues avec le dispositif d'aiguille fibrée afin de pouvoir le positionner non seulement comme une aide au diagnostic mais aussi comme une méthode prometteuse pour l’histopathologie in situ
The third Cancer Plan, launched in 2013, identifies early diagnosis as one of the major challenges for improving patient care. Despite the growth in medical imaging modalities and performance, challenges remain in diagnosis aid and optimizing the use of biopsy.Photonic imaging and especially spectrally resolved fluorescence has already been tested for the ex vivo characterization of breast and lung tumors, without contrast agent or sample processing. Our goal is to characterize the capabilities of an innovative medical device, developed in the laboratory, using a low-caliber fibered needle for the spectral analysis of the endogenous fluorescence of these lesions in situ. Our early work in preclinical and clinical studies showed significant differences in spectral signatures between benign and malignant tumors ex vivo and in vivo. Our results also highlighted the limits the device, in terms of specificity, for certain types of lesions.Another study was conducted on mammary tumors in order to identify the major tissue entities at the origin of the spectral signatures obtained with our fibered device. Spectral imaging in confocal and second harmonic microscopy (SHG), in multiphoton, has been implemented in order to establish a mapping of endogenous biomarkers of mammary tissues. We compare its results with the data obtained with the fibered needle device in order to position it not only as an aid to diagnosis but also as a promising method for in situ histopathology

L'hétérogénéité tumorale est un facteur important de progression et de résistance au traitement. L'imagerie multiparamétrique TEP-IRM offre des opportunités uniques de caractérisation biologique cellulaire, mais n’a jamais été évalué à l’échelle régionale intra-tumorale dans le cancer du poumon non à petites cellules (CBNPC), première cause de décès oncologique. Une approche multiparamétrique dynamique simultanée TEP-IRM au 18F-FDG a été développée en ce sens. Cette approche a nécessité l’implémentation « maison » de la méthode de référence de quantification TEP du métabolisme glucidique (modèle tri-compartimental de Sokoloff); le développement d’une méthode de correction inédite des distorsions géométriques en imagerie de diffusion, validée sur fantôme et testée cliniquement ; la validation sur fantôme de méthodes d’IRM quantitative (relaxométrie T1/T2), également testées cliniquement; et l’implémentation « maison » du modèle compartimental de Tofts (version étendue) pour l’évaluation de la vascularisation tumorale en IRM dynamique de perfusion. Les résultats de nos travaux expérimentaux effectués à l’échelle intra-tumorale régionale illustrent l’hétérogénéité des rapports entre métabolisme glucidique et vascularisation dans le CBNPC, deux caractéristiques biologiques fondamentales de progression tumorale, et montrent qu’un partitionnement tumoral non supervisé par modèle de mélange gaussien, intégrant l’ensemble des biomarqueurs TEP-IRM de ce projet, individualise 3 types de supervoxels, dont la signature biologique peut être prédite avec une exactitude de 97% par 4 biomarqueurs TEP-IRM dominants, révélés par méthodes métaheuristiques d'apprentissage machine
Tumor heterogeneity is an important factor of progression and resistance to treatment. Multiparametric PET-MRI imaging offers unique opportunities to characterize biological cellular processes, but has never been evaluated at the regional level in Non-Small Cell Lung Cancer (NSCLC), the leading cause of oncological death. A simultaneous dynamic multiparametric 18F-FDG PET-MRI approach has been developed to this end. This approach required the “in-house” implementation of the reference absolute PET quantitative method of glucose metabolism (Sokoloff's tri-compartmental model); the development of a method for correcting geometric distortions in diffusion weighted imaging, validated on phantom and clinically tested; the phantom validation of quantitative MRI methods (T1/T2 relaxometry), also clinically tested; and the "in-house" implementation of the Tofts compartmental model (extended version) for the evaluation of tumor vascularization by dynamic perfusion MRI. The results of our work, performed at the regional intra-tumor level, illustrate the heterogeneity of the regional interlinks between glucose metabolism and vascularization in NSCLC, two fundamental biological hallmarks of tumor progression, and show that an unsupervised tumor partitioning by Gaussian mixture model, integrating all the PET-MRI biomarkers of this project, individualizes 3 types of supervoxels, whose biological signature can be predicted with 97% accuracy by 4 dominant PET-MRI biomarkers, revealed by metaheuristic machine learning methods

Ventilation imaging using 4D-CT is a convenient and cost effective functional imaging methodology which might be of value in radiotherapy treatment planning to spare functional lung volumes. To calculate ventilation imaging from 4D-CT we must use deformable image registration (DIR). This study validates the DIR methods and investigates the dependence of calculated ventilation on DIR methods and ventilation algorithms. The first hypothesis is if ventilation algorithms are robust then they will be insensitive to the precise DIR used provided the DIR is accurate. The second hypothesis is that the change in Houndsfield Unit (HU) method is less dependent on the DIR used and depends more on the CT image quality due to the inherent noise of HUs in normal CT imaging. DIR of the normal end expiration and inspiration phases of the 4D-CT images was used to correlate the voxels between the two respiratory phases. All DIR algorithms were validated using a 4D pixel-based and point-validated breathing thorax model, consisting of a 4D-CT image data set along with associated landmarks. Three different DIR algorithms, Optical Flow (OF), Diffeomorphic Demons (DD) and Diffeomorphic Morphons (DM), were retrospectively applied to the same group of 10 esophagus and 10 lung cancer cases all of which had associated 4D-CT image sets that encompassed the entire lung volume. Three different ventilation calculation algorithms were compared (Jacobian, ΔV, and HU) using the Dice similarity coefficient comparison. In the validation of the DIR algorithms, the average target registration errors with one standard deviation for the DIR algorithms were 1.6 ± 0.7 mm, maximum 3.1 mm for OF, 1.3 ± 0.6 mm, maximum 3.3 mm for DM, 1.3 ± 0.6 mm, maximum 2.8 mm for DD, indicating registration errors were within 2 voxels. Dependence of ventilation images on the DIR was greater for the ΔV and the Jacobian methods than for the HU method. The Dice similarity coefficient for 20% of low ventilation volume for ΔV was 0.33 ± 0.03 between OF and DM, 0.44 ± 0.05 between OF and DD and 0.51 ± 0.04 between DM and DD. The similarity comparisons for Jacobian was 0.32 ± 0.03, 0.44 ± 0.05 and 0.51 ± 0.04 respectively, and for HU 0.53 ± 0.03, 0.56 ± 0.03 and 0.76 ± 0.04 respectively. Dependence of ventilation images on the ventilation method used showed good agreement between the ΔV and Jacobian methods but differences between these two and the HU method were significantly greater. Dice similarity coefficient for using OF as DIR was 0.86 ± 0.01 between ΔV and Jacobian, 0.28 ± 0.04 between ΔV and HU and 0.28 ± 0.04 between Jacobian and HU respectively. When using DM or DD as DIR, similar values were obtained when comparing the different ventilation calculation methods. The similarity values for 20% of the high ventilation volume were close to those found for the 20% low ventilation volume. Mean target registration error for all three DIR methods was within one voxel suggesting that the registration done by either of the methods is quite accurate. Ventilation calculation from 4D-CT demonstrates some degree of dependency on the DIR algorithm employed. Similarities between ΔV and Jacobian are higher than between ΔV and HU and Jacobian and HU. This shows that ΔV and Jacobian are very similar, but HU is a very different ventilation calculation method.

Cette thèse porte sur la prise en compte des incertitudes et du mouvement dans le traitement des tumeurs pulmonaires en radiothérapie, que ce soit par photons, par protons ou par ions légers (hadronthérapie). L’accent est mis sur les méthodes de prise en compte du mouvement dites "passives". Ces méthodes, ne nécessitant pas d’asservissement respiratoire pour la délivrance de la dose, sont moins lourdes à mettre en place, et limitent l’introduction de nouvelles sources d’incertitudes. Des contributions cliniques et méthodologiques sont proposées. Tout d’abord, l’imagerie tomodensitométrique (TDM) pour la planification des traitements doit faire l’objet d’une attention particulière dans le cas de tumeurs soumises aux mouvements respiratoires. Nous avons évalué l’influence de la présence d’artéfacts de mouvements dans les images TDM sur la qualité de la planification. Nous avons également proposé des méthodologies et des recommandations pour l’optimisation des paramètres d’acquisition ainsi qu’un algorithme original de détection automatique des artéfacts dans les images TDM 4D. L’une des principales sources d’incertitudes lors de la planification de traitements en radiothérapie concerne la délinéation des volumes cibles. Nous avons évalué la variabilité inter-observateur de délinéation du volume cible macroscopique (GTV) et du volume cible interne (ITV) via une méthode originale permettant de l’intégrer dans le calcul des marges de sécurité. La réduction des incertitudes dues au mouvement respiratoire peut être réalisée en associant au système de contention une compression abdominale afin de limiter l’amplitude du mouvement respiratoire. Nous avons proposé une étude visant à évaluer l’impact de l’utilisation d’un tel système en fonction de la localisation dans le poumon. En radiothérapie par photons, une stratégie appelée mid-position consiste à irradier la tumeur dans sa position moyenne pondérée dans le temps et permet de réduire les marges par rapport à une stratégie ITV tout en conservant une couverture dosimétrique correcte. Une partie du travail de la thèse a consisté à participer à l’élaboration d’une étude clinique visant à comparer les deux stratégies, ITV et mid-position. Dans la plupart des cas, le mouvement respiratoire a une distribution de probabilité non-gaussienne et asymétrique, pouvant invalider la recette de calcul de marges de van Herk pour des mouvements tumoraux fortement asymétriques et de grande amplitude. Nous avons proposé un modèle numérique afin de prendre en compte cette asymétrie. Enfin, la prise en compte du mouvement respiratoire en hadronthérapie par des marges de sécurité doit faire l’objet de considérations spécifiques, en particulier en raison de la sensibilité du dépôt de dose aux variations de densité sur la trajectoire du faisceau. Dans une dernière partie, la définition des marges de sécurité pour prendre en compte le mouvement respiratoire de manière optimale est discutée
This PhD thesis focuses on the uncertainties and motion management in lung radiation therapy and particle therapy. Passive motion management techniques are considered. They consist in delivering the dose without any respiratory beam monitoring which may be difficult to set up or may introduce additional uncertainties. Clinical and methodological contributions about different treatment steps are proposed. First of all, computed tomography (CT) images for treatment planning must be carefully acquired in the presence of respiration-induced tumor motion. We assessed the impact of motion artifacts on the quality of treatment planning. We also proposed methodologies and recommendations about the optimization of 4D-CT acquisition parameters and an original method for automated motion artifact detection in 4D-CT images. Target delineation introduces one of the main source of uncertainties during radiation therapy treatment planning. We quantified inter-observer variations in the delineation of the gross tumor volume (GTV) and the internal target volume (ITV) using an original method in order to incorporate them in margin calculation. Reduction of motion uncertainties can be achieved by combining an abdominal pressure device with the immobilization system to reduce the amplitude of respiratory motion. We proposed a study to evaluate the usefulness of such a device according to the tumor location within the lung. Delivering the dose to the ITV implies an important exposure of healthy tissues along the tumor trajectory. An alternative strategy consists in irradiating the tumor in its time-averaged mean position, the mid-position. Margins are reduced compared with an ITV-based strategy while maintaining a correct tumor coverage. One part of the work consisted in participating in the implementation of a clinical trial in photon radiation therapy to compare the two strategies, ITV and mid-position. In the margin recipe proposed by van Herk, a Gaussian distribution of all combined errors is assumed. In most cases, respiratory motion has an asymmetric non-Gaussian distribution and the assumption may not be valid for strongly asymmetric tumor motions with a large amplitude. We proposed a numerical population-based model to incorporate asymmetry and non-Gaussianity of respiratory motion in margin calculation. Finally, when taking respiratory motion into account in particle therapy with safety margins, one must consider various parameters, particularly the dose deposit sensitivity to density variations. The last part is dedicated to a discussion on the defining of safety margins in order to optimally take into account respiratory motion

In this thesis, we describe several imaging techniques specifically designed and developed for the assessment of pulmonary structure, function and pathology. We then describe the application of this technology within appropriate biological systems, including the identification, tracking and assessment of lung tumors in a mouse model of lung cancer. The design and development of a Large Image Microscope Array (LIMA), an integrated whole organ serial sectioning and imaging system, is described with emphasis on whole lung tissue. This system provides a means for acquiring 3D pathology of fixed whole lung specimens with no infiltrative embedment medium using a purpose-built vibratome and imaging system. This system enables spatial correspondence between histology and non-invasive imaging modalities such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET), providing precise correlation of the underlying 'ground truth' pathology back to the in vivo imaging data. The LIMA system is evaluated using fixed lung specimens from sheep and mice, resulting in large, high-quality pathology datasets that are accurately registered to their respective CT and H&E histology. The implementation of an in vivo micro-CT imaging system in the context of pulmonary imaging is described. Several techniques are initially developed to reduce artifacts commonly associated with commercial micro-CT systems, including geometric gantry calibration, ring artifact reduction and beam hardening correction. A computer controlled Intermittent Iso-pressure Breath Hold (IIBH) ventilation system is then developed for reduction of respiratory motion artifacts in live, breathing mice. A study validating the repeatability of extracting valuable pulmonary metrics using this technique against standard respiratory gating techniques is then presented. The development of an ex vivo laser scanning confocal microscopy (LSCM) and an in vivo catheter based confocal microscopy (CBCM) pulmonary imaging technique is described. Direct high-resolution imaging of sub-pleural alveoli is presented and an alveolar mechanic study is undertaken. Through direct quantitative assessment of alveoli during inflation and deflation, recruitment and de-recruitment of alveoli is quantitatively measured. Based on the empirical data obtained in this study, a new theory on alveolar mechanics is proposed. Finally, a longitudinal mouse lung cancer study utilizing the imaging techniques described and developed throughout this thesis is presented. Lung tumors are identified, tracked and analyzed over a 6-month period using a combination of micro-CT, micro-PET, micro-MRI, LSCM, CBCM, LIMA and H&E histology imaging. The growth rate of individual tumors is measured using the micro-CT data and traced back to the histology using the LIMA system. A significant difference in tumor growth rates within mice is observed, including slow growing, regressive, disappearing and aggressive tumors, while no difference between the phenotype of tumors was found from the H&E histology. Micro-PET and micro-MRI imaging was conducted at the 6-month time point and revealed the limitation of these systems for detection of small lesions ( < 2mm) in this mouse model of lung cancer. The CBCM imaging provided the first high-resolution live pathology of this mouse model of lung cancer and revealed distinct differences between normal, suspicious and tumor regions. In addition, a difference was found between control A/J mice parenchyma and Urethane A/J mice ‘normal’ parenchyma, suggesting a 'field effect' as a result of the Urethane administration and/or tumor burden. In conclusion, a comprehensive murine lung cancer imaging study was undertaken, and new information regarding the progression of tumors over time has been revealed.

Le principe de la radiothérapie est de délivrer le maximum de dose de rayons X à la tumeur en épargnant au mieux les tissus sains environnants. Dans le cas du cancer du poumon, les mouvements respiratoires représentent une difficulté majeure. L'imagerie tomodensitométrique (TDM) 4D fournit des informations de mouvement spécifique à chaque patient qui peuvent servir de base pour la construction de modèles de mouvement respiratoire. La disponibilité dans les salles de traitement d'imageurs tomographiques embarqués sur les accélérateurs linéaires permet une estimation direct du mouvement et offre des informations plus précises. Un tel système d'imagerie permet entre-autre d'acquérir des images fluoroscopiques : ensemble de projections radiographiques 2D acquises au cours du temps et sous le même angle de vue. Notre approche s'intègre dans des systèmes de synchronisation de l'irradiation avec la respiration. Actuellement, cette technique existe en utilisant pour signal de synchronisation soit un signal externe, soit un signal interne issu du mouvement de marqueurs implantés autour de la tumeur. Notre approche permet d'obtenir un signal de synchronisation obtenu à partir de données internes sans marqueurs implantés. Dans ce cadre, nous avons expérimenté, développé puis évalué 3 méthodes de détection du mouvement à partir de séquences fluoroscopiques. Ces méthodes sont basées respectivement sur la variation de l'intensité, l'extraction de la hauteur du diaphragme et le suivi de blocst. A partir d'un algorithme de mise en correspondance de blocs, nous avons étudié l'homogénéité du mouvement apparent et déterminé, sans a priori géométrique, des régions où le mouvement est uniforme. Nous avons ensuite étudié la corrélation entre le signal interne extrait sur des séquences fluoroscopiques, et un signal extrait d'une vidéo-caméra synchronisée aux séquences fluoroscopiques assimilable à un signal externe. Dans une dernière partie, nous proposons d'estimer le mouvement 3D de la tumeur à partir d'un modèle de mouvement a priori élaboré dans une étape de pré-traitement à l'aide d'images TDM 4D et du signal respiratoire acquis dans la salle de traitement. L'intérêt de notre approche est qu'elle ne nécessite pas de marqueurs implantés ce qui la rend moins invasive que de nombreuses autres techniques. D'autre part, nous proposons un suivi 2D donc potentiellement rapide, mais basé sur un modèle 3D sous-jacent permettant ainsi de retrouver le maximum d'information. Cliniquement, notre approche permettrait de réaliser une adaptation quotidienne aux mouvements inter-sessions. Une des limites de notre approche est qu'elle nécessite une prise d'images ionisantes en continue. Un système hybride basée sur la combinaison d'un signal interne et d'un signal externe permettrait de limiter la dose additionnelle. Des efforts supplémentaires sur la réduction du temps de calcul sont encore nécessaires pour espérer guider un traitement par une telle approche.

Le carcinome pulmonaire est une affection grave et fréquente dont la prise en charge a été bouleversée ces dernières années, tant sur le plan diagnostique que pronostique ou « théranostique » avec l’avènement des « thérapies ciblées ». Ces dernières permettent une nette amélioration de la survie et du confort des patients éligibles, mais ne sont pas sans compliquer le travail médical, depuis le diagnostic de la maladie jusqu’au suivi régulier du patient, sans oublier le choix des traitements ou les problèmes techniques posés par la multiplication arborescente des altérations moléculaires à rechercher à partir d’un tissu tumoral souvent peu abondant dans ce contexte particulier de l’oncologie thoracique. Ce travail de thèse collige 5 travaux de recherche selon deux angles d’approche : les marqueurs moléculaires pronostiques et « théranostiques » du cancer pulmonaire, et les procédures de diagnostic in vivo de cette pathologie. Le premier axe comporte 4 articles. Les deux premiers concernent l’évaluation d’une nouvelle technique moléculaire, la LD-RT-PCR, dans la détection des translocation géniques du cancer pulmonaire : la première étude est une étude de faisabilité, la deuxième est un travail de validation. Le troisième article explore l’association entre la présence d’une mutation STK11 dans les carcinomes pulmonaires et l’expression de PDL1. Enfin, le quatrième article est une étude de cas illustrant l’importance de l’approche morphologique du cancer pulmonaire. Le second axe est représenté par un travail comparant une technique d’imagerie in vivo par voie endoscopique utilisant la micro-endoscopie confocale par laser avec l’approche microscopique conventionnelle
Lung cancer is a serious and frequent condition for which the management strategies have been dramatically modified in recent years, from a diagnostic, prognostic and “theranostic” perspective, most notably with the introduction of “targeted therapies”. The latter have demonstrated dramatic improvement in both quality of life and survival rates of eligible patients, yet consequently highlight new complications in diagnosis, treatment options or technical considerations which can be attributed to the growing number of molecular alterations to be detected from limited tissue samples frequently encountered in thoracic oncology. This work combines 5 different research papers from 2 different angles: prognostic and “theranostic” molecular markers of lung cancer, as well as in vivo diagnostic procedures of lung cancer. The first angle encompasses 4 articles. The first two evaluate a new molecular technique, LD-RT-PCR, to detect gene translocation in lung cancer. The third article explores the association between STK11 mutations in lung cancer and the expression of PDL1. Finally, the fourth article is a case report illustrating the importance of a morphological approach to lung cancer. The second angle compares in vivo imaging techniques by endoscopy using confocal laser microendoscopy alongside a conventional microscopic approach

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-05-30T19:36:59Z No. of bitstreams: 1 GiovanniLucca.pdf: 1608375 bytes, checksum: 90ad0a568a12b861d1a2a3467c275a12 (MD5)
Made available in DSpace on 2017-05-30T19:36:59Z (GMT). No. of bitstreams: 1 GiovanniLucca.pdf: 1608375 bytes, checksum: 90ad0a568a12b861d1a2a3467c275a12 (MD5) Previous issue date: 2017-01-31
CAPES
Lung cancer is the leading cause of cancer death worldwide, which accounts for more than 17% percent of the total cancer related deaths. However, its early detection may help in a sharp drop in this mortality rate. Because of the arduous analysis process, alternatives such as computational tools that use image processing techniques and pattern recognition have been widely developed and explored for the early diagnosis of this disease, providing a second opinion to the specialist and making this process faster. Therefore, this work proposes a methodology for the diagnosis of slice-based lung nodules extracted from computed tomography images using evolutionary convolutional neural networks. Firstly, the nodules are divided into two sub-regions using the Otsu algorithm based on the particle swarm optimization algorithm. Then, the slices of the nodules and the slices of their sub-regions were resized to the 28 x 28 dimension and given as input simultaneously to the networks. The architecture of the model was composed of three convolutional neural networks sharing the same fully connected layer at the end. Being a parameterized model, the genetic algorithm was applied to optimize some parameters, such as the number of filters in the convolution layers and the number of neurons in the hidden layer. The proposed methodology was tested on the Lung Image Database Consortium and the Image Database Resource Initiative, resulting in a sensitivity of 94.66 %, specificity of 95.14 %, accuracy of 94.78 % and area under the ROC curve of 0.949.
O câncer de pulmão é a maior causa de morte por câncer em todo mundo, representando mais de 17% do total de mortes relacionadas com câncer. No entanto, sua detecçãao precoce pode ajudar em uma queda acentuada nesta taxa de mortalidade. Devido ao árduo processo na análise dos exames por imagens, alternativas como sistemas computacionais que utilizam técnicas de processamento de imagens e reconhecimento de padrões têm sido amplamente desenvolvidos e explorados para o diagnóstico precoce desta doen¸ca, provendo uma segunda opinião para o especialista e tornando esse processo mais rápido. Diante disso, este trabalho propõe uma metodologia para o diagnóstico de nódulos pulmonares baseado nas fatias extraídas da tomografia computadorizada usando as redes neurais convolucionais evolutivas. Primeiramente, os nódulos são divididos em duas sub-regiões utilizando o algoritmo de Otsu baseado no algoritmo de otimização por enxame de partículas. Em seguida, as fatias dos nódulos e as fatias das suas sub-regiões foram redimensionadas para a dimensão 28 x 28 e dadas como entrada simultaneamente às redes. A arquitetura do modelo foi composta por três redes neurais convolucionais compartilhando a mesma camada completamente conectada no final. Tratando-se de um modelo parametrizado, o algoritmo genético foi aplicado para otimização de alguns parâmetros, tais como a quantidade de filtros nas camadas de convolução e a quantidade de neurônios na camada oculta. A metodologia proposta foi testada na base de imagens Lung Image Database Consortium e a Image Database Resource Initiative, resultando em uma sensibilidade de 94,66%, especifidade de 95,14%, acurácia de 94,78% e área sob a curva ROC de 0,949.

Lung cancer is the leading cause of cancer-related deaths in the world and its main manifestation occurs due to the appearance of pulmonary nodules. Pulmonary nodule classification is a complex task because it is a subjective and qualitative process that might be compromised by interpretation errors. Therefore, it is important to integrate computational tools to the pulmonary nodule classification process, since they have the potential to characterize objectively and quantitatively the lesions, so they can aid lung cancer diagnosis process. Content-Based Image Retrieval (CBIR) has been described as one of the most promissing differential diagnosis tool, since it is capable of retrieving similar cases from large image databases that were previously diagnosed. However, CBIR has some limitations, like the image feature extraction process and the time to compare one reference image with an image database. In this context, the goal of this work is to develop an algorithm to aid the diagnosis of lung cancer and the pulmonary nodule classification, using CBIR with the integration of three methods: 3D Texture Analysis and 3D Margin Sharpness Analysis for nodule characterization, and optimization on the execution time of the nodule comparison with paralelism on a Graphics Processing Unit (GPU). Images used in this work were computed tomography scans provided by the Lung Image Database Consortium, which has pulmonary nodules identified and classified by specialists according to the lesion’s likelihood of malignancy. Texture Attributes (TA) were extracted from a co-occurrence matrix obtained from the nodule volume. Margin Sharpness Attributes (MSA) were extracted from perpendicular lines drawn over the borders on all nodule slices. Integrated Attributes (IA) were created by concatenating TA and MSA. Euclidean distance was employed as similarity metric between feature vectors. CBIR algorithm’s precision was evaluated on the 10 most similar cases according to the likelihood of malignancy and with Precision and Recall parameters, on single-core, multi-core and many-core architectures. Results showed that MSA obtained more efficiency on pulmonary nodule retrieval, in the majority of precision evaluation scenarios, with the precision increase of 2% compared with TA. Texture attributes obtained the same efficiency as IA and presented higher mean precision only on benign nodule retrieval with Precision vs. Recall metric, with the precision increase of 3% compared with MSA. Results also showed that GPU, represented by the many-core device, was able to decrease execution time on image feature vector comparison and increase Euclidean distance performance on pulmonary nodule retrieval, with speedups of 16x, 17x and 19x. Therefore, CBIR allied to 3D margin sharpness descriptors and GPU optimization have big potential as a computer-based tool on lung cancer diagnosis and pulmonary nodule classification.
Fundação de Amparo a Pesquisa do Estado de Alagoas
O câncer de pulmão é o tipo de câncer que mais causa mortes no mundo e sua principal manifestação ocorre devido ao aparecimento de nódulos pulmonares. A classificação dos nódulos pulmonares é uma tarefa complexa que está sujeita a diversos erros de interpretação. Dessa forma, é importante integrar ferramentas computacionais ao processo de classificação de nódulos pulmonares, pois elas têm potencial de auxiliar no processo de diagnóstico do câncer de pulmão. Técnicas de recuperação de imagens baseada em conteúdo (CBIR - Content-Based Image Retrieval) têm sido descritas como ferramentas de diagnóstico diferencial promissoras, pois elas são capazes de recuperar em grandes bases de dados casos similares previamente diagnosticados. Contudo, a CBIR possui algumas limitações, como o processo de extração de características das imagens e o tempo de execução na comparação de uma imagem de referência com uma base de imagens. Neste contexto, o objetivo deste trabalho é desenvolver um algoritmo para o auxílio computadorizado à classificação de nódulos pulmonares utilizando CBIR, com a integração das técnicas: Análise de Textura 3D e Análise de Nitidez de Borda 3D para a caracterização dos nódulos pulmonares, e otimização no tempo de execução da comparação entre os nódulos pulmonares com paralelismo em uma unidade de processamento gráfico (GPU - Graphics Processing Unit). As imagens utilizadas neste trabalho são de tomografia computadorizada provenientes do projeto público Lung Image Database Consortium, que possui nódulos pulmonares identificados e classificados por especialistas segundo a probabilidade de malignidade da lesão radiológica. Atributos de Textura (AT) foram extraídos a partir da matriz de coocorrência obtida sobre o volume do nódulo. Atributos de Nitidez de Borda (ANB) foram extraídos a partir de linhas ortogonais traçadas sobre as bordas da lesão em todas as fatias do volume. Atributos Integrados (AI) foram criados a partir da concatenação dos AT e ANB. Distância Euclidiana foi utilizada como métrica de similaridade entre os vetores de características. A avaliação do algoritmo de CBIR desenvolvido utilizou as métricas de Precisão vs. Revocação e precisão para os 10 casos mais similares segundo a probabilidade de malignidade dos nódulos, em arquiteturas single-core, multi-core e many-core. Os resultados mostraram que os ANB obtiveram maior eficiência na recuperação dos nódulos pulmonares, na maioria dos cenários da avaliação de precisão, com aumento de precisão de 2 pontos percentuais em relação aos AT e AI na recuperação dos 10 casos mais similares. Os AT obtiveram mesma eficiência que os AI e apresentaram maior precisão média apenas na recuperação de nódulos benignos, com aumento de precisão de 3 pontos percentuais em relação aos ANB, quando empregada Precisão vs. Revocação. Os resultados mostraram também que a GPU conseguiu diminuir o tempo de execução na comparação dos vetores de características e aumentar o desempenho da distância Euclidiana na recuperação dos nódulos pulmonares, com ganhos de performance de até 19x. Com isto, a CBIR aliada aos atributos de nitidez de borda 3D e otimização em GPU possuem grande potencial como ferramenta computacional ao diagnóstico do câncer de pulmão e classificação de nódulos pulmonares.

Lung cancer is the most common and deadly of all kinds of cancers. Nowadays, the best chances of cure rely on early diagnosis. During diagnosis, the specialist must perform segmentation of a suspect nodule and analyze its shape to classify the nodule as benign or malignant. Nodule segmentation is commonly performed manually, and because of that, it is considered tedious, time consuming and can be compromised from observer variations. It may still be subject to human fail, and aspects like eyestrain, emotional factors, and even the specialist experience can. Hence the use of a semiautomatic process of segmentation provides the specialist a tool capable of aiding the early detection, which improves the patient’s chance of survival. Current segmentation techniques based on fuzzy connectivity maps provide a satisfatory level of 84% of acceptance by specialists. Therefore, the purpose of this work is to combine an optimized version of an algorithm based on fuzzy connectivity maps with other segmentation algorithms obtain a robust method for lung nodules segmentation. The method combines three algorithms to obtain the nodule segmentation; a pre-processsing stage based on local adaptive segmentation, which removes vascular attachments and undesired structures, and forms an optimum mask containing the nodule and obtains an optimum seed for the method; a 3-D fuzzy connectivity map, formed by the evaluation of relevance through all image voxels according to the optimum seed; a contrast-based sphericity oriented region growing algorithm, which establishes an automated halting criteria for the fuzzy connectivity map. As a result, the method presents an optimal solution and several alternative solutions for the specialist. The method was evaluated using a benchmark of lung nodules, composed of images manually segmented by specialists, applied in several studies of lung cancer. The test used 140 nodules, divided in two groups according to its malignancy levels. The method showed capable of segmenting different kinds of nodules with a satisfactory precision between others works in literature that used the same dataset. The work has also found reference values of segmentation algorithms constants and evaluated the impact of the optimized fuzzy connectivity maps to the processing cost.
De todos os tipos, o câncer de pulmão é o mais incidente e letal do mundo. Atualmente, as melhores chances de cura residem no diagnóstico precoce. No processo de diagnóstico, o especialista deve segmentar o nódulo suspeito e analisar sua forma para classificá-lo como benigno ou maligno. A segmentação é tipicamente realizada de forma manual e, por isso, é considerada tediosa, consome muito tempo e sofre de variações do observador, estando sujeita à falha humana. Aspectos como cansaço visual, fatores emocionais e mesmo a experiência do médico, podem comprometer o diagnóstico. Assim, o uso de um processo semiautomático de segmentação fornece ao médico uma ferramenta capaz de auxiliá-lo na detecção precoce, favorecendo um aumento nas chances de sobrevivência do paciente. A literatura apresenta níveis satisfatórios de 84% de aceitação dos especialistas na segmentação semiautomática baseada em mapas de conexidade Fuzzy. Assim, este trabalho objetiva o uso de uma versão otimizada do algoritmo de formação dos mapas de conexidade Fuzzy 3-D, aplicada ao contexto de nódulos pulmonares, combinada com outros algoritmos de segmentação para formar um método robusto de segmentação de nódulos pulmonares. O método aplica uma combinação de três algoritmos para obter a segmentação nodular: um pré-processamento por segmentação adaptativa local que exclui anexos vasculares e outras estruturas indesejadas, ao mesmo tempo em que forma uma máscara ideal que contém o nódulo e obtém uma semente ideal para o método; em seguida, um mapa de conexidade Fuzzy 3-D é formado pela avaliação de pertinência de todos os voxels da imagem em relação à semente ideal obtida; por fim, um algoritmo de crescimento de região esférico baseado em contraste estabelece um critério de parada automatizado baseado no mapa de conexidade Fuzzy. Como resultado, o método apresenta uma solução ideal e diversas soluções alternativas para a escolha do especialista. O método foi avaliado utilizando uma base de referência de nódulos pulmonares manualmente segmentados por especialistas e utilizada em diversos estudos de câncer de pulmão. Foram utilizados 140 nódulos para os testes, separados em dois grupos de acordo com seus graus de malignância. O método foi capaz de segmentar nódulos de diferentes tipos, com uma média satisfatória entre os índices de precisão de outros trabalhos na literatura que também utilizaram a mesma base de nódulos manualmente segmentados. O trabalho também encontrou valores de referência para as constantes dos algoritmos de segmentação e avaliou o impacto da versão otimizada dos mapas de conexidade no custo de processamento.

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-06-23T21:24:53Z No. of bitstreams: 1 AntonioCarvalho.pdf: 2731250 bytes, checksum: 35369a74be0aec3dd6b29a792c37fc35 (MD5)
Made available in DSpace on 2017-06-23T21:24:53Z (GMT). No. of bitstreams: 1 AntonioCarvalho.pdf: 2731250 bytes, checksum: 35369a74be0aec3dd6b29a792c37fc35 (MD5) Previous issue date: 2016-10-10
Lung cancer has been identi ed as the leading cause of death among cancer patients worldwide. The high rates of deaths and instances of records of this type of cancer worldwide demonstrate the importance of the development and research in order to produce resources for the detection and early diagnosis of this disease. Because of the exhaustive analysis process, alternatives such as computational tools that use image processing techniques and pattern recognition have been widely explored. Therefore, to assist the expert in the identi cation and diagnosis of nodules, systems are developed Computer-Aided Detection (CAD) and Computer-Aided Diagnostic (CADx). This thesis proposes the development of methods that reduce false positives, and the diagnosis of volumes of interest in computed tomography. The proposed methods are based on image processing techniques and pattern recognition. For this, biology concepts have been adapted and applied to the study of the branch of the diversity of species; such concepts are the phylogenetic diversity indexes used in this thesis as texture descriptors. In another aspect, techniques that measure the properties of the shape of radiological ndings have been developed and adapted. Subsequently, an evolutionary methodology is used for the selection of the best models for training. Finally, a support vector machine is applied to perform the classi cation. Promising results were found in the 833 tests that we performed; these tests were divided into 80% for training and 20% for testing. In general, for the best results, we have false positive reduction methods, an accuracy of 99.57%, sensitivity of 99.45%, speci city of 99.61%, and an ROC curve of 0.992. The results obtained for the classi cation of the degree of malignancy and benignity are: accuracy of 93.46%, sensitivity of 92.95%, speci city of 93.49%, and an ROC curve of 0.931.
O câncer de pulmão é apontado como a principal causa de morte entre os pacientes com câncer. As altas taxas de mortes e registros de ocorrências desse câncer em todo o mundo demonstram a importância do desenvolvimento e investigação, a fi m de produzir meios para a detecção e o diagnóstico precoce dessa doença. Devido ao exaustivo processo de análise, alternativas como ferramentas de cunho computacional que utilizam técnicas de processamento de imagens e do reconhecimento de padrões têm sido amplamente exploradas. Assim, para auxiliar o especialista na identifi cação e diagnóstico de nódulos, são desenvolvidos sistemas Computer-Aided Detection (CAD) e Computer-Aided Diagnostic (CADx). Esta tese propõe o desenvolvimento de métodos para redução de falsos positivos em um sistema CAD e diagnóstico de nódulos em tomografi a computadorizada. Os métodos propostas baseiam-se em técnicas de processamento de imagens e reconhecimento de padrões. Para tanto, foram adaptados e aplicados os conceitos da biologia no ramo do estudo da diversidade entre espécies, sendo esses os índices de diversidade logenética, usados nesta tese como descritores de textura. Em outro aspecto, foram desenvolvidas e adaptadas técnicas capazes de mensurar propriedades de forma dos achados radiológicos. Seguindo, usou-se uma metodologia evolutiva genética para seleção dos melhores modelos de treinamento. E por fi m, foi aplicada a máquina de vetor de suporte para realizar a classificação . Resultados promissores foram encontrados em teste com 833 exames divididos em 80% para treino e 20% para testes. Em linhas gerais, para os melhores resultados tem-se, nos métodos de redução de falsos positivos: uma acurácia de 99,57%, sensibilidade de 99,45%, especificidade de 99.61% e uma curva ROC de 0,992. Já nos resultados para a classificação quanto a taxa de malignidade e benignidade, obtiveram-se os seguintes valores: acurácia de 93,46%, sensibilidade de 92,95%, especificidade de 93,49% e uma curva ROC de 0,931.

Made available in DSpace on 2016-08-17T14:52:36Z (GMT). No. of bitstreams: 1 dissertacao Bruno Rodrigues Froz.pdf: 1583465 bytes, checksum: f53ff1f85d91788fc7d52925b16f6794 (MD5) Previous issue date: 2015-02-02
Conselho Nacional de Desenvolvimento Científico e Tecnológico
The lung cancer is known for presenting the highest mortality rate and one of the lowest survival rate after diagnosis, which is mainly caused by the late detection and treatment. With the goal of assist the lung cancer specialists, computed aided diagnosis systems are developed to automate the detection and diagnosis of this disease. This work proposes a methodology to classify, with computed tomography images, lung nodules candidates and non-nodules candidates. The Lung Image Database Consortium (LIDC) image database is used to create an image database with nodules candidates and an image database with non-nodule candidates. Three techniques are utilized to extract texture measurements. The first one is the artificial life algorithm Artificial Crawlers. The second one is the use of Rose Diagram to extract directional measurements. The third and last one is an hybrid model to join the Artificial Crawlers and Rose Diagram texture measurements. In the classification, que Support Vector Machine classifier is used, with its radial basis kernel. The archived results are very promising. With 833 LIDC exams, divided in 60% for train and 40% for test, we reached na accuracy mean of 94,30%, sensitivity mean of 91,86%, specificity mean of 94,78%, variance coefficient of accuracy of 1,61% and ROC curves mean área of 0,922.
O câncer de pulmão é conhecido por apresentar a maior taxa de mortalidade e uma das menores taxas de sobrevida após o diagnóstico, o que é causado principalmente pela detecção e tratamento tardios. Para o auxílio dos especialistas em câncer pulmonar, são desenvolvidos sistemas de diagnósticos auxiliados por computador com o objetivo de automatizar a detecção e diagnóstico dessa doença. Este trabalho propõe uma metodologia para a classificação, através de imagens de tomografias computadorizadas, de candidatos a nódulos pulmonares e candidatos a não-nódulos. O banco de imagens Lung Image Database Consortium (LIDC) é utilizado para a criação de uma base de imagens de candidatos a nódulos e uma base de imagens de candidatos a não-nódulos. Três técnicas são utilizadas para a extração de medidas de textura. A primeira delas é o algoritmo de vidas artificiais Artificial Crawlers. A segunda técnica é a utilização do Rose Diagram para a extração de medidas direcionais. A terceira e última técnica é um modelo híbrido que une as medidas do Artificial Crawlers e do Rose Diagram. Para a classificação é utilizado o classificador Máquina de Vetor de Suporte (MVS), com o kernel de base radial. Os resultados alcançados são muito promissores. Utilizando 833 exames do LIDC divididos em 60% para treino e 40% para teste, alcançou-se uma média de acurácia de 94,30%, média de sensibilidade de 91,86%, média de especificidade de 94,78%, coeficiente de variância da acurácia de 1,61% e área média das curvas ROC de 0,922.

碩士
輔仁大學
統計資訊學系應用統計碩士班
107
According to the statistics of the Ministry of Health and Welfare for 106 years, cancer has been the leading cause of death for 36 consecutive years, causing 48,037 lives in one year. Lung Cancer is leading cause of death more than a decade.It was taking 9325 lives in 2017. In Lung Cancer, 80% to 90% of patients get non-small cell lung cancer (NSCLC). The artificial intelligence has developed rapidly in recent years, among which the fastest development is Deep Learning. Convolutional Neural Networks are most widely used in image recognition, text recognition, object detection, etc.; AlexNet was be rolled in 2012 the error rate of the neural network has been greatly revised. Today, the Convolutional Neural Networks has a good performance in many object classification problems. This study classified four broad categories of non-small cell lung cancer using public computed tomography images combined with clinical diagnosis as a source of analysis, through a the three dimensional Convolutional Neural Networks to build the model and this method can quickly classify non-small cell lung cancer.

We present a new x-ray imaging technique, Correlation Imaging (CI), for improved breast and lung cancer detection. In CI, multiple low-dose radiographic images are acquired along a limited angular arc. Information from unreconstructed angular projections is directly combined to reduce the effect of overlying anatomy - the largest bottleneck in diagnosing cancer with projection imaging. In addition, CI avoids reconstruction artifacts that otherwise limit the performance of tomosynthesis. This work involved assessing the feasibility of the CI technique, its optimization, and its implementation for breast and chest imaging.

First a theoretical model was developed to determine the diagnostic information content of projection images using a mathematical observer. The model was benchmarked for a specific application in assessing the impact of reduced dose in mammography. Using this model, a multi-factorial task-based framework was developed to optimize the image acquisition of CI using existing low-dose clinical data. The framework was further validated using a CADe processor. Performance of CI was evaluated on mastectomy specimens at clinically relevant doses and further compared to tomosynthesis. Finally, leveraging on the expected improvement in breast imaging, a new hardware capable of CI acquisition for chest imaging was designed, prototyped, evaluated, and experimentally validated.

The theoretical model successfully predicted diagnostic performance on mammographic backgrounds, indicating a possible reduction in mammography dose by as much as 50% without adversely affecting lesion detection. Application of this model on low-dose clinical data showed that peak CI performance may be obtained with 15-17 projections. CAD results confirmed similar trends. Mastectomy specimen results at higher dose revealed that the performance of optimized breast CI may exceed that of mammography and tomosynthesis by 18% and 8%, respectively. Furthermore, for both CI and tomosynthesis, highest dose setting and maximum angular span with an angular separation of 2.75o was found to be optimum, indicating a threshold in the number of projections per angular span for optimum performance.

Finally, for the CI chest imaging system, the positional errors were found to be within 1% and motion blur to have minimal impact on the system MTF. The clinical images had excellent diagnostic quality for potentially improved lung cancer detection. The system was found to be robust and scalable to enable advanced applications for chest radiography, including novel tomosynthesis trajectories and stereoscopic imaging.

Dissertation

Mounting evidence suggests that the superposition of anatomical clutter in x-ray chest radiography poses a major impediment to the detectability of subtle lung nodules. Through decomposition of projections acquired using different x-ray energy spectra, dual-energy (DE) imaging offers to dramatically improve lung nodule conspicuity. The development of a high-performance DE chest imaging system is reported, with design and implementation guided by fundamental imaging performance metrics. Analytical and experimental studies of imaging performance guided the optimization of key acquisition technique parameters, including x-ray filtration, allocation of dose between low- and high-energy projections, and peak-kilovoltage selection. To minimize anatomical misregistration between images, a cardiac gating system was designed and implemented to direct x-ray exposures to within the quiescent period of the heart cycle. The instrumentation and optimal imaging techniques have been incorporated in a DE imaging prototype system now deployed in a clinical study to evaluate the diagnostic performance of DE imaging.

碩士
國立陽明大學
放射醫學科學研究所
95
Objectives: Mutations in the kinase domain of epidermal growth factor receptor (EGFR) have higher levels of basal receptor phophorylation that are associated with clinical responsiveness to Iressa in patients with non-small cell lung cancer (NSCLC). High mutation rate for EGFR (55%) in Taiwanese patients of adenocarcinoma of lung suggests an urgent requirement of a non-invasive imaging tool for pre-treatment and during therapy evaluation of lung cancer patients using EGFR signaling inhibitor. This study aimed to assess the feasibility of morpholino-[124I]IPQA derivative PET as an in vivo imaging tool for the expression of different EGFR mutants in NSCLC. Methods: The morpholino-[*I]IPQA derivative selectively, irreversibly, and covalently binds to the ATP binding site of the activated (phosphorylated) EGFR kinase, but not to the inactive EGFR kinase. The morpholino-IPQA derivative was labeled with 124I or 131I using the method described previously. The radiochemical purity was determined by chromatography method. The uptake and washout study of the morpholino-[131I]IPQA derivative were conducted in H1299 NSCLC derivative cell lines. NOD/SCID mice were transplanted with the vector-transfected H1299 cells on the right shoulder and with different EGFR active H1299 cells on the left shoulder. The biodistribution study and the PET imaging of morpholino-[*I]IPQA derivative in tumor-bearing NOD/SCID mice were performed at 4th week. Results: The radiochemical purity of morpholino-[*I]IPQA derivative was ≧90% and the radiochemical yield was about 50%. In vitro radiotracer accumulation and washout studies demonstrated a rapid accumulation and progressive retention post washout of morpholino-[131I]IPQA derivative in H1299 NSCLC derivative cell lines (L858R and E746-A750 del cell lines), but not in EGFR-transfected H1299 cell line and vector-transfected H1299 cell line under serum-starved conditions. Using the morpholino- [124I]IPQA derivative we obtained noninvasive microPET images of EGFR activity in high basal EGFR expressing H1299 derivative cell lines (L858R and E746-A750 del cell lines) subcutaneous tumor xenografts in NOD/SCID mice, but not in subcutaneous tumor xenografts grown form control cell line (vector-transfected H1299 cell line). Conclusions: Different EGFR mutant (different EGFR activity) tumors have different morpholino-[*I]IPQA derivative uptake. In microPET study, L858R and E746-A750 del cells demonstrate high uptake. But the structure of morpholino-IPQA derivative still has to be modified to improve the water solubility and also the hepatobiliary clearance. Morpholino-[124I]IPQA derivative may be a potential probe for selecting the candidate patients suffering NSCLC for the small molecule TK inhibitor therapy in the future.

The developments of highly conformal and precise radiation therapy techniques promote the necessity of more accurate treatment target localization and tracking. On-board imaging techniques, especially the x-ray based techniques, have found a great popularity nowadays for on-board target localization and tracking. With an objective to improve the accuracy of on-board imaging for lung cancer patients, the dissertation work focuses on the investigations of using limited-angle on-board x-ray projections for image guidance. The limited-angle acquisition enables scan time and imaging dose reduction and improves the mechanical clearance of imaging.

First of all, the dissertation developed a phase-matched digital tomosynthesis (DTS) technique using limited-angle (<=30 deg) projections for lung tumor localization. This technique acquires the same traditional motion-blurred on-board DTS image as the 3D-DTS technique, but uses the planning 4D computed tomography (CT) to synthesize a phase-matched reference DTS to register with the on-board DTS for tumor localization. Of the 324 different scenarios simulated using the extended cardiac torso (XCAT) digital phantom, the phase-matched DTS technique localizes the 3D target position with an localization error of 1.07 mm (± 0.57 mm) (average ± standard deviation (S.D.)). Similarly, for the total 60 scenarios evaluated using the computerized imaging reference system (CIRS) 008A physical phantom, the phase-matched DTS technique localizes the 3D target position with an average localization error of 1.24 mm (± 0.87 mm). In addition to the phantom studies, preliminary clinical cases were also studied using imaging data from three lung cancer patients. Using the localization results of 4D cone beam computed tomography (CBCT) as `gold-standard', the phase-matched DTS techniques localized the tumor to an average localization error of 1.5 mm (± 0.5 mm).

The phantom and patient study results show that the phase-matched DTS technique substantially improved the accuracy of moving lung target localization, as compared to the 3D-DTS technique. The phase-matched DTS technique can provide accurate lung target localizations like 4D-DTS, but with much reduced imaging dose and scan time. The phase-matched DTS technique is also found more robust, being minimally affected by variations of respiratory cycle lengths, fractions of respiration cycle contained within the DTS scan and the scan directions, which potentially enables quasi-instantaneous (within a sub-breathing cycle) moving target verification during radiation therapy, preferably arc therapy.

Though the phase-matched DTS technique can provide accurate target localization under normal scenarios, its accuracy is limited when the patient on-board breathing experiences large variations in motion amplitudes. In addition, the limited-angle based acquisition leads to severe structural distortions in DTS images reconstructed by the current clinical gold-standard Feldkamp-Davis-Kress (FDK) reconstruction algorithm, which prohibit accurate target deformation tracking, delineation and dose calculation.

To solve the above issues, the dissertation further developed a prior knowledge based image estimation technique to fundamentally change the landscape of limited-angle based imaging. The developed motion modeling and free-form deformation (MM-FD) method estimates high quality on-board 4D-CBCT images through applying deformation field maps to existing prior planning 4D-CT images. The deformation field maps are solved using two steps: first, a principal component analysis based motion model is built using the planning 4D-CT (motion modeling). The deformation field map is constructed as an optimized linear combination of the extracted motion modes. Second, with the coarse deformation field maps obtained from motion modeling, a further fine-tuning process called free-form deformation is applied to further correct the residual errors from motion modeling. Using the XCAT phantom, a lung patient with a 30 mm diameter tumor was simulated to have various anatomical and respirational variations from the planning 4D-CT to on-board 4D-CBCTs, including respiration amplitude variations, tumor size variations, tumor average position variations, and phase shift between tumor and body respiratory cycles. The tumors were contoured in both the estimated and the `ground-truth' on-board 4D-CBCTs for comparison. 3D volume percentage error (VPE) and center-of-mass error (COME) were calculated to evaluate the estimation accuracy of the MM-FD technique. For all simulated patient scenarios, the average (± S.D.) VPE / COME of the tumor in the prior image without image estimation was 136.11% (± 42.76%) / 15.5 mm (± 3.9 mm). Using orthogonal-view 30 deg scan angle, the average VPE/COME of the tumors in the MM-FD estimated on-board images was substantially reduced to 5.22% (± 2.12%) / 0.5 mm (± 0.4 mm).

In addition to XCAT simulation, CIRS phantom measurements and actual patient studies were also performed. For these clinical studies, we used the normalized cross-correlation (NCC) as a new similarity metric and developed an updated MMFD-NCC method, to improve the robustness of the image estimation technique to the intensity mismatches between CT and CBCT imaging systems. Using 4D-CBCT reconstructed from fully-sampled on-board projections as `gold-standard', for the CIRS phantom study, the average (± S.D.) VPE / COME of the tumor in the prior image and the tumors in the MMFD-NCC estimated images was 257.1% (± 60.2%) / 10.1 mm (± 4.5 mm) and 7.7% (± 1.2%) / 1.2 mm (± 0.2mm), respectively. For three patient cases, the average (± S.D.) VPE / COME of tumors in the prior images and tumors in the MMFD-NCC estimated images was 55.6% (± 45.9%) / 3.8 mm (± 1.9 mm) and 9.6% (± 6.1%) / 1.1 mm (± 0.5 mm), respectively. With the combined benefits of motion modeling and free-form deformation, the MMFD-NCC method has achieved highly accurate image estimation under different scenarios.

Another potential benefit of on-board 4D-CBCT imaging is the on-board dose calculation and verification. Since the MMFD-NCC estimates the on-board 4D-CBCT through deforming prior 4D-CT images, the 4D-CBCT inherently has the same image quality and Hounsfield unit (HU) accuracy as 4D-CT and therefore can potentially improve the accuracy of on-board dose verification. Both XCAT and CIRS phantom studies were performed for the dosimetric study. Various inter-fractional variations featuring patient motion pattern change, tumor size change and tumor average position change were simulated from planning CT to on-board images. The doses calculated on the on-board CBCTs estimated by MMFD-NCC (MMFD-NCC doses) were compared to the doses calculated on the `gold-standard' on-board images (gold-standard doses). The absolute deviations of minimum dose (DDmin), maximum dose (DDmax), mean dose (DDmean) and prescription dose coverage (DV100%) of the planning target volume (PTV) were evaluated. In addition, 4D on-board treatment dose accumulations were performed using 4D-CBCT images estimated by MMFD-NCC in the CIRS phantom study. The accumulated doses were compared to those measured using optically stimulated luminescence (OSL) detectors and radiochromic films.

The MMFD-NCC doses matched very well with the gold-standard doses. For the XCAT phantom study, the average (± S.D.) DDmin, DDmax, DDmean and DV100% (values normalized by the prescription dose or the total PTV volume) between the MMFD-NCC PTV doses and the gold-standard PTV doses were 0.3% (± 0.2%), 0.9% (± 0.6%), 0.6% (± 0.4%) and 1.0% (± 0.8%), respectively. Similarly, for the CIRS phantom study, the corresponding values between the MMFD-NCC PTV doses and the gold-standard PTV doses were 0.4% (± 0.8%), 0.8% (± 1.0%), 0.5% (± 0.4%) and 0.8% (± 0.8%), respectively. For the 4D dose accumulation study, the average (± S.D.) absolute dose deviation (normalized by local doses) between the accumulated doses and the OSL measured doses was 3.0% (± 2.4%). The average gamma index (3%/3mm) between the accumulated doses and the radiochromic film measured doses was 96.1%. The MMFD-NCC estimated 4D-CBCT enables accurate on-board dose calculation and accumulation for lung radiation therapy under different scenarios. It can potentially be valuable for treatment quality assessment and adaptive radiation therapy.

However, a major limitation of the estimated 4D-CBCTs above is that they can only capture inter-fractional patient variations as they were acquired prior to each treatment. The intra-treatment patient variations cannot be captured, which can also affect the treatment accuracy. In light of this issue, an aggregated kilo-voltage (kV) and mega-voltage (MV) imaging scheme was developed to enable intra-treatment imaging. Through using the simultaneously acquired kV and MV projections during the treatment, the MMFD-NCC method enabled 4D-CBCT estimation using combined kV and MV projections.

For all XCAT-simulated patient scenarios, the average (± S.D.) VPE / COME of the tumor in the prior image and tumors in the MMFD-NCC estimated images (using kV + open field MV) was 136.11% (± 42.76%) / 15.5 mm (± 3.9 mm) and 4.5% (± 1.9%) / 0.3 mm (± 0.4 mm), respectively. In contrast, the MMFD-NCC estimation using kV + beam's eye view (BEV) MV projections yielded results of 4.3% (± 1.5%) / 0.3 mm (± 0.3 mm). The kV + BEV MV aggregation can estimate the target as accurately as the kV + open field MV aggregation. The impact of this study is threefold: 1. the kV and MV projections can be acquired at the same time. The imaging time will be cut to half as compared to the cases which use kV projections only. 2. The kV and MV aggregation enables intra-treatment imaging and target tracking, since the MV projections can be the side products of the treatment beams (BEV MV). 3. As the BEV MV projections originate from the treatment beams, there will be no extra MV imaging dose to the patient.

The above introduced 4D-CBCT estimation techniques were all based on limited-angle acquisition. Though limited-angle acquisition enables substantial scan time and dose reduction as compared to the full-angle scan, it is still not real-time and cannot provide `cine' imaging, which refers to the instantaneous imaging with negligible scan time and imaging dose. Cine imaging is important in image guided radiation therapy practice, considering the respirational variations may occur quickly and frequently during the treatment. For instance, the patient may experience a breathing baseline shift after every respiratory cycle. The limited-angle 4D-CBCT approach still requires a scan time of multiple respiratory cycles, which will not be able to capture the baseline shift in a timely manner.

In light of this issue, based on the previously developed MMFD-NCC method, an AI-FD-NCC method was further developed to enable quasi-cine CBCT imaging using extremely limited-angle (<=6 deg) projections. Using pre-treatment 4D-CBCTs acquired just before the treatment as prior information, AI-FD-NCC enforces an additional prior adaptive constraint to estimate high quality `quasi-cine' CBCT images. Two on-board patient scenarios: tumor baseline shift and continuous motion amplitude change were simulated through the XCAT phantom. Using orthogonal-view 6 deg projections, for the baseline shift scenario, the average (± S.D.) VPE / COME of the tumors in the AI-FD-NCC estimated images was 1.3% (± 0.5%) / 0.4 mm (± 0.1 mm). For the amplitude variation scenario, the average (± S.D.) VPE / COME of the tumors in the AI-FD-NCC estimated images was 1.9% (± 1.1%) / 0.5 mm (± 0.2 mm). The impact of this study is three-fold: first, the quasi-cine CBCT technique enables actual real-time volumetric tracking of tumor and normal tissues. Second, the method enables real-time tumor and normal tissues dose calculation and accumulation. Third, the high-quality volumetric images obtained can potentially be used for real-time adaptive radiation therapy.

In summary, the dissertation work uses limited-angle on-board x-ray projections to reconstruct/estimate volumetric images for lung tumor localization, delineation and dose calculation. Limited-angle acquisition reduces imaging dose, scan time and improves imaging mechanical clearance. Using limited-angle projections enables continuous, sub respiratory-cycle tumor localization, as validated in the phase-matched DTS study. The combination of prior information, motion modeling, free-form deformation and limited-angle on-board projections enables high-quality on-board 4D-CBCT estimation, as validated by the MM-FD / MMFD-NCC techniques. The high-quality 4D-CBCT not only can be applied for accurate target localization and delineation, but also can be used for accurate treatment dose verification, as validated in the dosimetric study. Through aggregating the kV and MV projections for image estimation, intra-treatment 4D-CBCT imaging was also proposed and validated for its feasibility. At last, the introduction of more accurate prior information and additional adaptive prior knowledge constraints also enables quasi-cine CBCT imaging using extremely-limited angle projections. The dissertation work contributes to lung on-board imaging in many aspects with various approaches, which can be beneficial to the future lung image guided radiation therapy practice.

Dissertation

Cancer is the second leading cause of death in the world, and therefore, there is an undeniable need to ensure early screening and detection systems worldwide. The aim of this project is to study the feasibility of a Cone Beam Computed Tomography (CBCT) scanner for simultaneous breast and lung imaging. Additionally, the development of reconstruction algorithms and the study of their impact to the image quality was considered. Monte Carlo (MC) simulations were performed using the PENELOPE code system. A MC geometry model of a CBCT scanner was implemented for energies of 30 keV and 80 keV for hypothetical scanning protocols. Microcalcifications were inserted into the breast and lung of the computational phantom (ICRP Adult Female Reference), used in the simulations for dose assessment and projection acquisition. Reconstructed images were analyzed in terms of the Contrast-to-Noise Ratio (CNR) and dose calculations were performed for two protocols, one with a normalization factor of 2 mGy in the breast and another with 5 mGy in the lungs. Both, MC geometry model and reconstruction algorithm were validated by means of on-field measurements and data acquisition in a clinical center. Dosimetric and imaging performances were evaluated through Quality Assurance phantoms (Computed Tomography Dose Index and Catphan, respectively). Results indicate that the best implementation of the reconstruction algorithm was achieved with 80 keV, using the Hanning filter and linear interpolation. More specifically, for a spherical lung lesion with a radius of 7 mm a 30% CNR gain was found when the number of projections varied from 12 to 36 (corresponding to a dose increase of a factor of 3). This research suggests the possibility of developing a CBCT modulated beam scanner for simultaneous breast and lung imaging while ensuring dose reduction. However further investigation regarding the number of projections needed for image reconstruction is required.

Tese mestrado integrado, Engenharia Biomédica e Biofísica (Engenharia Clínica e Instrumentação Médica) Universidade de Lisboa, Faculdade de Ciências, 2022
A Radioterapia é uma das modalidades principais para tratamentos de foro oncológico que visa destruir a ação proliferativa das células cancerígenas e reduzir o volume tumoral. A sua ação terapêutica através do uso de radiação ionizante tem, subjacente, a máxima de irradiar o tumor com uma elevada dose, ao mesmo tempo que os órgãos de risco (OARs) adjacentes, são tanto quanto possível protegidos. Quando um tumor se localiza no pulmão ou abdómen superior, como no fígado ou pâncreas, o seu movimento devido à respiração pode alcançar até 4 cm, especialmente na direção crânio-caudal, aumentando as incertezas relativas à posição do tumor. No Centro Clínico Champalimaud (CCC), o planeamento convencional dos tratamentos de radioterapia faz uso de uma tomografia computadorizada (CT) que é adquirida aquando da respiração livre do doente e que, por isso, apresenta geralmente artefactos que podem ser uma fonte de erro durante o planeamento. Nos casos em que o movimento do tumor é considerável, é ainda adquirida uma tomografia computadorizada quadrimensional (4DCT) que consiste entre 8 e 10 CTs que representam fases do ciclo respiratório. Posteriormente, a 4DCT é utilizada para delinear o volume interno do alvo (ITV) que engloba toda a extensão do movimento do tumor. Apesar da estratégia do ITV garantir uma adequada cobertura do volume-alvo, os OARs ficam expostos a doses de radiação desnecessárias e a um maior risco de toxicidade. Este efeito é ainda mais preocupante em tratamentos hipofracionados, onde doses mais elevadas são administradas num número reduzido de frações. Nos últimos anos têm sido desenvolvidas estratégias que visam tornar os tratamentos de radioterapia mais eficazes. Uma delas é a reconstrução de uma CT que representa a posição média do doente ao longo do ciclo respiratório (Mid-P CT). Esta estratégia resulta em volumes de tratamento menores do que a estratégia do ITV, possibilitando o aumento da dose e maior controlo tumoral local. O primeiro passo para a reconstrução do Mid-P CT é o registo deformável de imagens (DIR) entre uma das fases da respiração (uma CT da 4DCT), definida como a fase de referência, e as restantes fases. Deste processo resultam campos vetoriais deformáveis (DVF) que contém informação do deslocamento dos tecidos. Os DVFs são subsequentemente utilizados para transformar cada uma das fases da respiração para a posição média. O método do Mid-P foi implementado com sucesso no Instituto do Cancro Holandês (NKI) em 2008. Apesar dos bons resultados clínicos, o número de centros de radioterapia que utiliza esta técnica é muito reduzido. Tal deve-se, por um lado, à inexistência de soluções comerciais com esta funcionalidade e, por outro, ao esforço necessário alocar para implementar e validar soluções desenvolvidas internamente. O presente projeto teve como principal objetivo implementar a estratégia do Mid-P no CCC (Portugal). Para tal, foi otimizado um módulo – RunMidP – desenvolvido para o software 3D Slicer, que calcula o Mid-P CT e estima a amplitude do movimento do tumor e OARs com base nos DVFs. Considerando que a precisão do módulo e a qualidade de imagem do Mid-P CT devem atender os requisitos para o planeamento em radioterapia, foram realizados testes para validar o módulo. Sempre que possível, a sua performance foi comparada com outras aplicações desenvolvidas para a implementação da técnica do Mid-P, nomeadamente com um protótipo desenvolvido pela empresa Mirada Medical Ltd. (Reino Unido) – Mirada – e com o software desenvolvido no NKI (Holanda) – Wimp. Os testes foram divididos em três estudos diferentes, cada um com um conjunto de dados diferente. No primeiro estudo (estudo A), foram utilizadas 4DCT de 2 fantomas digitais, cuja função respiratória e cardíaca foi modelada de forma simplificada, e de 18 doentes com tumores localizados no pulmão (N = 8), no fígado (N = 6) e no pâncreas (N = 4). Neste estudo, foram comparados dois algoritmos DIR disponíveis no software 3D Slicer, o Plastimatch e o Elastix, em termos da precisão do registo e da qualidade de imagem do Mid-P CT reconstruído. Foi ainda avaliado a capacidade dos softwares RunMidP e Mirada representarem corretamente a posição média do doente e as diferenças das amplitudes do movimento do tumor estimadas pelos dois softwares. No estudo B, foram realizados testes de verificação semelhantes aos supre mencionados, em imagens sintéticas provenientes de 16 doentes, desta vez com a vantagem de se conhecer o “verdadeiro” Mid-P CT e as “verdadeiras” amplitudes do movimento do tumor. Estes foram comparados com os resultados obtidos com os softwares RunMidP e Mirada. Ainda, as unidades de Hounsfield (HU) no Mid-P CT reconstruído por RunMidP e Mirada foram comparadas com as HU na fase de referência, de modo a verificar se os Mid P CTs produziriam diferenças dosimétricas relevantes. No último estudo (estudo C), a qualidade de imagem do Mid-P CT foi avaliada quantitativamente e qualitativamente. Durante a análise qualitativa, foi pedido a dois médicos especialistas que avaliassem a viabilidade dos Mid-P CTs, reconstruídos pelos três softwares (RunMidP, Mirada e Wimp), para o planeamento dos tratamentos. O tempo da reconstrução do Mid-P CT a partir da 4DCT foi de cerca de 1h. Ambos os algoritmos, Plastimach e Elastix, demonstraram ser adequados para DIR de imagens do pulmão e abdómen superior, com diferenças estatisticamente não significativas (p > 0.05) em termos da precisão do registo. Contudo, o Mid-P CT reconstruído com Elastix apresentou uma melhoria na qualidade de imagem, sendo assim o algoritmo DIR escolhido para ser implementado no RunMidP. Em termos de métricas aplicadas a contornos definidos manualmente, tais como a distância de Hausdorf (HD) e coeficiente de Dice (DSC), o erro do registo de imagem foi menor que 1 mm, dentro do contorno do tumor, e 2 mm no pulmão. Os Mid-P CTs reconstruídos com o RunMidP e Mirada apresentaram maiores diferenças, relativamente ao “verdadeiro” Mid-P CT, na região do diafragma e zonas de maior homogeneidade como, por exemplo, no ar presente no intestino. Contudo, para a maioria dos doentes do estudo B, o Mid-P CT reconstruído com o software Mirada apresentou maior índice de similaridade estrutural (SSIM) relativamente ao “verdadeiro” Mid-P CT. Estes resultados podem estar na origem do uso de diferentes algoritmos DIR, mas deveram-se principalmente a uma falha na aplicação das transformações deformáveis pelo módulo RunMiP que foi corrigida posteriormente. Ainda, as diferenças entre as amplitudes estimadas e previstas foram menores que 1 mm para 37 tumores (78,9%), que resultam em diferenças menores que 0.3mm quando convertidas em margens de planeamento. Para além disso, as diferenças nos valores de HU dos Mid-P CTs comparativamente à fase de referência foram, em média, de 1 HU no tumor e OARs. Foram também observadas melhorias na qualidade de imagem do Mid-P CT, nomeadamente um aumento da relação sinal-ruído (SNR) e diminuição dos artefactos. Estes resultados estão de acordo com a avaliação dos médicos que, em geral, consideraram que os Mid-P CTs reconstruídos pelos três softwares são adequados para o planeamento dos tratamentos. No entanto, os Mid-P CTs reconstruídos com dados 4DCT provenientes do CCC apresentaram classificações inferiores aos reconstruídos com dados 4DCT do NKI. Em suma, as modificações do algoritmo DIR Plastimach para Elastix e a correção do método para aplicar as transformações deformáveis, permitiram uma melhoria na qualidade de imagem do Mid P CT e melhor performance do algoritmo, respetivamente. O módulo RunMidP, neste projeto otimizado e validado, apresenta um forte potencial para a reconstrução e implementação da estratégia do Mid-P na clínica, com performance comparável a outras aplicações existentes (Mirada e Wimp). Atenção especial deve ser dada aos dados 4DCT de input que parecem afetar a qualidade de imagem final do Mid-P CT. No futuro, valerá a pena otimizar os parâmetros de aquisição e reconstrução da 4DCT de modo a melhorar a qualidade de imagem e, ainda, o módulo RunMidP pode potencialmente ser otimizado no que respeita ao tempo de reconstrução do Mid-P CT e à precisão do DIR.
Radiotherapy for tumours in the thorax and upper abdomen is challenging since they move notably with breathing. To cover the whole extent of tumour motion, relatively large margins are added to treatment volumes, posing a higher risk of toxicity for surrounding organs-at-risk (OARs). The Mid Position (Mid-P) method accounts for breathing motion by using deformable image registration (DIR) to transform all phases of a 4DCT scan to a time-weighted average 3DCT scan (Mid-P CT). The Mid-P strategy results in smaller treatment volumes, potentially boosting the delivery of hypofractionated treatments. To bring the Mid-P approach to the Champalimaud Clinical Centre (CCC), an in-house Mid position software module – RunMidP – was optimized. The module reconstructs the Mid-P CT and estimates breathing motion amplitudes of tumours and relevant OARs. In addition, this project presents a set of experiments to evaluate the performance of the Mid-P method and its feasibility for clinical implementation. The experiments were conducted throughout three different studies using 4DCT data from 18 phantoms and 23 patients. In Study A, the accuracy and image quality of two DIR algorithms (Plastimatch and Elastix) were assessed using quantitative metrics applied on either warped images or manually delineated contours. The reproduction of the patient’s mean position by the Mid-P CT and the estimation of motion amplitudes were compared to a soon-to-be Mid-P commercial software developed by Mirada Medical Ltd. In Study B,similar experiments were performed, this time using a more rigorous reference – “true” Mid-P CT scans and “true” motion estimations. In Study C, the image quality of Mid P CT scans was assessed quantitatively and qualitatively. Both Plastimatch and Elastix registration showed comparable registration accuracy, although Elastix showed superior image quality of reconstructed Mid-P CTs. Based on contour metrics, the registration error was less than 2 mm. In-house Mid-P CTs showed a slightly lower match to ground truth Mid-P CTs than the ones reconstructed by the Mirada prototype due to differences in DIR methods and small shifts to the original image geometry. Higher image differences were found in the diaphragm lung interface, where the patient's anatomy moves faster due to breathing, and in homogeneous regions such as the air regions in the bowel. On the other hand, differences (estimated-predicted) in motion amplitudes smaller than 1 mm were observed in 37 moving tumours (78.7%), showing a good performance of the Mid-P algorithm. Regarding the image quality, improvements in the signal-to-noise ratio and removal of image artefacts in Mid-P CTs are great advantages for using them as the planning CT. Clinicians also gave a good assessment of the suitability of Mid-P CT scans for treatment planning. No significant differences were found in the performance of the RunMidP compared to other Mid-Position packages, although worse scores were given to the CCC dataset than the dataset from another hospital. The in-house Mid-position algorithm shows promising results regarding the use of the software module in radiotherapy for lung and upper abdomen cancer. Further exploration must be given to improve the registration accuracy, image quality of the input data, and speed up the reconstruction of the Mid-P CT scan.