Objective: To evaluate the efficacy of convolutional neural network (CNN) models combined with six-minute walk test (6MWT) data collected via wearable sensors for distinguishing early-stage Parkinson’s disease (PD) from healthy controls.

Background: PD, a progressive neurodegenerative disorder, presents diagnostic challenges in its early stages due to subtle motor symptoms [1]. Gait analysis using wearable sensors has emerged as a promising tool to capture early PD-specific characteristics, allowing for objective assessment when combined with advanced deep learning techniques [2,3].

Method: Participants included 78 early-stage PD patients and 50 healthy controls. Time-series data from six wearable sensors (upper arms, thighs, thoracic spine, and lumbar spine) during the 6MWT were segmented into 15-second intervals, encompassing straight walking and turning sections. Each 15-second segment of time-series gait data was preprocessed and converted into two-dimensional recurrence plots, resulting in a total of 24 accumulated image segments. A CNN model was then trained on these images to classify the groups [Figure 1]. Data processing and analysis were performed using MATLAB and Python.

Results: The CNN model demonstrated the highest classification accuracy with gyroscope data from the lumbar spine (83.5%), followed by the thoracic spine (83.1%) and right thigh acceleration data (79.5%) [Figure 2]. These findings highlight the potential of specific sensor placements in detecting early PD.

Conclusion: This study highlights the utility of wearable sensors and CNN models as cost-effective, non-invasive tools for early PD detection and progression monitoring. Such approaches may enhance clinical decision-making and patient outcomes.

Figure 1

Figure 2

References: [1] Rehman, R. Z. U., Buckley, C., Micó-Amigo, M. E., Kirk, C., Dunne-Willows, M., Mazzà, C., … & Del Din, S. (2020). Accelerometry-based digital gait characteristics for classification of Parkinson’s disease: what counts?. IEEE Open Journal of Engineering in Medicine and Biology, 1, 65-73.
[2] Pedrero-Sánchez, J. F., Belda-Lois, J. M., Serra-Añó, P., Mollà-Casanova, S., & López-Pascual, J. (2023). Classification of Parkinson’s disease stages with a two-stage deep neural network. Frontiers in Aging Neuroscience, 15, 1152917.
[3] Bailo, G., Saibene, F. L., Bandini, V., Arcuri, P., Salvatore, A., Meloni, M., … & Carpinella, I. (2024). Characterization of walking in mild Parkinson’s disease: reliability, validity and discriminant ability of the six-minute walk test instrumented with a single inertial sensor. Sensors, 24(2), 662.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/detection-of-early-stage-parkinsons-disease-using-convolutional-neural-network-models-and-wearable-sensors-from-the-six-minute-walk-test/