An automated convolutional neural network that uses artificial intelligence (AI) can successfully detect scaphoid fractures, a common type of wrist fracture, a new study finds.
This is important, the researchers say, because conventional X-ray may miss a fracture of the small bones of the wrist due to their overlap with the other bones of the wrist, potentially limiting visibility.
“Consequently, scaphoid fractures can be overlooked during initial X-ray examinations,” said study lead author Nils Hendrix, a Ph.D. candidate at the Jeroen Bosch Hospital and Jheronimus Academy of Data Science in the Netherlands.
In the study, “Development and Validation of a Convolutional Neural Network for Automated Detection of Scaphoid Fractures on Conventional Radiographs,” published online on April 28, 2021 in Radiology: Artificial Intelligence.
Hendrix and colleagues compared the performance of an automated convolutional neural network (CNN) to the performance of 11 radiologists in detecting scaphoid fractures on conventional radiographs. One hundred and ninety X-rays were used for the comparison.
The automated CNN is capable of identifying subtle patterns in images that is not discernible to the human eye. Previous studies found that it was actually inferior to the performance of radiologists, but these findings are based on larger datasets and a refinement of the AI algorithm.
For this study, the algorithm also employed class activation maps which show which region of the image the network’s predictions are based on.
Overall, the system had a sensitivity of 78% for detecting fractures with a positive predictive value of 83%. This was comparable to the performance of the 11 radiologists.
Hendrix believes the system can improve clinical care of wrist fractures by reducing the delay of treatment and the onset of complications.
“The system may be able to assist residents, radiologists or other physicians by acting either as a first or second reader, or as a triage tool that helps prioritize worklists, potentially reducing the risk of missing a fracture,” Hendrix said.
“The convolutional neural network may also reduce unnecessary wrist immobilization, performed out of precaution, in more than half of the patients with clinical suspicion for having a scaphoid fracture,” he said.
Hendrix and his colleagues plan on refining the CNN further so it can combine multiple x-ray views for its predictions. They hope the system can eventually be used to detect fractures in other bone structures.
