Regression Concept Vectors for Bidirectional Explanations in Histopathology

Event Sponsor: 
Computing, Environment and Life Sciences Seminar
Start Date: 
Aug 23 2018 - 11:00am
Building 240/Room 1416
Argonne National Laboratory
Mara Graziani
Speaker(s) Title: 
HES-SO Valais
Thomas Brettin

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.​