DeepMiner: Discovering Interpretable Representations for Mammogram Classification and Explanation
This work addresses the need for interpretability in AI-driven medical diagnosis, specifically for mammogram analysis, though it appears incremental as it builds on existing CNN probing methods.
The authors tackled the problem of interpreting deep neural network decisions in mammogram classification by developing DeepMiner, a framework that identifies interpretable units in CNNs aligned with BI-RADS concepts and generates explanations consistent with radiology reports.
We propose DeepMiner, a framework to discover interpretable representations in deep neural networks and to build explanations for medical predictions. By probing convolutional neural networks (CNNs) trained to classify cancer in mammograms, we show that many individual units in the final convolutional layer of a CNN respond strongly to diseased tissue concepts specified by the BI-RADS lexicon. After expert annotation of the interpretable units, our proposed method is able to generate explanations for CNN mammogram classification that are consistent with ground truth radiology reports on the Digital Database for Screening Mammography. We show that DeepMiner not only enables better understanding of the nuances of CNN classification decisions but also possibly discovers new visual knowledge relevant to medical diagnosis.