Machine learning to quantify habitual physical activity in children with cerebral palsy

Abstract

Aim To investigate whether activity‐monitors and machine learning models could provide accurate information about physical activity performed by children and adolescents with cerebral palsy (CP) who use mobility aids for ambulation. Method Eleven participants (mean age 11y [SD 3y]; six females, five males) classified in Gross Motor Function Classification System (GMFCS) levels III and IV, completed six physical activity trials wearing a tri‐axial accelerometer on the wrist, hip, and thigh. Trials included supine rest, upper‐limb task, walking, wheelchair propulsion, and cycling. Three supervised learning algorithms (decision tree, support vector machine [SVM], random forest) were trained on features in the raw‐acceleration signal. Model‐performance was evaluated using leave‐one‐subject‐out cross‐validation accuracy. Results Cross‐validation accuracy for the single‐placement models ranged from 59% to 79%, with the best performance achieved by the random forest wrist model (79%). Combining features from two or more accelerometer placements significantly improved classification accuracy. The random forest wrist and hip model achieved an overall accuracy of 92%, while the SVM wrist, hip, and thigh model achieved an overall accuracy of 90%. Interpretation Models trained on features in the raw‐acceleration signal may provide accurate recognition of clinically relevant physical activity behaviours in children and adolescents with CP who use mobility aids for ambulation in a controlled setting. What this paper adds Machine learning may assist clinicians in evaluating the efficacy of surgical and therapy‐based interventions. Machine learning may help researchers better understand the short‐ and long‐term benefits of physical activity for children with more severe motor impairments.

Machine Learning to Quantify Physical Activity in Children with Cerebral Palsy: Comparison of Group, Group-Personalized, and Fully-Personalized Activity Classification Models

“…Third, the current study only trained RF classifiers and did not benchmark the performance with other supervised or unsupervised learning algorithms. RF classifiers were chosen in the current study because they are ensemble learning models which have been shown to provide accurate activity recognition in children with CP [ 22 , 54 ], as well as those with typical development [ 55 , 56 ]. As the aim of the current study was to evaluate the influence of personalization on the classification accuracy, it was important that models were trained using the same supervised learning algorithm.…”

A Framework for User Adaptation and Profiling for Social Robotics in Rehabilitation

“…Decision trees form a group of algorithms that are composed of logical rules that allow for easily extracting human knowledge, in comparison to other algorithms that are not self-explanatory (for example, support vector machines). Specifically, we have used a decision tree classifier executed with the scikit-learn library for Python [52], an implementation of the C4.5 algorithm [53], one of the most studied and used decision tree algorithms. This algorithm generates a tree-shaped model that, depending on the angles of the different joints, reaches different leaves related with specific labels, deciding whether the pose performed by the patient is correct or not.…”

“…Although there has been important progress in the symbiosis between Machine Learning and rehabilitation with social robots, there is still a long way to go. Previous research has focused on studying the use of Machine Learning techniques for recognising certain behaviours in children and adolescents with cerebral palsy with information for different accelerometers placed on different parts of the body [ 54 ]. This research showed promising results of this combination.…”

Machine learning for monitoring and evaluating physical activity in cerebral palsy

“…Goodlich et al 3 have applied three supervised learning algorithms (decision tree, support vector machine, random forest) to investigate whether activity monitors and machine learning models could provide accurate information about physical activity performed by children and adolescents with CP who use mobility aids for ambulation. The algorithms were trained on features in the raw acceleration signal on the wrist, hip, and thigh.…”

“…The present study reminds us of the undeniable need for collaboration between clinicians and machine learning experts. 3,4 To establish personalized diagnoses and treatments, doctors need to analyse a large amount of data that can be acquired from movement sensors, 1,3 images, assessment tools, and medical and biological records. 2,4 A machine learning model could be the appropriate support in analysing these heterogeneous clinical data collected from patient's electronic health records.…”