Sensorized Motor and Cognitive Dual Task Framework for Dementia Diagnosis: Preliminary Insights From a Cross-Sectional Study.

BACKGROUND: This study explores the use of novel motor and cognitive dual task (MCDT) approaches, based on upper limb motor function (ULMF) and lower limb motor function (LLMF), to discern individuals with mild cognitive impairment (MCI) or subjective cognitive impairment (SCI) from older adults who are cognitively healthy (OA). OBJECTIVE: The study objectives encompass (1) the exploration of alternatives to the traditional walking MCDT; (2) the examination of various ULMF and LLMF MCDT modalities, incorporating different exercises with varying motor difficulties; and eventually, (3) the assessment of OA in comparison with people with MCI and SCI to acquire more nuanced insights into different stages of the diseases. METHODS: The upper and lower limb motor performances of 44 older adults were evaluated using a wearable inertial system during 5 MCDTs comprising 2 ULMF tasks (forefinger tapping [FTAP] and thumb-forefinger tapping [THFF]) and 2 LLMF tasks (toe tapping heel pin [TTHP] and heel tapping toe pin [HTTP]). The gold standard for MCDT, 10-meter walking (GAIT), was included. We incorporated 5 pooled indices based on MCDTs, demographic data, and clinical scores into logistic regression models. RESULTS: In 2-class classification models (MCI vs OA), HTTP showed the highest accuracy, at 93%; TTHP and TTHF models reached 89% accuracy; and FTAP and GAIT achieved 85% accuracy in distinguishing between the 2 groups of participants. In 3-class classification models (MCI vs SCI vs OA), transitioning from FTAP to THFF improved participant characterization by +5%. TTHP outperformed HTTP by +9%. Furthermore, models effectively identified individuals with MCI, with HTTP achieving 76% recall and TTHP achieving 88% recall. CONCLUSIONS: This study emphasizes the potential of an integrated, sensorized MCDT framework that combines a broader theoretical foundation and task selection with neuropsychological and behavioral data. This approach can enhance our understanding of dementia and provide clinicians with valuable diagnostic tools. Although these tasks demonstrated ease and efficiency, validation in subsequent clinical studies is necessary.