Deep Learning and Microsatellite Instability in Colorectal Cancer

Universal testing of colorectal cancer (CRC) for microsatellite instability (MSI) or mutation mismatch repair (MMR) deficiency is recommended, but infrastructure requirements and cost have hampered widespread adoption. Deep learning is capable of detecting MSI or MMR deficiency based on specific histological findings on hematoxylin and eosin-stained slides, such as tumor-infiltrating lymphocytes, mucinous differentiation, heterogeneous morphology, and poor differentiation. However, this has not been validated on a large scale.

To address this validation need, investigators trained a deep learning system on over 8800 CRCs that had undergone molecular analyses, as part of an international consortium. Several experiments were conducted to asses…