We describe a fully-automated system for analysing X-rays of the wrist to identify possible fractures. Fractures of the distal radius in the wrist are estimated to be about 18% of the fractures seen in adults and 25% of those seen in children. Unfortunately such fractures are amongst the most frequently missed by doctors in Emergency Departments (EDs). A system which can identify suspicious areas could reduce the number of misdiagnoses. We automatically locate the outline of the radius in both posteroanterior (PA) and lateral (LAT) radiographs, then use shape and texture features to classify abnormalities. We show for the first time that fractures can be better identified in the lateral view, and that combining information from both views leads to an overall improvement in performance.