Assessing cardiovascular risks from a mid-thigh CT image: a tree-based machine learning approach using radiodensitometric distributions

Abstract

The nonlinear trimodal regression analysis (NTRA) method based on radiodensitometric CT distributions was recently developed and assessed for the quantification of lower extremity function and nutritional parameters in aging subjects. However, the use of the NTRA method for building predictive models of cardiovascular health was not explored; in this regard, the present study reports the use of NTRA parameters for classifying elderly subjects with coronary heart disease (CHD), cardiovascular disease (CVD), and chronic heart failure (CHF) using multivariate logistic regression and three tree-based machine learning (ML) algorithms. Results from each model were assembled as a typology of four classification metrics: total classification score, classification by tissue type, tissue-based feature importance, and classification by age. The predictive utility of this method was modelled using CHF incidence data. ML models employing the random forests algorithm yielded the highest classification performance for all analyses, and overall classification scores for all three conditions were excellent: CHD (AUCROC: 0.936); CVD (AUCROC: 0.914); CHF (AUCROC: 0.994). Longitudinal assessment for modelling the prediction of CHF incidence was likewise robust (AUCROC: 0.993). The present work introduces a substantial step forward in the construction of non-invasive, standardizable tools for associating adipose, loose connective, and lean tissue changes with cardiovascular health outcomes in elderly individuals.

Figure 1

Workflow of the present study with nonlinear trimodal regression analysis parameters Gaussian distribution: from a mid-thigh CT scan, 11 radiodensitometric distributions parameter are extracted and used as features for assessing cardiovascular risks through three tree-based algorithms.

Figure 2

Mean Hounsfield Unit distributions for male and female subjects, with and without chronic heart failure (CHF).

Figure 3

ROC curves for coronary heart disease (CHD), cardiovascular disease (CVD) and chronic heart failure (CHF) classification with K-fold cross-validation and nonlinear trimodal regression analysis by k = 12.

Figure 4

Results from tissue-based machine learning feature importance. (A) Example of a segmented false-color CT cross-section to illustrate the morphology of fat (orange), loose connective (blue), and lean muscle (red) tissue. (B) Total model accuracy (%) for each algorithm and cardiac pathophysiology, visually illustrating (with analogous colors) the compositional accuracy of each model with respect to tissue type. (C) Compositional accuracy (%) for each model with respect to tissue type.