Sphygmopalpation at specific locations of human wrists has been used as a medical measurement technique in China since the Han Dynasty (202 BC–220 AD). Traditional Chinese medicine (TCM) practitioners can now decipher at least 28 basic pulse patterns using their fingertips. This TCM technique of examining individual arterial pulses by palpation has recently undergone an upsurge in popularity as a low-cost and non-invasive diagnostic technique for monitoring a patient health status. However, none of these 28 supposed pulse patterns have a standard digital representation. I have developed a pulse sensing platform for studying and digitalizing arterial pulse patterns via a TCM approach. This platform consists of a robotic hand with three fingers for pulse measurement and an artificial neural network together with pulse signal preprocessing for pulse pattern recognition. The platforms previously reported by other research groups or current commercially available systems exhibit one or more of the following imperfections: only a single channel for data acquisition, low sensitivity and rigid sensors, lack of control of the applied force, and lack of an intelligent data analytical system in many reported works.

The platform presented here features up to three tactile sensing channels (each with 4×6 sensing pixel arrays) for recording data, pressure-feedback-controlled robotic fingers, and machine learning algorithms. The machine learning algorithms that target on classifying/recognizing TCM pulse patterns are explored to pursue the most suitable one in mimicking TCM doctors’ diagnostics logic based on the data collected by our developed platform. The algorithms range from nonparametric classification using k-means clustering algorithm to deep learning classifiers using a convolutional neural network. On the basis of these studies, I proposed a methodology for obtaining “X-ray” images (i.e., 2D grey-scale images) of pulse information constructed on the basis of the sensing data from three locations and three applied pressures (i.e., mimicking TCM doctors); it contains all arterial pulse information in spatial and temporal spans and could be used as an input to deep learning algorithms. Three types of consistent pulse patterns, namely, “HUA” (滑), “XI” (細), and “CHEN” (沉), as described by TCM doctors, could be identified in a selected group of subjects who were diagnosed by TCM practitioners by applying our developed platform and algorithms. The classification rates are 98.6% in the training process and 97.7% in testing result for these three basic pulse patterns. The high classification rate of the developed platform could lead to further development of a high-level artificial intelligence system incorporating knowledge from TCM — the robotics finger system could become a standard clinical equipment for digitalizing and visualizing human arterial pulses.

The necessity to develop a flexible sensing array for palpation is highlighted after validating the platform and algorithms to digitalize and recognize the human arterial pulse via a TCM approach. Given that the current commercialized pressure sensors do not meet the requirements for high sensitivity, repeatability, resolution, and flexibility, a pressure sensor for sphygmopalpation necessitates thinner and completely flexible materials. A flexible tactile pressure sensor using capacitive sensing mechanism is fabricated using mainly 3D printing technology. An all-in-one printing process of producing a prototype capacitive sensing array is presented using accumulated expertise in diverse material-deposition processes, including micro-dispensing, sputtering, and 3D printing. The fabricated sensor array has 256 sensing elements over a 1600 mm2 sensing area on a flexible substrate of 0.3 mm thickness. The spatial and temporal pressure distribution are demonstrated to show the sensor’s capability for human physiological monitoring in the form of finger pressing and arterial pulse signal. According to the results, the fabricated sensing array can detect approximately 0.5–2.0 psi static pressure with the corresponding maximum capacitance change of 69.31 pF to 191.40 pF, and the estimated sensitivity is 0.17 kPa⁻¹. The arterial pressure is also captured by using 9 out of the 256 elements on the sensing area with the normalized peak-to-peak value of 6.67 pF. Finally, a circuit control system to collect the output signal of the large number of sensing elements is also implemented, and the acquired signal is wirelessly transferred to a computational platform via Bluetooth.