Monocular Optical-Flow Method for Gait Event Detection and Foot-Strike Pattern Evaluation in Parkinsonian Gait

View at publisher Summarise with VIRAT
Authors: Luca Ceriola, Diego Torricelli, Francisco Molina Rueda, Fabrizio Patanè, Marco Donati, Ilaria Mileti
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Gait event detection and foot strike pattern (FSP) evaluation are critical in clinical, rehabilitation applications, particularly in monitoring pathologies such as Parkinson's disease (PD). While traditional gait analysis methods such as force platforms and optoelectronic systems offer high accuracy, they are limited by cost, setup complexity, and environmental constraints. This paper presents a novel, low-cost, vision-based system that uses a single sagittal-view RGB camera and neural network modules to identify and track subjects, segment the foot closest to the camera and estimate optical flow. The proposed modular pipeline, named Foot Otsu Optical Flow (FOOF), can automatically detect initial contact and toe-off events, as well as classify foot strike patterns (heel, midfoot, and toe strikes) while walking. The method was validated with respect to an optoelectronic system involving 25 participants, 15 with PD and 10 controls, walking on flat terrain and on irregular terrain. The results demonstrate high sensitivity 0.99 ± 0.04, and specificity 1.00 ± 0.00, and temporal errors up to 0.02 ± 0.01 s for gait event detection. Foot strike pattern analysis on flat terrain indicated excellent heel strike identification (sensitivity: 0.93, precision: 1.00). However, midfoot and toe strike detection presented greater challenges due to misclassifications, particularly in complex walking scenarios and among participants with PD. Heel strike identification remained robust on irregular terrain (sensitivity: 0.83, precision: 0.96), though midfoot (sensitivity: 0.95, precision: 0.67) and toe strikes (sensitivity: 0.68, precision 0.59) were less reliably detected. This study demonstrates a new method for accurately detecting gait events and evaluating FSP with a single camera across different terrains in multiple people environments.

Fields of Study & Subjects

Gait Recognition and Analysis Balance Gait and Falls Prevention Lower Extremity Biomechanics and Pathologies

Publication details

Journal / source
IEEE Sensors Journal
Volume
26
Issue
10
Pages
15662–15672
Publication date
1 April 2026
Language
EN