At most hospitals in Taiwan, the radiology report is not immediately available, especially at night, according to presenter Dr. Chi-Tung Cheng, a trauma surgeon at Linkou Chang Gung Memorial Hospital. As a result, primary physicians -- typically emergency or surgical specialists -- must decide on their own if the patient has a fracture or not.
"We all have some experience of having a [missed] or delayed diagnosis for hip/pelvic fracture in trauma patients, especially in a busy trauma environment or facing a patient multiple trauma," Cheng told AuntMinnie.com.
Seeking to utilize AI to help avoid these missed hip and pelvic fractures, the researchers developed a deep-learning model that could simultaneously handle both diagnostic tasks on pelvic radiographs. It also shows the user the area of the image that led to its findings. In testing, the algorithm outperformed previous methods, according to Cheng.
"The most exciting result of this study is the possibility that AI and human physicians can be complementary," he noted. "Although the performance of our neural network is not outperforming all the doctors, it still can cover those fractures that human physicians missed."
The hospital is now working on building a computer-aided diagnosis system based on the algorithm, Cheng said. Learn more about their method by sitting in on this presentation.