Description
Chronic arthritis is a heterogeneous group of diseases characterized by long-lasting inflammation of joints. The objective of the project is to create a computer aid diagnostic system that can automatically assess the activity of synovitis. This system will eliminate human dependent discrepancies and also will reduce the cost of ultrasound images evaluation, save physicians’ time and will result in increased availability of health care personnel at lower costs. Novel methods for automated analysis and measurement of ultrasound images will improve the quality and the repeatability of assessments and will support better control of the disease. New research initiatives, but also new multidisciplinary research works related to Medicine and Computer Science are expected. In a series of work packages under the project, the donor project partners will, in collaboration, focus on developing technology for automated assessment of joint synovitis activity which will support medical ultrasound and Power Doppler examinations used in rheumatology.
Summary of project results
Rheumatoid arthritis is the most common rheumatic disease with arthritis and causes substantial functional disability in approx. 50% of patients after 10 years. Accurate measurement of the disease activity is crucial in providing treatment and care to patients. The MEDUSA project is focused on developing a computer-aided diagnostic system that will support the assessment of the severity of synovitis. The system is related to the automated assessment of the severity of synovitis. The prototype of a computer-aided diagnostic system and the algorithms that are essential for the analysis of the ultrasonic images of the finger joints is one of the scientific outcomes of the MEDUSA project. The Medusa Evaluation System prototype uses bone, skin, joint and synovitis area detectors to create a mutual structural model that is based on the evaluation of synovitis. Several algorithms, which support the semi-automatic or automatic detection of the bone region, were prepared and a system that uses the statistical data processing approach in order to automatically localise the regions that are of interest. The detection has been improved to 55% after further development. We have assessed the agreement ratio between all of the human observers with the Fleiss’ kappa coefficient and in pairs between all of the human assessors, automated assessments and standards using the Cohen’s kappa statistic. The agreement ratio was high in the human/standard pairs, moderate between all human of the assessors and slight between the standard and automated assessments. However, in all of the models, p was lower than 0.05, which means that the agreement ratio was significant and not random. The Spearman’s correlation was calculated between all of the observers and the standard. There was a high correlation between all of the pairs of human observers and the human/standard. The correlation was moderate between the automated assessments and all of the human assessors and the standard. In all of the cases, the correlation was significant. Within the MEDUSA project, we developed the first working automated synovitis assessment tool in history. Although it needs to be improved to provide a more adequate assessment, it is a significant breakthrough in musculoskeletal ultrasound. We foresee the following possible development areas: 1.Further improving the detection of the synovitis localisation and classification 2.Adding power Doppler analysis 3.Developing detections for other joints
Summary of bilateral results
The project has been a successful bilateral collaboration where two sides complemented each other within their respective competences. The Norwegian partners provided medical background knowledge and expertise regarding the significance and diagnosis of synovitis and examining ultrasound images. They also brought into the project proficiency in data acquisition, management, organization and labelling. They conducted medical studies, collected large amount of data needed for training and testing the automation algorithms, annotated and organized the data. They also conducted the final blind tests and studies that have demonstrated the significance of the achieved results. The Polish partners was responsible for the algorithmic and software engineering part of the projects. They created novel algorithms and methods in image recognition, pattern analysis and machine learning, developed software, conducted optimizations, studies and tests. Both partner organized scientific conferences concerning synovitis arthritis.