Relation of HE staining section quality with AI-assistedpathological diagnosis
GAO Haili
YU Yan
LUO Wenying
LI Bowen
YANG Jianghui
Abstract：Objective To investigate the impact of digital hematoxylin-eosin (HE) stained section quality on the accuracy of artificial intelligence (AI)-assisted pathological diagnosis. Methods A total of 168 cases (1,213 HE sections) were collected from the Department of Pathology, Beijing Tsinghua Changgong Hospital, affiliated with Tsinghua University, from January to March 2025. After digitization using a digital slide scanner (20x objective lens), the slides were evaluated by both pathologists and the AI-assisted pathological diagnosis system Miis in a double-blind manner. The effects of digital HE staining section quality issues (such as incomplete tissue, uneven thickness, tremor marks, etc.) on AI-assisted pathological diagnosis were analyzed, along with their correlation. Results When there were quality issues such as incomplete tissue, uneven thickness, or tremor marks in the pathological sections, leading to the inability of pathologists to make a diagnosis, pathologists would request technicians to re-prepare the sections to ensure accurate diagnosis. However, the AI-assisted pathological diagnosis system could not identify these quality issues, and analyzing substandard sections could lead to risks of missed or incorrect diagnoses. Conclusion High-quality digital sections can provide a solid and reliable data foundation for AI diagnostic systems, helping AI achieve more accurate and error-free judgments in the field of pathological diagnosis.
Keywords：HE sectionsSection qualityAI-assisted pathological diagnosisDigital sectionsAI diagnosis
Publication Date：2025-12-28
Online Publishing Date：2026-01-28(First online date of this platform, not the publication date of the document)
Pages：4( 1701-1704 )