An algorithm detects signs of osteoarthritis on an MRI scan before the first symptoms appear. With artificial intelligence, treatments could rely on drugs rather than surgery.
An algorithm has been developed through a cooperation between the University of Pittsburgh School of Medicine and Carnegie Mellon University College of Engineering. It detects signs of osteoarthritis on an MRI scan before the first symptoms appear. With the help of artificial intelligence, the world's most common joint disease could in the future be treated preventively with drugs rather than through surgery.
"The gold standard for the diagnosis of arthritis is the x-ray," said Kenneth Urish, Associate Professor of Orthopaedic Surgery (University of Pittsburgh, USA) and Deputy Medical Director of the Bone and Joint Centre (UPMC Magee-Women's Hospital, Pittsburgh, USA). "When the cartilage deteriorates, the space between the bones diminishes. When you see arthritis on x-rays, the damage is already done. It is much easier to prevent the cartilage from disintegrating than to try to make it grow again".
About five million people in Germany suffer from symptoms caused by osteoarthritis. Until now, arthrosis has been treated by inserting a new joint into the affected joint. In the USA, knee joint replacement is the most common surgery for people over 45 years of age. Drugs that prevent pre-symptomatic osteoarthritis from developing into a full-blown joint deterioration are not yet available. Rheumatoid arthritis, on the other hand, can already be treated with drugs. The same is desired for osteoarthritis. Several drugs are in the preclinical phase.
In order to feed the algorithm with data, the scientists drew on data from the Arthrosis Initiative, which had observed thousands of people developing knee osteoarthritis for seven years. In particular, the data were collected from patients who had hardly any signs of cartilage damage at the beginning of the study. Today the researchers know which test persons later developed osteoarthritis. Using this data, the algorithm can detect the first patterns on the MRI scans of pre-symptomatic people at risk of osteoarthritis.
"When physicians look at these images of cartilage, there is no pattern visible to the naked eye, but that does not mean that there is no pattern there. It just means you can't see it with conventional means," said Shinjini Kundu, M.D., Ph.D., lead author of the study, who completed this project as part of her graduate training under the Pitt Medical Scientist Training Programme and the Carnegie Mellon Department of Biomedical Engineering.
To validate this approach, Kundu trained the model on part of the knee MRI data and then tested it on patients from whom the computer had no data. The algorithm predicted osteoarthritis with 78% accuracy from MRIs taken three years before the onset of symptoms.
"Instead of recruiting 10,000 people and observing them for 10 years, we can easily recruit 50 people we know will develop osteoarthritis in two or five years," Urish said. "Then we can give them the experimental drug and see if it stops the development of the disease.
Sources:
1. https://www.upmc.com/media/news/092120-Kundu-Urish-OA
2. enabling early detection of osteoarthritis from presymptomatic cartilage texture maps via transport-based learning Shinjini Kundu, Beth G. Ashinsky, Mustapha Bouhrara, Erik B. Dam, Shadpour Demehri, Mohammad Shifat-E-Rabbi, Richard G. Spencer, Kenneth L. Urish, and Gustavo K. Rohde. PNAS first published on September 21, 2020. https://doi.org/10.1073/pnas.1917405117