A machine learning model for multi-night actigraphic detection of chronic insomnia: development and validation of a pre-screening tool

Abstract

We propose a novel machine learning-based method for analysing multi-night actigraphy signals to objectively classify and differentiate nocturnal awakenings in individuals with chronic insomnia (CI) and their cohabiting healthy partners. We analysed nocturnal actigraphy signals from 40 cohabiting couples with one partner seeking treatment for insomnia. We extracted 12 time-domain dynamic and nonlinear features from the actigraphy signals to classify nocturnal awakenings in healthy individuals and those with CI. These features were then used to train two machine learning classifiers, random forest (RF) and support vector machine (SVM). An optimization algorithm that incorporated the predicted quality of each night for each individual was used to classify individuals into CI or healthy sleepers. Using the proposed actigraphic signal analysis technique, coupled with a rigorous leave-one-out validation approach, we achieved a classification accuracy of 80% (sensitivity: 76%, specificity: 82%) in classifying CI individuals and their healthy bed partners. The RF classifier (accuracy: 80%) showed a better performance than SVM (accuracy: 75%). Our approach to analysing the multi-night nocturnal actigraphy recordings provides a new method for screening individuals with CI, using wrist-actigraphy devices, facilitating home monitoring.

Modeling Sleep Quality Depending on Objective Actigraphic Indicators Based on Machine Learning Methods

“…Previous research [ 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 ] proves the relationship between various sensory data and sleep parameters. However, in contrast to this literature, the present study links not only objective sensory measurements and sleep characteristics but also subjective (personal) metrics of perceived sleep quality (PSQI survey results).…”

“…The accuracy difference is because various mathematical variables were used in the models. A previous study [ 39 ] developed models to classify nocturnal awakenings based on actigraph data. The authors used statistical data, entropy, Poincaré plot features, total sleep time, wake after sleep onset, sleep-wake ratio, sleep latency, and sleep efficiency obtained during the experiment, with cohabiting couples where one of the participants had insomnia disorder.…”

A systematic review on automatic identification of insomnia

“…It is typically employed in conjunction with other diagnostic tools and a clinical evaluation to understand the broader context of a patient's sleep disorder. Only a few authors used other signals like actigraphy used by Kusmakar et al (2021), wrist temperature, and body position (Rodriguez-Morilla et al 2019, Sharma et al 2022d). The method used by Hamida et al (2016) utilizing power and ratio of relative power as a feature using the Hjorth parameter showed attractive results.…”

“…In ML, SVM has achieved the highest accuracy as shown in figure 9. An asymptomatic study utilizing natural language processing of electronic medical records (EMR) and a phenotyping study using clustering of time series data derived from wearable devices were both conducted (Park et al 2019, Kusmakar et al 2021. In addition, studies on intervention utilizing smartphone applications that reflect the COVID-19 isolation era were conducted (Philip et al 2020a, 2020b, Ge et al 2020, Philip et al 2020.…”

“…Since the monitoring of PSG signals requires the visitation to sleep centres, a few researchers have come up with theories to detect insomnia using actigraphy data from wearable devices such as Fitbit (Behar 2014, Ramesh et al 2021. Angelova et al (2020) and Kusmakar et al (2021) have also used actigraphy signals as data for the classifiers. Identification using actigraphy signals is beneficial as monitoring can be conducted in a more natural environment and can be made more accessible.…”