Machine learning techniques for computer-aided classification of active inflammatory sacroiliitis in magnetic...

Area of Science:

• Medical Imaging
• Machine Learning in Healthcare
• Rheumatology

Background:

• Current diagnosis of active inflammatory sacroiliitis in axial spondyloarthritis (axSpA) relies on MRI, but interpretation by experts shows significant variability.
• This variability necessitates the exploration of objective diagnostic tools to improve consistency and accuracy in axSpA assessment.

Purpose of the Study:

• To investigate the efficacy of machine learning algorithms in identifying active inflammatory sacroiliitis on MRI in patients with axSpA.
• To assess the performance of different machine learning classifiers in detecting sacroiliac joint (SIJ) inflammation.

Main Methods:

• A retrospective analysis of 56 sacroiliac joint MRI exams was conducted, with data split into training (80%) and external test (20%) sets.
• Manual segmentation and feature extraction were performed, followed by classification using Support Vector Machine, Multilayer Perceptron (MLP), and Instance-Based Algorithm.
• Feature selection was enhanced using Relief and Wrapper methods, with performance evaluated via 10-fold cross-validation and external validation.

Main Results:

• The Multilayer Perceptron (MLP) classifier, using 6 features selected by the Wrapper method, achieved 100% sensitivity, 95.6% specificity, and 84.7% accuracy on the training dataset.
• On the external test dataset, the MLP classifier demonstrated 100% sensitivity, 66.7% specificity, and 80% accuracy.
• These results indicate strong performance of the MLP model in identifying sacroiliac joint subchondral bone marrow edema.

Conclusions:

• Machine learning methods, particularly the Multilayer Perceptron (MLP), show significant potential for aiding in the detection of active inflammatory sacroiliitis on MRI STIR sequences in axSpA patients.
• The developed MLP model offers a promising tool to enhance diagnostic objectivity and accuracy for sacroiliitis in axSpA.