Auswahl der wissenschaftlichen Literatur zum Thema „Histopathological tumor segmentation“

Tumor segmentation is a fundamental task in histopathological image analysis. Creating accurate pixel-wise annotations for such segmentation tasks in a fully-supervised training framework requires significant effort. To reduce the burden of manual annotation, we propose a novel weakly supervised segmentation framework based on sparse patch annotation, i.e., only small portions of patches in an image are labeled as ‘tumor’ or ‘normal’. The framework consists of a patch-wise segmentation model called PSeger, and an innovative semi-supervised algorithm. PSeger has two branches for patch classification and image classification, respectively. This two-branch structure enables the model to learn more general features and thus reduce the risk of overfitting when learning sparsely annotated data. We incorporate the idea of consistency learning and self-training into the semi-supervised training strategy to take advantage of the unlabeled images. Trained on the BCSS dataset with only 25% of the images labeled (five patches for each labeled image), our proposed method achieved competitive performance compared to the fully supervised pixel-wise segmentation models. Experiments demonstrate that the proposed solution has the potential to reduce the burden of labeling histopathological images.

(1) Background: Histopathological assessment of Wilms tumors (WT) is crucial for risk group classification to guide postoperative stratification in chemotherapy pre-treated WT cases. However, due to the heterogeneous nature of the tumor, significant interobserver variation between pathologists in WT diagnosis has been observed, potentially leading to misclassification and suboptimal treatment. We investigated whether artificial intelligence (AI) can contribute to accurate and reproducible histopathological assessment of WT through recognition of individual histopathological tumor components. (2) Methods: We assessed the performance of a deep learning-based AI system in quantifying WT components in hematoxylin and eosin-stained slides by calculating the Sørensen–Dice coefficient for fifteen predefined renal tissue components, including six tumor-related components. We trained the AI system using multiclass annotations from 72 whole-slide images of patients diagnosed with WT. (3) Results: The overall Dice coefficient for all fifteen tissue components was 0.85 and for the six tumor-related components was 0.79. Tumor segmentation worked best to reliably identify necrosis (Dice coefficient 0.98) and blastema (Dice coefficient 0.82). (4) Conclusions: Accurate histopathological classification of WT may be feasible using a digital pathology-based AI system in a national cohort of WT patients.

In recent years, improved deep learning techniques have been applied to biomedical image processing for the classification and segmentation of different tumors based on magnetic resonance imaging (MRI) and histopathological imaging (H&E) clinical information. Deep Convolutional Neural Networks (DCNNs) architectures include tens to hundreds of processing layers that can extract multiple levels of features in image-based data, which would be otherwise very difficult and time-consuming to be recognized and extracted by experts for classification of tumors into different tumor types, as well as segmentation of tumor images. This article summarizes the latest studies of deep learning techniques applied to three different kinds of brain cancer medical images (histology, magnetic resonance, and computed tomography) and highlights current challenges in the field for the broader applicability of DCNN in personalized brain cancer care by focusing on two main applications of DCNNs: classification and segmentation of brain cancer tumors images.

Perineural invasion (PNI), a sign of poor diagnosis and tumor metastasis, is common in a variety of malignant tumors. The infiltrating patterns and morphologies of tumors vary by organ and histological diversity, making PNI detection difficult in biopsy, which must be performed manually by pathologists. As the diameters of PNI nerves are measured on a millimeter scale, the PNI region is extremely small compared to the whole pathological image. In this study, an efficient deep learning-based method is proposed for detecting PNI regions in multiple types of cancers using only PNI annotations without detailed segmentation maps for each nerve and tumor cells obtained by pathologists. The key idea of the proposed method is to train the adopted deep learning model, U-Net, to capture the boundary regions where two features coexist. A boundary dilation method and a loss combination technique are proposed to improve the detection performance of PNI without requiring full segmentation maps. Experiments were conducted with various combinations of boundary dilation widths and loss functions. It is confirmed that the proposed method effectively improves PNI detection performance from 0.188 to 0.275. Additional experiments were also performed on normal nerve detection to validate the applicability of the proposed method to the general boundary detection tasks. The experimental results demonstrate that the proposed method is also effective for general tasks, and it improved nerve detection performance from 0.511 to 0.693.

The segmentation and classification of cell nuclei are pivotal steps in the pipelines for the analysis of bioimages. Deep learning (DL) approaches are leading the digital pathology field in the context of nuclei detection and classification. Nevertheless, the features that are exploited by DL models to make their predictions are difficult to interpret, hindering the deployment of such methods in clinical practice. On the other hand, pathomic features can be linked to an easier description of the characteristics exploited by the classifiers for making the final predictions. Thus, in this work, we developed an explainable computer-aided diagnosis (CAD) system that can be used to support pathologists in the evaluation of tumor cellularity in breast histopathological slides. In particular, we compared an end-to-end DL approach that exploits the Mask R-CNN instance segmentation architecture with a two steps pipeline, where the features are extracted while considering the morphological and textural characteristics of the cell nuclei. Classifiers that are based on support vector machines and artificial neural networks are trained on top of these features in order to discriminate between tumor and non-tumor nuclei. Afterwards, the SHAP (Shapley additive explanations) explainable artificial intelligence technique was employed to perform a feature importance analysis, which led to an understanding of the features processed by the machine learning models for making their decisions. An expert pathologist validated the employed feature set, corroborating the clinical usability of the model. Even though the models resulting from the two-stage pipeline are slightly less accurate than those of the end-to-end approach, the interpretability of their features is clearer and may help build trust for pathologists to adopt artificial intelligence-based CAD systems in their clinical workflow. To further show the validity of the proposed approach, it has been tested on an external validation dataset, which was collected from IRCCS Istituto Tumori “Giovanni Paolo II” and made publicly available to ease research concerning the quantification of tumor cellularity.

Background: Lung cancer is a fatal disease that kills approximately 85% of those diagnosed with it. In recent years, advances in medical imaging have greatly improved the acquisition, storage, and visualization of various pathologies, making it a necessary component in medicine today. Objective: Develop a computer-aided diagnostic system to detect lung cancer early by segmenting tumor and non-tumor tissue on Tissue Micro Array Analysis (TMA) histopathological images. Method: The prototype computer-aided diagnostic system was developed to segment tumor areas, non-tumor areas, and fundus on TMA histopathological images. Results: The system achieved an average accuracy of 83.4% and an F-measurement of 84.4% in segmenting tumor and non-tumor tissue. Conclusion: The computer-aided diagnostic system provides a second diagnostic opinion to specialists, allowing for more precise diagnoses and more appropriate treatments for lung cancer.

Oral squamous cell carcinoma is most frequent histological neoplasm of head and neck cancers, and although it is localized in a region that is accessible to see and can be detected very early, this usually does not occur. The standard procedure for the diagnosis of oral cancer is based on histopathological examination, however, the main problem in this kind of procedure is tumor heterogeneity where a subjective component of the examination could directly impact patient-specific treatment intervention. For this reason, artificial intelligence (AI) algorithms are widely used as computational aid in the diagnosis for classification and segmentation of tumors, in order to reduce inter- and intra-observer variability. In this research, a two-stage AI-based system for automatic multiclass grading (the first stage) and segmentation of the epithelial and stromal tissue (the second stage) from oral histopathological images is proposed in order to assist the clinician in oral squamous cell carcinoma diagnosis. The integration of Xception and SWT resulted in the highest classification value of 0.963 (σ = 0.042) AUCmacro and 0.966 (σ = 0.027) AUCmicro while using DeepLabv3+ along with Xception_65 as backbone and data preprocessing, semantic segmentation prediction resulted in 0.878 (σ = 0.027) mIOU and 0.955 (σ = 0.014) F1 score. Obtained results reveal that the proposed AI-based system has great potential in the diagnosis of OSCC.

We present a novel method to assess the variations in protein expression and spatial heterogeneity of tumor biopsies with application in computational pathology. This was done using different antigen stains for each tissue section and proceeding with a complex image registration followed by a final step of color segmentation to detect the exact location of the proteins of interest. For proper assessment, the registration needs to be highly accurate for the careful study of the antigen patterns. However, accurate registration of histopathological images comes with three main problems: the high amount of artifacts due to the complex biopsy preparation, the size of the images, and the complexity of the local morphology. Our method manages to achieve an accurate registration of the tissue cuts and segmentation of the positive antigen areas.

Abstract Increasing implementation of whole slide image (WSI) and advances in computing capacity enable the use of artificial intelligence (AI) in pathology, such as quantification of biomarkers, aids in diagnosis and detection of lymph node metastasis. However, predicting therapy response in cancer patients from pre-treatment histopathologic images remains a challenging task, limited by poor understanding of tumor immune microenvironment. In this study, we aimed to develop AI models using multi-source histopathologic images to predict neoadjuvant chemotherapy (NAC) response in HER2-positive (HER2+) breast cancers. First, pretreatment tumor tissues were stained with Hematoxylin and Eosin (H&E) and multiplex immunohistochemistry (IHC) including tumor immune microenvironment markers (PD-L1: immune checkpoint protein; CD8: marker for cytotoxic T-cells; and CD163: marker for type 2 macrophages). Next, we developed an AI-based pipeline to automatically extract histopathologic features from H&E and multiplex IHC WSIs. The pipeline included: A) H&E tissue segmentation based on DeepLabV3 model to generate stroma, tumor, and lymphocyte-rich regions. B) IHC marker segmentation to segment CD8, CD163, and PD-L1 stained cells. C) H&E and IHC non-rigid registration to match H&E and IHC images since they were stained from different levels of tissue. D) Image-based registration and segmentation histopathologic feature construction. A total of 36 histopathological features were constructed to represent tumor immune microenvironment characteristics such as ratios of PD-L1, CD8 and CD163 in tumoral, stromal or lymphocyte-rich regions. They were used to train machine learning (ML) models to predict NAC response in a training dataset with 62 HER2+ breast cancers (38 with complete and 24 with incomplete response). The ML model using logistic regression demonstrated the best performance with an area under curve (AUC) of 0.8975. We also tested ML models using pathologists-derived histopathologic features, but the best performed model showed an AUC of 0.7880. Finally, the developed logistic regression ML model was tested on an external validation dataset with 20 HER2+ breast cancers (10 with complete and 10 with incomplete response) and yielded an AUC of 0.9005. In summary, we described an automatic, accurate and interpretable AI-based pipeline to extract histopathologic features from H&E and IHC WSIs and then used these features to develop machine learning model to accurately predict NAC response in HER2+ breast cancers. The ML model using AI-based extracted features outperformed the model using features manually generated by pathologists. Citation Format: Zhi Huang, Anil V. Parwani, Kun Huang, Zaibo Li. Developing artificial intelligence algorithms to predict response to neoadjuvant chemotherapy in HER2-positive breast cancer. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5436.

Summary Introduction: The standard treatment for locally advanced rectal cancer (RC) consists of neoadjuvant chemoradiation followed by radical surgery. Regardless the extensive use of SUVmax in 18F-FDG PET tumor uptake as representation of tumor glycolytic consumption, there is a trend to apply metabolic volume instead. Thus, the aim of the present study was to evaluate a noninvasive method for tumor segmentation using the 18F-FDG PET imaging in order to predict response to neoadjuvant chemoradiation therapy in patients with rectal cancer. Method: The sample consisted of stage II and III rectal cancer patients undergoing 18F-FDG PET/CT examination before and eight weeks after neoadjuvant therapy. An individualized tumor segmentation methodology was applied to generate tumor volumes (SUV2SD) and compare with standard SUVmax and fixed threshold (SUV40%, SUV50% and SUV60%) pre- and post-therapy. Therapeutic response was assessed in the resected specimens using Dworak's protocol recommendations. Several variables were generated and compared with the histopathological results. Results: Seventeen (17) patients were included and analyzed. Significant differences were observed between responders (Dworak 3 and 4) and non-responders for SUVmax-2 (p<0.01), SUV2SD-2 (p<0.05), SUV40%-2 (p<0.05), SUV50%-2 (p<0.05) and SUV60%-2 (p<0.05). ROC analyses showed significant areas under the curve (p<0.01) for the proposed methodology with sensitivity and specificity varying from 60% to 83% and 73% to 82%, respectively. Conclusion: The present study confirmed the predictive power of the variables using a noninvasive individualized methodology for tumor segmentation based on 18F-FDG PET/CT imaging for response evaluation in patients with rectal cancer after neoadjuvant chemoradiation therapy.

L’anatomopathologie est la discipline médicale responsable du diagnostic et de la caractérisation des maladies par inspection macroscopique, microscopique, moléculaire et immunologique des tissus. Les technologies modernes permettent de numériser des lames tissulaire en images numériques qui peuvent être traitées par l’intelligence artificielle pour démultiplier les capacités des pathologistes. Cette thèse a présenté plusieurs approches nouvelles et puissantes qui s’attaquent à la segmentation et à la classification pan-cancer des images de lames numériques. L’apprentissage de modèles de segmentation pour des lames numériques est compliqué à cause de difficultés d’obtention d’annotations qui découlent (i) d’une pénurie de pathologistes, (ii) d’un processus d’annotation ennuyeux, et (iii) de différences majeurs entre les annotations inter-pathologistes. Mon premier axe de travail a abordé la segmentation des tumeurs pan-cancéreuses en concevant deux nouvelles approches d’entraînement faiblement supervisé qui exploitent des annotations à l’échelle de la lame qui sont faciles et rapides à obtenir. En particulier, ma deuxième contribution à la segmentation était un algorithme générique et très puissant qui exploite les annotations de pourcentages de tumeur pour chaque lame, sans recourir à des annotations de pixels. De vastes expériences à grande échelle ont montré la supériorité de mes approches par rapport aux méthodes faiblement supervisées et supervisées pour la segmentation des tumeurs pan-cancer sur un ensemble de données de plus de 15 000 lames de tissus congelés. Mes résultats ont également démontré la robustesse de nos approches au bruit et aux biais systémiques dans les annotations. Les lames numériques sont difficiles à classer en raison de leurs tailles colossales, qui vont de millions de pixels à plusieurs milliards de pixels, avec un poids souvent supérieur à 500 mégaoctets. L’utilisation directe de la vision par ordinateur traditionnelle n’est donc pas possible, incitant l’utilisation de l’apprentissage par instances multiples, un paradigme d’apprentissage automatique consistant à assimiler une lame comme un ensemble de tuiles uniformément échantillonnés à partir de cette dernière. Jusqu’à mes travaux, la grande majorité des approches d’apprentissage à instances multiples considéraient les tuiles comme échantillonnées de manière indépendante et identique, c’est-à-dire qu’elles ne prenaient pas en compte la relation spatiale des tuiles extraites d’une image de lame numérique. Certaines approches ont exploité une telle interconnexion spatiale en tirant parti de modèles basés sur des graphes, bien que le véritable domaine des lames numériques soit spécifiquement le domaine de l’image qui est plus adapté aux réseaux de neurones convolutifs. J’ai conçu un cadre d’apprentissage à instances multiples puissant et modulaire qui exploite la relation spatiale des tuiles extraites d’une lame numérique en créant une carte clairsemée des projections multidimensionnelles de patches, qui est ensuite traitée en projection de lame numérique par un réseau convolutif à entrée clairsemée, avant d’être classée par un modèle générique de classification. J’ai effectué des expériences approfondies sur trois tâches de classification d’images de lames numériques, dont la tâche par excellence du cancérologue de soustypage des tumeurs, sur un ensemble de données de plus de 20 000 images de lames numériques provenant de données publiques. Les résultats ont mis en évidence la supériorité de mon approche vis-à-vis les méthodes d’apprentissage à instances multiples les plus répandues. De plus, alors que mes expériences n’ont étudié mon approche qu’avec des réseaux de neurones convolutifs à faible entrée avec deux couches convolutives, les résultats ont montré que mon approche fonctionne mieux à mesure que le nombre de paramètres augmente, suggérant que des réseaux de neurones convolutifs plus sophistiqués peuvent facilement obtenir des résultats su
Anatomic pathology is the medical discipline responsible for the diagnosis and characterization of diseases through the macroscopic, microscopic, molecular and immunologic inspection of tissues. Modern technologies have made possible the digitization of tissue glass slides into whole slide images, which can themselves be processed by artificial intelligence to enhance the capabilities of pathologists. This thesis presented several novel and powerful approaches that tackle pan-cancer segmentation and classification of whole slide images. Learning segmentation models for whole slide images is challenged by an annotation bottleneck which arises from (i) a shortage of pathologists, (ii) an intense cumbersomeness and boring annotation process, and (iii) major inter-annotators discrepancy. My first line of work tackled pan-cancer tumor segmentation by designing two novel state-of-the-art weakly supervised approaches that exploit slide-level annotations that are fast and easy to obtain. In particular, my second segmentation contribution was a generic and highly powerful algorithm that leverages percentage annotations on a slide basis, without needing any pixelbased annotation. Extensive large-scale experiments showed the superiority of my approaches over weakly supervised and supervised methods for pan-cancer tumor segmentation on a dataset of more than 15,000 unfiltered and extremely challenging whole slide images from snap-frozen tissues. My results indicated the robustness of my approaches to noise and systemic biases in annotations. Digital slides are difficult to classify due to their colossal sizes, which range from millions of pixels to billions of pixels, often weighing more than 500 megabytes. The straightforward use of traditional computer vision is therefore not possible, prompting the use of multiple instance learning, a machine learning paradigm consisting in assimilating a whole slide image as a set of patches uniformly sampled from it. Up to my works, the greater majority of multiple instance learning approaches considered patches as independently and identically sampled, i.e. discarded the spatial relationship of patches extracted from a whole slide image. Some approaches exploited such spatial interconnection by leveraging graph-based models, although the true domain of whole slide images is specifically the image domain which is more suited with convolutional neural networks. I designed a highly powerful and modular multiple instance learning framework that leverages the spatial relationship of patches extracted from a whole slide image by building a sparse map from the patches embeddings, which is then further processed into a whole slide image embedding by a sparse-input convolutional neural network, before being classified by a generic classifier model. My framework essentially bridges the gap between multiple instance learning, and fully convolutional classification. I performed extensive experiments on three whole slide image classification tasks, including the golden task of cancer pathologist of subtyping tumors, on a dataset of more than 20,000 whole slide images from public data. Results highlighted the superiority of my approach over all other widespread multiple instance learning methods. Furthermore, while my experiments only investigated my approach with sparse-input convolutional neural networks with two convolutional layers, the results showed that my framework works better as the number of parameters increases, suggesting that more sophisticated convolutional neural networks can easily obtain superior results

Oral cancer (OC) is among the top ten cancers worlwide, with more than 90% being squamous cell carcinoma. Despite diagnostic and therapeutic development in OC patients’ mortality and morbidity rates remain high with no advancement in the last 50 years. Development of diagnostic tools in identifying pre-cancer lesions and detecting early-stage OC might contribute to minimal invasive treatment/surgery therapy, improving prognosis and survival rates, and maintaining a high quality of life of patients. For this reason, Artificial Intelligence (AI) algorithms are widely used as a computational aid in tumor classification and segmentation to help clinicians in the earlier discovery of cancer and better monitoring of oral lesions. In this paper, we propose an AI-based system for automatic segmentation of the epithelial and stromal tissue from oral histopathological images in order to assist clinicians in discovering new informative features. In terms of semantic segmentation, the proposed AI system based on preprocessing methods and deep convolutional neural networks produced satisfactory results, with 0.878 ± 0.027 mIOU and 0.955 ± 0.014 F1 score. The obtained results show that the proposed AI-based system has a great potential in diagnosing oral squamous cell carcinoma, therefore, this paper is the first step towards analysing the tumor microenvironment, specifically segmentation of the microenvironment cells.