AI breast cancer diagnostic software performs with high accuracy and helps pathologists improve accuracy and efficiency

Investigators at The Ohio State University Wexner Medical Center in the USA published a retrospective study in Clinical Breast Cancer. Pathologists reviewed 104 breast pathology cases using the Ibex Breast software (Ibex Medical Analytics) which identifies areas of tissue with suspicious pathology including invasive carcinoma, non-invasive lesions, lobular neoplasia and microcalcification. Accuracy, immunohistochemistry (IHC) test use, and time spent were compared to unassisted review, and the standalone performance of Ibex Breast was also assessed. Using AI assistance pathologists improved accuracy 97.1% to 100%, with a statistically significant average reduction of review time of 16.5% with 33% reduction in use of IHC (IHC reduction not statistically signicant). The standalone performance of the AI was high, with AUC of 0.976, 0.976 and 0.953 in detection of invasive carcinoma, lobular neoplasia, and microcalcifications, respectively. This study provides basis for validating routine use in prospective cases.

Read full study