Our poster explores how visual read, SUVR thresholding, and a deep learning AI model agree, and differ, in predicting amyloid status from PET imaging. While both quantitative SUVR and AI show strong alignment with expert visual reads, they misclassify different borderline cases, suggesting they capture complementary information.
We discuss how AI amyloid status predictions may help identify subtle, spatial uptake patterns that global SUVR can miss and how combining these approaches could better support reader confidence. This has direct relevance for clinical trials, where accurate and consistent amyloid assessment is critical for participant selection, monitoring disease progression, and evaluating treatment effects.
IXICO’s ongoing research into integrating AI-derived predictions with SUVR seeks to provide standardized, quantitative support to visual reads, improving confidence in borderline cases and streamlining clinical trial workflows.
