Abstract Background Immuno-inflammation and systemic alterations are key features of chronic diseases. While PET molecular imaging is widely used in precision medicine, conventional analyses are lesion-centric, focusing on detection, localization, and quantification. Such approaches overlook disease-induced homeostatic changes occurring at the whole-body level. Recently, PET connectomics has emerged as a graph-based method to characterize metabolic crosstalk between organs. In this study, we introduce a framework for generating individualized PET-based connectomes, enabling robust assessment of personalized systemic homeostasis. Methods We analyzed routine PET imaging data from a tertiary care center, including patients with advanced systemic disease ( N = 22 highly selected patients with Group I advanced pulmonary arterial hypertension) and 46 matched controls. Our computational framework captures the voxel-wise distributional profile of radiotracer uptake within organs, rather than relying on summary measures. Pairwise metabolic distances between organ distributions were used to construct subject-specific, whole-body metabolic networks - termed connectomes. Machine learning and statistical modeling were applied to evaluate the ability of these networks to distinguish disease states and map multi-organ metabolic interactions. Results Here we show that this framework successfully generates stable, individualized metabolic networks from a single PET scan. A graph-based classifier differentiates patients from controls with 75% accuracy. Notably, metabolic connections involving the right heart emerge as the primary drivers of disease discrimination, consistent with the known pathophysiology of advanced pulmonary arterial hypertension. Group-level analyses corroborate these findings, revealing specific alterations in network connectivity. Conclusions Personalized PET-based connectomics can detect individual-level homeostatic perturbations using standard imaging protocols. This non-invasive approach offers a promising strategy to characterize the systemic impact of chronic diseases and represents a shift from population-level analyses toward truly personalized metabolic phenotyping.
Labarthe et al. (2026) studied this question.