Inconsistent Performance of Deep Learning Models on Mammogram Classification

ConclusionOur results demonstrate performance inconsistency across the data sets and models, indicating that the high performance of deep learning models on one data set cannot be readily transferred to unseen external data sets, and these models need further assessment and validation before being applied in clinical practice.
Source: Journal of the American College of Radiology - Category: Radiology Source Type: research