New AI Model Detects Diseases and Explains Itself to Doctors and Patients
-
An AI model developed at Beckman Institute can accurately identify tumors and diseases in medical images and explain its reasoning via visual maps. This enhances doctor-patient communication and early detection.
-
The model uses deep neural networks to make nuanced decisions like a human. However, these networks are usually "black boxes" that can't explain their reasoning.
-
This new model produces an "equivalency map" highlighting regions that contributed most to a diagnosis. Doctors can check the AI's work and explain it to patients.
-
In tests detecting breast tumors, eye disease, and heart problems, the model was as accurate as other AI systems but more transparent.
-
Researchers hope future versions can detect anomalies all over the body to further improve medical diagnoses and build doctor-patient trust.