Genomics-based Survival Analysis for Lung Cancer using Multimodal Data

Bachelor’s Project-II (Collaboration with TATA Cancer Research Hospital) at the Medical Deep Learning and Artificial Intelligence Lab, IIT Bombay, Guide: Prof. Amit Sethi

Summary: In this project, we combined genetic information, chemotherapy drug properties, patient demographics and imaging data of different kinds to predict tumor progression free interval and effectiveness of chemotherapies.

Paper submitted to IEEE ISBI 2025 (International Symposium on Biomedical Imaging

Paper link

  • Created neural Cox Proportional Hazards models for genomic data using graph neural networks and neural ranking
  • Employed discriminator-based domain adaptation and transfer learning to use mouse data to fortify human TCGA data
  • Attained 0.88 concordance index by including clinical and image data, and gene subset selection with L1 regularization
  • Created a novel cross modal alignment-based genomics token learning network that surpassed the TransMIL baseline