In-vehicle Sensing and Data Analysis for Older Drivers with Mild Cognitive Impairment

Moshfeghi, Sonia; Jan, Muhammad Tanveer; Conniff, Joshua; Ghoreishi, Seyedeh Gol Ara; Jang, Jinwoo; Furht, Borko; Yang, Kwangsoo; Rosselli, Monica; Newman, David; Tappen, Ruth; Smith, Dana · 2023 · arXiv

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of detecting mild cognitive impairment (MCI) in older drivers by analyzing real-world driving behavior through unobtrusive in-vehicle sensing. As the population of adults aged 65 and older grows, distinguishing between normal age-related changes and pathological cognitive decline becomes critical for road safety. Traditional clinical evaluations are limited in scope and frequency, often failing to capture subtle, day-to-day functional changes. The authors aim to bridge this gap by developing a low-cost, programmable sensing system that collects longitudinal telematics and vision data to identify early warning signs of cognitive impairment using machine learning. The research involved 236 participants aged 65 or older, including those with and without MCI, over a three-year period. The authors installed custom telematics units based on Raspberry Pi 4 hardware, equipped with GPS, inertial measurement units, and On-Board Diagnostics connectors, into participants' vehicles. These devices collected 7,794 data points across 19 driver behavior indexes, including demographic factors (age, education, BMI) and driving metrics (trip frequency, duration, speed, harsh acceleration, braking, and turning). Participants underwent quarterly clinical assessments, such as the Montreal Cognitive Assessment, to validate cognitive status. Data preprocessing involved quantile normalization and outlier treatment to reduce skewness. The authors employed Random Forest supervised machine learning models to classify MCI status, testing various input combinations ranging from age alone to comprehensive demographic and driving variables. Statistical analysis revealed that drivers with MCI exhibited smoother and safer driving patterns compared to their non-MCI counterparts, suggesting a compensatory mechanism where impaired drivers consciously avoid erratic behaviors. Trip distribution analysis indicated that participants predominantly drove during daylight hours, with half of all trips occurring in the afternoon. The Random Forest models demonstrated varying performance levels; models incorporating driving variables alongside demographic data achieved the highest accuracy (0.86) and area under the curve (0.95). Feature importance analysis identified the number of night trips, total number of trips, and education level as the most influential predictors for distinguishing MCI status. The findings suggest that in-vehicle sensing combined with machine learning offers a viable, unobtrusive method for monitoring cognitive decline in older adults. By identifying subtle behavioral changes in naturalistic driving settings, this approach enables earlier detection of MCI than traditional clinical methods alone. The study highlights the potential for intelligent systems to provide real-time interventions, such as adaptive driving assistance, thereby enhancing safety and prolonging mobility for older drivers. This work contributes to the development of digital biomarkers for cognitive health, supporting both clinical assessment and autonomous vehicle safety protocols.

Key finding

Older drivers with MCI showed smoother, more cautious naturalistic driving than non-MCI peers, and Random Forest models identified night-trip count, total-trip count, and education as the strongest features distinguishing them.

Methodology

naturalistic

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via discover_arxiv on 2026-05-04 (4 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified_with_issues.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• mci dementia driving
• cognitive impairment
• exposure measurement
• telematics crash prediction
• older drivers
• cognitive capacity variation

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Methodological Resource: tool software, validation psychometrics
• Theoretical Contribution: computational model