Regression Concept Vectors for Bidirectional Explanations in Histopathology
This work provides a method for bidirectional explanations in histopathology, which is important for building confidence in AI-assisted medical decision-making, though it appears incremental as it builds on existing concept-based explanation techniques.
The authors tackled the problem of explaining deep neural network predictions in medical histopathology by proposing Regression Concept Vectors (RCVs) to measure network sensitivity to continuous concepts like nuclei texture, and found that nuclei texture is relevant for detecting tumor tissue in breast lymph node samples.
Explanations for deep neural network predictions in terms of domain-related concepts can be valuable in medical applications, where justifications are important for confidence in the decision-making. In this work, we propose a methodology to exploit continuous concept measures as Regression Concept Vectors (RCVs) in the activation space of a layer. The directional derivative of the decision function along the RCVs represents the network sensitivity to increasing values of a given concept measure. When applied to breast cancer grading, nuclei texture emerges as a relevant concept in the detection of tumor tissue in breast lymph node samples. We evaluate score robustness and consistency by statistical analysis.