Deep Learning Attacks Joint Degeneration and Osteoarthritis: Musculoskeletal Imaging Research Published in ‘Radiology’

Deep learning has become a powerful tool in radiology in recent years. Researchers at the UC San Francisco Department of Radiology and Biomedical Imaging have started using deep learning methods to characterize joint degeneration and osteoarthritis, which will ultimately reduce the number of total joint replacements. In a recent paper published in Radiology (PubMed) they demonstrate that it is possible to automatically identify (segment) cartilage and meniscus tissue in the knee joint and extract measures of tissue structure such as volume and thickness, as well as tissue biochemistry, by a method know as MR relaxometry. Cartilage and meniscus morphological and biochemical changes are tissue-level symptoms of joint degeneration.

To perform automatic segmentation of cartilage and meniscus, they developed a deep learning model based on the U-Net convolutional network architecture using 638 image datasets. Performance of the automatic segmentation was evaluated using the Dice coefficient overlap with manual segmentation done by many radiologists, which took more than an hour for each data set. The models averaged five seconds to generate automatic segmentations with excellent agreement with those done by radiologists. The precision and agreement between measures of cartilage thickness and biochemistry provided by the deep learning models and those using manual methods was also excellent. Measure of relaxation times (biochemical information) and morphologic characterization of joint tissues (such as thickness and volume of cartilage) are not available in the clinic today, due to the long analysis times required. These advances bring new hope for extracting quantitative information from magnetic resonance images, thus standardizing and making the staging the extent of joint degeneration precise and accurate. These advances will have significant impact in the monitoring and diagnosis of osteoarthritis (OA), a debilitating condition affecting the quality of life for millions of adults and the leading cause of total joint replacements.

Berk Norman, a research data analyst with the Majumdar/Link Lab, was the lead author on the study along with UCSF Department of Radiology and Biomedical Imaging faculty members Valentina Pedoia, PhD, assistant professor and Sharmila Majumdar, PhD, professor and vice chair of research and director of the Musculoskeletal Research Interest Group (RIG).

Their research was published last month in Radiology, the journal of the Radiology Society of North American (RSNA). The study was supported by GE Healthcare IT Business and the Arthritis Foundation (ACL Proof of Feasibility, Trial 6157). Dr. Majumdar’s research is supported by grants from the National Institute of Arthritis and Musculoskeletal and Skin Diseases and the National Institutes of Health. This work has led to translational projects that will directly impact the clinic, and is currently being conducted in collaboration with the Center for Digital Health Innovations, UCSF and GE Healthcare.

Here is a link to the full study on PubMed.