AI in the diagnosis of hip fractures

Introduction

The emerging research into artificial intelligence (AI) over recent years has demonstrated impressive advancements, with great potential for radiology applications¹. With improvements in chip processing, lower hardware costs and advances in machine learning algorithms; many complex cognitive tasks previously accomplished by humans can now be performed by machines².

Deep learning is a branch of AI inspired by the biological networks of the human brain, which allow artificial neural networks to "learn" from massive amounts of data³. Convolutional neural networks (CNNs) are a class of artificial neural network used to analyse visual imagery by automatically and adaptively learning spatial hierarchies of features through a connected and multi-layered approach⁴.

Hip fractures are a common cause of increased morbidity and mortality in elderly individuals. In 2020, 63,283 hip fractures were reported in the UK National Hip Fracture Database⁵...