Scholarlink

Testing for mismatch repair deficiency (dMMR) and high-grade microsatellite instability (MSI-H) has become an integral part of the routine diagnostic workup for colorectal cancer (CRC) and endometrial cancer (EC) in clinical practice, and using MSI to predict the therapeutic response to immune checkpoint inhibitors (ICIs) has now become a hot focus. Recent reports have shown that patients with dMMR/MSI-H tumors in CRC and EC can benefit from ICIs. So far, MSI is the only widely recognized biomarker that can be used to evaluate the effectiveness of immunotherapy for CRC and EC. The humanized monoclonal anti-PD-1 antibody pembrolizumab is the first ICI to have its clinical activity investigated in patients with dMMR/MSI-H CRC and EC. Immunohistochemical testing for dMMR has recently become a standard testing item in pathology laboratories, and next-generation sequencing is also a powerful method for analyzing MSI. Although the examination of MSI is important for immunotherapy, high cost and time constraints have limited the widespread adoption of MSI testing. Here, we developed an ensemble transformer multiple instance deep learning (DL) system and evaluated the system to predict MSI status using histopathological whole slide images (WSIs) from multiple datasets of Tri-Service General Hospital and the Cancer Genome Atlas in MSI-related EC and CRC. The results demonstrate that the presented system has great potential to predict MSI status directly from H&E WSIs in EC and CRC, enabling personalized targeted immunotherapy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transformer Deep Learning to Detect Microsatellite Instability Using Histopathological Slides to Guide Colorectal and Endometrial Cancer Immunotherapy

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ching-Wei Wang,
  • Tai-Kuang Chao

摘要

Testing for mismatch repair deficiency (dMMR) and high-grade microsatellite instability (MSI-H) has become an integral part of the routine diagnostic workup for colorectal cancer (CRC) and endometrial cancer (EC) in clinical practice, and using MSI to predict the therapeutic response to immune checkpoint inhibitors (ICIs) has now become a hot focus. Recent reports have shown that patients with dMMR/MSI-H tumors in CRC and EC can benefit from ICIs. So far, MSI is the only widely recognized biomarker that can be used to evaluate the effectiveness of immunotherapy for CRC and EC. The humanized monoclonal anti-PD-1 antibody pembrolizumab is the first ICI to have its clinical activity investigated in patients with dMMR/MSI-H CRC and EC. Immunohistochemical testing for dMMR has recently become a standard testing item in pathology laboratories, and next-generation sequencing is also a powerful method for analyzing MSI. Although the examination of MSI is important for immunotherapy, high cost and time constraints have limited the widespread adoption of MSI testing. Here, we developed an ensemble transformer multiple instance deep learning (DL) system and evaluated the system to predict MSI status using histopathological whole slide images (WSIs) from multiple datasets of Tri-Service General Hospital and the Cancer Genome Atlas in MSI-related EC and CRC. The results demonstrate that the presented system has great potential to predict MSI status directly from H&E WSIs in EC and CRC, enabling personalized targeted immunotherapy.