Towards prediction of morphological heart age from computed tomography angiography

TL;DR

This study demonstrates that a morphology-based biomarker of heart age can be derived from CTA images by registering hearts to a common template, extracting dense supervoxel-level features (median density, local volume via the Jacobian determinant, and robust variances), and predicting chronological age with PCA-reduced linear regression. The approach yields a mean absolute error around $2.7$ years and $R^2$ values near $0.38$–$0.44$, with strong agreement between age predictions from whole-heart and sub-regions, supporting the robustness of the morphological heart age signal. Saliency analyses map tissue-density and volumetric features to positive or negative associations with age, enhancing interpretability and highlighting both known and novel regions contributing to aging patterns. The framework presents a promising, interpretable morphological biomarker that could complement chronological age for assessing cardiovascular aging and future health risk, with future work aimed at broader SCAPIS validation and clinical outcome correlations.

Abstract

Age prediction from medical images or other health-related non-imaging data is an important approach to data-driven aging research, providing knowledge of how much information a specific tissue or organ carries about the chronological age of the individual. In this work, we studied the prediction of age from computed tomography angiography (CTA) images, which provide detailed representations of the heart morphology, with the goals of (i) studying the relationship between morphology and aging, and (ii) developing a novel \emph{morphological heart age} biomarker. We applied an image registration-based method that standardizes the images from the whole cohort into a single space. We then extracted supervoxels (using unsupervised segmentation), and corresponding robust features of density and local volume, which provide a detailed representation of the heart morphology while being robust to registration errors. Machine learning models are then trained to fit regression models from these features to the chronological age. We applied the method to a subset of the images from the Swedish CArdioPulomonary bioImage Study (SCAPIS) dataset, consisting of 721 females and 666 males. We observe a mean absolute error of $2.74$ years for females and $2.77$ years for males. The predictions from different sub-regions of interest were observed to be more highly correlated with the predictions from the whole heart, compared to the chronological age, revealing a high consistency in the predictions from morphology. Saliency analysis was also performed on the prediction models to study what regions are associated positively and negatively with the predicted age. This resulted in detailed association maps where the density and volume of known, as well as some novel sub-regions of interest, are determined to be important. The saliency analysis aids in the interpretability of the models and their predictions.

Figures (15)

• Figure 1: Illustration of the main components of the analysis pipeline. Step 1: The images are registered to a common template space. Step 2: A supervoxel segmentation mask is generated, followed by feature extraction of robust statistics (median and robust standard deviation) from each supervoxel. Step 3: Principal component analysis (PCA) is applied to the supervoxel-derived features. Step 4: Linear regression is applied to these spatially standardized low-dimensional features to predict the target value. Step 5: Saliency analysis is performed by mapping the saliency of each feature back into image space to visualize the strength (and sign) of the association between each feature and the target. The entire analysis is applied to males and females separately.
• Figure 2: Illustration of the supervoxels delineated over an axial slice of the female template (a), and features (b,c,d,e) for a randomly selected female subject, mapped to and displayed in the template space.
• Figure 3: Illustration of the first 6 PCA components for the Supervoxel-PCA representations of the female sub-group. Gray (as seen in the background) represents no contribution to the PCA component, black a negative contribution, and white a positive contribution.
• Figure 4: Scatter plot of chronological age prediction results using the supervoxel-based features from the whole heart.
• Figure 5: Pearson correlation between the age prediction of various spatial sub-ROIs, as well as chronological age for the female sub-group.