Participant Study Data for Cardiovascular Metric Estimation with a Developed PPG Device

With the ongoing advancements of photoplethysmography (PPG) devices for clinical and domestic applications, there exists a distinct lack of publicly available literature addressing the best possible hardware design for these devices. In this project, this question was answered by investigating the hardware design configuration of a PPG device, especially focussing on the light source wavelength, light source brightness, number of light sources, photodetector lens shape, and sensor-to-skin contact pressure. This device was evaluated on the quality of the PPG signals it measures, as well as its ability to accurately estimate the user’s heart rate, blood oxygen saturation, and blood pressure over different skin tones. A participant study was also conducted to collect cardiovascular metric data from 110 participants with varying skin tones, which was used to develop a neural network for blood pressure estimation as well as determine the best hardware configuration of the PPG device. This resulted in the hardware configuration that consisted of a red-and-infrared light-emitting diode (LED) system, a flat lens photodiode, and high sensor-to-skin contact pressure. This configuration was setup with its two red LEDs at a medium level of brightness and four infrared LEDs at a high level of brightness and was primarily used to measure data on the user’s fingertip. The consequent estimations’ mean absolute errors were 3.511 BPM for heart rate, 0.615% for blood oxygen saturation, and 11.279 mmHg for the systolic blood pressure and 6.802 mmHg for the diastolic blood pressure estimations, respectively. Finally, the selected PPG device’s bias to skin tone was determined. It was concluded that the device had very little bias to skin tone, due to it having a 3.817 dB variance over all the skin tones tested. With the successful identification of the most optimal hardware design configuration, future devices are expected to perform with improved accuracy, usability, and versatility in clinical settings. Furthermore, it provides the necessary groundwork for future research to explore additional applications of PPG devices.

随着光电容积脉搏波描记法（PPG）设备在临床和家庭应用领域的持续发展，关于这些设备最佳硬件设计的公开文献却鲜有问世。在本项目中，我们通过研究PPG设备的硬件设计配置，特别是聚焦于光源波长、光源亮度、光源数量、光电探测器镜头形状以及传感器与皮肤接触压力等方面，解答了这一问题。该设备在测量PPG信号质量以及准确估计用户心率、血氧饱和度和血压在不同肤色上的能力方面进行了评估。此外，还开展了一项参与者研究，收集了110名不同肤色的心血管计量数据，这些数据被用于开发血压估计的神经网络，并确定PPG设备的最佳硬件配置。该配置包括红光和红外发光二极管（LED）系统、平面镜头光电二极管以及高传感器与皮肤接触压力。该配置中，两个红光LED以中等亮度、四个红外LED以高亮度设置，主要用于测量用户指尖的数据。随后估计的均方根误差分别为心率3.511 BPM、血氧饱和度0.615%、收缩压11.279 mmHg以及舒张压6.802 mmHg。最后，确定了所选PPG设备对肤色的偏差。结果表明，由于该设备在所有测试的肤色上具有3.817 dB的方差，其对肤色的偏差非常小。在成功识别出最优化硬件设计配置后，未来设备有望在临床环境中实现更佳的准确性、易用性和多功能性。此外，这为未来研究探索PPG设备的其他应用提供了必要的理论基础。