Transfer Learning Based Classification of MSI and MSS Gastrointestinal Cancer

EasyChair Preprint 7392

Abstract

Gastrointestinal and Colorectal cancers are treated with chemotherapy and its other forms which are not able to provide higher survival rates [1]. Immunotherapy is increasingly becoming popular due to its promising response especially to mutated tumors such as MicroSatellite Instability (MSI) cancers with deficient DNA Mismatch-Repair system (dMMR). Generally, 85% of all the cases related to gastrointestinal and colorectal cancers have proficient DNA Mismatch-Repair system (pMMR) which are also called MicroSatellite Stability (MSS). Only about 15% of the gastrointestinal and colorectal cancer patients have deficient DNA Mismatch-Repair system (dMMR) causing MicroSatellite Instability (MSI) in their tumors. While Immunotherapy responds well to patients with MSI tumors, it is resistant to MSS tumors [2]. Hence, it’s important to classify MSI vs. MSS tumors so that appropriate treatment can be given to the patients. Clinically MSI cancers are difficult to be detected after stage III due to  their sensitivity to pembrolizumab inhibitors [3][4]. In this work, deep learning based transfer learning approach is detailed that can accurately classify MSI vs. MSS cancers using histological images which are derived from formalin-fixed paraffin-embedded (FFPE).

Keywords: MicroSatellite Instability (MSI), MicroSatellite Stability (MSS), Deep learning, Deep convolutional neural network, Gastrointestinal Cancer, Colorectal cancer, Transfer Learning.

Keyphrases: Deep Convolutional Neural Network, MicroSatellite Instability (MSI), MicroSatellite Stability (MSS)

@booklet{EasyChair:7392,
  author    = {Zabiha Khan and R Loganathan},
  title     = {Transfer Learning Based Classification of MSI and MSS Gastrointestinal Cancer},
  howpublished = {EasyChair Preprint 7392},
  year      = {EasyChair, 2022}}