Related Experiment Video
Updated: Aug 25, 2025
Clinical-oriented Three-dimensional Gait Analysis Method for Evaluating Gait Disorder
Published on: March 4, 2018
A deep-learning approach for automatically detecting gait-events based on foot-marker kinematics in children with
Yong Kuk Kim1, Rosa M S Visscher1, Elke Viehweger2
1Laboratory for Movement Biomechanics, Institute for Biomechanics, ETH Zürich, Zürich, Switzerland.
Insights
A deep-learning approach accurately detects initial contact (IC) events during gait analysis for cerebral palsy patients, offering an automated alternative to manual methods. This method uses foot-marker kinematics to improve the classification of motor deficits.
Area of Science:
- Biomechanics
- Neurology
- Artificial Intelligence
Background:
- Neuromotor pathologies significantly impact locomotion and quality of life.
- Clinical gait analysis is crucial for classifying motor deficits and understanding walking patterns.
- Current methods for detecting gait events like initial contact (IC) and toe-off (TO) rely on manual annotation or force plates, which have limitations.
Purpose of the Study:
- To develop and evaluate a deep-learning Long Short-Term Memory (LSTM) model for automatic detection of IC and TO events.
- To assess the model's performance using foot-marker kinematics from cerebral palsy subjects.
- To compare different input marker combinations for optimal gait event detection across subgroups.
Main Methods:
- A deep-learning LSTM model was trained to detect IC and TO using 3D positions and velocities of four foot markers (hallux, calcaneus, distal second metatarsal, proximal fifth metatarsal).
- Data from 363 cerebral palsy subjects were retrospectively analyzed.
- Various combinations of foot markers were tested across three IC subgroups (heel, midfoot, forefoot).
Main Results:
- The LSTM approach achieved 89.7% accuracy in detecting IC events within 16ms, with an 18.5% false alarm rate for the overall group.
- For TO events, detection accuracy was 71.6% with a 33.8% false alarm rate.
- The TOE|HEE marker combination was optimal for IC detection across subgroups, while TO detection required subgroup-specific markers.
Conclusions:
- Deep-learning LSTM-based detection of IC events using TOE|HEE markers provides an automated, reliable alternative to manual annotation and force plate methods.
- This automated approach overcomes limitations of existing methods, such as operator dependency and restricted measurement capabilities.
- The findings support the use of AI in clinical gait analysis for improved classification of neuromotor deficits.
Abstract:
Neuromotor pathologies often cause motor deficits and deviations from typical locomotion, reducing the quality of life. Clinical gait analysis is used to effectively classify these motor deficits to gain deeper insights into resulting walking behaviours. To allow the ensemble averaging of spatio-temporal metrics across individuals during walking, gait events, such as initial contact (IC) or toe-off (TO), are extracted through either manual annotation based on video data, or through force thresholds using force plates. This study developed a deep-learning long short-term memory (LSTM) approach to detect IC and TO automatically based on foot-marker kinematics of 363 cerebral palsy subjects (age: 11.8 ± 3.2). These foot-marker kinematics, including 3D positions and velocities of the markers located on the hallux (HLX), calcaneus (HEE), distal second metatarsal (TOE), and proximal fifth metatarsal (PMT5), were extracted retrospectively from standard barefoot gait analysis sessions. Different input combinations of these four foot-markers were evaluated across three gait subgroups (IC with the heel, midfoot, or forefoot). For the overall group, our approach detected 89.7% of ICs within 16ms of the true event with a 18.5% false alarm rate. For TOs, only 71.6% of events were detected with a 33.8% false alarm rate. While the TOE|HEE marker combination performed well across all subgroups for IC detection, optimal performance for TO detection required different input markers per subgroup with performance differences of 5-10%. Thus, deep-learning LSTM based detection of IC events using the TOE|HEE markers offers an automated alternative to avoid operator-dependent and laborious manual annotation, as well as the limited step coverage and inability to measure assisted walking for force plate-based detection of IC events.
