The goal of this work is to establish a scientific basis for treating crouch gait, one of the most common movement abnormalities among children with cerebral palsy. Crouch gait is characterized by persistent flexion of the knee. It is an inefficient means of locomotion;if not corrected it often leads to bone deformities and serious, life-long physical limitations. We will develop predictive tools to determine the probability that each of the common treatments for crouch gait will improve a subject's excess knee flexion. For each treatment, we will identify a set of predictive variables. We will then identify all subjects from a large database of patients who received the surgery and use these subjects'outcome data to choose a weighting coefficient for each of the biomechanics based predictive variables. We will assess the predictive accuracy when applied to a different set of subjects using cross validation. These tools will predict whether or not a particular surgery will improve a patient's excessive knee flexion. Our preliminary work suggests that 80% prediction accuracy is feasible The success of this project will result in rigorously tested methods to aid treatment planning for crouch gait, and will hopefully produce better, more predictable treatment outcomes. Although multi-joint movement abnormalities such as crouch gait are exceptionally complex, the development of statistical models that predict which patients will benefit from specific surgical treatments is an important and necessary step toward designing more effective treatments.
PROJECT RELEVANCE Crouch gait is one of the most common movement abnormalities among children with cerebral palsy. Surgeries are frequently performed in an effort to improve crouch gait, but it is difficult to predict which patients will benefit from these procedures because the biomechanical factors that cause crouch gait are unknown. This project will result in simulations and statistical models to aid treatment planning for crouch gait, which will hopefully lead to better, more predictable treatment outcomes.
Showing the most recent 10 out of 47 publications
