A Systematic Review of Bone Fracture Detection Models using Convolutional Neural Network Approach
Keywords:Bone Fracture, Digital X-Ray, Machine Learning (ML), Deep Learning, Convolution Neural Network (CNN).
Purpose: Automated processing of digital x-rays, specific to bone fractures, may reduce the involved diagnosis cost as well as provide support to a non-orthopaedic or a small clinician at a remote place to identify a bone fracture and take corrective action.
Method: Machine Learning (ML) approach is establishing itself as a viable technique to automate its diagnosis using digital x-rays. ML approach involving Convolution Neural Network (CNN) is also extensively used in the domain of computer vision. This study follows PRISMA guidelines prescribed to conduct a systematic review. It reviews existing CNN approaches used in the bone fracture detection domain and summarises the findings with respect to specific bone fractures.
Result: This review revealed the existence of different approaches to apply CNN and transfer learning to detect fractures in different types of bones as currently there is no universal approach applicable to detect fractures in different types of bones.
Discussion: The present status of ML domain does not provide a universal method to support all bone fracture scenarios. This limitation may be addressed through synchronization of different CNN approaches into a synthesized hybrid pipeline.