RSNA: Deep Learning Model Has Identified Imaging Biomarker for Chronic Stress

TUESDAY, Nov. 25, 2025 (HealthDay News) — A deep learning model has identified an imaging biomarker of chronic stress, according to a study to be presented at the annual meeting of the Radiological Society of North America, being held from Nov. 30 to Dec. 4 in Chicago.

Elena Ghotbi, M.D., from Johns Hopkins University School of Medicine in Baltimore, Maryland, and colleagues obtained data on 2,842 adult participants from the Multi-Ethnic Study of Atherosclerosis whose adrenal glands were fully visualized in the inferior slices of noncontrast chest computed tomography (CT). Deep learning models were applied to the CT scans to calculate total adrenal volume; the adrenal volume index (AVI) was defined as volume divided by height (cm³/m²). Salivary cortisol was collected over two days; total exposure was represented by cortisol area under the curve (AUC).

The researchers found that the deep learning model achieved a Dice score of 0.81 ± 0.09; median adrenal volume was 9.6 ± 4.6 cm³. There was an association seen for higher AVI with greater cortisol AUC, peak cortisol, and allostatic load (β = 0.06, 0.07, and 0.07, respectively). Compared with those with low stress, participants with high perceived stress had 0.23 cm³/m² higher AVI. There was also an association noted for AVI with higher left ventricular end-diastolic mass index (β = 2.65 g/m²). Each 1 cm³/m² increase in AVI was associated with higher risks for heart failure and mortality (hazard ratios, 1.044 and 1.045, respectively).

“This is the very first imaging marker of chronic stress that has been validated and shown to have an independent impact on a cardiovascular outcome, namely, heart failure,” Ghotbi said in a statement.

Press Release

More Information