Session MD.4

Early Detection of Falling Asleep at the Wheel: A Heart Rate Variability Approach

G Dorfman Furman*, A Baharav, C Cahan, S Akselrod

Tel Aviv University
Tel Aviv, Israel

The long tern experience of our Medical Physics group at Tel Aviv University in characterizing sleep-wake states has allowed us to develop a very sensitive tool for detection of sleep onset and arousals during clinical sleep tests. The heart of the new method is based on real time monitoring of ECG and derivation of the Autonomic Nervous System (ANS) activity time–depended Heart Rate Variability (HRV) decomposition, previously developed by our group. The measurements can be performed by applying existing ECG monitoring technologies, which can be easily installed in a vehicle without disturbing the driver. In this study we check the feasibility of a new approach, able to detect driver's tendency to fall asleep at the wheel, and can alert the drowsy driver before he/she falls asleep (FA). Drowsy drivers do not FA instantaneously. There is a preceding period of measurable performance deterioration. A close functional and anatomical link between the brain centers which govern sleep and ANS activity exists. We found well defined changes which precede FA in bed at night, and developed original and complex algorithms to detect it within seconds. We aim to adapt our sleep algorithms to a simulated driver environment, as a first step in the development of a novel way to predict FA at the wheel based on ECG only. The tests are performed at Shaare Zedek Medical Center on healthy volunteers under sleep deprivation conditions and include a normal night sleep followed by 36 hours of sleep deprivation with alternate simulated drives and Maintenance of Wakefulness Test (MWT). During the entire experiment ECG, EEG, EMG, eyes movement and video are recorded. Preliminary results on 20 FA events tracked from the first 3 volunteers provide promising results. FA events during sleep deprivation in a simulator are detected simultaneously by video and EEG analysis and HRV time-frequency decomposition; there is a consistent and reproducible decrease in the VLF of more than 15% from a baseline, an increase of the LF. HF increases and returns rapidly to its baseline within a few seconds, before each FA. Changes in the time domain show that the HR decreases more than 2 STD from its regional mean value. The described changes in ECG parameters are reproducible and based on a robust and easy to acquire signal. Thus we suggest that the completion of the present study can provide an useful tool for monitoring drivers drowsiness and preventing traffic accidents.

(Abstract Control Number: 183)