Deep Clustering Activation Maps for Emphysema Subtyping

TL;DR

This paper introduces a deep learning clustering approach using dense features from segmentation networks to improve emphysema subtyping in CT scans, providing high-resolution interpretability and competitive accuracy.

Contribution

The novel method leverages dense features for high-resolution visualization and interpretability in emphysema subtyping, outperforming baseline clustering methods.

Findings

Achieved 43% unsupervised clustering accuracy

Outperformed baseline with 41% accuracy

Comparable to supervised classification at 45% accuracy

Abstract

We propose a deep learning clustering method that exploits dense features from a segmentation network for emphysema subtyping from computed tomography (CT) scans. Using dense features enables high-resolution visualization of image regions corresponding to the cluster assignment via dense clustering activation maps (dCAMs). This approach provides model interpretability. We evaluated clustering results on 500 subjects from the COPDGenestudy, where radiologists manually annotated emphysema sub-types according to their visual CT assessment. We achieved a 43% unsupervised clustering accuracy, outperforming our baseline at 41% and yielding results comparable to supervised classification at 45%. The proposed method also offers a better cluster formation than the baseline, achieving0.54 in silhouette coefficient and 0.55 in David-Bouldin scores.