Evaluating the effectiveness of technical adjustments for improving the accuracy of photoplethysmography-based heart rate monitors in cardiac patients

Abstract Background Heart rate (HR) tracking with wrist-worn photoplethysmography (PPG)-based monitors could assist in following up physical activity (PA) in cardiac patients, but the accuracy can vary, particularly during intensive exercise. Currently, limited research has investigated the influence of technical factors such as device positioning, sensor cleanliness and interference from body hair, on the PPG accuracy. Purpose This study assessed whether technical adjustments can improve PPG-based HR accuracy in cardiac patients with low PPG accuracy during exercise, and evaluated the impact of different exercise types on PPG accuracy. Methods A total of 30 patients following a cardiac rehabilitation (CR) programme were monitored with both a PPG-based wrist HR monitor (Fitbit Inspire 2) and an ECG-based reference chest-strap monitor (Polar H10) during their training sessions. For baseline accuracy assessment, each patient completed three training sessions with standard device guidance: wearing the PPG monitor positioned three fingers above the wrist joint, tightened securely but comfortably. Patients with low baseline accuracy – defined as a mean absolute percentage error (MAPE) ≥10% during ≥70% of the training time – entered an optimisation phase involving three technical adjustments: degreasing the PPG sensor, shaving part of the forearm to reduce interference, and fixing the position of the monitor on the wrist with tape and gauze to ensure consistent placement and reduce movement-related artefacts. After three further training sessions, accuracy was reassessed. Subanalyses were conducted for each of the subtypes of exercise performed during trainings. Results Of the 30 patients (mean age: 58.1 years, 33.3% female), 20 (66.7%) demonstrated high PPG accuracy at baseline, with a mean of 88.4±5.2% accurate training time (MAPE <10%). The 10 patients with low baseline accuracy had an average of 54.3±15.7% accurate training time. Following optimisation, three of these patients (30%) improved above the threshold, with an average increase of 17.4±7.1% in accurate measurements (mean post-optimisation: 83.5±5.4%). The remaining seven patients (70%) retained low accuracy, with a mean post-optimisation of 56.4±17.8%. Patients who improved had a higher baseline percentage of accurate training time (mean: 66.0±3.8%) compared to those who did not (mean: 49.3±16.3%), though this difference was not statistically significant (p=0.08) (Figure 1). As shown in Figure 2, exercises involving intense arm movements (i.e. rowing and arm biking) had a lower percentage of accurate PPG measurements (56.6% ± 4.3%) compared to other exercises (78.2% ± 8.8%) during the baseline assessment (p=0.03). Conclusions Reliable PPG-based PA monitoring is only feasible for a subgroup of cardiac patients. Moreover, technical adjustments contribute minimally to improving PPG accuracy, with only 30% of patients with low baseline accuracy showing improvement post-intervention.