Please login to view this media
- Talk
- 20/09/2022
- UK
Determining Image Usability and Femoral Head Coverage Using Artificial Intelligence: A Future Avenue in Newborn Hip Screening
Description
Abhinav Singh, an orthopedic registrar and PhD student at the University of Oxford, addresses the audience at the BOA about his research focusing on enhancing newborn hip screening through artificial intelligence (AI). He expresses gratitude for the opportunity to present, especially with his father in attendance.
His research titled "Determining Images, Ability and Femoral Head Coverage Using Artificial Intelligence, A Future Avenue for Newborn Screening," aims to improve early diagnosis of hip dysplasia, a common condition in newborns. Abhinav explains the importance of early diagnosis, highlighting that failure to do so can lead to significant complications requiring surgical intervention. Despite existing selective screening processes in the UK, the rates of late diagnoses have not lessened since 1986, prompting the need for improved methodologies, as identified by leading health authorities.
Abhinav discusses the role of AI in diagnostics, expounding on the use of a convolutional neural network (CNN) for analyzing a set of 192 ultrasound images from infants under 12 weeks. The initial anatomical segmentation tasks were performed by two experts, identifying key structures such as the ilium, labrum, and femoral head. The AI demonstrated impressive accuracy, with the ability to predict critical anatomical structures comparably to expert opinions.
Results indicate that the AI discerned the ilium and femoral head with 88% and 98.2% accuracy respectively, and it agrees with human experts in 90% of cases. The model exhibits a sensitivity of 90% and a specificity of 88%, marking it as a potentially superior method of diagnosis compared to existing models. Abhinav concludes by outlining future steps for the research, including upscaling and validation for clinical application, and opens the floor for questions.