Surface based deep learning to define HIV-related brain aging. Included investigators

With prolonged survival duration from combination antiretroviral therapy (cART), the percentage of older adults living with HIV (PLWH) will continue to rise in coming years. Despite the viral suppression that can be achieved on cART, PLWH remain at considerable risk of brain injury and cognitive decline. Many lines of evidence indicate accelerated aging in HIV infection, thought to be driven by chronic immune activation. Determining HIV’s impact on the brain and its function has proven challenging due to difficulties evaluating individual differences in brain aging relative to chronological age in PLWH. This has been further hampered by the heterogeneities within the PLWH population, as well as those of the normal aging process. The purpose of this proposal is to advance our understanding of how HIV infection affects brain aging. In order to achieve this purpose, we propose to use our newly developed method, surface-based deep learning (SBDL), which harnesses the computational power of deep learning using convolutional neural network architectures applied to surface meshes of geometric surfaces of the brain. This method provides a data-driven approach to determine an individual’s “brain age” using T1 weighted brain MRI data. Our team is the first group to apply geometric deep learning to image-based applications and this approach has proven successful in the accurate prediction of brain age. We have previously estimated brain age with SBDL in a large sample of healthy participants ranging in age from 6-86 years, achieving state-of-the-art performance (average error less than 4.5 years with a coefficient of correlation of 0.95 when compared to chronologic age). Additional investigations in individuals with dementia as well in Super Agers (individuals who demonstrate relative preservation of cognitive functioning well into advanced years), have shown that “brain age” more accurately reflects cognitive and functional status than chronologic age (manuscript in preparation). Given the similarities between cognitive decline in aging and the similar phenotype observed in PLWH, we hypothesize that SBDL prediction of brain age can serve as a valuable surrogate biomarker of individual differences in cognitive status in the setting of HIV infection. This approach provides a powerful new tool for investigating how HIV infection may accelerate the rate of brain aging and contribute to cognitive decline in PLWH. To test our hypotheses we will implement SBDL brain age prediction in imaging data that has been acquired in the NIH-funded Chicago Early HIV Infection Study. This study has previously acquired extensive imaging data in participants in initial stages of HIV infection and in age-matched seronegative controls. This cohort is ideal for testing our hypotheses because participants have extensive imaging available for SBDL analysis at baseline (early HIV infection) and at well-defined follow-up timepoints. Identification of factors associated with accelerated brain aging in HIV infection will help illuminate potential pathophysiologic mechanisms and inform new therapeutic targets and interventions to minimize long-term cognitive complications and preserve quality of life in PLWH.