Preprint · Jan 2, 2026

Wearable IMU-based MCI Screening: Optimal Sensor Placement and Static Balance Tasks

By Jamshed M, Shahzad A, KIM K

Mild Cognitive Impairment (MCI) represents an early stage of dementia where subtle motor deficits often accompany cognitive changes. The proposed study systematically evaluates sensor placements and static balance conditions to define a practical minimal-sensor wearable framework for effective MCI screening. Sixty older adults (30 MCI, 30 cognitively normal) completed four stance tasks: eyes-open, eyes-closed, and single-leg stands, while wearing inertial measurement units (IMUs) on the waist, bilateral thighs and bilateral shanks. From the acceleration and gyroscope signals of the sensors, 76 time-domain and frequency-domain features were extracted, from each sensor, to quantify postural sway. A multi-step feature selection pipeline identified a compact subset of the most discriminatory features. Several machine learning classifiers were evaluated in a leave-one-subject-out (LOSO) crossvalidation. The results show that a single IMU on the waist or shank yields the highest individual accuracy of 76.7%, and a dual-sensor fusion (waist+shank) further improves performance to 79.2% in eyes-open condition. Overall, the proposed system consistently distinguished MCI from healthy controls across all evaluated tasks. The results underscore the feasibility of a simple wearable system as an effective, non-invasive tool for early MCI detection.

doi.org/10.36227/techrxiv.176739792.21676372/v1Open accessBiomarkers