Machine learning applications for multi-scale computed tomography of skeletal tissues

In his doctoral dissertation, Santeri Rytky developed artificial intelligence methods for experimental research on osteoarthritis and computed tomography (CT), three-dimensional X-ray imaging, of calcified tissues. Osteoarthritis is the most common joint disease in the world, affecting millions of people globally. For example, in Finland, the disease is estimated to cost the society nearly two billion euros in expenses and lost income. Dental diseases are also very common and, when left untreated, lead to expensive procedures. Precise hospital CT devices can image the joints and dentition, but even the latest equipment cannot distinguish microscopic tissue damage.

In the initial stages of the research, artificial intelligence methods were developed to analyze images from a precise micro-CT device. The first method assessed disease changes in tissue samples from the knee joint and could be useful in the development of osteoarthritis drugs. The second part focused on characterizing the rabbit knee joint calcified cartilage, a thin tissue area that attaches joint cartilage to bone. A better understanding of calcified cartilage could help in the treatment of certain forms of osteoarthritis, such as cartilage damage extending to the bone.

The latest artificial intelligence method in the dissertation is intended for improving the quality of bone and dental images taken with clinical CT devices. The artificial intelligence model is trained to recognize microscopic structures in human knee specimens and micro-CT images of extracted teeth. In the testing phase, the model refined CT images of the knee, wrist, ankle, and dentition with hospital devices, and the predicted images look promising. Since the goal is to predict more accurate images than is possible with hospital CT devices, evaluating the test images is challenging and requires further research. Few device manufacturers have released the first commercial image-enhancement software, and these methods may have many future applications in more precise disease diagnostics and treatment planning, as well as reducing patient radiation exposure. In summary, the results of the doctoral dissertation take a step towards understanding osteoarthritis and more accurate CT imaging of mineralized tissues.