janghana/breast-cancer-classification-using-vit

Comparison of Lesion Interpretation Assistance Performance in Brest MRI Images According to the Configuration of Vision Transformer(ViT)

Enhancing Breast Cancer MRI Classification Through Vision Transformer(ViT) and Multi-Marker

Project Overview

This research project, running from September 2023 to December 2023, investigates the effectiveness of Vision Transformer (ViT) configurations in assisting the interpretation of lesions in breast MRI images.

Background

As an Undergraduate Research Student, I contributed to a collaborative effort to refine breast cancer classification in MRI imaging. This study is particularly significant as it compares the efficiency of a multi-marker model against a single-marker model in breast cancer detection.

Objectives

The primary aim is to validate the enhanced performance of a multi-marker model. This model integrates both the label of breast cancer presence and information on protein markers, including Estrogen Receptor (ER), Progesterone Receptor (PR), and Human Epidermal Growth Factor Receptor 2 (HER2). Our work seeks to demonstrate that these critical protein markers can be identified through non-invasive MRI, offering potential for improved diagnostic methods.

Methodology

• Employed Vision Transformer (ViT) techniques for deep learning-based image analysis.
• Developed models that utilize complex markers to predict breast cancer presence with higher accuracy compared to traditional methods.
• Conducted comparative analysis to evaluate the performance enhancements achieved through the use of multi-marker models.