Files
Dewandre_64181300_2020.pdf
UCLouvain restricted access - Adobe PDF
- 4.51 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Treatment selection for patients after hip or knee arthroplasty is time-consuming and repetitive for physiotherapists. This study gathers the experience and work of health professionals in order to structure and automate treatment selection using data collected by the moveUP company. Forty different exercises can be suggested by the physiotherapists on a daily basis. In order to assign the most suitable treatment to a given patient, several machine learning models such as the decision tree, K-nearest neighbours, the multi-layer perceptron, the support vector machine and the random forest will be investigated and tested for each suggested exercise. This study shows that the best suited models to predict hip and knee treatments are the decision tree and the random forest. This work eventually leads to the development of a production-ready tool that provides a list of exercises tailored to the patient's needs. This new selector considers parameters such as relevance and redundancy of the treatment from one day to the next. Furthermore, safety rules are implemented to take a special care of patients with comorbidities such as osteoporosis.