Daichi Hayashi, MD, PhD, an associate professor of Clinical Radiology at the Renaissance School of Medicine at Stony Brook University, is co-investigator and senior author of a research paper, “Improving Radiographic Fracture Recognition Performance and Efficiency Using Artificial Intelligence,” that has been published in the journal Radiology.
The study demonstrates that AI-assisted X-ray interpretation for fracture diagnosis can help improve diagnostic accuracy and efficiency for human readers, and at the same time potentially improve patient care by reducing patient waiting time and diagnostic errors. The research involves authors from many institutions, and is led by Dr. Ali Guermazi, a professor of Radiology at Boston University School of Medicine.
For this study, a large number of X-rays were collected, with and without fractures, from multiple institutions across the United States. The team then used these X-rays to train its AI algorithm to detect fractures. Expert human readers — musculoskeletal radiologists who are subspecialized radiology doctors that receive focused training on reading bone X-rays — defined the gold standard in the study and compared the performance of human readers with and without AI assistance. The team used a variety of readers to simulate real-life scenarios that included radiologists, orthopedic surgeons, emergency physicians and physician assistants, rheumatologists, and family physicians, all of whom read X-rays in real clinical practice to diagnose fractures in their patients. Each reader’s diagnostic accuracy of fractures, with and without AI assistance, was compared against the gold standard. They also assessed the diagnostic performance of AI alone against the gold standard.
It was determined that AI can help physicians and physician assistants in interpreting X-rays in the setting of acute trauma, such as when patients present to the Emergency Room after an injury and suspected fracture. AI-assisted computer programs can automatically detect fractures on X-rays, alert human readers and prioritize the reading of X-rays with positive fractures. Since AI can annotate the X-rays with the suspected location of a fracture, it takes a shorter amount of time for human readers to detect the fracture when compared to reading without AI assistance. The study is specific to adult patients, and more research is needed for fracture detection in pediatric patients.
Emergency Room and Urgent Care Clinics are typically very busy, and patients often have long wait times before they can be seen, receive a diagnosis, and be treated. This AI algorithm can potentially contribute to less waiting time for patients at these facilities.
“It is exciting to realize that AI can be a powerful tool to help radiologists and other physicians to improve diagnostic performance and increase efficiency, while potentially improving patient experience at the time of a hospital or clinic visit,” said Dr. Hayashi. “Our study was focused on fracture diagnosis, but a similar concept can be applied to other diseases and disorders. Our ongoing research interest is how to best utilize AI to help healthcare providers improve patient care.”