To track vital signs on the wrist, a wearable system, named WrisTee, has been developed for remote health monitoring. Using three optical sensors with four light sources, WrisTee allows users to measure 12 photoplethysmogram (PPG) signals at three locations on the radial artery. Two types of PPG signals with opposite polarities were discovered and designated as inphase and invert-phase signals. We provided a unified viewpoint regarding their differences based on the Beer−Lambert law, and showed that both signals can be used for heart monitoring using data analyzed from a selected subject. Using reflective pulse-transition time (R-PTT) and the standard deviation of R-PTT (σR-P T T ), we proposed a method for selecting the optimal wavelength to achieve the best quality signal, thus minimizing storage requirements, power resources, and computational costs. We conducted an experiment on ten subjects to evaluate the feasibility of the proposed method. Our results demonstrated that WrisTee is capable of finding the optimal positions and wavelengths for monitoring vital signs. To automatically detect the PPG phases, six machine learning (ML) models were explored to assess their accuracy for PPG-phase classification. The experimental results show that a convolutional neural network can be the best candidate for phase classification. Hence, it can be integrated into WrisTee for noninvasive health monitoring such as heart rate, heart rate variability, or blood pressure. Our work paves a new direction in bio-signal medical researches by adopting in-phase and invert-phase PPGs for healthcare monitoring.

Introduction Cardiovascular diseases (CVD), diabetes and chronic kidney disease are in among of the most society concerns with a high correlation to blood pressure (BP) factor. As a result, developing a wearable device with BP monitoring function are highly demanded for health caring and monitoring. For measuring BP, traditional methods based on a sphygmomanometer is a gold standard, but it is not suitable for wearable application. So, a solution employed some sensors place along artery paths of a human body and estimated BP from time transfer delay or Pulse Transit Time (PTT) of blood volume through these paths [1]. In practice, it is comfortable if all sensors are located centrally in specifically region and wrist hand is considered the best. For detecting the pulse wave signal of blood, photoplethysmography (PPG) is the king reigning the wearable device market for healthcare because of small form factor, none-electrode-contact requirement and multi-wavelength applications in extracting health indexes such as heart rate, SPO2, Glucose, Hydration levels etc. For estimating BP, a custom PPG sensor in this paper is designed to optimize signal strength collecting and then applied it to measure PTT on radial artery at three wrist’s location as Fig. 1. Custom PPG Sensor Four LEDs with different wavelength (530nm, 660nm, 850nm and 940nm) are employed. The distances between LEDs and a single photo detector (PD) are considered to find the optimized places to detect blood pulse as well as the intensity of light through controlling the electric current on LEDs. The PD is a broadband sensitivity sensor permitting it absorbs all reflecting lights from LEDs on blood flow. Method First, TracePro simulation software is used to find how multi-wavelength lights penetrates on skin and select the optimized distance. Here, we follow the procedure has been deployed in some researches from Paul C.-P. Chao et al. [2]. The distance of LEDs to PD changed from 1.2mm to 4 mm with a 0.1mm step. After that, three well-known locations on radial artery in Traditional Chinese Medicine at wrist namely Cun, Guan and Chi are considered to measure PTT from peak to peak of pulse wave signal and compare each other’s to find which pair is the most suitable for estimating BP. The position of these points and the depth of radial artery had been widely studied in [3]. For BP validation, a 24-hour ambulatory BP monitoring of Oscar 2TM from SunTech Mediacal® is used to extract heart rate, systolic BP and diastolic BP. Six healthy subjects with average age 30 ± 5, height 170 ± 8 cm are recruited. Finally, commercial PPG sensors are employed and compared to our custom PPG sensor. Results and Conclusions From our simulation and experiment, the optimized distance for ratio AC/DC between Red and Green LEDs to PD are 1.8 mm and 2.2 mm, respectively. Likewise, two ratios of Near Infrared LEDs 850nm and 940nm are equivalent at 2.6 mm. Next, for measuring PTT between locations PTT between Cun-Guan, Chi-Guan and Cun-Chi 23 ± 5 ms, 27 ± 3 ms and 42 ± 8 ms respectively. The correlation values of the PTT to BP are r = 0.9 (Cun-Guan), 0.85 (Chi-Guan) and 0.78 (Cun-Chi). Comparing to employ commercial PPG sensor, the ratio AC/DC of signal is higher 1.32%. As a result, our design PPG sensor is suitable to the outcomes of the studies in [2] and [3]. References [1] Mukkamala, Ramakrishna, Jin-Oh Hahn, Omer T. Inan, Lalit K. Mestha, Chang-Sei Kim, Hakan Töreyin, and Survi Kyal, Toward ubiquitous blood pressure monitoring via pulse transit time: theory and practice, IEEE Trans. on Biomedical Engineering 62, no. 8 (2015): 1879-1901. [2] Kao, Yung-Hua, Paul C-P. Chao, and Chin-Long Wey. "Design and validation of a new ppg module to acquire high-quality physiological signals for high-accuracy biomedical sensing." IEEE Journal of Selected Topics in Quantum Electronics 25, no. 1 (2018): 1-10. [3] Kim, Jaeuk U., Yu Jung Lee, Jeon Lee and Jong Yeol Kim, Differences in the Properties of the Radial Artery between Cun, Guan, Chi, and Nearby Segments Using Ultrasonographic Imaging: A Pilot Study on Arterial Depth, Diameter, and Blood Flow, Evidence-based complementary and alternative medicine : eCAM (2015). Figure 1