A study of relationship between Social Determinant of Health and Imaging based Age Estimation using Head CT

Abstract

AbstractBackgroundThe biological age of a person represents their cellular level health in terms of biomarkers like inflammation, oxidative stress, telomere length, epigenetic modifications, and DNA damage. Biological age may be affected by extrinsic factors like environmental toxins and poor diet indicating socioeconomic disadvantage. While biological age can provide a much more accurate risk estimate for age-related comorbidities and general decline in functioning than chronological age, it requires well-established laboratory tests for estimation.MethodologyAs an alternative to laboratory testing for biological age estimation, Incidental medical imaging data may demonstrate biomarkers related to aging like brian tissue atrophy. In this study, we designed a deep learning based image processing model for estimation of biological age from computed tomography scans of the head. We then analyzed the relation between gap in biological and chronological age and socioeconomic status or social determinants of health estimated by social deprivation index (SDI).ResultsOur CNN based image processing regression model for biological age estimation achieves mean absolute error of approximately 9 years between estimated biological and chronological age with -0.11 correlation coefficient with SDI. With the fusion of imaging and SDI in the process of age estimation, mean absolute error is reduced by 11%.ConclusionThe results of our experiments clearly establish a correlation between social determinants of health and the gap between biological and chronological ages.