Student thesis: Phd


Fractures of the wrist are usually identified in Emergency Departments (ED) by doctors examining lateral (LAT) and posterioanterior (PA) radiographs. Unfortunately missing such fractures is one of the most common diagnostic errors in EDs, leading to delayed treatment and more suffering for the patient. This is mainly because the majority of patients attending EDs are seen by less experienced junior doctors. This problem is widely acknowledged, so in many hospitals X-rays are reviewed by an expert radiologist at a later date - however this can lead to significant delays on missed fractures which can have an impact on the eventual outcome. There is an urgent need for automated methods to analyse radiographs of the wrist in order to identify abnormalities and thus prompt clinicians, hopefully reducing the number of errors. This project developed the first fully automated system to analyse the wrist in the two standard views (i.e. PA and LAT). The system achieves an encouraging fracture detection rate, with an AUC of 0.93 from LAT view, of 0.95 from PA view, and of 0.96 from both views combined. The project also worked on improving the state-of-art technique Random Forest Regression Voting Constrained Local Model (RFCLM) in order to perform better on overlapping structures in radiographs and showed significant performance improvements when segmenting the radius and ulna in wrist radiographs, and femoral condyles in lateral knee radiographs.
Date of Award1 Aug 2019
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorTimothy Cootes (Supervisor) & Carole Twining (Supervisor)

Cite this