Abstract
PurposeObject classification and localization is a key task of computer-aided diagnosis (CAD) tool. Although there have been numerous generic deep learning (DL) models developed for CAD, there is no work in the literature to evaluate their effectiveness when utilized in diagnosing fractures in proximity of joint implants. In this work, we aim to assess the performance of existing classification systems on binary and multi-class problems (fracture types) using plain radiographs. In addition, we evaluated the performance of object detection systems using the one- and two-stage DL architectures.MethodsA data set of 1272 X-ray images of Peri-prosthetic Femur Fracture PFF was collected. The fractures were annotated with bounding boxes and classified according to the Vancouver Classification System (type A, B, C) by two clinical specialists. Four classification models such as Densenet161, Resnet50, Inception, VGG and two object detection models such as Faster RCNN and RetinaNet were evaluated, and their performance compared. Six confusion matrix-based measures were reported to evaluate fracture classification. For localization of the fracture, Average Precision and localization accuracy were reported.ResultsThe Resnet50 showed the best performance with 95% accuracy and 94% F1-score in the binary classification: fracture/normal. In addition, the Resnet50 showed 90% accuracy in multi-classification (normal, Vancouver type A, B and C).ConclusionsA large data set of PFF images and the annotations of fracture features by two independent assessments were created to implement a DL-based approach for detecting, classifying and localizing PFFs. It was shown that this approach could be a promising diagnostic tool of fractures in proximity of joint implants.
Highlights
In 1991 it was suggested that total hip replacement (THR) may be the operation of the century that can provide excellent pain relief and an improved quality of life for patients with severe arthritis [18]
Following a primary THR, Prosthetic Femur Fractures (PFFs) accounts for 10.5% of revision hip arthroplasties [36] and it is predicted that 4.6% of THR patients can be affected by PFF [1]
Two classification experiments of PFFs were evaluated— binary classification to distinguish between fracture and no fracture X-ray images and classification according to Vancouver Classification System (VCS)
Summary
In 1991 it was suggested that total hip replacement (THR) may be the operation of the century that can provide excellent pain relief and an improved quality of life for patients with severe arthritis [18]. With a growing elderly population, the rates of THRs is increasing (approximately 90,000 procedures per year in the UK) [36] accompanied by an unavoidable rise in associated post-operative complications such as Peri-Prosthetic Femur Fractures (PFFs) that occur in 3.5% of patients who undergo THR [1]. PFFs are usually caused by low energy falls in elderly patients, but can be due to implant loosening, osteolysis or stress from an adjacent implant. International Journal of Computer Assisted Radiology and Surgery implant and surgical approach taken, and the fracture image to assess the fracture characteristics and the implant for loosening and osteolysis [27]. The management of PFF varies from non-operative treatment to open reduction and internal fixation (ORIF) to revision of the prosthesis [19]. VCS considers three main fracture features: fracture location, implant loosening and bone quality [4]
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call