Dissertations / Theses on the topic 'Breast cancer prediction'

A major issue in the management of breast cancer is to identify patients who are less likely to be cured after primary treatment and would benefit from adjuvant chemotherapy. Of great importance is also identification of patients with only local disease who traditionally would be given chemotherapy but would survive without. In this thesis we have validated the utility of the two-gene ratio HOXB13:IL17BR, which previously has been demonstrated to predict disease-free survival in tamoxifen-treated breast cancer patients. We have also studied the prognostic and predictive utility of a single gene as a biomarker in breast cancer medicine. We could confirm that HOXB13:IL17BR may classify patients with different treatment benefit; only patients with a low value showed benefit from prolonged duration of tamoxifen therapy, whereas for the group with high ratios, the long-term recurrence rate did not improve with longer treatment duration. The combination of HOXB13:IL17BR and the molecular grade index (MGI), another prognostic marker, has been shown to outperform either alone in predicting risk of breast cancer recurrence. We validated the prognostic utility of HOXB13:IL17BR+MGI in a large randomized patient cohort and found that this risk classification identified more than 50% of the tamoxifen-treated lymph node-negative patients as having a less than 3% risk of distant recurrence and breast cancer death. Furthermore, we developed and tested a continuous risk model of HOXB13:IL17BR+MGI called Breast Cancer Index (BCI), for estimation of recurrence risk at the individual level. Our study shows that BCI has the ability to identify more than 50% of patients with a low risk of recurrence more accurately than using traditional risk assessment. These results suggest that BCI may help clinicians to make better informed treatment decisions and spare toxic chemotherapy for a large group of breast cancer patients. The protein expression of HOXB13 was also shown to be a valuable predictor in postmenopausal patients. High expression was associated with worse outcome after tamoxifen therapy. In a premenopausal cohort, patients with hormone receptor-positive tumors showed benefit from tamoxifen regardless of HOXB13 expression. Further analysis indicated that estrogen receptor β (ERβ) modified the performance of HOXB13 as a predictor of treatment effect and should be taken into account when identifying patients less likely to respond to the therapy given. In conclusion, BCI identifies patients with a very low risk of distant recurrence. It may be utilized in the management of breast cancer patients to optimize the use of chemotherapy. HOXB13 protein expression may be used as a marker for tamoxifen benefit, but its performance in premenopausal patients might be modified by ERβ.

Thesis (Ph. D.)--University of North Carolina at Chapel Hill, 2006.
Title from electronic title page (viewed Oct. 10, 2007). "... in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Curriculum in Genetics and Molecular Biology." Discipline: Genetics and Molecular Biology; Department/School: Medicine.

Axillary lymph node (ALN) metastasis status is an important prognostic marker in breast cancer and is widely employed for tumour staging and defining an adjuvant therapy. In an attempt to avoid invasive procedures which are currently employed for the diagnosis of nodal metastasis, several markers have been identified and tested for the prediction of ALN metastasis status in recent years. However, the nonlinear and complex relationship between these markers and nodal status has inhibited the effectiveness of conventional statistical methods as classification tools for diagnosing metastasis to ALNs. The aim of this study is to propose a reliable artificial neural network (ANN) based decision support system for ALN metastasis status prediction. ANNs have been chosen in this study for their special characteristics including nonlinear modelling, robustness to inter-class variability and having adaptable weights which makes them suitable for data driven analysis without making any prior assumptions about the underlying data distributions. To achieve this aim, the probabilistic neural network (PNN) evaluated with the .632 bootstrap is investigated and proposed as an effective and reliable tool for prediction of ALN metastasis. For this purpose, results are compared with the multilayer perceptron (MLP) neural network and two network evaluation methods: holdout and cross validation (CV). A set of six markers have been identified and analysed in detail for this purpose. These markers include tumour size, oestrogen receptor (ER), progesterone receptor (PR), p53, Ki-67 and age. The outcome of each patient is defined as metastasis or non-metastasis, diagnosed by surgery. This study makes three contributions: firstly it suggests the application of the PNN as a classifier for predicting the ALN metastasis, secondly it proposes a the .632 bootstrap evaluation of the ANN outcome, as a reliable tool for the purpose of ALN status prediction, and thirdly it proposes a novel set of markers for accurately predicting the state of nodal metastasis in breast cancer. Results reveal that PNN provides better sensitivity, specificity and accuracy in most marker combinations compared to MLP. The results of evaluation methods’ comparison demonstrate the high variability and the existence of outliers when using the holdout and 5-fold CV methods. This variability is reduced when using the .632 bootstrap. The best prediction accuracy, obtained by combining ER, p53, Ki-67 and age was 69% while tumour size and p53 were the most significant individual markers. The classification accuracy of this panel of markers emphasises their potential for predicting nodal spread in individual patients. This approach could significantly reduce the need for invasive procedures, and reduce post-operative stress and morbidity. Moreover, it can reduce the time lag between investigation and decision making in patient management.

Introducció: S'ha demostrat que el cribratge mamogràfic redueix la mortalitat per càncer de mama. Seguint les recomanacions de la Comissió Europea, els països europeus han establert programes poblacionals de cribratge que ofereixen mamografies biennals a dones d'entre 50 i 69 anys d'edat. No obstant això, el cribratge de càncer de mama no està lliure de controvèrsia ja que existeix un debat en relació a l'equilibri entre la reducció de la mortalitat i els efectes adversos. Per a millorar aquest equilibri, l'evidència científica actual dóna suport al cribratge personalitzat. Els estudis de modelització han demostrat que modificar l'interval de cribratge, la prova de cribratge o el rang d'edat de la població objectiu en funció del risc individual de les dones produeix un major benefici que les estratègies convencionals. Per tant, és necessari ampliar la informació actual sobre els factors de risc d'aquesta malaltia i crear models de predicció del risc individual mitjançant l'anàlisi de grans bases de dades poblacionals. Objectiu: L'objectiu general d'aquesta tesi és aprofundir en l'anàlisi del cribratge poblacional del càncer de mama. En concret, aquesta tesi pretén avaluar diferents factors de risc de càncer de mama per a desenvolupar i validar un model de predicció de risc individual d'aquesta malaltia. Es va analitzar com la densitat mamària afecta als diferents indicadors del cribratge en el context de la mamografia digital. A continuació, es van avaluar les diferències en el risc de càncer de mama en funció de si una lesió benigna de mama es va diagnosticar en un cribratge prevalent o un cribratge incident. També es va analitzar la interacció entre la densitat mamària i les lesions benignes en el risc de desenvolupar càncer de mama. Posteriorment, es va realitzar una revisió sistemàtica per a actualitzar l'evidència existent, dur a terme una valoració crítica i una avaluació del risc de biaix i resumir els resultats dels models de risc individualitzats que s'utilitzen per a estimar el risc de càncer de mama en les dones de la població general. Finalment, es va dissenyar un model de predicció individual del risc de càncer de mama i es va validar internament, a partir d'informació fàcilment accessible en un episodi de cribratge. Conclusions: i) Els diferents indicadors de cribratge es veuen afectats negativament per la densitat mamària, disminuint la sensibilitat i el valor predictiu positiu de la prova a mesura que augmenta la densitat mamària. ii) El risc de càncer de mama conferit per una lesió benigna difereix segons la mena de cribratge (prevalent o incident). Fins on sabem, aquest és el primer estudi que analitza l'impacte del tipus de cribratge en el pronòstic de la lesió benigna. iii) El risc de càncer de mama augmenta de manera independent amb la presència d'una lesió benigna i amb una major densitat mamària i es manté elevat durant més de 15 anys. iv) Els models de predicció són eines prometedores per a implementar polítiques de cribratge basades en el risc individualitzat. No obstant això, és un repte recomanar qualsevol d'ells per a la personalització del cribratge ja que necessiten millorar la seva qualitat i capacitat discriminatòria. v) Es va dissenyar i validar internament un model de predicció de risc capaç d'estimar el risc de càncer de mama a curt i llarg termini utilitzant la informació recollida de manera rutinària en el cribratge mamogràfic. El model inclou edat, antecedents familiars de càncer de mama, antecedents de lesió benigna i patrons mamogràfics previs, que van resultar estar relacionats amb un augment del risc de càncer de mama. El model ha de ser validat externament i actualitzat amb noves variables.
Introducción: Se ha demostrado que el cribado mamográfico reduce la mortalidad por cáncer de mama. Siguiendo las recomendaciones de la Comisión Europea, los países europeos han establecido programas poblacionales de cribado que ofrecen mamografías bienales a mujeres de entre 50 y 69 años de edad. Sin embargo, el cribado de cáncer de mama no está libre de controversia ya que existe un debate en cuanto al equilibrio entre la reducción de la mortalidad y los efectos adversos. Para mejorar este equilibrio, la evidencia científica actual apoya el cribado personalizado. Los estudios de modelización han demostrado que modificar el intervalo de cribado, la prueba de cribado o el rango de edad de la población objetivo en función del riesgo individual de las mujeres produce un mayor beneficio que las estrategias convencionales. Por lo tanto, es necesario ampliar la información actual sobre los factores de riesgo de esta enfermedad y crear modelos de predicción del riesgo individual mediante el análisis de grandes bases de datos poblacionales. Objetivo: El objetivo general de esta tesis es profundizar en el análisis del cribado poblacional del cáncer de mama. En concreto, esta tesis pretende evaluar diferentes factores de riesgo de cáncer de mama para desarrollar y validar un modelo de predicción de riesgo individual de esta enfermedad. Se analizó cómo la densidad mamaria afecta a los distintos indicadores de cribado en el contexto de la mamografía digital. A continuación, se evaluaron las diferencias en el riesgo de cáncer de mama en función de si una lesión benigna de mama se diagnosticó en un cribado prevalente o un cribado incidente. También se analizó la interacción entre la densidad mamaria y las lesiones benignas en el riesgo de cáncer de mama. Posteriormente, se realizó una revisión sistemática para actualizar la evidencia existente, llevar a cabo una valoración crítica y una evaluación del riesgo de sesgo y resumir los resultados de los modelos de riesgo individualizados que se utilizan para estimar el riesgo de cáncer de mama en las mujeres de la población general. Por último, se diseñó un modelo de predicción individual del riesgo de cáncer de mama y se validó internamente, basado en información fácilmente accesible en un episodio de cribado. Conclusiones: i) Los distintos indicadores de cribado se ven afectados negativamente por la densidad mamaria, disminuyendo la sensibilidad y el valor predictivo positivo de la prueba a medida que aumenta la densidad mamaria. ii) El riesgo de cáncer de mama conferido por una lesión benigna difiere según el tipo de cribado (prevalente o incidente). Hasta donde sabemos, este es el primer estudio que analiza el impacto del tipo de cribado en el pronóstico de la lesión benigna. iii) El riesgo de cáncer de mama aumenta de forma independiente con la presencia de una lesión benigna y con una mayor densidad mamaria y se mantiene elevado durante más de 15 años. iv) Los modelos de predicción son herramientas prometedoras para implementar políticas de cribado basadas en el riesgo individualizado. Sin embargo, es un reto recomendar cualquiera de ellos para la personalización del cribado ya que necesitan mejorar su calidad y capacidad discriminatoria. v) Diseñamos y validamos internamente un modelo de predicción de riesgo capaz de estimar el riesgo de cáncer de mama a corto y largo plazo utilizando la información recogida de forma rutinaria en el cribado mamográfico. El modelo incluye edad, antecedentes familiares de cáncer de mama, antecedentes de lesión benigna y patrones mamográficos previos, que resultaron estar relacionados con un aumento del riesgo de cáncer de mama. El modelo debe ser validado externamente y actualizado con nuevas variables.
Background: Mammographic screening has been shown to reduce mortality from breast cancer. Following the recommendations of the European Council, European countries have started population-based screening programs that offer biennial mammograms to women aged between 50 and 69 years. The results of the effectiveness of population-based screening are controversial in terms of the balance between mortality reduction and adverse effects. To improve this balance, current evidence supports personalized screening. Modeling studies have shown that modifying the screening interval, screening modality, or age range of the target population based on women's individual risk yields a greater benefit than conventional standard strategies. Several risk models have been designed to estimate women's individual breast cancer risk based on their personal characteristics. However, most of these models have not been specifically developed to estimate the risk of women targeted for breast cancer screening. There is therefore a need to broaden current information on risk factors for breast cancer and the estimation of individual risk prediction models through the analysis of large population-based databases. Aims: The general objective of the thesis is to deepen the analysis of population-based breast cancer screening. Specifically, the aim of this thesis is to assess different breast cancer risk factors in order to develop and validate an individualized breast cancer risk prediction model. We evaluated how breast density affects screening performance indicators in a digital mammography context. Then, we assessed differences in breast cancer risk across benign breast disease diagnosed at prevalent or incident screens. To our knowledge, this is the first time that such an approach has been used. We also evaluated the interaction between breast density and benign breast disease. Subsequently, we performed a systematic review to update the existing evidence, conduct a critical appraisal and risk of bias assessment and summarize the results of the individualized risk models that are used to estimate the risk of breast cancer in women in the general population. Finally, a breast cancer risk prediction model was designed and internally validated, based on information easily accessible at screening. Conclusions: i) Performance screening measures are negatively affected by breast density, with sensitivity and positive predictive value decreasing as breast density increases. ii) The risk of breast cancer conferred by benign breast disease differed according to type of screen (prevalent or incident). To our knowledge, this is the first study to analyze the impact of screening type on the prognosis of benign breast disease. iii) The risk of breast cancer independently increased with the presence of benign breast disease and with greater breast density and remained elevated for over 15 years. iv) Individualized risk prediction models are promising tools for implementing risk-based screening policies. However, it is a challenge to recommend any of them since they need further improvement in their quality and discriminatory capacity. v) We designed and internally validated a risk prediction model able to estimate short- and long-term breast cancer risk using information routinely reported at screening participation. The model included age, family history of breast cancer, benign breast disease and previous mammographic findings, which were found to be related to an increase in breast cancer risk. The model should be externally validated and updated with new variables.
Universitat Autònoma de Barcelona. Programa de Doctorat en Metodologia de la Recerca Biomèdica i Salut Pública

Different forms of cancer have been widely studied and documented in various studies across the world. However, there have not been many similar studies in the developing countries - particularly those on the African continent (Parkin, et al., 2005). This thesis seeks to uncover the geo-demographic occurrence patterns of the disease by applying three Data mining Techniques, namely Logistic Regression (LR), Neural Networks (NNs) and Decision Trees (DTs), to learn the underlying rules in the overall behaviour of breast cancer. The data, 3,057 observations on 29 variables obtained from four cancer treatment centres in Libya (2004-2008), were interrogated using multiple K-folds cross validation. The predictive strategy yielded a list of breast cancer predictor factors ordered according to their importance in predicting the disease. Comparison between our results and those obtainable from conventional LR, NN and DT models shows that our strategy out-performs the conventional variable selection. It is expected that the findings from this thesis will provide an input into comparative geo-ethnic studies of cancer and provide informed intervention guidelines in the prevention and cure of the disease, not only in Libya but also in other parts of the world.

L'objectif de cette thèse est d'étudier la sensibilité des lignes cellulaires du cancer du sein BTCL aux agents stabilisants des microtubules (taxanes et ixabépilone) afin de: 1 - identifier la pharmaco-génomique prédictif de la réponse (résistance / sensibilité) comme une signature moleculaire, et de valider cette signature sur d'autres études dont les données génomiques sont disponibles en ligne, donc mis l'expression des gènes prédictifs de GES pour Tax- sensibilité (333 gènes ) et Ixa-sensibilité (79 gènes) ont été définis, et les Taxanes prédicateurs GES a considérablement prédit Pac-sensibilité dans BTCL, et pathologiques réponse complète à base de Pac-chimiothérapie néoadjuvante chez les patients du cancer du sein. 2 - étudier le rôle des cellules souches du cancer (ALDH +) sur la réponse thérapeutique aux Taxanes et donc, Nous identifions quatre lignes BTCL qui présentent un enrichissement significative dans le pourcentage et le nombre absolu de ALDELFUOR cellules positives dans chacun de ces quatre BTCLs après 5 jours de traitement par le paclitaxel, en contraste avec les résultats précédents, nous avons constaté que dans ces autres 3 BTCLs le phénomène est inversé avec la diminution significative du pourcentage et le nombre absolu de cellules positives ALDELFUOR trouve dans chacun de ces trois BTCLs après 5 jours du traitement par le paclitaxel. Une signature moléculaire de SCC résistant / sensible de 243 pb avec 179 gènes dont 152 gènes sont régulés à la hausse et 27 gènes régulés à la baisse au CSC résistantes au paclitaxel, une sorte prédicteurs génomiques pour Tax - sensibilité au CSC résistantes au paclitaxel peut être dérivée à partir BTCL et peut être utile pour mieux comprendre les mécanismes de résistance aux taxanes et de l'implication de la CSC dans cette résistance, afin de mieux sélectionner des traitements cytotoxiques chez les patients du cancer du sein et l'identification des d'autres marqueurs potentiels de thérapies ciblées dans l'avenir. 3 - Nous avons testé l'impact de l'altération des paramètres génomiques et protéiques ou les mutations de certains gènes comme tau (MAPT), K-alpha tubuline (TUB A1B) tubuline alpha-6 (A1C TUB) tubuline beta 3 (TUBB3) et stathmine (STMN1), malheureusement nous n'avions jamais identifier une mutation pour être corrélée à la réponse des BTCL aux Taxanes. 4 - Nous essayons d'étudier au niveau de protéines par immunohistochimie sur le tissu de micro-array et cyto-micro-array pour certains paramètres qui ont été déjà prouvé (in vitro) pour être corrélée à la réponse aux Taxanes, (cette partie est en fait en cours)
The aim of this thesis is to study the sensitivity of breast cancer cell lines BTCL to microtubule-stabilizing agents (Taxanes and ixabepilone) in order to:1- identify pharmaco-genomic predictor of response (resistance /sensitivity) as a molecular signature, and to validate this signature on others studies of which the genomic data are available on line, so gene expression set GES predictors for Tax-sensitivity (333 genes) and Ixa-sensitivity (79 genes) were defined, and the Taxanes GES predictors has significantly predicted Pac-sensitivity in BTCL, and pathological complete response to Pac-based neo-adjuvant chemotherapy in BC patients.2- study the role of cancer stem cell (ALDH+) on the therapeutic response to Taxanes and their we identify 4 BTCLs which present a significant enrichment in the percentage and the absolute numbers of ALDELFUOR-positive cells found in each of these 4 BTCLs after 5 days of treatment by Paclitaxel , In contrast to the previous results we found that in others 3 BTCLs these phenomenon is inversed with the significant decrease of the percentage, and the absolute numbers of ALDELFUOR-positive cells found in each of these 3 BTCLs after 5 days of treatment by Paclitaxel.A molecular signature of CSC resistant /sensitive of 243 pb with 179 genes of which 152 genes are up- regulated and 27 genes down-regulated in CSC resistant to Paclitaxel, so a genomic predictors for Tax-sensitivity in CSC resistant to Paclitaxel can be derived from BTCL and may be helpful for better understanding the mechanisms of resistance to Taxanes and the implication of CSC in this resistance in order to better select of cytotoxic treatment in breast cancer patients and identification of others potential markers for targeted therapies in the future .3- we tested the impact of the alteration of genomic and proteic parameters or the mutations of some genes like tau (MAPT),Tubulin K- ALPHA (TUB A1B) Tubulin alpha-6 (TUB A1C) Tubulin beta 3 (TUBB3) and Stathmin (STMN1), unfortunately we did'nt identify a mutations to be correlated to BTCL response to Taxanes .4- we try to study at the level of proteins by immunohistochemistry on the tissue micro- array and cyto-micro-array for some parameters which have been already proved (in vitro) to be correlated with response to Taxanes , ( this part is actually ongoing)

This research study investigated the association of microRNA related single nucleotide polymorphisms (miRSNPs) with breast cancer susceptibility in Australian Caucasian women. The thesis then progressed with developing an in silico methodology for miRNA-target identification followed by the validation of miRNA-target relationships regarding the distinctive molecular subtypes of human breast cancers.

This thesis focuses on the modelling and prediction of factors affecting attendance to screening invitations of the NHS Breast Screening Programme. The analysis is based on data collected by the Warwickshire, Solihull and Coventry Breast Screening Unit from 1989 up to 2001 with respect to invitation to screening for the prevention of breast cancer in non-symptomatic women. Using a novel approach to the analysis of the data, from the perspective of the screening episode of each woman, rather than the usual analysis from the perspective of the screening round of the units, a statistical analysis is carried out on the whole registered population for the first time. Amendments to the current formulae for coverage calculations, the introduction of a new parameter (invitation rate) and the proposal for a reduction of the invitation period (period of time between two consecutive invitations) follows from the analysis. A preliminary analysis of predictive methodologies, including traditional statistical methods and artificial intelligent methods, gives the foundation to the formulation of two new algorithms; the first, for the prediction of attendance of women to screening invitations, and the second for the prediction of occurrence of screening variation (change of appointment dates) of women to invitations. Both algorithms are based on neural network generated models able to learn from the previous screening behaviour history of the woman, a technique not previously explored for the prediction of attendance. The accuracy of the new proposed algorithm for the prediction of attendance to invitation is tested on a blind study using data not previously seen by the predictive system, and for which results were unknown at the time when the predictions were made. From the obtained results, it is concluded to recommend the implementation by the NHS Breast Screening Unit of the two algorithms proposed for the prediction of the women’s attendance and screening variation to their invitation for screening. With these predictions, women likely not to attend, or change appointment date, can be identified and appropriately targeted with the aim of increasing their attendance in the short term, and in the long term, reducing breast cancer mortality.

The growth and development of the breast is to a large extent regulated by oestrogens through the oestrogen receptor (ER). Activation of the ERα triggers transcription of genes that are important for cell proliferation and stimulates entry into the G1 phase of the cell cycle. In breast cancer the ERα is often upregulated and is therefore a suitable target for adjuvant therapies such as tamoxifen. Although tamoxifen is an effective treatment in most cases, tumours sometimes acquire resistance to the drug. The aim of this thesis was to investigate the impact of G1 phase proteins and 11q13 amplification on prognosis and treatment response in breast cancer. The material used was from a clinical trial in which postmenopausal breast cancer patients were randomised to chemotherapy or radiotherapy and tamoxifen or no adjuvant treatment. We studied the expression of cyclin D1, cyclin E and Rb with immunohisochemistry and amplification of CCND1 and PAK1 with real time PCR. We found that among patients with high tumour expression of cyclin D1, overexpression of ErbB2 was associated with reduced recurrence-free survival. Both cyclin D1 and cyclin E overexpression were associated with reduced tamoxifen response. High expression of cyclin D1 has been found to induce ligand independent activation of ERα in breast cancer cells and might also switch tamoxifen from acting as an antagonist to an agonist. Overexpression of cyclin E has been shown to be associated with expression of low molecular weight isoforms of the protein that possess an increased kinase activity and are insensitive to p21 and p27 inhibition. Furthermore, amplification of 11q13, and in particular the gene PAK1, was a strong predictor of tamoxifen resistance. The pak1 protein is involved in phosphorylation and ligand independent activation of the ERα. We also found that lost expression of either p53 or Rb reduced the patients benefit from radiotherapy compared with patients with normal expression of both proteins. Normally, ionizing radiation upregulates p53 resulting in G1 arrest or apoptosis. If either functional p53 or Rb is missing the cells can proceed from G1 to the S phase despite damaged DNA. The expression of the microRNA, miR-206, was analysed with real time PCR, and the results showed that high expression of miR-206 correlated to low expression of ERα and 11q13 amplification. In vitro studies have shown that miR-206 negatively regulates the expression of ERα. Taken together the G1 regulators and amplification of 11q13 seem to have an important role in predicting the patient’s response to adjuvant therapy.

This study was undertaken to evaluate some important prognostic factors in human breast cancer. The prognostic value of accepted clinicopathological factors such as the presence of axillary lymph node métastasés, histologic grade, clinical and pathological stage was confirmed. In a cohort of stage T3,T4,M0 breast cancer with 91 patients (paper I) DNA ploidy by static cytometry (SCM) turned out to be the most important prognostic factor. In a cohort of stage T2,M0 breast cancer with 99 patients (paper III) the presence of involved axillary nodes and low histologic grade were independent prognostic factors. According to life-table analyses DNA ploidy by flow cytometty (FCM) and SCM were significant prognostic predictors for survival but S-phase fraction (SPF) was not. The significant discrimination between euploid and aneuploid tumours was seen also among the node-negative patients. In a patient material with 158 tumours of predominantly low stages (73% T0,T1, papers IV and V) and calculated mammographie tumour volume doubling time (DT) DNA ploidy by FCM gave no significant prognostic information. A computer program was used to calculate SPF from the histograms obtained by FCM. SPF with a cut-off value of 7.5% between tumours with high and low proliferation rate was a highly significant and independent prognostic factor for survival. The other independent prognostic predictors were low histologic grade, the presence of involved axillary nodes and stage II and III (versus stage I). DT values for 158 patients (papers IV and V) varied between 0.6 and 65.8 months (mean 10.9 months) and 11 tumours showed no growth at all between mammographies. The median value of 9.0 months was chosen as cut-off point between slow and fast growing tumours. The prognostic power of DT was however low, and the difference between slow and fast growing tumours was significant only for distant disease-free survival. Seventy-one of the 158 tumours were detected by mammographie screening. The screening detected carcinomas with predominantly long DT:s were discovered at an early stage and showed favourable characteristics concerning DNA ploidy and SPF. FCM was a rapid and reliable method for DNA analysis with a better prognostic discrimination between euploid and aneuploid groups than SCM (papers II and III). SPF, DNA ploidy and histologic grade are significantly correlated to one another but show no strong correlation to the presence of axillary lymph node métastasés. There is also a significant correlation between DT on one hand and DNA ploidy and SPF on the other hand. In conclusion the classic prognostic factors are still valuable. DNA ploidy as a single prognostic factor seems to have a relatively low prognostic power and seems to be of limited clinical value. SPF is a highly significant prognostic predictor for breast cancer of low stage, but the clinical value is not defined.

S. 3-38: sammanfattning, s. 39-94: 5 uppsatser

digitalisering@umu

Pour les patients diagnostiqués avec un cancer au stade précoce, les décisions de traitement dépendent de l’évaluation du risque de rechute métastatique. Les outils de pronostic actuels sont fondés sur des approches purement statistiques, sans intégrer les connaissances disponibles sur les processus biologiques à l’oeuvre. L’objectif de cette thèse est de développer des modèles prédictifs du processus métastatique en utilisant une approche de modélisation mécaniste et la modélisation à effets mixtes. Dans la première partie, nous étendons un modèle mathématique du processus métastatique pour décrire la croissance de la tumeur primaire et de la masse métastatique totale chez des souris traitées avec le sunitinib (un inhibiteur de tyrosine kinase ayant une action anti-angiogénique) administré comme traitement néoadjuvant (i.e. avant exérèse de la tumeur primaire). Le modèle est utilisé pour tester des hypothèses expliquant les effets différentiels du sunitinib sur la tumeur primaire et les métastases. Des algorithmes d’apprentissage statistique sont utilisés pour évaluer la valeur prédictive des biomarqueurs sur les paramètres du modèle.Dans la deuxième partie de cette thèse, nous développons un modèle mécaniste pour la prédiction du temps de rechute métastatique et le validons sur des données cliniques des patientes atteintes d’un cancer du sein localisé. Ce modèle offre des prédictions personnalisées des métastases invisibles au moment du diagnostic, ainsi que des simulations de la croissance métastatique future, et il pourrait être utilisé comme un outil de prédiction individuelle pour aider à la gestion des patientes atteintes de cancer du sein
For patients diagnosed with early-stage cancer, treatment decisions depend on the evaluation of the risk of metastatic relapse. Current prognostic tools are based on purely statistical approaches that relate predictor variables to the outcome, without integrating any available knowledge of the underlying biological processes. The purpose of this thesis is to develop predictive models of the metastatic process using an established mechanistic modelling approach and the statistical mixed-effects modelling framework.In the first part, we extend the mathematical metastatic model to describe primary tumour and metastatic dynamics in response to neoadjuvant sunitinib in clinically relevant mouse models of spontaneous metastatic breast and kidney cancers. The calibrated model is then used to test possible hypothesis for the differential effects of sunitinib on primary tumour and metastases, and machine learning algorithms are applied to assess the predictive power of biomarkers on the model parameters.In the second part of this thesis, we develop a mechanistic model for the prediction of the time to metastatic relapse and validate it on a clinical dataset of breast cancer patients. This model offers personalised predictions of the invisible metastatic burden at the time of diagnosis, as well as forward simulations of metastatic growth, and it could be used as a personalised prediction tool to assist in the routine management of breast cancer patients

Data sets with missing values are a pervasive problem within medical research. Building lifetime prediction models based solely upon complete-case data can bias the results, so imputation is preferred over listwise deletion. In this thesis, artificial neural networks (ANNs) are used as a prediction model on simulated data with which to compare various imputation approaches. The construction and optimization of ANNs is discussed in detail, and some guidelines are presented for activation functions, number of hidden layers and other tunable parameters. For the simulated data, binary lifetime prediction at five years was examined. The ANNs here performed best with tanh activation, binary cross-entropy loss with softmax output and three hidden layers of between 15 and 25 nodes. The imputation methods examined are random, mean, missing forest, multivariate imputation by chained equations (MICE), pooled MICE with imputed target and pooled MICE with non-imputed target. Random and mean imputation performed poorly compared to the others and were used as a baseline comparison case. The other algorithms all performed well up to 50% missingness. There were no statistical differences between these methods below 30% missingness, however missing forest had the best performance above this amount. It is therefore the recommendation of this thesis that the missing forest algorithm is used to impute missing data when constructing ANNs to predict breast cancer patient survival at the five-year mark.

[Truncated abstract] For this thesis, the molecular biology of cancer was approached from two directions. Firstly, an investigation was conducted on the role of growth factor receptor-bound protein 7 (Grb7) in breast cancer. Grb7 is an adapter molecule that binds to a variety of proteins, including the growth factor receptor and proto-oncogene, ErbB2, and mediates signalling to downstream pathways. It has been linked to cell migration and an invasive phenotype, and is of interest as a therapeutic target. To investigate the role of Grb7 in breast cancer, preliminary experiments were performed that, firstly, determined the expression of wild-type Grb7 and a splice variant, Grb7V, in a range of cell lines, and secondly, aided the development of a protocol for treating cells with short interfering RNA (siRNA) against Grb7 and the ErbB ligand, heregulin (HRG), in a cell system appropriate for measuring the functional outcomes. Using this protocol in conjunction with CellTitre (CT) proliferation assays, it was demonstrated that Grb7 does not play a role in the proliferation of either unstimulated or HRG-stimulated SK-BR-3 breast cancer cells. Furthermore, using the protocol in conjunction with Boyden chamber migration assays, it was shown that inhibition of Grb7 expression has a slight stimulatory effect on HRG-stimulated SK-BR-3 cell migration. Thus, Grb7 was found to play only a minor role in the migration of SK-BR-3 cells, suggesting that it is not an ideal anti-cancer target for breast cancers modelled by this cell system. Concurrently, a second investigation was conducted, which similarly sought insight into the molecular biology of cancer, but adopted a more strategic approach. ... These results provide evidence for a biologically significant role for the miR-7-mediated regulation of EGFR expression. A microarray experiment was also performed to identify genes that were down-regulated following treatment with miR-7 compared to NS precursor. Of 248 down-regulated genes, including EGFR, 37 promising new miR-7 target candidates were identified. Functional clustering of down-regulated genes and promising target candidates suggested that miR-7 may have functionally-related targets involved in processes including cell motility and brain-associated functions. This investigation thus yielded a program capable of accurately predicting a miRNA target not predicted by any other target prediction program, verified a previously unknown miRNA:target interaction with functional consequences in cancer cells and provided the first steps towards investigating miR-7-mediated regulation in greater depth. Furthermore, EGFR was, to our knowledge, the first example of a verified miRNA target with target sites that are not conserved across mammals, an observation with important implications for computational target prediction and the evolution of miRNA regulatory systems. In addition, the demonstrated growth inhibitory and cytotoxic effects of miR-7 on lung cancer cells raise the possibility of a miR-7-based therapeutic for the treatment of EGFR-overexpressing tumours.

Mammographic density, the amount of radiodense tissue on a mammogram, is a strong risk factor for breast cancer, with properties that could be an asset in screening and prevention programmes. Its use in risk prediction contexts is currently limited, however, mainly due to di culties in measuring and interpreting density. This research investigates rstly, the properties of density as an independent marker of breast cancer risk and secondly, how density should be measured. The rst question was addressed by analysing data from a chemoprevention trial, a trial of hormonal treatment, and a cohort study of women with a family history of breast cancer . Tamoxifen-induced density reduction was observed to be a good predictor of breast cancer risk reduction in high-risk una ected subjects. Density and its changes did not predict risk or treatment outcome in subjects with a primary invasive breast tumour. Finally absolute density predicted risk better than percent density and showed a potential to improve existing risk-prediction models, even in a population at enhanced familial risk of breast cancer. The second part of thesis focuses on density measurement and in particular evaluates two fully-automated volumetric methods, Quantra and Volpara. These two methods are highly correlated and in both cases absolute density (cm3) discriminated cases from controls better than percent density. Finally, we evaluated and compared di erent measurement methods. Our ndings suggested good reliability of the Cumulus and visual assessments. Quantra volumetric estimates appeared negligibly a ected by measurement error, but were less variable than visual bi-dimensional ones, a ecting their ability to discriminate cases from controls. Overall, visual assessments showed the strongest association with breast cancer risk in comparison to computerised methods. Our research supports the hypothesis that density should have a role in personalising screening programs and risk management. Volumetric density measuring methods, though promising, could be improved.

Сучасні дослідження доводять особливу функціональну роль ММП-9 у створенні ніш у місцях віддаленого метастазування. Клінічні дослідження останніх років спрямовані на застосування ММП як маркерів прогнозування захворювання. Метою дослідження було провести кореляцію між рівнем експресії МПП-2, -9 із випадками інвазивної протокової карциноми грудної залози в групах з наявними метастазами та без них. На основі отриманих даних встановити кількісні показники для прогнозування метастазування

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2015-03-24T12:29:18Z No. of bitstreams: 2 license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Tese - Augusto Ribeiro Gabriel - 2014.pdf: 1977753 bytes, checksum: 126a201173259e6ab5741dfeadcb8ba9 (MD5)
Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2015-03-24T14:07:53Z (GMT) No. of bitstreams: 2 license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Tese - Augusto Ribeiro Gabriel - 2014.pdf: 1977753 bytes, checksum: 126a201173259e6ab5741dfeadcb8ba9 (MD5)
Made available in DSpace on 2015-03-24T14:07:53Z (GMT). No. of bitstreams: 2 license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Tese - Augusto Ribeiro Gabriel - 2014.pdf: 1977753 bytes, checksum: 126a201173259e6ab5741dfeadcb8ba9 (MD5) Previous issue date: 2014-08-05
With the exception of skin cancer, breast cancer remains the most common malignant neoplasm affecting women both in Brazil and worldwide, inflicting severe economic, social and emotional consequences on patients and their families. Despite advances made in diagnostic and therapeutic techniques over recent decades, mortality rates from breast cancer remain expressive. To be able to treat tumors appropriately, not only profound knowledge of the cell mechanisms involved in their genesis, but also knowledge of the mechanisms involved in the success or failure of treatment is crucial. Various methods have been developed for this purpose, including the evaluation of biological tumor markers and the genetic studies. The objective of the present study was to evaluate some biological markers involved in the differentiation and evolution of breast cancer, using immunohistochemistry on tissue arrays. A retrospective study was conducted between 2006 and 2012 in which clinical data were obtained from patient charts, and tissue samples conserved in paraffin blocks were prospectively analyzed and correlated with each other and with the patient’s response to neoadjuvant chemotherapy. The results were presented in two papers. In the first article, biomarker expression was evaluated in biopsy specimens obtained at diagnosis and then following treatment with adjuvant chemotherapy, with correlations being drawn between the differences found. Statistically significant differences were found in Ki-67, IGF-1, topoisomerase II-alpha and CK5/6 marker expression, indicating the effect of chemotherapy on the proliferation index of the malignant breast tumors. On the other hand, no statistically significant differences were found in HER2, estrogen and progesterone receptors, PTEN or EGFR. In the second paper, correlations were sought between biological marker expression and the patient’s outcome response to previous chemotherapy, with results showing significant correlations between the HER2 and topoisomerase II-alpha markers and pathologic complete response despite the fact that the sample was small. No other statistically significant correlations were found with any of the other markers evaluated. When molecular subtypes were analyzed, the study showed a greater frequency of pathologic complete response for the HER2 subtype and this difference was statistically significant. Another important result was the correlation between the tendency towards a reduction in mean Ki-67 values and a clinical benefit from the treatment implemented a finding that led to the preparation of a third paper, which consisted of an integrative review of the Ki-67 marker. This review concluded that further studies need to be conducted on the Ki-67 marker and that its expression should be analyzed dynamically to establish whether a correlation exists between this marker and patients’ prognosis and whether Ki-67 is a predictor of treatment response.
Tanto no Brasil quanto no mundo, excluindo-se o câncer de pele, o câncer de mama ainda é a neoplasia maligna que mais acomete as mulheres, trazendo prejuízo econômico, social e emocional para elas e suas famílias. Apesar dos avanços diagnósticos e terapêuticos observados nas últimas décadas, o câncer de mama ainda carrega taxas de mortalidade expressivas. Para que se possa tratar de forma adequada os tumores, é imprescindível o conhecimento profundo dos mecanismos celulares envolvidos na sua gênese, bem como dos mecanismos envolvidos no sucesso ou fracasso do tratamento. Vários métodos foram desenvolvidos neste sentido, como o estudo de marcadores biológicos dos tumores e estudos genéticos. O presente estudo teve como proposta avaliar alguns marcadores biológicos envolvidos na diferenciação e evolução do câncer de mama através da técnica de imunohistoquímica em amostras preparadas em matrizes de arranjo teciduais. Realizou-se um estudo retrospectivo no período compreendido entre 2006 e 2012, quando foram obtidos dados clínicos de prontuários e amostras de tecidos conservados em blocos de parafina, prospectivamente analisados e correlacionados entre si e com os desfechos de resposta à quimioterapia neoadjuvante. Os resultados foram apresentados em dois artigos. No primeiro artigo avaliou-se a expressão dos marcadores nas biópsias quando da realização do diagnóstico e após o tratamento com a quimioterapia adjuvante, correlacionando-se as diferenças encontradas. Diferenças de expressão dos marcadores Ki67, IGF-1, Topoisomerase II-alfa e CK5/6 foram observadas, com significado estatístico indicando o efeito da quimioterapia no índice de proliferação dos tumores malignos de mama. Por outro lado, os marcadores HER2, receptores de estrógenos, receptores de progesterona, PTEN e EGFR não apresentaram diferenças significativas. No segundo artigo, correlacionou-se a expressão dos marcadores biológicos com os desfechos de resposta à quimioterapia prévia e concluiu-se que, embora em uma amostra pequena, os marcadores HER2 e Topoisomerase II-alfa apresentaram correlação significativa com a resposta patológica completa, o que não aconteceu com os demais marcadores. Analisando subtipos moleculares este estudo evidenciou, de forma estatisticamente significativa, maior frequência de resposta patológica completa para o subtipo HER2. Outro achado importante foi evidenciado na correlação entre a tendência na redução da média dos valores de Ki67 e o benefício clínico do tratamento realizado, fato este que levou à elaboração do terceiro artigo, uma revisão integrativa acerca do marcador Ki67. Esta revisão permitiu concluir que o marcador em análise ainda precisa ser objeto de estudos e que sua expressão deve ser analisada de forma dinâmica, para avaliar se a mesma correlaciona-se com o prognóstico das pacientes e a predição de resposta ao tratamento.

Contexte : Le cancer est la seconde cause de mortalité chez la femme en France, et la première chez les femmes âgées de 35 à 84 ans. Le cancer du sein est le cancer le plus fréquemment diagnostiqué, représentant 35 % des cas chez les femmes en France en 2012. De multiples facteurs, comportementaux et non-Comportementaux, augmentant le risque de cancer, tant en incidence qu’en mortalité, ont été identifiés dans la littérature, tandis que leur influence conjointe est très peu évaluée. Dans le cas du cancer du sein, certains facteurs diffèrent selon le statut ménopausique des femmes, suggérant une étiologie différente entre les cancers du sein diagnostiqués avant et après la ménopause. Objectif : Les données de la cohorte prospective française E3N ont été utilisées pour évaluer l’influence des facteurs comportementaux et non-Comportementaux sur le risque de cancer et de mortalité chez les femmes avant et après la ménopause. Nous avons également cherché à estimer leur impact relatif sur la population et à identifier les facteurs à forts pouvoirs prédictifs.Résultats : Nos résultats suggèrent que le mode de vie a une influence modeste sur le risque de cancer et de mortalité lors de l’adhésion à une seule recommandation de santé publique. En revanche, elle est conséquente lors d’une adhésion conjointe à plusieurs recommandations. Les facteurs comportementaux jouent ainsi un rôle non négligeable dans la survenue de cancer et sur le risque de décès prématuré. Dans le cas du cancer du sein, ces facteurs influencent particulièrement le risque après la ménopause, tandis qu’avant la ménopause leur impact est plus faible que les facteurs qui ne relèvent pas du mode de vie ou de choix personnels. Ces observations sont retrouvées lorsque l’on cherche à prédire le risque de cancer du sein avant et après la ménopause. En effet, la prédiction du risque de cancer du sein en préménopause s’établit principalement à partir de facteurs non-Comportementaux, alors que la prédiction du risque en postménopause est également déterminée par des facteurs comportementaux.Conclusion : Nous avons montré que l’étiologie du cancer du sein diffère selon la nature de la tumeur, et en particulier selon le statut ménopausique des femmes. À tout âge, le mode de vie a une influence sur le risque de cancer et de mortalité prématurée, particulièrement après la ménopause lorsque leur impact est supérieur à celui des facteurs non-Comportementaux. Ces résultats demandent, cependant, à être reproduits dans des études prospectives portant sur des femmes plus jeunes
Background: Cancer is the second leading cause of mortality among women in France, and the leading cause of mortality among women aged between 35 and 84. Breast cancer is the most frequently diagnosed cancer, with 35% of cases among women in France in 2012. Multiple behavioural and non-Behavioural factors have been associated with increases in cancer incidence and mortality. However, the literature about their combined impact is scarce. Regarding breast cancer, some risk factors differed according to the menopausal status, suggesting a different etiology between premenopausal and postmenopausal breast cancers.Objectives: Data from the E3N prospective cohort of French women were used to evaluate the influence of behavioural and non-Behavioural factors on cancer risk before and after the menopause and on mortality. In addition, we aimed at estimating their relative impact on the population and identifying factors with the highest predictive power.Results: Our results suggest a modest influence of the lifestyle on cancer risk and mortality when adhering to only one public health recommendation. However, the influence is substantial with a combined adherence to several recommendations. Behavioural factors play a key role in the occurrence of cancer and mortality risk. Regarding breast cancer, these factors influence particularly the risk after the menopause, while before, their impact is lower than non-Behavioural factors. These observations were retrieved when aiming at predicting breast cancer risk according to menopausal status. Prediction was established by non-Behavioural factors in premenopause, while the prediction in postmenopause was driven by behavioural factors.Conclusion: We have shown that the etiology of breast cancer differs according to the nature of the tumour, and particularly according to the menopausal status of women. Whatever the age, lifestyle influence the risk of cancer and mortality, especially after the menopause when their impact is higher than the non-Behavioural factors’ one. New results from prospective study on younger women are warranted to confirm the results

Thesis (M.S.)--West Virginia University, 2008.
Title from document title page. Document formatted into pages; contains viii, 85 p. : col. ill. Includes abstract. Includes bibliographical references (p. 81-85).

Les résultats des projets actuels apportent une information, sur différents aspects des rôles de la voie PI3K, dans le développement du cancer du sein, et la réponse au traitement. Les projets particuliers couvrent des sujets liés à la voie aux niveaux concernant les récepteurs de la famille HER, activant la voie PI3K, ainsi que PI3K et les effecteurs en découlant. Les effets pronostic et prédictif de la dérégulation de PI3K sont les sujets centraux de la recherche décrite ici. Une baisse d’expression de PI3KR1 est associée à une survie réduite dans notre cohorte de patients. Une attention particulière a été portée aux mutations de PIK3CA communes dans le cancer du sein. Tandis que les mutations de PIK3CA agissent comme des marqueurs de bon pronostic chez les patients anti-HER2-naïfs, ces mutations agissent au contraire comme prédicteurs négatifs de la réponse au traitement par trastuzumab. Les résultats décrits mènent un peu plus vers l’implication de plusieurs voies moléculaires altérées, en particulier la voie de signalisation Wnt, dans la tumorigénèse des cancers du sein PIK3CA mutés. De plus, nous avons testé les taux de lapatinib plasmatique montrant une augmentation pertinente dans les périodes d’état d’équilibre du traitement. Par ailleurs, nous avons démontré des incohérences dans l’évaluation de l’EGFR et proposé des approches pour l’interprétation des comptages d’immunohistochimie et de FISH. Tous ces sujets sont connectés par la 170 voie PI3K, et le besoin d’approfondir les connaissances actuelles, et d’apporter de nouvelles informations utiles applicables dans le futur dans les pratiques cliniques
Results of the presented research projects bring information about several aspects of the PI3K signaling pathway roles in breast cancer development and treatment response. The particular projects covered the subjects connected with the signaling pathway, ranging from the HER family receptors activating the pathway, and PI3K to the downstream levels of signalisation. The prognostic and predictive effect of PI3K deregulation was the central subject of the described research. The decreased expression of PIK3R1 associated with reduced survival of our patients. A special focus was put on the PIK3CA mutations which are common in breast cancer. Whereas the PIK3CA mutations act as a good prognostic marker in patients non-treated with the HER2 inhibitors, these mutations predict a negative response to trastuzumab treatment. The described results, furthermore, draw attention to the role of several altered molecular signaling pathways in breast cancer development, especially to the Wnt signaling pathway. The lapatinib plasma levels showing the relevant increase in comparison with the already described efficient steady-state levels were also described in one of the projects. Moreover, various modifications to EGFR status assessment were compared and showed that EGFR FISH and IHC count interpretation depended significantly on method and thresholds used. All these subjects are connected by the PI3K pathway, the need to deepen current knowledge and bring new useful information applicable in future clinical practice

In 2009, 713,220 new cases of cancer were diagnosed for women in the United States with more than a quarter million deaths. African American women and lesbians exhibit behavioral risk factors as well as diminished access to and utilization of breast cancer screening that reduces opportunities for early detection. This secondary analysis of a national convenience-based study examined screening compliance among 647 African American lesbian and bisexual women. Barriers to accessing screening represented the theoretical framework for this study. Bivariate chi square analysis was used to assess the association between independent variables: sociodemographic characteristics; participation in wellness activities; sexual orientation/gender identity; and experience with health care providers and the three dependent breast cancer screening compliance variables: breast self-examination (BSE), clinical breast examination (CBE), and mammography screening. Statistically significant associations between dependent and independent variables at the .05 level were further analyzed with logistic regression. Results of the ten regression models found that BSE was predicted by socioeconomic characteristics and participation in wellness activities. Compliance with CBE guidelines was predicted by sociodemographic characteristics, wellness activities, sexual orientation/gender identity and provider experience. Sociodemographic variables and provider experience also predicted mammography screening. Overall compliance was predicted by sociodemographic characteristics, namely insurance status. The social change implications of this research are an improved understanding of African American lesbian and bisexual women's screening behavior and guidance toward interventions that can improve and breast cancer screening compliance with guidelines.

The disparity between survival rates for Black and White women with breast cancer is well documented and has been examined in terms socioeconomics, environment, tumor type, and genetics. However, there is little examination of the role of health care facilities in cancer disparities. Health care facilities are representative of societal norms and beliefs that include location, quality of care, finance, policies, and staffing; therefore, they are a proxy for social justice and social change. The purpose of this study was to examine correlations between health care facility type; social determinants of cancer such as poverty, culture, and social justice; and breast cancer survival rates. Using the social determinants of cancer theoretical framework, the breast cancer survival rate of 4,087 Black and White women in Georgia between the ages of 45 and 69 was studied. The relationship between breast cancer survival and predictors including race, income, health care facility type, grade, and tumor type (4 sub-variables) were examined using the Kaplan-Meier Method, log-rank test, and Cox proportional hazard model. The log-rank test suggested no statistically significant difference in the survival functions among patients in different health care facilities (Ï?2(2) = 0.0150, p = 0.9926). The Cox proportional hazard model suggested no statistically significant relationship between breast cancer survival and health care facility type, after controlling for other predictors (Ï?2(2) = 0.3647, p = 0.8333). This result indicates that healthcare facilities do not influence breast cancer survival rates, however, given the persistent health outcome disparities further research in the area is warranted.

Skin reactions are unavoidable side effects of radiotherapy for breast cancer that may limit the amount of treatment a patient is able to receive. As well, the discomfort associated with the treatment may negatively affect the patient's quality of life and their willingness to complete a course of treatment that typically extends over seven weeks. Prior literature suggests that variations in patients' tissue reactions to radiation may be related to Individual patient characteristics. Before health care providers can intervene to prevent or minimise skin reactions, a clinical model that helps predict which patients will experience more skin reactions is needed. The purpose of the study was twofold: firstly, to test the theoretical relationships between factors that impair healing and the severity of radiation skin reactions; and secondly, to develop a model to predict the severity of radiation skin reactions in women being treated for breast cancer. The theoretical framework for the study was based on two bodies of knowledge, radiobiology and wound healing. This framework specified three sets of potential predictors of radiation induced skin reactions. These were radiation factors (e.g. dose, fractionation), genetic factors (e.g. personal and family history or cancer, radiosensitive conditions) and personal factors (e.g. age, smoking history, nutritional status). It was hypothesised that the severity of the skin reaction was a function of the relationship between these constructs. A sample of 126 women was recruited to the study over a 14-month data collection period. All the women had undergone lumpectomy and were commencing a standard radiation protocol of 45 Gray to the whole breast delivered in daily fractions of 1.8 Gray over five weeks, and a 20 Gray electron boost to the lumpectomy site delivered in daily fractions of 2 Gray over two weeks. After obtaining written informed consent, data on potential factors were collected by interview at the commencement of treatment and from the medical records. Weekly observations of the skin using the Radiation Therapy Oncology Group scoring system were recorded throughout the seven weeks of treatment. The breast was divided into eight anatomical sites to increase specificity in the final analysis. The mean inter-rater reliability of RTOG scoring between the three observers was 0.85. Chi square analysis revealed that several factors were associated with a more severe reaction. Significant factors from the "personal construct" included smoking, chemotherapy, history of skin cancer, reaction of the skin to UV radiation, lymphocele aspiration, condition of the lumpectomy scar at the commencement of treatment, weight, and the size of the breast. Stepwise logistic regression analysis revealed the relative risk and predictive value of the factors. A predictive model was developed for each of the eight anatomical Sites of the breast for weeks three to seven of radiation treatment. The principal predictors were a large breast size, smoking during the treatment period, and having had a lymphocele aspirated on at least one occasion prior to radiotherapy. The results show that it is possible to predict the severity of skin reactions in individual patients. The research contributes to theory development in radiation skin reactions and to the practice of radiation oncology nursing. Practice implications centre on individualising the preparation, education and management of women undergoing radiation therapy for breast cancer. Further research with larger samples and using different anatomical sites will contribute to the development of a skin reaction risk assessment tool for general use in radiation oncology nursing.

About one in eight women in the United States will develop breast cancer over the course of her lifetime. Moreover, patient-to-patient variability in disease progression continues to complicate clinical decisions in diagnosis and treatment for breast cancer patients. Early detection of tumors is a key factor influencing patient survival, and advancements in diagnostic and imaging techniques has allowed clinicians to spot smaller sized lesions. There has also been an increase in premature treatments of non-malignant lesions because there is no clear way to predict whether these lesions will become invasive over time. Patient variability due to genetic polymorphisms has been investigated, but studies on variability at the level of cellular activity have been extremely limited. An individual’s biochemical milieu of cytokines, growth factors, and other stimuli contain a myriad of cues that pre-condition cells and induce patient variability in response to tumor progression or treatment. Circulating white blood cells called monocytes respond to these cues and enter tissues to differentiate into monocyte-derived macrophages (MDMs) and osteoclasts that produce cysteine cathepsins, powerful extracellular matrix proteases. Cathepsins have been mechanistically linked to accelerated tumor growth and metastasis. This study aims to elucidate the variability in disease progression among patients by examining the variability of protease production from tissue-remodeling macrophages and osteoclasts. Since most extracellular cues initiate multiple signaling cascades that are interconnected and dynamic, this current study uses a systems biology approach known as cue-signal-response (CSR) paradigm to capture this complexity comprehensively. The novel and significant finding of this study is that we have identified and predicted donor-to-donor variability in disease modifying cysteine cathepsin activities in macrophages and osteoclasts. This study applied this novel finding to the context of tumor invasion and showed that variability in tumor associated macrophage cathepsin activity and their inhibitor cystatin C level mediates variability in cancer cell invasion. These findings help to provide a minimally invasive way to identify individuals with particularly high remodeling capabilities. This could be used to give insight into the risk for tumor invasion and develop a personalized therapeutic regime to maximize efficacy and chance of disease free survival.

the majority of patients with early breast cancer receive some form of systemic adjuvant therapy (chemo-, endocrine, and/or targeted therapy). Despite the increase in adjuvant therapy prescription, little progress has been made with respect to assisting oncologists to determine which breast cancer patients, particularly those deemed at “lower risk” of relapse, require chemotherapy or other systemic therapy and which women can safely be treated with loco-regional treatment alone. For these reasons, the identification of prognostic and predictive markers that will assist the clinician in selecting the most suitable form of medical therapy has become very high priority as well as a real challenge in translational research.

\
Doctorat en Sciences biomédicales et pharmaceutiques
info:eu-repo/semantics/nonPublished

Dissertação apresentada para obtenção do Grau de Doutor em Engenharia Electrotécnica e de Computadores – Sistemas Digitais e Percepcionais pela Universidade Nova de Lisboa, Faculdade de Ciências e Tecnologia
There are several clinical prognostic models in the medical field. Prior to clinical use, the outcome models of longitudinal cohort data need to undergo a multi-centre evaluation of their predictive accuracy. This thesis evaluates the possible gain in predictive accuracy in multicentre evaluation of a flexible model with Bayesian regularisation, the (PLANN-ARD), using a reference data set for breast cancer, which comprises 4016 records from patients diagnosed during 1989-93 and reported by the BCCA, Canada, with follow-up of 10 years. The method is compared with the widely used Cox regression model. Both methods were fitted to routinely acquired data from 743 patients diagnosed during 1990-94 at the Christie Hospital, UK, with follow-up of 5 years following surgery. Methodological advances developed to support the external validation of this neural network with clinical data include: imputation of missing data in both the training and validation data sets; and a prognostic index for stratification of patients into risk groups that can be extended to non-linear models. Predictive accuracy was measured empirically with a standard discrimination index, Ctd, and with a calibration measure, using the Hosmer-Lemeshow test statistic. Both Cox regression and the PLANN-ARD model are found to have similar discrimination but the neural network showed marginally better predictive accuracy over the 5-year followup period. In addition, the regularised neural network has the substantial advantage of being suited for making predictions of hazard rates and survival for individual patients. Four different approaches to stratify patients into risk groups are also proposed, each with a different foundation. While it was found that the four methodologies broadly agree, there are important differences between them. Rules sets were extracted and compared for the two stratification methods, the log-rank bootstrap and by direct application of regression trees, and with two rule extraction methodologies, OSRE and CART, respectively. In addition, widely used clinical breast cancer prognostic indexes such as the NPI, TNM and St. Gallen consensus rules, were compared with the proposed prognostic models expressed as regression trees, concluding that the suggested approaches may enhance current practice. Finally, a Web clinical decision support system is proposed for clinical oncologists and for breast cancer patients making prognostic assessments, which is tailored to the particular characteristics of the individual patient. This system comprises three different prognostic modelling methodologies: the NPI, Cox regression modelling and PLANN-ARD. For a given patient, all three models yield a generally consistent but not identical set of prognostic indices that can be analysed together in order to obtain a consensus and so achieve a more robust prognostic assessment of the expected patient outcome.

A Thesis submitted to The University of Arizona College of Medicine - Phoenix in partial fulfillment of the requirements for the Degree of Doctor of Medicine.
Objective and Hypothesis We hypothesize that tumor heterogeneity or tissue complexity, as measured by quantitative texture analysis (QTA) on mammogram, is a marker of tumor aggressiveness in breast cancer patients. Methods Tumor heterogeneity was assessed using QTA on digital mammograms of 64 patients with invasive ductal carcinoma (IDC). QTA generates six different values – Mean, standard deviation (SD), mean positive pixel value (MPPV), entropy, kurtosis, and skewness. Tumor aggressiveness was assessed using patients’ Oncotype DX® Recurrence Score (RS), a proven genomic assay score that correlates with the rate of remote breast cancer recurrence. RS and hormonal receptor status ‐ estrogen receptor (ER) and progesterone receptor (PR) ‐ were collected from pathology reports. Data were analyzed using statistical tools including Spearman rank correlation, linear regression, and logistic regression. Results Linear regression analysis showed that QTA parameter, SD, was a good predictor of RS (F=6.89, p=0.0108, R2=0.0870) at SSF=0.4. When PR status was included as a predictor, PR status and QTA parameter Skewness‐Diff, achieved linear model of greater fit (F=15.302, p<0.0001, R2=0.2988) at SSF=1. Among PR+ patients, Skewness‐Diff was a good linear predictor of RS (F=9.36, p=0.0034, R2=0.1320) at SSF=0.8. Logistic regression analysis showed that QTA parameters were good predictors of high risk RS probability, using different cutoffs of 30 and 25 for high risk RS; these QTA parameters were Entropy‐Diff for RS>30 (chi2=10.98, p=0.0009, AUC=0.8424, SE=0.0717) and Mean‐Total for RS>25 (chi2=9.98, p=0.0016, AUC=0.7437, SE=0.0612). When PR status was included, logistic models of higher log‐likelihood chi2 were found with SD‐Diff for RS>30 (chi2=18.69, p=0.0001, AUC=0.9409, SE=0.0322), and with Mean‐Total for RS>25 (chi2=25.56, p<0.0001, AUC=0.8443, SE=0.0591). For PR+ patients, good predictors were SD‐Diff for RS>30 (chi2=6.87, p=0.0087, AUC=0.9212, SE=0.0515), and MPP‐Diff and Skewness‐Diff for RS>25 (chi2=16.17, p=0.0003, AUC=0.9103, SE=0.0482). Significance Quantitative measurement of breast cancer tumor heterogeneity using QTA on digital mammograms may be used as predictors of RS and can potentially allow a non‐invasive and cost‐effective way to quickly assess the likelihood of RS and high risk RS.

As cancer tumours develop, competition between cells will favour those with some mutations over others, creating a dynamic heterogeneous system made up of different cell populations, called sub-clones. This heterogeneity poses a challenge for treatment, as this variety serves to ensure there is almost always a portion of the cells which are resistant to any one targeted therapy. This can be avoided by combining therapies, but finding viable combinations experimentally is expensive and time-consuming. However, there is also cooperation between sub-clones, and being able to better model and predict these dynamics could allow this interdependence to be exploited. In order to investigate how best to tackle tumour heterogeneity, while avoiding acquired resistance, I have developed the first comprehensive computational model of the gene regulatory network in breast cancer focused on the c-myc oncogene and the differences between sub-clones. I model the system as a discrete, qualitative network, which can reproduce the conditions in heterogeneous tumours, as well as predict the effect of perturbations mimicking mutations or application of therapy. Together with experimental collaborators, I apply my computational model to an in vivo mouse model of MMTV-Wnt1 driven breast cancer, which has high and low c-myc expressing sub-clones. I show that the computational model is able to reproduce the behaviour of this system, and predict how best to target either one sub-clone individually or the tumour as a whole. I show how combination therapies offer more paths to attack the tumour, and how two drugs can work synergistically. For example, I predict how Mek inhibition will preferentially affect one sub-clone, but the addition of COX2 inhibition improves effectiveness across the tumour as a whole. In this thesis, I show how a computational network model can predict treatments in breast cancer, and assess the effects on different clones of different treatment combinations. This model can be easily extended with new data, as well as adapted to different types of cancer. This therefore represents a novel method to find viable combination therapies computationally and speed up the development of new cancer treatments.

Gene expression profiling of breast cancer delineates a particularly aggressive subtype referred to as “basal-like”, which comprises ~15% of all cases, afflicts younger women and is refractory to endocrine and anti-HER2 therapies. Immunohistochemical surrogate definitions for basal-like breast cancer, such as the ER/PR/HER2 triple negative phenotype and models incorporating positive expression of cytokeratin 5 (CK5/6) and/or epidermal growth factor receptor are more amenable to implementation in a clinical setting. Despite this and the fact that basal-like breast carcinomas are being increasingly recognized as a distinct clinical entity, there is no diagnostic method used and reported in routine practice. Without a reproducible test to identify this aggressive subtype in the clinic there will be no ability to establish clearly defined intake criteria for subtype-specific clinical trials, translating to no progress in the management of this form of the disease and little change in breast cancer survival rates for the foreseeable future. A first evaluation of performance of the triple negative definition and various surrogate immunopanels for basal-like breast cancer in clinical laboratories is described in the initial chapters of this dissertation. Considerable staining variability of individual biomarkers included in immunopanels typically led to only moderate concordance with a gene expression gold standard for identification of basal-like breast carcinomas. Lack of standardization was the underlying reason for all of the observed variability, supporting the notion that further standardization efforts through continual participation in external quality assurance programs are needed before routine diagnosis of basal-like breast carcinomas could be made in a clinical setting. In light of this, we sought to identify more easy-to-interpret and robust biomarkers for this disproportionately deadly type of breast cancer. A parallel comparison of 46 proposed immunohistochemical biomarkers of basal-like breast cancer was performed against a gene expression profile gold standard. Results from that survey determined that loss of expression of INPP4B and positive expression of nestin had the strongest associations with this aggressive subtype. Paving the way for further studies, this comprehensive immunohistochemical biomarker survey is a necessary step to determine an optimized surrogate immunopanel that best defines basal-like breast cancer in a practical and clinically-accessible way.
Medicine, Faculty of
Pathology and Laboratory Medicine, Department of
Graduate

碩士
國立成功大學
醫學資訊研究所
102
In recent years, with the rapid advances in science and technology, people have paid more attentions to self-health conditions by using health examination. The health examination can avoid people missing the best time of disease diagnosis and treatment. The medical records of patients and mammogram diagnosis are contributory factors of breast cancer. Instead of using medical records or mammogram apart, the proposed method combines features automatically extracted from mammograms and medical records of patients to build a breast cancer prediction model. In preprocessing step of imaging data, the proposed method uses fast and adaptive bidimensional empirical mode decomposition (FABEMD) to segment the mammograms for glandular tissue. After integrating imaging data and clinical data, the proposed method uses search constraints to select significant features. The proposed approach solves the problem of the traditional decision tree which has complicated branches, not only saves time but also effectively improves the accuracy of prediction model of breast cancer. Our method was applied to real dataset which consists of 579 patients, and the results show that the proposed method attains high accuracy of 98%.

Dissertação de Mestrado em Engenharia Informática apresentada à Faculdade de Ciências e Tecnologia da Universidade de Coimbra
Breast Cancer (BC) is the second most frequently diagnosed cancer and the fth cause of cancer mortality worldwide. Among women, it is the leading cause of cancer deaths, with more than 500 000 registered deaths in 2012, and Portugal also re ects that reality. Survival prediction plays a crucial role in diseases with associated high mortality rates, since it has the power to help clinicians to de ne each patient's prognosis, thus allowing to personalize the corresponding treatments. Particularly for BC, prognosis is related to the patterns of recurrence (cancer that reappears after treatment), and it even di ers depending on the local involved. This work analyses the data of a cohort of 97 patients, with a total of 27 characteristics, more than 50% of them incomplete. Therefore, the rst step is to handle Missing Data (Imputation or Deletion), to perform Classi cation afterwards. The purpose is to study the prognostic factors that de ne recurrence of female BC, to try to build a model that accurately predicts recurrence patterns, which would create the possibility of more targeted treatments. The application of machine learning algorithms to the prediction of recurrence in di erent sites seems to be a novel application of these methodologies, and the results can lead the way to a better understanding of the pathways of BC recurrence.

碩士
亞洲大學
生物資訊學系碩士班
96
MicroRNAs (miRNAs) are non-coding RNA with a site of about 22 nucleotides long. They are tailored from the hairpin stem-loop miRNA precursors. MiRNAs do not translate into proteins, but taking part in many kinds of in vivo biological process. Many biological researches indicated that miRNAs are involved in the tumor developing processes, and they could possibly regulate tumor suppressor genes (TSG) or oncogenes. In the first part of this thesis, we investigate the possibility that miRNAs could possibly targeting human breast cancer TSG. In the second part, a large-scale search of over- and under-expressed breast mRNAs is performed by using the ArrayExpress microarray data. The binding strength of miRNA and mRNA is obtained by using miRanda, and the correlation of miRNAs and mRNAs expression levels in different tissues is obtained by computing the PEARSON correlation coefficients. Certain miRNAs are identified that could possibly regulating TSG and OG. It is hoped the result of this research could provide useful information for a better understanding of breast cancer induction.

博士
國立陽明大學
公共衛生研究所
96
In recent years, Data Mining attracts great concern from information industries, its main reason is that a large amount of extant materials can be used extensively, and there are urgent demands to be changed these materials into useful information and knowledge. The information and knowledge obtained are admissible to improve and promote efficiency, the field used includes very much, and the application case that Data Mining in the medical field increases gradually. According to records from Department of Health, Breast cancer and Liver cancer were major manifestations among Taiwanese population leading to deaths of top ten causes in Taiwan. These two indications had some characteristics in common as increasing risk with increasing age and sharing the same pool of risk factors in our living environment. The central role of data mining uses artificial intelligence and statistical methods to extract meaningful information from puzzles of variables and data. The present study focused on the investigation of the application of artificial intelligence and data mining techniques to the prediction models of breast cancer and liver cancer. The artificial neural network, decision tree, logistic regression, and genetic algorithm were used for the comparative studies and the accuracy and positive predictive value of each algorithm were used as the evaluation indicators. 699 records acquired from the breast cancer patients, 729 records acquire from the liver cancer patient. In breast cancer data, 9 predictor variables, and 1 outcome variable were incorporated for the data analysis followed by the 10-fold cross-validation. The results revealed that the accuracies of logistic regression model were 0.9637 (sensitivity 0.9716 and specificity 0.9482), the decision tree model 0.9435 (sensitivity 0.9615, specificity 0.9105), the neural network model 0.9502 (sensitivity 0.9628, specificity 0.9273), the genetic algorithm model 0.9878 (sensitivity 1, specificity 0.9802). The accuracy of the genetic algorithm was significantly higher than the average predicted accuracy of 0.9612. The predicted outcome of the logistic regression model was higher than that of the neural network model but no significant difference was observed. The average predicted accuracy of the decision tree model was 0.9435 which was the lowest of all 4 predictive models. The standard deviation of the 10-fold cross-validation was rather unreliable. On other hand, liver cancer data include 12 predictor variables, and 1 outcome variable were incorporated for the data analysis followed by the 10-fold cross-validation. The results revealed that the accuracies of logistic regression model were 0.7658 (sensitivity 0.7682 and specificity 0.7630, the decision tree model 0.7636 (sensitivity 0.7497, specificity 0.7793), the neural network model 0.7760 (sensitivity 0.7875, specificity 0.7679), the genetic algorithm model 0.8072 (sensitivity 0.8444, specificity 0.0.763). The accuracy of the genetic algorithm was significantly higher than the average predicted accuracy of 0.7684. The predicted outcome of the neural network model was higher than that of the logistic regression model and decision model but no significant difference was observed. The present study indicated that the genetic algorithm model yielded better results than other data mining models for the analysis of the data of breast cancer and liver cancer patient in terms of the overall accuracy of the patient classification, the expression and complexity of the classification rule. The results showed that the genetic algorithm described in the present study was able to produce accurate results in the classification of breast cancer data/liver cancer data and the classification rule identified was more acceptable and comprehensible.

碩士
南台科技大學
企業管理系
95
The incidence and mortality rate of breast cancer in Taiwanese Women have increased gradually due to the urban life style and westen style food.In the recent 5 years, the incidence of breast cancer in Taiwanese Women became the first in all cancers. The highest perioid of incidence of breast cancer is between 45 to 55 years old. In the early stage of breast cancer, it is almost asymptoatic and keep the patients from medical help.When the breast cancer was diagnosed, many of them aleady have lymph node metastasis. This situation also lifts the recurrent rate. Due to the progress of information technology and medical information system, hospitals also have accumlated a large amount of data in the database of information systems. Therefore, much useful medical knowledge could be mined from the history data. The prediction of breast cancer relapse is very helpful for post-operative treatment and followup. The statistical methods had been applied to predict breast cancer relapse. However, this study employed data mining techciques, including C4.5 decision tree and SVM, to construct recurrence prediction models of breast cancer. To improve the prediction efficiency, this study also applied committee machine methods, including AdaBoost and Bagging, to increase the relapse prediction accuracy. The empirical results show that AdaBoost mechanism can ehance prognosis accuracy of C4.5 and SVM models on breast cancer relapse. Keyword：Breast cancer relapse、C4.5 decision tree、Support Vector Machine、Committee Machine

碩士
高雄醫學大學
醫學研究所
101
Background and Purpose Breast Cancer is the most common cancer in the world women. Also, the breast cancer is causing the highest mortality rate the main reason. The breast cancer is the number one killer for women and become the highest ranking in Asia. This study is therefore comparing artificial neural network (ANN) and logical regression (LR) prediction models to find the best of important effect factors. The purposes of this research are as follows: Ⅰ、To investigate long-term trend analysis of the breast cancer patient after surgery in 5-year mortality; Ⅱ、To compare the accuracy of different predict models for breast cancer patients after surgery in 5-year mortality; Ⅲ、To conduct the global sensitivity analyze and to estimate the significant predictors for breast cancer patients after surgery in 5-year mortality. Research Methods This study subject of "National Health Insurance Research Database" is the research framework. The study design used the retrospective method. The study period is from 1996 to 2010. The study subjects of the breast cancer patients are above sixteen years old after surgery. The samples of study are total 3,632 people. The use of the diagnostic codes ICD-9-CM174x (174.0-174.9) and disposal code 85.20-23,85.33-36,85.4 x, 85.5x, 85.6x, 85.7x, 85.8x, 85.95 find the predictor factors of meaningful. Investigate the important effect factors are by the breast cancer patient analysis after surgery in 5-year mortality and use of the patient''s characteristics, hospital characteristics and time characteristics, separately. In addition, the significant predictors are used in the ANN and LR to build model and to compare the accuracy. SPSS 19.0 statistical software was employed for the data collection and analysis. The main statistical methods include: descriptive statistics and inferential statistics (trend analysis, university analysis, and multivariate analysis - including logistic regression and neural networks and sensitivity analysis). Results In this study, the model build are used data mining technology of ANN and LR and seven important variables (age, charlson comorbidity index (CCI), hospital level, hospital volume, surgeon volume, chemotherapy, radiotherapy and hormone therapy). The results show that ANN model is better than LR model. The sensitivity of ANN and LR is 5.66% and 3.77%, respectively. The 1-specificity, positive predictive value, negative predictive value and accuracy are good performance. The AU-ROC curve is 0.70 and 0.52, respectively. Generally, the performance of ANN is better than LR model. The first three important predictors of ANN are surgeon volume, chemotherapy and age. The LR model are the surgeon volume, age, charlson comorbidity index (CCI). Conclusions and Recommendations The research results found that the distribution trend for the breast cancer patient after surgery in 5-year mortality is a significant change with time in patients’ characteristics and hospital characteristics. The means show that the correlation significant factors are worth applied in clinically and become improvement factors even if as standard of treatment guidelines. To compare the different predict models, found that the neural network can be used to expand predict variable items. The various diseases are easy to analyze and investigate, systematically. The death predicts model is use of appropriate research methods with development. In the future, this prediction model can applied to other cancers. In addition, the most important factor is surgeon volume. So that, how to enable health authorities to assist and train surgeons relevant experience reference. Secondary the impact factors such as age are increased with mortality rate. Therefore, the breast cancer should repeated think the relation with age and pay attention in how to promote and improve screening rates issues and to reach early detection and early treatment. Finally, this study expect can provide clinical reference value for medical staff and integrate medical treatment and to establish a predict model of medical decision.

碩士
國立中正大學
電機工程研究所
91
The objective of this thesis is to build a diagnosis system for breast cancer classification and pathological section prediction. In the classification part, the tumor images were first sampled by two methods: taking (1) the largest rectangular region inside the tumor image, and (2) the 64×64 rectangular part that has the darkest grey-levels. The features were extracted from the co-occurrence matrix and multi-resolution decomposition by wavelet transform. Linear discrimant analysis and k-NN methods were used to classify the tumors into benign and malignant ones. Best results were found in the combination of using the largest rectangular segment of the tumor image, extracting the contrast of the co-occurrence matrix as features, and classifying by linear discriminant analysis. The recognition rates were 84.6 % for the benign and 78.38 % for the malignant tumors, respectively. The second part of this study is trying to predict the possible pathological section appearance by analyzing the ultrasound images. Regions of interest (ROIs) of the ultrasound images were marked by the doctors. Two neighbor small (32×32) image sections were segmented from the marked region and one is used in training the classifier and the other for testing. Based on the studies in the classification part, contrast and entropy calculated from the co-occurrence matrix were chosen as the prominent features for the image query, and the first five images in the database that are nearest in the features space were retrieved. From the 34 image query tests, 24 (70.58%) of their corresponding images were retrieved and were ranked in the first order, and 31 (91.18%) were found in the first five nearest images.

碩士
國立臺灣大學
資訊工程學研究所
106
Breast cancer is the most commonly diagnosed cancer and is the second leading cause of death among women after lung cancer. Patients who suffer from breast cancer can still recuperate with early diagnosis and proper treatment, despite the fact that breast cancer possesses a high mortality rate. If untreated, breast cancer can surreptitiously spread and invade other organs by transporting its cells via nearby hematogeneous or lymphatic routes. This type of breast cancer is classified as metastatic breast cancer. Whereas, the non-metastatic cancer does not possess this ability. This thesis presents a breast cancer classification between the metastasis and non-metastasis using the densely connected convolutional neural network (DenseNet). Several studies have also successfully revealed the presence of suspicious tissue surrounding the tumor region (peritumor). Inspired by a previous study that utilized image matting to obtain the peritumor from the trimap, the peritumor can further be extracted at different pixel thicknesses by adjusting the unknown region thickness of the trimap. Thus, this study trained the neural network using peritumor images (instead of the tumor only) at different pixel thicknesses: 5, 10, 15, 20, 25, and 30 pixels. This study finds that the peritumor 15 pixels achieved the best performance with an accuracy of 84.8%, a sensitivity of 88.8%, a specificity of 81.3%, and an area under the curve (AUC) of receiver operating characteristic (ROC) of 0.926. In addition, based on the results, the peritumor 10 pixels may not be too farfetched from being considered for future studies as it scored an accuracy of 84.1%, a sensitivity of 83.2%, a specificity of 84.8%, and an AUC score of 0.906.