Afstudeerproject AI-voorspelling distale radiusfractuur op CT

MSc graduate project

Can we predict distal radius fracture displacement on CT using artificial intelligence?

Background
Displaced distal radius fractures (DRFs) are common and the incidence is increasing due to the ageing population(1, 2). In the Netherlands, the incidence is estimated at 20 per 10.000 persons per year (3).  Two-thirds of the DRFs are displaced and therefore closed reduction and cast immobilization is required at the emergency care department (4).  Since 32-64% of the DRF displace again in the first treatment weeks despite the initial closed reduction, patients are often radiographically reevaluated after 1, 2 and 6 weeks. Often CT’s are performed, amongst others, to evaluate displacement since CT’s are more sensitive and accurate for evaluating the fracture characteristics. If redisplacement occurs after closed reduction, surgical treatment is delayed 2-3 weeks after the trauma. Delayed surgical treatment has an influence on physical functioning and pain. Patients who were immediately operated on compared to initially treated with a cast and delayed operated due to redisplacement, reported less pain and improved physical function after one year (5). This study concludes it is cost-effective to perform immediate surgery in these patients, especially when patients were salaried employees. However, this results in unnecessary surgical treatment of patients with a distal radius fracture that will not redisplace. Identifying patients who are at low risk of redisplacement could avoid unnecessary surgery.

Aim
Deep learning can be used to identify missed fractures at the emergency department (6). Although it is mostly used on conventional radiographs, Computed Tomography (CT) has been used too (7).  Therefore, we aim to predict which distal radius fractures will displace based on CT, using deep learning. We hypothesize if we could identify these patients, patients could be operated on immediately to improve long-term pain and physical functioning, with lower costs. Furthermore, unnecessary surgery can be avoided. 

Project
For this project, we include patients with an initially acceptably reduced distal radius fracture on CT with follow-up imaging to measure redisplacement. Based on the follow-up imaging patients are allocated to either the group with or without pre-defined fracture-redisplacement. The goal of the project is to train a deep learning model to predict redisplacement based on fracture characteristics using CT scan images. This project is a collaboration with the Australian Institute for Machine Learning, Adelaide, Australia, who are worldwide known for their expertise.

Work environment
The MSc project will be conducted at the Department of Orthopaedics and Sports Medicine in collaboration with the Department of Trauma Surgery at the Erasmus MC. Data collection occurs at the Erasmus University MC and University Medical Center Groningen. The study project is a collaboration of the Machine Learning Consortium. You will collaborate with a computer scientist at the Australian Institute for Machine Learning, Adelaide, Australia, who will help with the creation of the deep learning model. Also, for co-supervision of the project and its machine learning aspects, experts from the Biomedical Imaging Group Rotterdam at Erasmus MC TU Delft will be involved.  You will be working in a multi-disciplinary team including image analysis experts, scientists and clinicians.

If you are interested, it will be possible to

  • Attend surgery and other clinical activities during your MSc project

  • Work with world-leading experts in image analysis and orthopedic trauma surgery

  • Improve patient care

  • Visit and collaborate with the Australian Institute for Machine Learning, Adelaide, Australia

Requirements
If you are an MSc student with an affinity for developing and validating advanced techniques for the analysis of medical image data sets, machine learning or deep learning, we are looking for you!

Information and application

Supervisors:

  • Abigael Cohen, PhD student, Department of orthopaedics and sports medicine, Erasmus MC

  • Dr. Mathieu Wijffels, Trauma Surgeon, Erasmus MC

  • Prof. Dr. Wiro Niessen, Biomedical Imaging TU Delft / Erasmus MC

  • Dr. Joost Colaris, Orthopaedic Trauma Surgeon, Erasmus MC

To inform or to apply, please email (your CV along with a motivation letter to) a.cohen.1@erasmusmc.nl