Detection and Assessment of Abnormal Health Condition by Analyzing Circadian Rhythm and Physiologic Reserve

• Main Authors: Li-Ren Hou, 侯力仁
• Other Authors: 傅立成
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/39805966473597282364

碩士 === 國立臺灣大學 === 資訊工程學研究所 === 102 === Modern society shows an increasing proportion of lifestyle diseases because of un-healthy modern lifestyle. Cancer is one of them and main cause of death in Taiwan. Therefore, self-health management and monitoring is more and more important for modern people. In this research, we focus on two kinds of body status, circadian rhythm and physiologic reserve because they are complementary to each other (short-term and long-term health condition) and related to cancer which is the most serious lifestyle dis-ease. A system is designed to detect and assess abnormal health condition by analyzing circadian rhythm and physiologic reserves. According to the literature, rest-activity cycle is highly correlated with circadian function. Moreover, physiologic reserves can also be inferred by physical activity except for weight. Hence, the wrist accelerometer is used to record physical movement continuously, and signal magnitude area (SMA) of tri-axis acceleration is accumulated every minute to estimate intensity. Besides, load cell is ap-plied to measure weight changes. Totally six rhythm features and five physiologic re-serve features are extracted from 24-h SMA time series data and weight data. They are used to represent status of circadian rhythm and physiologic reserve. Then support vector data description (SVDD) is applied to generate a hyper-sphere boundary which encloses the feature of normal situations. The outliers of hyper-sphere boundary will be detected as abnormal situations. Moreover, the distance between outli-ers and sphere center is defined as deviation degree of abnormal health condition. Fi-nally, deviation degree of circadian rhythm and physiologic reserves will be combined to form a score to represent the current health alert level and then it alerts users or doctors. Through this system we expect to help people to maintain healthy lifestyles more easily and perceive abnormal health condition early to catch the golden hour for treatment.