Deciphering Hand Arthritis: A Machine Learning Approach with Quantitative SPECT-CT Analysis

Arthritis is an inflammation of the joints, which are made up of cartilage, muscles, and ligaments that absorb shock from movement so that the bones can move smoothly between each other, and is accompanied by swelling and heat. Traditionally, radiography has been utilized to pinpoint areas of inflammation, with modern medical imaging modalities like PET/CT and SPECT/CT offering enhanced quantitative data for more precise arthritis diagnosis. This investigation employed a machine learning algorithm to analyze the Maximum standardized uptake value (SUV
max
) of finger and wrist joints derived from SPECT/CT, alongside demographic factors such as age and gender, to detect arthritis presence. The Extra Tree's classification outcomes for distinguishing between arthritis and normal hand conditions achieved 69.04%, 65.55%, and 76.36% for accuracy, AUC, and F1-score, respectively. These findings underscore the potential of SUV
max
computer-aided diagnosis in enhancing the diagnostic accuracy of hand arthritis.