Microsatellite Instability
Microsatellite instability (MSI) is a genomic feature indicating errors in DNA mismatch repair, impacting cancer development and treatment response. Current research heavily focuses on developing accurate and efficient MSI prediction models using artificial intelligence, particularly deep learning architectures like convolutional neural networks (CNNs), transformers (e.g., Swin Transformer), and attention-based multiple instance learning, often leveraging whole-slide images from hematoxylin and eosin (H&E) stained biopsies and integrating data from other modalities like CT scans. These advancements aim to replace or supplement costly and time-consuming traditional MSI testing methods, ultimately improving cancer diagnosis and treatment personalization. The improved accuracy and efficiency of AI-based MSI prediction holds significant promise for streamlining clinical workflows and optimizing patient care.
Papers
Patient-level Microsatellite Stability Assessment from Whole Slide Images By Combining Momentum Contrast Learning and Group Patch Embeddings
Daniel Shats, Hadar Hezi, Guy Shani, Yosef E. Maruvka, Moti Freiman
Predicting microsatellite instability and key biomarkers in colorectal cancer from H&E-stained images: Achieving SOTA predictive performance with fewer data using Swin Transformer
Bangwei Guo, Xingyu Li, Jitendra Jonnagaddala, Hong Zhang, Xu Steven Xu