Reproducible PET/CT Radiomic Features Track Treatment Response in TB-Infected Marmosets

Radiomic analysis of preclinical tuberculosis imaging can yield quantitative features that may act as biomarkers of lesion evolution and treatment response, but such features are notoriously sensitive to lesion size, segmentation variability, and acquisition parameters. A new study in marmosets infected with Mycobacterium tuberculosis set out to define the methodological considerations affecting radiomic feature reliability and to identify a reproducible subset of features associated with longitudinal treatment-associated changes.

The researchers analysed longitudinal PET/CT scans from infected marmosets that were either left untreated or treated with the standard of care regimen of isoniazid, rifampicin, pyrazinamide, and ethambutol. Lesions were independently segmented by multiple readers, and features were tested for inter-reader reproducibility using Lin’s concordance correlation coefficient. Potential scanner-associated differences were screened with the Mann-Whitney U-test, and linear mixed-effects models were then applied to characterise within-group temporal trajectories in treated and untreated lesions.

Lesion size proved critical to reliability. Lesions smaller than 16 cubic millimetres yielded no reproducible shape features and substantially fewer reproducible second-order features than larger lesions, underscoring how strongly feature robustness depends on the volume being measured.

Sequential filtering for inter-reader reproducibility and detectable scanner-associated differences excluded 58.9 percent of the initial feature set. Linear mixed-effects models identified 176 features exhibiting significant time-dependent trajectories after multiple-testing correction, a number further reduced to 110 after excluding remaining scanner-level effects. Hierarchical clustering then condensed this set to 26 representative features capturing distinct patterns of treatment-associated change.

Variance decomposition indicated that lesion-level variability exceeded subject-level variability for the evaluated longitudinal measurements, suggesting that individual lesions within an animal can differ more from one another than animals differ from each other over time. The resulting 26-feature set was further evaluated in three additional cohorts of animals treated with alternative regimens or agents.

The authors conclude that the study defines an inter-reader reproducible subset of predominantly PET-derived features that captures treatment-associated changes, supporting standardised radiomic pipelines for longitudinal assessment of therapeutic response in preclinical tuberculosis models. The work is published as open access original research in EJNMMI Research.

Subject of Research: Radiomic feature reproducibility and treatment response in PET/CT imaging of tuberculosis-infected marmosets

Article Title: Defining an inter-reader reproducible radiomic feature set associated with treatment-related changes in PET/CT images of TB-infected marmosets

Article References: Brooks, C. J., Muñoz-Barrutia, A., Abdi, A., Fleegle, J. D., Gomez, F., Weiner, D. M., Pérez-Benito, D., Gutiérrez de Pablo, L., Via, L. E., & Vaquero, J. J. (2026). Defining an inter-reader reproducible radiomic feature set associated with treatment-related changes in PET/CT images of TB-infected marmosets. EJNMMI Research. https://doi.org/10.1186/s13550-026-01530-9

Image Credits: AI Generated

DOI: 10.1186/s13550-026-01530-9

Keywords: radiomics, PET/CT, tuberculosis, marmosets, treatment response, inter-reader reproducibility, lesion segmentation, linear mixed-effects models, preclinical imaging, feature filtering, Defining, inter-reader