Yin and Yang Study
The study is consisting of 3 components
Case identification and ground truthing
Multi-site staining + scanning, and slide+ data collection
AI Performance analysis comparison
Project 1. To identify relevant and representative cases alongside a ground truth from experts
The plan is to obtain 3 anonymized prostate samples from resection specimens from 3 different patients. The histopathological findings will be assessed by two board certified pathologists who will assign a preliminary Gleason grade. The three cases should have distinct glandular morphologies. Once selected, the laboratory will cut an additional series of standardized 5µm, blank sections on charged glass slides followed by a final section (H&E stained). The H&E stained first, and last section will be scanned and reviewed by a panel of experts – who will confirm that the Gleason grade of the first and last section is identical. This will assure comparable grades on all intervening blank sections. The total number of blank sections is 30 composed of n=12 blanks (one for each participating institution), plus 18 extra sets.
Project 2. To distribute cases to multiple institutions for staining and scanning and return digital files and slide
Blank sections will be sent to 12 institutions for staining according to their routine clinical protocol. The participating institution is asked to scan the slide and answer a brief questionnaire. The scanned file will be uploaded to a USB stick and shipped back with the stained slide to the central laboratory. Returned slides will be scanned at the central laboratory and all files and information will be hosted
Project 3. To obtain AI-driven Gleason grades on the study set and assess the relationship of staining and scanning on algorithm performance
The anonymized files alongside relevant metadata for data entry is hosted. Participant can download the files and perform analysis. The result of the algorithms is provided back to the central laboratory. Analysis entails anonymization of the submitted results. The primary endpoint of the study is overall percent agreement of predicted Gleason grades. Secondary endpoints are the influence of staining variability, scanning variability, and institutional variability for color reproducibility and absolute deviation from a reference scan (FDA-authorized scanning system).