Resume of Jikai Yu
Education Experience
2023.9-nowadays HuZhou University,Faculty of Information Engineering
I majored in computer technology, and my research focused on the intersection of computer vision and pathology.
Many colorectal cancer patients need to undergo MMR (Mismatch Repair) or MSI (Microsatellite Instability) testing to select subsequent treatments. The MMR system includes proteins such as MLH1, MSH2, MSH6, and PMS2, which are responsible for identifying and repairing areas of DNA replication errors. There are short tandem repeat sequences in the genome known as microsatellites, which are prone to slippage errors during replication and are highly dependent on the MMR system for repair. When any of these four proteins are abnormal, it can lead to MMR deficiency, failure to repair the misaligned DNA, resulting in MSI, causing abnormalities in tumor-related genes, and thereby inducing the occurrence of cancer. The induction of colorectal cancer is closely related to microsatellite instability, so the prediction of microsatellite instability is very important for the prognosis and diagnosis of colorectal cancer.
In recent years, WSI (Whole Slide Image, WSI) has been widely used for cancer diagnosis due to its characteristic of efficiently storing high-resolution pathological images in a pyramid structure, replacing traditional microscope examinations by digitizing pathological tissues, and has become an indispensable part of daily research in pathology.
Similar to traditional supervised learning, pathologists provide corresponding labels for WSI. However, the size of WSI is usually in the order of billions of pixels, making it impractical to directly use GPU for computation on WSI. Considering these limitations, patch-based training methods have begun to be widely used. However, due to the large size and high resolution of WSI, obtaining labels for patches is difficult, so multiple instance learning has begun to be widely applied to the analysis and prediction of WSI.
In the field of medical image analysis, a WSI is referred to as a “bag,” and each image in the WSI is referred to as an “instance,” and we usually only have “bag” level labels. Therefore, multiple instance learning can be fully utilized for the prediction of microsatellite instability to promote the development of computer-aided diagnosis (CAD).