Machine learning technique in application and comparison in pediatric fracture healing time / Kedija Seid
Machine learning methods have been used in this study to analyze and predict the required healing time among pediatric orthopedic patients particularly for lower limb fracture. Random forest (RF), Self-Organizing Feature map (SOM), decision tree (DT), support vector machine (SVM) and Artificial Neur...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/9375/ http://studentsrepo.um.edu.my/9375/1/Kedija_Seid.pdf http://studentsrepo.um.edu.my/9375/6/kedija.pdf |
| Summary: | Machine learning methods have been used in this study to analyze and predict the required healing time among pediatric orthopedic patients particularly for lower limb fracture. Random forest (RF), Self-Organizing Feature map (SOM), decision tree (DT), support vector machine (SVM) and Artificial Neural Network (ANN) were used to analyze the data obtained from the pediatric orthopedic unit in University Malaya Medical Centre. Radiographs of long bones of lower limb fractures involving the femur, tibia and fibula from children under twelve years, with ages recorded from the date and time of initial injury. Inputs assessment included the following features: type of fracture, angulation of the fracture, contact area percentage of the fracture, age, gender, bone type, type of fracture, and number of bone involved; all of which were determined from the radiographic images. Leave one out method was used to enhance machine learning models as dataset that was available for this project were limited in numbers. RF is used to select variables affecting bone healing time. To our best knowledge there is no study reported using machine learning method to predict paediatric orthopaedics fracture healing time. Findings from this study identified contact area percentage of fracture, type of fracture, number of fractured bone and age as important variables in explaining the fracture healing pattern. SVM model for predicting fracture healing time outperformed ANN and RF models. Based on the outcomes obtained from the models it is concluded that RF, Decision Tree, SVM, ANN and SOM techniques can be used to assist in analysis of the healing time efficiently. |
|---|