Abstract
Auricular acupuncture, a key technique in integrative medicine, was widely utilized for diagnosing and treating diverse diseases based on the auricular acupuncture map (AAM), which shows the somatotopic organization of different parts of the human body according to the reflex properties of specific auricular points (APs). Auricular electrical skin resistance (AESR) is a known diagnostic indicator used to reflect the body health status. However, current clinical detection tools essentially utilized manually operated, pressure-sensitive, single-point electrical measurement devices and consequently acquired low-repeatability AESR data. Hence, reliable and rapid bio-signals collection at multiple APs has been challenging. This thesis presents a personalized diagnostic device with three-dimensional (3D) distributed electrodes for mapping spatiotemporal ASER signals. Design of the sensor, data processing and analysis from human studies are introduced as follows.First, the delicately designed 3D personalized auricular sensor (3D-PAS) mainly includes two parts: conformable auricular mold and distributed 3D electrodes. The entire sensor fabrication process (which was carried out for 30 different human subjects) includes auricular shape transferring by ear impression molding and 3D-scanning, electrodes layout designing by locating APs and scaling geometric parameters, and one-step prototyping by multi-material 3D-printing. The result of these steps is a stable 3D sensing interface conforming tightly to the ultra-curved human auricular skin. Mechanical analysis, including demonstration of location-specific auricular skin curvature distribution and geometric design (diameter and orientation scaling) of electrode pathways, were also performed to achieve consistent sensing area among multiple electrodes, which enabled excellent measurement repeatability.
Second, using this 3D-PAS device, AESR signals could be simultaneously and stably monitored at multiple pre-located APs on the auricle in real time and were directly read from a self-developed user interface. This device showed superior AESR measurement repeatability with the coefficient of variation (CV) of only 4.9%, i.e., achieving 7 times improvement as compared to commercial detectors. With AESR signals collected at multiple APs, a 3D AESR contour was generated for the first time by a numerical interpolation algorithm to visualize the overall AESR distribution across the entire auricle.
Third, human studies comprising of multi-point AESR measurements were performed to compare ASER signal distribution across different subjects and investigate AESR distribution changes under different body conditions after stationary cycling exercise using 3D-PAS tools prototyped subject by subject. From AESR data collected at 10 APs on both left- and right auricles of 30 volunteers, human subject-specific AESR distributions were observed for the first time. There are mainly four different types of AESR distribution patterns among these subjects. 3D AESR contours show high consistency in the spatial AESR distribution among selected subjects, as well as the large variability among across-class subjects. On the other hand, from the AESR data monitored at 13 APs on 17 human volunteers respectively before intensity-fixed stationary cycling exercise (7km within 20 minutes) and 5, 30, 60 minutes after exercise under the same experimental condition, auricular region-specific AESR changes were observed in 98% of the total 51 tests. 3D AESR contours indicated the AESR distribution changing process during the whole test. Furthermore, to investigate the relation between changes of ASER levels and changes of body conditions, the corelations of AESR with heart rate (HR) and blood pressure (BP) which certainly reflected body physiological changes were also studied by statistical analysis.
Forth, unsupervised machine learning techniques including K-means clustering and Principal Component Analysis (PCA) were utilized to further validate the observations from human studies. 30 human subjects were classified into 4 clusters with a high silhouette score (S-score) of 0.764 during the AESR-based K-means clustering; Datasets from cycling tests and control tests on 17 volunteers were clearly classified into two clusters with an S-score of 0.88 on average, and three clusters just matched to different statuses of body conditions were obtained in analysis of 85.1% of datasets from all cycling tests on volunteers.
In summary, a novel 3D electrical sensing platform for real-time, multi-region AESR monitoring across the entire auricle was designed; data processing in 3D contours generation for spatiotemporal AESR mapping was demonstrated; observations including human-specific AESR distributions, after-exercise auricular region-specific AESR changes, and correlations of AESR with HR and BP were presented for the first time; data analysis by unsupervised machine learning for AESR classification was performed. This platform could facilitate auricular diagnosis and be developed for mapping of more bio-signals in the future. With further investigations, ASER signals have the potential to serve as a new biological marker for healthcare and biomedical applications, including diagnosing abnormal physiological changes, early disease prediction, and therapeutic progress monitoring.
| Date of Award | 31 Aug 2021 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Wen Jung LI (Supervisor) |