．Speaker．

Dr. Chiao-Min Chen

Data Scientist of Artificial Intelligence Development Center, Changhua Christian Hospital

Speaker's Biography

Dr. Chen is currently the Data Scientist of the Artificial Intelligence Development Center of Changhua Christian Hospital (CCH). She received her doctoral degree in Computer Science and Information Engineering from the National Taiwan University in 2020. Before that, Dr. Chen has been dedicating herself to the biomedical engineering field for over ten years. She joined the CCH AI development team in 2021. She was actively involved in machine deep learning and data alignment, including AI researches, development, and application in the medical system and service. 

Topic

AI-based Computer-aided System for Histopathology Images

Abstract

Pathology is a significant field in modern medical diagnosis for disease research. As an adjunct to pathological disease research, a digital pathology image has been developed for evaluating suspicious abnormalities in the tissue specimen. In order to establish efficient diagnostic procedures, various computer-aided diagnosis (CAD) systems have been prompted to assist pathologists. Prostate cancer is a significant health concern in aging males, and proper management of the disease depends on accurately interpreting pathology specimens. However, reading prostatectomy histopathology slides, which is basically for staging, is usually time-consuming and differs from reading small biopsy specimens, which are mainly used for diagnosis. With the aim of reducing the workload of pathologists, an AI-based CAD system is developed for whole-slide histopathology images to outline the malignant regions. The system achieves 0.726 of the Dice coefficients. The sensitivity and specificity are observed to be 96.7% and 93.9%, respectively. The development of the system will be effective in assisting pathologists in automatic diagnostic routine tasks. Accurate and precise alignment of histopathology tissue sections is crucial in interpreting the proteome topology and cell-level 3D reconstruction of diseased tissues. However, the realization of an automated and robust method for aligning non-globally stained immunohistochemical sections is still challenging. The AI-based image registration system is successfully automated to align most of the images. The Hausdorff distance is 48.93 μm, showing a significant improvement. Further advancements in the screening of the proteome topology and 3D tissue reconstruction could be expected. In this seminar, the significant research results will be shared.