The use of deep learning to predict HER2 status in breast cancer directly from histopathology slides

Simple item page
dc.contributor.advisorCoetzer, Johannesen_ZA
dc.contributor.authorSmith, Alexandra Nicoleen_ZA
dc.contributor.otherStellenbosch University. Faculty of Science. Dept. of Applied Mathematics.en_ZA
dc.date.accessioned2024-03-01T12:40:11Z
dc.date.accessioned2024-04-26T11:25:35Z
dc.date.available2024-03-01T12:40:11Z
dc.date.available2024-04-26T11:25:35Z
dc.date.issued2024-03
dc.descriptionThesis (MSc)--Stellenbosch University, 2024. en_ZA
dc.description.abstractENGLISH ABSTRACT: The treatment of breast cancer is significantly influenced by the identification of various molecular biomarkers, including Human Epidermal Growth Factor Receptor 2 (HER2). Current techniques for determining HER2 status involve immunohistochemistry (IHC) and in-situ hybridisation (ISH) methods. HER2 testing, which is routinely applied in cases of invasive breast cancer, serves as the primary biomarker guiding HER2-targeted therapies. HE-stained whole slide images, which are more cost-effective, time-efficient, and routinely produced during pathological examinations, present an opportunity for leveraging deep learning to enhance the accuracy, speed, and affordability of HER2 status determination. This thesis introduces a deep learning framework for predicting HER2 status directly from the morphological features observed in histopathological slides. The proposed system has two stages: initially, a deep learning model is employed to differentiate between benign and malignant tissues in whole slide images, using annotated regions of invasive tumours. Following this, the effectiveness of Inception-v4 and Inception-ResNet-v2 architectures in biomarker status prediction is explored, comparing their performance against previous model architectures utilised for this task, namely Inception-v3 and ResNet34. The study utilises a dataset comprising whole slide images from 147 patients, sourced from the publicly available Cancer Genome Atlas (TCGA). Models are trained using 256 ◊ 256 patches extracted from these slides. The best-performing model, Inception-v4, achieved an area under the receiver operating characteristic curve (AUC) of 0.849 (95% confidence interval (CI): 0.845 ≠ 0.853) per-tile and 0.767 (CI:0.556 ≠ 0.955) per-slide in the test set. This research demonstrates the capability of deep learning models to accurately predict HER2 status directly from histopathological whole slide images, offering a more cost- and time-efficient method for identifying clinical biomarkers, with the potential to inform and accelerate the selection of breast cancer treatments.en_ZA
dc.description.abstractAFRIKAANSE OPSOMMING: Die behandeling van borskanker word grootliks beïnvloed deur die identifikasie van verskeie molekulêre biomerkers, soos Human Epidermal Growth Factor Receptor 2 (HER2). Die huidige metodes vir die bepaling van HER2-status behels immunohistochemie (IHC) en in-situ hibridisasie (ISH). HER2-toetsing word gereeld gebruik vir gevalle van indringende borskanker en is die belangrikste biomerker wat gebruik word om HER2-gerigte terapieë te lei. HE-gekleurde heelskyfiebeelde, wat koste-effektief, tyddoeltreffend en algemeen tydens patologiese ondersoeke vervaardig is, bied ’n geleentheid om diepleer te gebruik om die akkuraatheid, spoed en bekostigbaarheid van die bepaling van HER2-status te verbeter. Hierdie tesis bied ’n diepleerraamwerk aan wat HER2 status direk kan voorspel vanaf die morfologiese kenmerke wat in histopatologiese skyfies waargeneem word. Die voorgestelde stelsel bestaan uit twee fases: eerstens word ’n diepleermodel gebruik om te onderskei tussen goedaardige en kwaadaardige weefsels in heelskyfiebeelde deur gebruik te maak van geannoteerde streke van indringende gewasse. Daarna word die doeltreffendheid van Inception-v4 en Inception-ResNet-v2 argitekture in die voorspelling van biomerkerstatus ondersoek en vergelyk met vorige modelargitekture wat vir hierdie taak gebruik is, naamlik Inception-v3 en ResNet34. Die studie gebruik ’n datastel met volledige skyfiebeelde van 147 pasiënte, verkry vanaf die publiek beskikbare Cancer Genome Atlas (TCGA). Modelle word opgelei deur gebruik te maak van 256 ◊ 256 kolle wat uit hierdie skyfies onttrek is. Die beste presterende model, Inception-v4, bereik ’n oppervlakte onder die ontvanger se bedryfskenmerkkurwe (AUC) van 0.849 (95% vertrouensinterval: 0.845- 0.853) per teël en 0.767 (CI: 0.556-0.955) per skyfie in die toetsstel . Hierdie navorsing demonstreer die vermoë van diepleermodelle om HER2-status akkuraat te voorspel direk vanaf histopatologiese heelskyfiebeelde, wat ’n meer koste- en tyddoeltreffende metode bied om kliniese biomerkers te identifiseer en moontlik die keuse van borskankerbehandelings te bespoedig.af_ZA
dc.description.versionMastersen_ZA
dc.format.extent89 pages : illustrations (some color)en_ZA
dc.identifier.urihttps://scholar.sun.ac.za/handle/10019.1/130268
dc.language.isoen_ZAen_ZA
dc.language.isoen_ZAen_ZA
dc.publisherStellenbosch : Stellenbosch Universityen_ZA
dc.rights.holderStellenbosch Universityen_ZA
dc.subject.lcshBreast -- Cancer -- Treatmenten_ZA
dc.subject.lcshCancer -- Histopathologyen_ZA
dc.subject.lcshHER-2 proteinen_ZA
dc.subject.lcshImmunohistochemistryen_ZA
dc.subject.nameUCTDen_ZA
dc.titleThe use of deep learning to predict HER2 status in breast cancer directly from histopathology slidesen_ZA
dc.typeThesisen_ZA
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
smith_use_2024.pdf
Size:
50.03 MB
Format:
Adobe Portable Document Format
Description:
Download
Collections
Masters Degrees (Applied Mathematics)