Adam Yala

@Radiology_AI Vendors could do better to support local validation studies. You could even imagine doing tiny n-1 trials to build confidence. Results build trust #RadAIchat

T3. How can #AI overcome the lack of trust between some mammographers and CAD? #RadAIchat

@Radiology_AI 3/ Efficient evidence generation is critical #RadAIchat

@Radiology_AI 2/ And for a new AI tool, the old trial doesn't apply. This creates an evidence disincentive, since you can always wait for the next thing. We need better evidence frameworks that are nimbler. This something we're actively working on #RadAIchat

T2. What are some of the biggest challenges facing the practical use of #AI in breast imaging? #RadAIchat

@heacockmd There are also cool opportunities to improve workflow, figure out who could do a same-day workup, how many diag views might be needed, etc. Maggie Chung also had some interest working trying predict contrast from pre-contrast MR series, and there are cool GenAI opportunities

5/ Multimodal & foundation models are emerging, combining breast imaging & clinical data to guide screening, imaging steps, and predict cancer subtypes and treatment responses. #RadAIchat

T1: 1/ Use of AI in screening #mammo/#DBT is one of the best radiology AI use-cases. Recent European trials (MASAI, ScreenTrustCAD) show AI can serve as a 2nd reader without dropping CDR or raising recalls. But what’s next? #RadAIchat