Pulse of Asia 2022

Speaker's biography and Meeting abstract

Kelvin Tsoi, BSc, PhD
Associate Professor
JC School of Public Health and Primary Care, The Chinese University of Hong Kong, Shatin, Hong Kong
Prof. Kelvin Tsoi is a Digital Epidemiologist, currently serving in both School of Public Health and Primary Care and Big Data Decision Analytics Research Centre under The Chinese University of Hong Kong. His research interests focus on digital innovation in chronic disease management, including mobile and telecare application for hypertension management, technological implementation and social engagement for cognitive screening, artificial intelligent application on electronic health records. He also works as the traditional epidemiologist on evidence-based medicine and population cohort studies.

Prof. Tsoi is the President of The International Society for Digital Health (ISDH) since 2019. It aims to encourage interdisciplinary research with innovative technology between medicine and engineering. The Society has organized annual symposium in connecting experts from different sectors in the world and to promote digital health.  His project entitled “Dementia Screening in 3 Seconds with 1 Simple Drawing” has recently won a Bronze Medal at the 2021 Special Edition of the Geneva International Exhibition of Inventions.


Meeting abstract

Lecture title:
Artificial Intelligence for Classification of Blood Pressure Variability: A New Approach for an Old Idea
Kelvin Tsoi1,2
1. School of Public Health and Primary Care, The Chinese University of Hong Kong 
2. Stanley Ho Big Data Decision Analytics Research Centre, The Chinese University of Hong Kong

Literature shows that the visit-to-visit blood pressure variability (BPV) is one of the risk factors for cardiovascular disease (CVD). The visit-to-visit BPV is generally described by using the standard deviations, the coefficients of variation, the difference between maximum and minimum and the average of BP. However, there is no consensus on BPV severity and classification. Our aim of this study was to apply different machine learning algorithms for BPV level classification to compare with the traditional quantile clustering. Our BP records were extracted from the SPRINT study in the United States and the eHealth cohort in HK. One-third of participants were assigned as high BPV level in traditional quantile classification. However, when we use machine learning algorithms, only 10%-27% are assigned as high BPV levels. Across the different machine learning algorithms, k-means clustering is the most stable with good clustering similarities. This presentation will demonstrate how BPV classification can be improved by machine learning methods for better CVD prediction.