Machine learning application in colorectal cancer screening from histopathology whole slide images

Background: Colorectal cancer is the second prevalent malignancy in view of both incidence and mortality in HK, and account for 15.8% and 15.9% respectively of all cancers in 2021. Over 90% of colorectal cancer are adenocarcinoma. The low mortality rate highlighted the importance of early cancer detection, and confirmation of diagnosis. The gold standard of cancer diagnosis is based on the cancer cell identification from the histopathological slides by experienced pathologists with advanced training on cancer screening. The recent development of digital whole slide scanners allows the histopathological image analysis feasible and facilitate cancer diagnosis by pathologists. This study aimed to build an image feature-based model using histopathology whole slide images to differentiate patients with colorectal cancer from normal.<br/><br/>Methods: 822 H&amp;E images were collected from 30 patients with adenocarcinoma of colon from the United Christian Hospital in HK. Another 821 H&amp;E images were collected from 30 healthy subjects as control (HC). 22 Gray level Co-occurrence Matrix (GLCM) image features and 11 Gray Level Run Length Matrix (GLRLM) image features were retrieved from the collected images by an in-house developed software. Both the diseased and HC groups were divided into a training and a validation group on 70:30 basis. Support Vector Machines (SVM) classifier was used to build the model using training group and validated the model by the validation group using the Matlab classifier learner.<br/><br/>Results: The receiver operating characteristics (ROC) curve showed that the average class and overall accuracy achieved 94.8%. The area under the curve (AUC) was 0.98, with sensitivity of 93% and specificity of 96%. <br/><br/>Conclusions: The machine learning model can accurately differentiate colorectalcancer patient from normal one in histopathology whole slide images. The application can be fully automatic, and it is practical in applying for colorectal cancer screening.