A deep-learning approach for automatically detecting gait-events based on foot-marker kinematics in children with cerebral palsy -- which markers work best for which gait patterns?
Table of Contents
-
About The Project
-
Getting Started
- License
Normalisation of gait cycles is a key towards using clinical gait analysis as a tool to monitor neuromotor pathologies. Most commonly, this normalisation is achieved by detecting gait events, such as initial contact (IC) or toe-off (TO), through either manually annotated video data, or based on thresholds of ground reaction forces. This study developed a deep-learning long short-term memory approach to automatically detect IC and TO based on the foot-marker kinematics of $363$ Cerebral Palsy subjects (age: $11.8\pm{3.2}$). Different input combinations of four foot-markers (HLX, HEE, TOE, PMT5) were evaluated across three subgroups exhibiting different gait patterns (IC with the heel, midfoot, or forefoot). Overall, 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 based detection of IC events using the TOE|HEE marker 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.
This git repository consists python codes for training the network. MATLAB was used for performance analysis.
- Clone/download this repo
git clone https://github.com/ykukkim/W04_DL.git
- Trainin_LSTM -> consists python scripts from training to validating the performance of models.
Python version: 3.9
conda install pytorch==1.8.1 torchvision==0.9.1 torchaudio==0.8.1 cudatoolkit=10.2 -c pytorch