Please login to view this media
- Talk
- 20/09/2022
- UK
Emerging Technologies in Hand Surgery: Automatic Occult Scaphoid Fracture Detection in Wrist MRI
Description
This VTT transcript captures a day at a medical conference, where Dr. Xin Chen from the University of Nottingham presents his research on utilizing artificial intelligence (AI) for automatic detection of scaphoid fractures in wrist imaging through MRI. The presentation is chaired by Alexia Karantana, a clinical academic and surgeon, who introduces the session focusing on technology, data, and recovery in the field of hand surgery.
Dr. Chen explains the limitations of traditional X-ray imaging, particularly highlighting the difficulties in detecting scaphoid fractures, which often remain unseen. He proposes the use of MRI, which is more sensitive for identifying such fractures, but acknowledges the challenge posed by a scarcity of radiologists to review MRI results promptly.
To address this issue, Dr. Chen discusses the development of an AI-based system designed to automate the detection of wrist fractures. He elaborates on the evolution of AI, introducing concepts like rule-based systems, machine learning, and deep learning. Through the use of a convolutional neural network (CNN), Dr. Chen's team has developed a method that not only segments the bones in an image but also identifies fractures with impressive sensitivity and specificity rates, suggesting its potential for integration into clinical practice.
During the session, the audience engages with Dr. Chen, raising questions regarding the balance of sensitivity and specificity in fracture detection, the implications for broader applications beyond scaphoid fractures, and the reliability of the machine learning model given the limited data set. Dr. Chen emphasizes the pilot nature of the study and proposes that with more data, the accuracy of the AI model will only improve. The discussion highlights the promising intersection of technology and medicine, showcasing the potential of AI to enhance diagnostic processes.