A new computer-based method developed to identify gait patterns in children with cerebral palsy can help clinicians use gait for diagnosis and to assess treatment outcomes, according to researchers.
The varied gait features in kids with cerebral palsy pose challenges related to disease management, therapy, evaluation of treatment outcomes, and diagnosis.
Because these gait patterns are difficult to distinguish by the human eye, computerized methods were developed to help classify them.
The study from India, “Optimized Clustering Techniques for Gait Profiling in Children with Cerebral Palsy for Rehabilitation,” published in The Computer Journal, developed a “clustering” method that would enable the diagnosis, assessment, and evaluation of treatment outcomes based on gait patterns in children with cerebral palsy.
Several past studies have attempted to use computer analyses to classify gait patterns based on a technique known as supervised machine learning. But this method is limited since it depends on the number of patients used in the process.
An alternative method, called unsupervised learning or clustering, allows gait information to be classed into groups that have common features. However, it’s difficult to determine the best number of clusters or groups.
An analogy would be trying to classify a large group of apples based on their shapes and sizes. There might be many ways of classifying or clustering them; a group might be large and even, and another might be small and even, and so on.
But once a classification system is established, a computer faced with a new unclassified fruit can organize it based on different clusters.
Researchers analyzed the gait patterns of 156 individuals and used their newly developed clustering technique to find gait profiles. They then tested their method using an unseen test sample.
The results showed that the newly developed method is more powerful at diagnosis and assessing treatment outcomes compared to previously developed gait classification techniques.
“Optimized based gait profile clusters could assist quantitatively in clinical rehabilitation evaluation for the children affected by [cerebral palsy],” researchers wrote.