Advancing Dementia Screening with Sensorized Motor-Cognitive Dual Tasks
The search for accessible, noninvasive biomarkers to identify cognitive decline earlier than traditional neuropsychological tools is driving researchers toward digital health solutions. A recent cross-sectional study investigates a comprehensive sensorized motor-cognitive dual task (MCDT) framework that combines upper and lower limb movements with cognitive challenges to distinguish healthy aging (OA) from subjective cognitive impairment (SCI) and mild cognitive impairment (MCI). By deploying wearable inertial sensors and a robust analytical pipeline, the work explores a broader MCDT repertoire beyond the conventional walking task, aiming to reveal subtle motor-cognitive interference linked to early dementia pathology.
What is the Sensorized MCDT Framework?
At the heart of the framework is a battery of five motor tasks. Two upper-limb tasks involve finger tapping (forefinger tapping and thumb-forefinger tapping), while two lower-limb tasks involve toe tapping and heel tapping. A gold-standard gait assessment (GAIT) completes the set. The study extends the typical dual-task paradigm by integrating different motor loads (upper vs. lower limb) under controlled cognitive interferences of varying difficulty (counting backwards by 1, 3, or 7). Each task is performed both as a single motor task and under dual-task conditions, with cognitive load designed to stress executive functions such as working memory and attention. Wearable MEMS-IMU devices capture precise kinematic data, enabling objective biomarkers to be extracted for analysis.
Methods in Brief: Participants, Protocol, and Analysis
The Marco-Sens multicentric project recruited 44 older adults from memory centers in Nice, France. Participants underwent comprehensive neuropsychological evaluation, including MMSE, FCSRT, Digit Span, verbal fluency, and Trail Making Tests, along with a frontal assessment battery (FAB). Based on clinical criteria, participants were classified as OA (n=10), SCI (n=17), or MCI (n=17). The experimental protocol employed five MCDTs (FTAP, THFF, TTHP, HTTP, GAIT) with 4 cognitive-load levels (CL0–CL3). Data processing used a 100 Hz sampling rate and a four-stage pipeline: signal preprocessing, segmentation, feature extraction, and a weighted dual-task cost (DTC) that accounts for cognitive performance and engagement via a proxy of correct responses (Nc).
A key methodological contribution is the pooling of DTC metrics into five indices per exercise, reducing the feature space while preserving discriminative information. These pooled indices feed a multivariate logistic regression framework, with age and FAB scores included to control for demographic and executive-function influences. Leave-one-out cross-validation provided estimates of model performance across two- and three-class classification problems: OA vs MCI and OA vs SCI vs MCI.
Key Findings: MCDT Tasks and Diagnostic Accuracy
The results indicate that novel MCDT tasks, especially lower-limb protocols, can outperform the traditional GAIT task in certain discrimination scenarios. Notably, two LLMF tasks (toe tapping and heel tapping) achieved higher accuracy than GAIT in distinguishing MCI from OA in the two-class model, with HTTP (heel tapping) reaching up to 93% accuracy. For multiclass challenges (OA, SCI, MCI), lower-limb tasks again showed robust performance, with TTHP (toe tapping) often providing strong discrimination. Upper-limb tasks demonstrated that increasing motor complexity (from FTAP to THFF) can improve detection of MCI against OA, albeit with more mixed results when SCI is included. Overall, the framework suggests that a diverse, task-rich MCDT battery offers more nuanced insights than walking alone.
Among methodological advances, the weighted DTC approach, which factors cognitive efficiency and commitment into the cost computation, yielded more accurate classification than traditional methods that rely solely on motor performance. This refinement reduces false negatives/positives by acknowledging that cognitive effort and engagement influence motor output under dual-task conditions.
Implications for Clinical Practice and Research
Several implications emerge from these preliminary findings. First, expanding MCDT protocols beyond gait can improve early detection of cognitive decline and provide more sensitive biomarkers for MCI and SCI. Second, lower-limb tasks, when embedded in a wearable platform, offer a practical and scalable route for home-based or remote assessment, potentially increasing screening reach in aging populations. Third, integrating cognitive-performance proxies into DTC calculations provides a more faithful representation of participants’ cognitive reserves and engagement, which could improve longitudinal tracking of disease progression or response to interventions.
Limitations include the modest sample size and cross-sectional design, which constrain causal inferences and generalizability. The study underscores the need for larger, longitudinal studies to validate the robustness of pooled MCDT indices across diverse populations and disease subtypes. Future work should also explore optimal motor loads and task order to maximize diagnostic yield, as well as potential combinations of MCDT indices with other digital biomarkers, neuroimaging, and genetic data.
Conclusion: A Flexible, Noninvasive Tool for Dementia Screening
Sensorized MCDT protocols hold promise as a low-cost, noninvasive augmentation to current dementia screening. By leveraging a spectrum of upper- and lower-limb tasks and incorporating cognitive performance into the analysis, clinicians may gain early, actionable insights into cognitive decline trajectories. As the field moves toward home-based monitoring and continuous assessment, this framework could become a cornerstone of a continuum of care, enabling timely intervention and closer tracking of therapeutic effects.
