Vigorous Physical Activity Predicts Deep Learning Derived Brain Volumes in 8823 Individuals
Cyrus Raji1, Somayeh Meysami2, Sam Hashemi3, Saurabh Garg3, Nasrin Akbari3, Thanh Duc Nguyen3, Ahmed Gouda3, Yosef Gavriel Chodakiewitz3, Rajpaul Attariwala3
1Washington University in St Louis, 2Pacific Neuroscience Institute, Providence Saint Johns Health Center, 3Prenuvo
Objective:
We explored relationship between vigorous physical activity and deep learning determined brain volumes.
Background:
Physical activity reduces dementia risk and brain volume loss on MRI is an important biomarker for neurodegeneration. It is unknown if vigorous physical activity relates to improved brain structure and neuroprotection.
Design/Methods:
Overall, 8823 healthy participants across 4 sites were scanned on 1.5T MR with a short whole-body MR imaging protocol: whole body coronal T1, STIR from vertex to feet, whole-body axial DWI from vertex to proximal-thighs and axial T2 TSE without fat suppression from skull base to pelvis. Brain sequences were sagittal isotropic MP-RAGE and 2D FLAIR. Deep learning with FastSurfer trained 134 participants aged 27-66 segmented 96 brain regions. Partial correlations modeled vigorous physical activity, defined as sports, fitness and recreational activities that increased respiration and pulse rate for at least 10 continuous minutes, with brain volumes. Analyses controlled for age, sex, and total intracranial volume. Benjamini-Hochberg False Discovery Rate of 5% accounted for multiple comparisons.
Results:
This cohort had an average age of 52.90 ± 13.06 years ranging from 23-84 years with 52.5% men and 47.5% women. Overall, 4269 individuals (48.3%) reported vigorous physical activity at least once per week with 1140 (27%) engaging for 3 days per week. Persons with vigorous activity were younger (mean age 52.08 versus 54.6 for non-vigorous physical activity (p < .00001) and 51.6% men versus 48.3% women undertook vigorous physical activity. After accounting for co-variates and multiple comparisons, vigorous physical activity predicted larger volumes in multiple regions including: gray and white matter volumes (Partial R=.11, p=1.01e-22), hippocampus (Partial R=.07, p= 9.31e-11), frontal cortex (Partial R=.08, p=6.84e-12) temporal lobes (Partial R=.11, 9.72e-22), left entorhinal cortex (Partial R=.07, p=1.62e-9).
Conclusions:
Vigorous physical activity predicted larger deep learning segmented brain volumes, including those relevant for cognition and vulnerable to neurodegeneration.