Abstract:Abstract Introduction Insomnia is highly prevalent and difficulty falling asleep is the most common complaint. The transition from wakefulness to sleep is characterized by a shift from sympathetic to parasympathetic regulation, which can be assessed by heart rate dynamics. Using a machine learning approach, we aimed to explore whether heart rate dynamics during sleep onset can indicate the presence of insomnia. Methods We used data from a multi-center cohort study (Sleep Heart Health Study) and included 1835 subjects (age 40-90 years, mean age 62.2±11.2 years, 50.8% male) with overnight polysomnography-based sleep studies and validated sleep scorings. Heart rate data during sleep onset was analyzed using heart rate variability time domain measures (NN, SDNN, RMSSD, pNN20) and a nonlinear measure (multiscale entropy). These measures along with basic demographic features (age, sex and BMI) were used as features in machine learning. We ran four commonly used machine learning approaches, including support vector machines (SVMs), decision tree, random forest, and K-nearest neighbors (KNN), with 80% data for training and 20% for testing. Results Based on the mean of test accuracy of five iterations for each model, the highest accuracy for detecting insomnia was achieved by random forest (mean accuracy of 0.902), followed by SVM (mean accuracy 0.894), decision tree (mean accuracy 0.892), and KNN (mean accuracy 0.891). When using the same features to predict self-reported difficulty falling asleep during the study night, SVM achieved the highest accuracy (0.68) while the accuracy of the other approaches were slightly lower. Conclusion Machine learning models may effectively predict insomnia using demographic features and heart rate data during sleep onset, but unbiased analysis is needed for the next step. With wearable technology, easily accessible heart/pulse rate data may provide valuable detection of physiological alteration during sleep onset and offer an evaluation approach for the efficacy of interventions targeting sleep onset difficulties. Support (if any) This research project was made possible by a grant from the American Academy of Sleep Medicine Foundation (PI: Yan Ma). We acknowledge the National Sleep Research Resource as the resource through which the Data/Datasets were obtained. PW and GY were supported by NIH K24 AT009282 and K24 AT009465 respectively.

Prenatal Cortisol Levels Estimation Using Heart Rate and Heart Rate Variability: A Weak Supervised Learning Based Approach

Development and validation of a machine learning algorithm for predicting the risk of postpartum depression among pregnant women

Using Electronic Health Records and Machine Learning to Predict Postpartum Depression.

Machine Learning Based Stress Monitoring in Older Adults Using Wearable Sensors and Cortisol as Stress Biomarker

Strategies for Reliable Stress Recognition: A Machine Learning Approach Using Heart Rate Variability Features

Detection of maternal and fetal stress from the electrocardiogram with self-supervised representation learning

Identification of major depression patients using machine learning models based on heart rate variability during sleep stages for pre-hospital screening

Automatic classification of fetal heart rate based on a multi-scale LSTM network

Classification of Stress via Ambulatory ECG and GSR Data

Real-Time Psychological Stress Detection According to ECG Using Deep Learning

Personalized Stress Monitoring using Wearable Sensors in Everyday Settings

Exploring Unsupervised Machine Learning Classification Methods for Physiological Stress Detection

Coupling analysis of heart rate variability and cortical arousal using a deep learning algorithm

Learning to Predict Human Stress Level with Incomplete Sensor Data from Wearable Devices

ReLearn: A Robust Machine Learning Framework in Presence of Missing Data for Multimodal Stress Detection from Physiological Signals

0425 Detecting Insomnia and Predicting Difficulty Falling Asleep: Machine Learning of Heart Rate Data During Sleep Onset

A Novel Method for Mental Stress Assessment Based on Heart Rate Variability Analysis of Electrocardiogram Signals

Sleep‐phasic heart rate variability predicts stress severity: Building a machine learning‐based stress prediction model

DeepHeart: Semi-Supervised Sequence Learning for Cardiovascular Risk Prediction

A Physiological Data-Driven Model for Learners' Cognitive Load Detection Using HRV-PRV Feature Fusion and Optimized XGBoost Classification

Emotion and Stress Recognition Utilizing Galvanic Skin Response and Wearable Technology: A Real-time Approach for Mental Health Care