Related Experiment Video
Updated: Aug 5, 2026
07:42
Dual-Task Stroop Paradigm for Detecting Cognitive Deficits in High-Functioning Stroke Patients
Published on: December 16, 2022
Identifying cognitive impairment in older adults using machine learning on combined fNIRS and motion data during an
Kelsi Petrillo1, Nima Toosizadeh1,2,3
1Department of Rehabilitation and Movement Sciences, School of Health Professions, Rutgers University, 65 Bergen St, Newark, NJ 07107.
Research Square
|July 29, 2026
Summary
Early dementia detection is crucial. This study shows combining motor and brain activity during a dual task can accurately screen cognitive impairment in older adults using machine learning models.
Area of Science:
- Neuroscience
- Gerontology
- Biomedical Engineering
Background:
- Early dementia identification is vital for effective treatment, but distinguishing early cognitive impairment from normal aging is challenging.
- Objective screening tools are needed to aid clinicians in diagnosing cognitive decline in older adults.
- Combining motor and neural data offers a multimodal approach to enhance diagnostic accuracy.
Purpose of the Study:
- To explore the efficacy of classification models integrating motor and functional near-infrared spectroscopy (fNIRS) data for early cognitive impairment screening.
- To assess the performance of machine learning algorithms in differentiating cognitively normal older adults (CNOA) from cognitively impaired older adults (CIOA).
- To evaluate the utility of an upper extremity dual task (UEF) in an objective dementia screening protocol.
Main Methods:
- Participants (CNOA, n=43; CIOA, n=32) underwent a 3-minute rest followed by a 3-minute UEF dual task (serial subtraction and elbow flexion).
- Motor variability (gyroscope) and fNIRS-derived anterior prefrontal cortex connectivity were extracted as features.
- Logistic regression, support vector machine (SVM), and bootstrap aggregated decision trees were employed for cognitive classification.
Main Results:
- Support vector machine (SVM) models demonstrated superior performance in cognitive classification.
- SVM models achieved an average accuracy of 76%, a Receiver Operating Characteristic - Area Under Curve (ROC-AUC) of 0.86, and an F1 score of 69.
- The UEF dual task, when analyzed with classification algorithms, showed promise for objective early dementia identification.
Conclusions:
- The combination of motor and fNIRS-based features during a UEF dual task provides an objective method for screening cognitive impairment.
- SVM classification models show significant potential for accurate early dementia detection in older adults.
- This multimodal approach may improve the timeliness and accuracy of dementia diagnosis, facilitating earlier clinical interventions.
