Unobtrusive sleep monitoring using bed-sheet pressure sensors

Abstract

At least 50% of the world’s elderly population, whose range is fast growing, experience disturbed sleep. Sleep studies have become an extensive approach serving as a diagnostic tool for health-care professionals. Currently, the gold-standard is Polysomnography (PSG) recorded in a sleep laboratory. However, it is obtrusive, requires qualified technicians, is time consuming, and expensive. With the introduction of commercial off-the-shelf technologies in the medical field, alternatives to the conventional methods which may be now used at home on several nights have been conceived to ensure sleep stages and sleep quality detection. However, the lack of validation or scientific consensus regarding the reliability of these devices remains a challenge for researchers and the industry. Cardio-respiratory and physical activities remain the most promising physiological measurements to detect sleep stages without complete PSG. The statistically proven impacts and budgets related to sleep disorders are phenomenal, showing that the field needs more research. This thesis aims at providing the reader with a multidimensional research perspective on sleep studies and physiological parameters monitoring during sleep using unobtrusive data acquisition techniques and apparatus. In this vein, we present an exhaustive review of developments made in unobtrusive sleep assessment. Additionally, a categorization of current approaches is presented based on methodological considerations, from data acquisition frameworks and physiological measurements, to information processing. We discuss the three main physiological functions that could potentially be explored to advance unobtrusive sleep studies based on autonomous physiological functions, mainly cardiac, breathing, and movements activities. The latter review helped us achieve our three contributions. First we propose an autonomous method for classifying the four state-of-art human body lying postures (HBLP) in healthy adults subjects: supine, prone, left and right lateral, with no sensors or cables attached on the body and no constraints imposed on the subject, using a pressure sensor mattress. In contrast to the majority of previous similar works, prone and supine postures were successfully separated in the classification. We found that using the body weight distribution along with the shape and edges contributes to a better classification performance, and hence, helps separate supine and prone postures. The results are satisfactorily promising towards unobtrusively monitoring the posture for ulcer prevention. The method can be used in sleep studies, post-surgical procedures or applications requiring HBLP identification. Second, we leverage the reliable results of posture classification in order to develop an unobtrusive posture-adaptive in bed breathing rate (BR) monitoring system using bed-sheet pressure sensors. Throughout this contribution, we demonstrated, that with proper signal processing, pressure sensor mattresses could be used interchangeably with respiratory belts, which have been approved for medical use by the American Association of Sleep Medicine (AASM), providing a more convenient solution for both subjects and health professionals. Third, we propose and clinically validate a deep learning based classification method for unobtrusive identification of sleep stages using bed-sheet pressure sensors. Although the results presented in this paper are primary and not yet satisfactory to claim an eventual adoption of bed-sheet pressure sensors in sleep clinics, we believe that the potential of such applications is worth being recognized and further explored. We argue that the proposed method could be a step towards unobtrusive sleep studies that require less resources. Subsequently, we give recommendations and practical steps for future endeavors seeking to bring contributions to unobtrusive sleep studies using pressure sensor mattresses. We discuss limitations and challenges facing current solutions, and we highlight open research areas, which we hope would pave the way for future research endeavors addressing the question: how to assess sleep stages and sleep quality less intrusively, and reliably?

Date	8 Sept 2020
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Jean-Marc Lina (Supervisor) & Georges Kaddoum (Co-supervisor)