Phase Wise Classification of Abnormal Gait in Children With Cerebral Palsy Using Hybrid Neural TEMPODE Deep Learning Techniques

The neurological condition known as Cerebral Palsy which affects motor control, frequently results in abnormal gait patterns that have a major negative influence on mobility and quality of life. It is essential to analyse and categorise these abnormalities of gait in order to implement tailored therapeutic interventions. The proposed study presents two Hybrid Neural TEMPODE (Temporal Ordinary Differential Equations) architectures for phase-wise classification of abnormal gait in children with cerebral palsy that combines Temporal Convolutional Networks (TCN) with Neural Ordinary Differential Equations (NODE) and Temporal Fusion Transformers (TFT) with Neural Ordinary Differential Equations. The suggested models classified the abnormal gait classes when the gait data is swing, stance and complete gait data, outperforming TCN, TFT and traditional deep learning methods like Long Short Term Memory model (LSTM) and Gated Recurrent Unit (GRU). Phase-specific feature modelling and strong hybrid architectures are two important advancements that achieved maximum accuracies of 94.61%, 94.40% and 94.29% for stance, swing and full gait cycle data, respectively. The findings highlight how well hybrid neural TEMPODE captures minute changes in gait, opening the door for more sophisticated clinical diagnostic treatments and a deeper comprehension of gait abnormalities in CP.