Cluster Activation Mapping with Applications to Medical Imaging

TL;DR

This paper introduces CLAM, a novel method combining deep clustering with visualization techniques to interpret what features influence cluster assignments in medical imaging, aiding trust and discovery.

Contribution

The work develops CLAM, a new approach for visualizing deep clustering results in medical images, enabling better understanding and validation of cluster-based findings.

Findings

Successfully visualized cluster features in lung CT scans

Identified new sarcoidosis subtypes based on imaging data

Validated approach with simulation and real patient data

Abstract

An open question in deep clustering is how to understand what in the image is creating the cluster assignments. This visual understanding is essential to be able to trust the results of an inherently complex algorithm like deep learning, especially when the derived cluster assignments may be used to inform decision-making or create new disease sub-types. In this work, we developed novel methodology to generate CLuster Activation Mapping (CLAM) which combines an unsupervised deep clustering framework with a modification of Score-CAM, an approach for discriminative localization in the supervised setting. We evaluated our approach using a simulation study based on computed tomography scans of the lung, and applied it to 3D CT scans from a sarcoidosis population to identify new clusters of sarcoidosis based purely on CT scan presentation.