Speaker: Ibrahim Abukhiran
Description: This presentation explores the evolving role of digital pathology, image analysis, and artificial intelligence (AI) in modern diagnostic practice and biomarker discovery. The session focuses on the transition from subjective, manual assessment toward quantitative, AI-enabled workflows that improve diagnostic reproducibility, efficiency, and clinical decision-making. Key topics include the distinction between generative AI (used for tasks such as automated surgical pathology report formatting) and non-generative AI (used for quantitative analysis, classification, and prediction), with real-world examples demonstrating safe, secure, and PHI-free clinical integration.
At the end of the presentation, participants will be able to:
– Differentiate between generative AI and non-generative AI and identify appropriate clinical use cases for each within digital pathology workflows (e.g., report formatting vs. biomarker quantification).
– Apply image analysis principles to select an appropriate technology (conventional IA, ML, or DL) based on task complexity, data availability, and institutional resources.
– Assess quantitative pathology outputs (e.g., TILs, Ki-67, ER/PR/HER2, PD-L1) for technical validity by recognizing sources of error, including scanner variability, thresholding issues, and color calibration drift.
-Implement a structured quality control and oversight approach for AI-enabled pathology workflows by identifying bias, stochastic behavior, data drift, and the need for human-in-the-loop review.