SCIENCE
We Don't Start With AI.
We Start With Medicine.
The question should not begin with what an algorithm can do. It should begin with what medicine needs.
CLINICALLY LED DEVELOPMENT
Clinical need before computational capability.
UNMET CLINICAL NEED→CLINICAL QUESTION→MEDICAL EXPERTISE→SCIENTIFIC EVIDENCE→CLINICAL MODEL→COMPUTATIONAL INTELLIGENCE→VALIDATION→CLINICAL APPLICATION→LIFECYCLE LEARNING
BEYOND ACCURACY
Medical AI must be evaluated in context.
Who is the intended patient?
Who is the intended user?
What evidence informs the output?
How are limitations and uncertainty communicated?
How does the healthcare professional understand the output?
What happens when conditions change?