Deep learning improves radiologist workflow, efficiency determining musculoskeletal MRI protocol

Deep learning and artificial intelligence (AI) are often associated with identifying nodules and classifying images, but a recent study found convolutional neural networks (CNNs) can be utilized in radiology workflows to determine musculoskeletal MRI protocols.

The research was published online April 4 in the Journal of Digital Imaging, by the study’s sole author, Young Han Lee, with the department of radiology at YUHS-KRIBB Medical Convergence Research Institute and Center for Clinical Imaging Data Science at Yonsei University College of Medicine in South Korea.

“These results support using deep learning to assist radiologists in their work by providing timely and highly accurate protocol determinations that only require rapid confirmation,” Lee wrote.

A total of 5258 musculoskeletal MRI exams were collected from a hospital information system’s electronic medical records between January and December 2016. Patient ages, genders, referring departments, exam names and use of contrast agent that matched each test were gathered.

The deep learning CNN model analyzed routine or tumor protocols containing word combinations of the referring department, region, contrast media (or not), gender and age. Results were evaluated using the receiver operating characteristic (ROC) curve and evaluated by a radiologist-confirmed protocol for reference.

Results showed the optimal cut-off values for protocol determination between routine protocols and tumor protocols was 0.5067 with a sensitivity of 92.1 percent, a specificity of 95.7 percent, and an area under curve (AUC) of 0.97. Overall accuracy was 94.2 percent.

Additionally, all MRI protocols were correct in pelvic bone, upper arm, wrist and lower leg MRIs.

“In the field of radiology, machine learning can be applied not only for image analysis but also for patient safety, improving work efficiency, and optimization of radiology workflow,” Lee wrote. “One of the possible applications is to minimize human errors in radiology, and this includes laterality errors of radiologic reports.”