ResearchPod Summary
While the dangers of driving under the influence of active substance intoxication are widely studied, the driving performance of individuals currently undergoing rehabilitation has received far less attention. Chronic stimulant dependence impairs attention, executive function, and motor control, meaning that the recovery and withdrawal phases can also pose significant traffic safety risks. Traditional roadside testing and police enforcement are insufficient for capturing these subtle, ongoing impairments.
To address this gap, this study investigated whether physiological and behavioral monitoring could identify high-risk drivers in rehabilitation. The researchers utilized a low-fidelity driving simulator to evaluate 12 male drivers undergoing methamphetamine rehabilitation (divided into early-stage and long-term treatment groups) and compared them with 12 matched healthy control drivers. To induce varying cognitive workloads and unmask subtle neuro-motor differences, participants navigated driving scenarios under three distinct auditory conditions: upbeat music, calm music, and a baseline silence condition.
During the simulator trials, the system continuously gathered data from two primary non-invasive sources: an eye-tracker to monitor oculomotor dynamics and a Kinect sensor to capture biomechanical postural movements. The raw data underwent rigorous preprocessing, including statistical outlier removal and a moving average filter to smooth trajectories, such as tracking the 3D spatial movements of the driver's left elbow joint. Oculomotor metrics focused on pupil angular velocity and visual fixation durations across various spatial and temporal thresholds.
Statistical analyses revealed distinct behavioral differences between the groups. Specifically, the combined mean displacement of the left elbow joint for the recovering patient groups was more than double that of the healthy control group, pointing to increased physical restlessness or altered steering control. Furthermore, early-stage treatment patients exhibited significantly faster pupil angular movements than controls, while long-term treatment patients showed slower eye movements, suggesting potential long-term physiological adaptations during rehabilitation.
Using four selected behavioral and physiological features, a K-Nearest Neighbors (KNN) classification model was trained and validated through randomized data shuffling. By evaluating various feature combinations and neighborhood configurations, the optimized model achieved an overall classification accuracy of 90%.
Integrating such machine-learning models into Advanced Driver Assistance Systems (ADAS) enables a proactive safety strategy. By continuously monitoring a driver's oculomotor and biomechanical parameters, in-vehicle systems can issue real-time alerts to the driver, passengers, and external networks before a collision occurs, effectively mitigating traffic risks associated with substance rehabilitation.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.