NORMA eResearch @NCI Library

Using Self-Supervised Learning Models to Predict Invasive Ductal Carcinoma from Histopathological Images

Poonawala, Taher Abbas (2023) Using Self-Supervised Learning Models to Predict Invasive Ductal Carcinoma from Histopathological Images. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
Preview
PDF (Master of Science)
Download (9MB) | Preview
[thumbnail of Configuration manual]
Preview
PDF (Configuration manual)
Download (3MB) | Preview

Abstract

The process of learning representations without annotated data is called selfsupervised learning (SSL). SSL has found most of its applications and proof of concept in the field of natural images since its debut. These strategies have a lot of potential, especially in situations when data is scarce. While these algorithms would almost certainly benefit immensely from being researched and benchmarked on sparse medical datasets such as microscope images, they have gotten little attention so far. This research focuses on creating a framework for applying the SSL approach to low data-regime scientific datasets. The dataset consisted of histopathological scans of invasive ductal carcinoma (IDC) with 96-pixel resolution images. Methods for adapting SSL protocols to operate with this data collection were thoroughly investigated. Employing the SimCLR framework, which uses a contrastive approach, to learn data representations with a focus on cropping algorithms. The fascinating structure of the dataset allows for significant changes. After training on an unlabeled dataset, the SimCLR achieved 84% accuracy as a prelude to logistic regression for image classification. The ResNet18 model, considered the baseline, could only properly predict 75% of the images. It outperformed a model trained only from supervision with 500 images per label by 8% using just a tenth of the labeled data.

Item Type: Thesis (Masters)
Supervisors:
Name
Email
Basilio, Jorge
UNSPECIFIED
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer)
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 23 May 2023 16:54
Last Modified: 23 May 2023 16:54
URI: https://norma.ncirl.ie/id/eprint/6635

Actions (login required)

View Item View Item