An AI Algorithm for Lymphocyte Focus Score of Minor Salivary Gland Biopsy Samples for Diagnosing Sjogren's Syndrome

Overview

The aim of this research is to discover an artificial intelligence (AI) algorithm for lymphocyte focus score in whole slide images of labial minor salivary gland (SG) biopsy samples for diagnosing Sjogren's Syndrome, in order to enhance the precision of pathological interpretation of labial minor SG biopsy samples in patients with suspected Sjogren's syndrome and aid clinicians make an accurate diagnose. A remote AI-assisted pathological interpretation platform for lymphocyte focus score in labial SG will be built for the global based on the research results. The research will propose the AI-assisted pathological interpretation of lymphocyte focus score in labial minor SG biopsy samples in the future guidelines for the diagnosis and treatment of Sjogren's syndrome. The research will： 1. Develop and debug the AI algorithm for lymphocyte focus score in whole slide images of labial minor SG biopsy samples for diagnosing Sjogren's Syndrome; 2. Internal test of the AI algorithm; 3. Clinical validation of the AI algorithm with blind method in multiple centers; 4）Built a remote AI-assisted pathological interpretation platform for lymphocyte focus score in labial SG for the global and Explore its clinical application.

1. Develop and debug the AI algorithm for lymphocyte focus score in whole slide images of labial minor SG biopsy samples for diagnosing Sjogren's Syndrome; A total of 200 H\&E staining slides of labial minor SG biopsy samples are collected from Sun Yat-sen Memorial Hospital of Sun Yat-sen University and scanned into digital pathological images. The ground truth of gland tissue area and lymphocyte foci numbers in each image is interpreted by three senior pathologists with over 5 years of related experience. 2. Internal test of the AI algorithm; A total of 500 additional digital pathological images of labial gland biopsy tissues are collected from Sun Yat-sen Memorial Hospital of Sun Yat-sen University. The ground truth of gland tissue area and lymphocyte foci numbers in each images is interpreted by three senior pathologists with over 5 years of related experience. The AI algorithm's accuracy, specificity, sensitivity, positive predictive value and negative predictive value in evaluating the area of labial gland and the number of lymphocyte foci are calculated. Comparison of whether the image meets the criteria for Sjögren's syndrome (focus score greater than 1) is also conducted between the AI algorithm and the ground truth. 3. Clinical validation of the AI algorithm with blind method in multiple centers; A total of 600 additional digital pathological images of labial gland biopsy tissues are collected from six external centers. The ground truth of gland tissue area and lymphocyte foci numbers in each images is interpreted by three senior pathologists with over 5 years of related experience. The AI algorithm's accuracy, specificity, sensitivity, positive predictive value and negative predictive value in evaluating the area of labial gland and the number of lymphocyte foci are calculated. Comparison of whether the image meets the criteria for Sjögren's syndrome (focus score greater than 1) is also conducted between the AI algorithm and the ground truth. 4）Built a remote AI-assisted pathological interpretation platform for lymphocyte focus score in labial SG for the global and Explore its clinical application. Digital pathological images of labial gland biopsy tissue can be uploaded to the Labial Gland Pathological Focus Score Remoting platform. AI-assisted pathological interpretation on gland tissue area, lymphocyte foci numbers, and whether meeting the criteria for Sjögren's syndrome (focus score greater than 1) is compared with the ground truth.