Abstract
Purpose :
Brief physical exercise causes immediate and vast changes to various health metrics. One’s immediate response and recovery may be indicative of their ocular or systemic health. Maximizing data collection within a one-hour window provides a practical way of obtaining a deeper understanding of how the body’s response is related to health status. Further, it has been shown that large data sets can be used to train deep learning artificial intelligence (AI) algorithms, as already demonstrated for diabetic retinopathy grading using AI. We have initialized a scalable, one-hour protocol in which various non-invasive health metrics are recorded before and after brief exercise, and our preliminary findings are presented.
Methods :
Twenty four healthy participants initially rested for 15 minutes prior to the first blood pressure (BP) measurement. Each 6-minute following cycle comprised a BP measurement and intraocular eye pressure (IOP) two minutes later via non-contact tonometry. This cycle was iterated three times before exercise, the third concluding with a blood glucose test. Exercise consisted of 5 minutes of cycle ergometry at 70-75% of max heart rate. Post-exercise, 6-minute cycles included glucose measurements at the end of the 1st and 3rd cycles. Other metrics collected for future analysis include heart rate monitoring (www.polar.com), periocular imagery (www.blinkframes.com), and corneal hysteresis (www.reichert.com).
Results :
In all participants, exercise increased systolic blood pressure (p=<0.0001) and remained elevated until Post 3 timepoint; timelines are shown in Fig. 1. Both diastolic blood pressure (p=0.02) and IOP (p<0.0001) were reduced with exercise. Participants with baseline IOP < 15 mmHg exhibited a 15.1% reduction in mean IOP, contrasting with a substantial 23.4% reduction in those with baseline IOP > 15 mmHg. The lower IOP group showed a higher reduction in diastolic blood pressure: 8.1% vs 4.3%. Glucose levels (mg/dL) reduced with exercise, which was similar in both the high and low IOP groups (12.1% and 10.7%).
Conclusions :
A new protocol may be scalable for large data collection needed to train AI algorithms to better identify health conditions. Preliminary data indicate the effect of IOP from exercise depends on initial IOP.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.