Introduction: Expanding the Dementia Diagnostic Toolkit
As the global population ages, dementia, including Alzheimer’s disease (AD), presents a growing clinical and economic burden. Early identification and prognosis are essential, yet current biomarkers remain imperfect, often requiring costly or invasive methods. A promising direction is the use of digital biomarkers derived from motor and cognitive assessments. Recent work emphasizes motor function assessment alongside standard neuropsychological tests to detect mild cognitive impairment (MCI) and dementia earlier in the disease course.
The study summarized here investigates a sensorized motor-cognitive dual task (MCDT) framework that integrates upper and lower limb activities with cognitive load. By combining inertial measurement units (IMUs) and a diverse set of motor tasks, researchers aim to generate objective digital motor biomarkers that clinicians can deploy in routine care, research labs, and potentially at home.
What is MCDT and Why It Matters
MCDT leverages the motor-cognitive interference effect: performing a motor task while counting or naming tasks taxes neural networks, exposing subtle cognitive impairment. Traditionally, gait-based MCDTs have dominated this field. However, the current work broadens the toolbox by including upper-limb (ULMF) and lower-limb (LLMF) tasks, extending beyond walking to tapping and fine motor movements. This broader approach may accommodate older adults with mobility constraints while maintaining diagnostic sensitivity.
Study Design at a Glance
The cross-sectional investigation enrolled 44 older adults at memory centers in Nice, France, as part of the Marco-Sens multicentric project. Participants were categorized as OA (healthy), SCI (subjective cognitive impairment), or MCI, with 17 in the MCI group, 17 in SCI, and 10 OA. All underwent neuropsychological testing (MMSE, FCSRT, Digit Span, verbal fluency, Trail Making) and MRI/blood tests. A 5-task MCDT battery included two upper-limb tapping tasks (forefinger tapping FTAP and THFF), two lower-limb tapping tasks (toe tapping TTHP and heel tapping HTTP), and the gold-standard GAIT walking task. Each task was performed under single-task and three dual-task conditions with escalating cognitive load (counting backwards by 1, 3, or 7).
Sensor System and Data Pipeline
The study used a low-cost MEMS-based wearable system with four IMUs (two on the hands, two on the feet). The devices collected data at 100 Hz and transmitted wirelessly for offline analysis. Signal processing involved Butterworth filtering, event detection for each task, and feature extraction across 45 motor parameters per cognitive load level, plus 13 GAIT parameters. A key innovation was the weighted DTC (dual-task cost) that accounts for cognitive performance by incorporating the number of correct responses, yielding a measure that reflects both motor impairment and cognitive engagement.
From Features to Diagnosis: Pooled Indices and Models
To handle the small sample, researchers reduced dimensionality by creating five pooled DTC indices per exercise and then inputting these into logistic regression models. Age and frontal executive function (FAB) were included to adjust for confounding factors. The models compared two-group (OA vs MCI) and three-group (OA vs SCI vs MCI) classifications using upper and lower limb MCDTs, including GAIT as a benchmark.
Key Findings and Clinical Implications
Two striking outcomes emerged:
- Lower-limb MCDTs (LLMF) generally outperformed GAIT in distinguishing MCI from OA and SCI in the two-class problem. Notably, the heel tapping (HTTP) and toe tapping (TTHP) tasks achieved the highest accuracy and even surpassed GAIT in some models.
- Upper-limb MCDTs (ULMF) showed that increasing motor complexity (e.g., THFF vs FTAP) enhanced discrimination between MCI and OA, but did not consistently outperform GAIT in multiclass (OA, SCI, MCI) analyses. The finger-based tasks were more sensitive for certain distinctions, whereas hand movements could sometimes misclassify SCI.
Overall, a sensorized MCDT framework that blends diverse motor tasks with cognitive loading presents a promising, noninvasive, and cost-effective approach to early dementia screening. The 3-class problem (OA, SCI, MCI) remains challenging, signaling the need for larger samples and possibly more nuanced motor tasks to capture the SCI-to-MCI transition with higher fidelity.
Limitations and Future Directions
The authors note several limitations, including modest sample size and group imbalances. They advocate for larger longitudinal studies to validate the predictive power of MCDT pooled indices, explore additional motor loads, and stratify MCI by subtypes. Expanding to home-based monitoring could support a continuum of care from hospital to living room, enabling timely interventions and monitoring of therapeutic effects.
Conclusion: Toward Practical, Home-Integrable Diagnostics
This study supports a sensorized MCDT framework as a feasible, scalable addition to dementia care. By integrating upper- and lower-limb tasks with cognitive challenges, and by leveraging weighted dual-task costs, clinicians may detect subtle cognitive decline earlier and more reliably. The approach aligns with a broader movement toward digital, noninvasive biomarkers that can be deployed across clinical and home settings, ultimately contributing to earlier intervention and improved patient trajectories.
