Keywords: Deep Clustering, Emphysema Clustering, Class Activation Maps
TL;DR: Deep Clustering Activation Maps for Emphysema Subtyping
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 COPDGene study, 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, achieving 0.54 in silhouette coefficient and 0.55 in David-Bouldin scores.
Paper Type: methodological development
Primary Subject Area: Unsupervised Learning and Representation Learning
Secondary Subject Area: Application: Radiology
Paper Status: original work, not submitted yet
Source Code Url: being prepared
Data Set Url: http://www.copdgene.org/
Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.
Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.