HistoGenomics

Expanding on the DeepSMILE algorithm's foundation, we aim to validate its approach for homologous recombination deficiency (HRD) testing in extensive retrospective cohorts of breast and ovarian cancer patients using H&E slides.

DeepSMILE, a breakthrough by Schirris et al., employs self-supervised learning and tumor heterogeneity awareness to predict HRD and Microsatellite instability (MSI) without detailed annotations. Our project extends this innovation, aiming to demonstrate the algorithm's potential in streamlining diagnostics for personalized therapy, particularly in breast and ovarian cancers. By leveraging AI to interpret cancer tissue slides, we envision a future where HRD testing is more accessible, quicker, and cost-efficient, significantly impacting patient treatment plans and outcomes.

This initiative marks a pivotal step toward employing AI in refining diagnostic accuracy and expediting treatment decisions across cancer types.