Expert identification of visual primitives used by CNNs during mammogram classification
This work addresses the need for explainable AI in medical imaging for radiologists, though it is an incremental step toward broader interpretability.
The researchers tackled the problem of interpreting internal representations in CNNs for mammogram classification by developing an expert-in-the-loop method, resulting in the identification of visual patterns correlated with medical phenomena like mass tissue and calcificated vessels.
This work interprets the internal representations of deep neural networks trained for classification of diseased tissue in 2D mammograms. We propose an expert-in-the-loop interpretation method to label the behavior of internal units in convolutional neural networks (CNNs). Expert radiologists identify that the visual patterns detected by the units are correlated with meaningful medical phenomena such as mass tissue and calcificated vessels. We demonstrate that several trained CNN models are able to produce explanatory descriptions to support the final classification decisions. We view this as an important first step toward interpreting the internal representations of medical classification CNNs and explaining their predictions.