Heart rate variability with circadian rhythm removed achieved high accuracy for stress assessment across all times throughout the day

Yafei Shen ^1,2 Zihan Fang ^1,2 Tao Zhang ³ Feng Yu ^1,2 Ying Xu ^4* Ling Yang ^1,2*

• ¹ School of Mathematical Sciences, Soochow University, Suzhou, Jiangsu Province, China
• ² Center for Systems Biology, Soochow University, Suzhou, Jiangsu Province, China
• ³ Jiangsu Province Engineering Research Center of Molecular Target Therapy and Companion Diagnostics in Oncology, Suzhou Vocational Health College, Suzhou, China
• ⁴ Jiangsu Key Laboratory of Neuropsychiatric Diseases and Cambridge-Su Genomic Resource Center, Suzhou medical college of Soochow University, Suzhou, China

Assessing real-time stress in individuals to prevent the accumulation of stress is necessary due to the adverse effects of excessive psychological stress on health. Since both stress and circadian rhythms affect the excitability of the nervous system, the influence of circadian rhythms needs to be considered during stress assessment. Most studies train classifiers using physiological data collected during fixed short time periods, overlooking the assessment of stress levels at other times. In this work, we propose a method for training a classifier that can identify stress and resting states throughout the day, based on 10 short-term heart rate variability (HRV) feature data obtained from morning, noon, and evening. Collecting and analyzing the heartbeat interval data of all volunteers over three consecutive days, we show that all HRV features exhibit 12-hour and 24-hour circadian rhythms. In addition, the circadian rhythm differences across different days for individuals are relatively small. After removing the circadian rhythm trends from the HRV feature data using the Smoothness Priors Approach (SPA), we train XGBoost models to assess stress.The results show that using detrended data as the training set can improve the overall accuracy of stress assessment across all time periods. When combining data from different time periods as the training dataset, the accuracy of the classifier trained on detrended data increases by 13.67%. These indicate that using HRV features with circadian rhythm trends removed is an effective method for assessing stress at all times throughout the day.