DeepMiner: Discovering Interpretable Representations for Mammogram Classification and Explanation

TL;DR

DeepMiner is a framework that uncovers interpretable units in CNNs trained on mammograms, enabling explanations aligned with radiology reports and potentially revealing new diagnostic visual features.

Contribution

The paper introduces DeepMiner, a novel method for identifying and annotating interpretable units in CNNs for mammogram classification, enhancing interpretability and medical insight.

Findings

Units respond to BI-RADS concepts

Generated explanations match radiology reports

Potential discovery of new visual diagnostic features

Abstract

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.