Fig. 1

Example of biased outcome in a diabetic retinopathy automatic screening program. Patients with the same disease stage with different outcomes secondary to slightly different classifications