View in EDS

PDGV Dataset: Investigating Deep Learning CNN for Gait-Based Parkinson's Disease Recognition.

Saved in:
Bibliographic Details
Title: PDGV Dataset: Investigating Deep Learning CNN for Gait-Based Parkinson's Disease Recognition.
Authors: Mahfouf, Zohra1 (AUTHOR) z.mahfouf@univ-soukahras.dz, Jarraya, Islem2 (AUTHOR) islem.jarraya@regim.usf.tn, Dhieb, Thameur2,3 (AUTHOR) thameur.dhieb@regim.usf.tn, Neji, Mohamed2,4 (AUTHOR) mohamed.neji@regim.usf.tn, Farhat, Nouha5,6 (AUTHOR) nouha.farhat15@gmail.com, Smaoui, Emna5,6 (AUTHOR) emna.smaoui04@gmail.com, M. Hamdani, Tarek2,7 (AUTHOR) tarek.hamdani@regim.usf.tn, Damak, Mariem5,6 (AUTHOR) mariem_benamar@yahoo.fr, Mhiri, Chokri5,6 (AUTHOR) mhiri.chokri@gmail.com, M. Alimi, Adel2,8 (AUTHOR) adel.alimi@regim.usf.tn
Source: Arabian Journal for Science & Engineering (Springer Science & Business Media B.V. ). Mar2026, Vol. 51 Issue 5, p6897-6917. 21p.
Subjects: Parkinson's disease, Convolutional neural networks, Deep learning, Animal locomotion, Computer vision, Diagnosis, Neurodegeneration, Machine learning
Abstract: Parkinson's disease (PD) is a neurodegenerative disorder characterized by progressive motor impairments, prominently reflected in gait abnormalities. This study presents a preliminary investigation using the newly developed Parkinson's Disease Gait Video (PDGV) dataset to evaluate the discriminative capacity of gait features to distinguish PD patients from healthy controls. The goal is to support the development of noninvasive, vision-based diagnostic tools through effective feature extraction and classification techniques. Experiments conducted on Gait Energy Images derived from PDGV demonstrate that appearance-based representations combined with deep convolutional neural networks can effectively capture pathological gait patterns. The investigation of the performance of VGG, AlexNet, DenseNet, and ResNet from 0 ∘ , 90 ∘ , and 180 ∘ view angles showed that ResNet-101 is the most performant model, achieving accuracy, precision, recall, and F1-score of 0.81, 0.84, 0.64, and 0.73, respectively, from the 90 ∘ view angle. These findings underscore the potential of gait-based analysis for automated PD screening and provide a foundation for future intelligent assistive diagnostic systems. [ABSTRACT FROM AUTHOR]
Copyright of Arabian Journal for Science & Engineering (Springer Science & Business Media B.V. ) is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:Parkinson's disease (PD) is a neurodegenerative disorder characterized by progressive motor impairments, prominently reflected in gait abnormalities. This study presents a preliminary investigation using the newly developed Parkinson's Disease Gait Video (PDGV) dataset to evaluate the discriminative capacity of gait features to distinguish PD patients from healthy controls. The goal is to support the development of noninvasive, vision-based diagnostic tools through effective feature extraction and classification techniques. Experiments conducted on Gait Energy Images derived from PDGV demonstrate that appearance-based representations combined with deep convolutional neural networks can effectively capture pathological gait patterns. The investigation of the performance of VGG, AlexNet, DenseNet, and ResNet from 0 ∘ , 90 ∘ , and 180 ∘ view angles showed that ResNet-101 is the most performant model, achieving accuracy, precision, recall, and F1-score of 0.81, 0.84, 0.64, and 0.73, respectively, from the 90 ∘ view angle. These findings underscore the potential of gait-based analysis for automated PD screening and provide a foundation for future intelligent assistive diagnostic systems. [ABSTRACT FROM AUTHOR]
ISSN:2193567X
DOI:10.1007/s13369-025-10921-4