The neural network learns to differentiate between healthy and inflamed bones using the joints of the fingers.
Researchers have been able to teach artificial intelligence neural networks to distinguish between two different types of arthritis and healthy joints. The neural network was able to detect 82% of healthy joints and 75% of cases of rheumatoid arthritis. When combined with a doctor’s expertise, it could lead to more accurate diagnoses. Researchers intend to further investigate this approach in another project.
This discovery of a team of doctors and computer scientists was published in the journal Frontiers in medicine.
There are many different varieties of arthritis and it can be difficult to determine what type of inflammatory disease affects the patient’s joints. Computer scientists and physicians at Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) and Universitätsklinikum Erlangen have now taught artificial neural networks to distinguish between rheumatoid arthritis, psoriatic arthritis, and healthy joints in an interdisciplinary research effort.
Within the BMBF-funded project “Molecular Characterization of Arthritis Remission (MASK),” a team led by Prof. Andreas Maier and Lukas Folle from the Department of Computer Science 5 (Pattern Recognition) and PD Dr. Arnd Kleyer and Prof. Dr. Georg Schett of the Department of Medicine 3 at the Universitätsklinikum Erlangen was tasked with investigating the following questions: Can artificial intelligence (AI) recognize different forms of arthritis based on joint shape patterns? Is this strategy useful for making more accurate diagnoses of undifferentiated arthritis? Is there any part of the joint that should be inspected more carefully during a diagnosis?
Currently, the lack of biomarkers makes it difficult to correctly classify the relevant form of arthritis. X-ray images used to aid in diagnosis are also not completely reliable because their two-dimensionality is insufficiently accurate and leaves room for interpretation. This is in addition to the challenge of placing the joint under examination for X-ray imaging.
Artificial networks learn using finger joints
To find answers to his questions, the research team focused its investigations on the metacarpophalangeal joints of the fingers - regions of the body that are very often affected early in patients with autoimmune diseases such as rheumatoid arthritis or psoriatic arthritis. A network of artificial neurons was trained using finger scans from high-resolution quantitative peripheral computed tomography (HR-pQCT) to differentiate between “healthy” joints and those of patients with rheumatoid arthritis or psoriasis.
HR-pQCT was selected because it is currently the best quantitative method of producing three-dimensional images of human bones at the highest resolution. In the case of arthritis, changes in bone structure can be detected very precisely, which makes a precise classification possible.
Neural networks could make more targeted treatment possible
A total of 932 new HR-pQCT scans from 611 patients were then used to see if the artificial network could really implement what it had learned: Can it provide a correct assessment of previously classified finger joints?
The results showed that AI detected 82% of healthy joints, 75% of cases of rheumatoid arthritis and 68% of cases of psoriatic arthritis, which is a very high probability of hitting without further information. When combined with the expertise of a rheumatologist, it could lead to more accurate diagnoses. In addition, when cases of undifferentiated arthritis were reported, the network was able to classify them correctly.
“We are very pleased with the results of the study, as they show that artificial intelligence can help us classify arthritis more easily, which could lead to a faster and more targeted treatment for patients. However, we are aware that there are other categories that need to be included in the network. We also intend to transfer the AI method to other imaging methods, such as ultrasound or MRI, which are more readily available, ”explains Lukas Folle.
Hotspots could lead to faster diagnostics
While the research team was able to use high-resolution computed tomography, this type of imaging is only rarely available to physicians under normal circumstances due to space and cost restrictions. However, these new findings are still useful because the neural network has detected certain areas of the joints that provide the most information about a particular type of arthritis, known as intra-articular hot spots. “In the future, this could mean that doctors could use these areas as another piece in the diagnostic puzzle to confirm suspicious cases,” says Dr. Kleyer. This would save time and effort during diagnosis and is already possible, for example, using ultrasound. Kleyer and Maier plan to further investigate this approach in another project with their research groups.
Reference: “Classification based on deep learning of inflammatory arthritis by identifying joint shape patterns — How can neural networks tell us where to do“ clinical deepening ”by Lukas Folle, David Simon, Koray Tascilar, Gerhard Krönke, Anna-Maria Liphardt, Andreas Maier, Georg Schett and Arnd Kleyer, March 10, 2022, Frontiers in medicine.
DOI: 10.3389 / fmed.2022.850552