Spectral image clustering on dual-energy CT scans using functional regression mixtures

TL;DR

This paper introduces novel functional data analysis and clustering techniques tailored for dual-energy CT scans, leveraging spectral information for improved tissue characterization and tumor delineation.

Contribution

It develops the first FDA-based spectral clustering method for DECT data, integrating spatial context and energy decay curves for enhanced analysis.

Findings

Effective clustering of DECT scans demonstrated on head and neck cancer data.

Outperforms baseline algorithms in tumor delineation accuracy.

Potential to improve clinical outcome prediction models.

Abstract

Dual-energy computed tomography (DECT) is an advanced CT scanning technique enabling material characterization not possible with conventional CT scans. It allows the reconstruction of energy decay curves at each 3D image voxel, representing varying image attenuation at different effective scanning energy levels. In this paper, we develop novel functional data analysis (FDA) techniques and adapt them to the analysis of DECT decay curves. More specifically, we construct functional mixture models that integrate spatial context in mixture weights, with mixture component densities being constructed upon the energy decay curves as functional observations. We design unsupervised clustering algorithms by developing dedicated expectation maximization (EM) algorithms for the maximum likelihood estimation of the model parameters. To our knowledge, this is the first article to adapt statistical FDA tools and model-based clustering to take advantage of the full spectral information provided by DECT. We evaluate our methods on 91 head and neck cancer DECT scans. We compare our unsupervised clustering results to tumor contours traced manually by radiologists, as well as to several baseline algorithms. Given the inter-rater variability even among experts at delineating head and neck tumors, and given the potential importance of tissue reactions surrounding the tumor itself, our proposed methodology has the potential to add value in downstream machine learning applications for clinical outcome prediction based on DECT data in head and neck cancer.