AI Can Accurately Interpret Contrast-Enhanced Breast MRI

Background: Use of artificial intelligence (AI) methods to analyze radiology images is rapidly developing. One of these areas being evaluated for its clinical potential is the use of AI to evaluate contrast-enhanced (CE) breast MRI. Objective: To determine whether AI methods to analyze CE breast MRI images could provide reliable results for clinical use. Methods: Subjects with known normal and/or benign breast findings and pathological proven small invasive breast cancer Results: 78 CE breast MRI examinations were considered normal or benign, 36 CE breast MRI examinations had benign lesions, and 44 CE breast MRI examinations has biopsy-proven invasive carcinomas Of 44 CE breast MRI examinations with known carcinomas, 40 were correctly identified using the AI methodology, indicating a 90.9% sensitivity. Of 4 carcinomas not correctly identified by the AI methodology, 2 were 5 mm, 1 was 6 mm, and 1 was 8 mm. The AI methodology resulted in a true-negative result in 72 of 78 normal or ben more...