ABSTRACT
OBJECTIVE: This study was conducted to measure radiographic joint space width and to estimate erosion in the hands of patients with rheumatoid arthritis. It showed that joint space width, homogeneity, and invariant moments are parameters to discriminate between the normal and the rheumatoid joint. METHODS: In order to measure the joint space width and to estimate erosion in the finger joint, 32 radiographic images were used - 16 images for training and 16 images for testing. The joint space width was measured in order to quantify the joint space narrowing. Also, homogeneity and invariant moments was computed in order to quantify erosion. Finally, artificial neural networks were constructed and tested as a classifier distinguishing between the normal and the rheumatoid joint. RESULTS: The joint space width of normal was 1.04+/-0.15 mm and the width of patients with rheumatoid arthritis was 0.94+/-0.15 mm. The Homogeneity of normal was 16568.83+/-2669.83 and invariant moments were 6843.45+/-2937.55. They were statistically difference (p<.05). Using these characteristics, artificial neural networks showed that they discriminate between normal and rheumatoid arthritis (AUC=0.91). CONCLUSION: Measuring joint space width, estimating homogeneity, and invariant moments provide the capability to distinguish between a normal joint and a rheumatoid joint.