Iliopectineal Line Fracture Detection for Computer-Aided Acetabular Fracture Classification

Proper recognition, description and classification of acetabular fractures allow orthopedic surgeons to develop a treatment plan for patients with acetabular trauma. The most widely used classification system of acetabular fractures was developed in the 1960s by Judet and Letournel. It relies on recognizing anatomical landmarks of the pelvis using three conventional radiographs. Despite the simplicity of Judet and Letournel classification system, accurate classification of acetabular fracture remains difficult and challenging even for expert surgeons. A literature review have shown that several authors have developed diagnostic algorithms to assist students, residents, radiologists, and surgeons in classifying acetabular fractures according to Judet and Letournel system, but the literature does not mention, as far as we know, any application that automatically diagnoses and recognizes the pattern of acetabular fractures based on conventional radiographs. This paper concerns the first step towards such application which consists of recognizing one major anatomical landmark on conventional radiographs, the iliopectineal line. In the study, two machine learning algorithms (Support Vector Machine and Neural Network) were used on anteroposterior view x-ray images in order to recognize and diagnose if the iliopectineal line is disrupted or not. Accuracy in both methods exceeds 91%.