Validity of a computer-assisted manual segmentation software to quantify wrist erosion volume using computed tomography scans in rheumatoid arthritis

BACKGROUND To investigate the performance of conventional radiography (CR) for the detection of bone erosions of wrist in rheumatoid arthritis (RA) using multidetector computed tomography (CT) as the reference method and to evaluate the validity of a computer-assisted manual segmentation (outlining) technique to quantify erosion volume on CT scans. METHODS Twenty five RA patients and six controls underwent CT and radiographic evaluation of the dominant wrist on the same day. CT was performed by using a 64 GE light Speed VCT power. Wrists images were evaluated separately and scored for the presence of erosions according to the Outcome Measures in Rheumatology Rheumatoid Arthritis MRI Scoring System (RAMRIS) and the Sharp/van der Heijde scoring method. Measurements of bone erosion volumes were obtained using OsiriX medical imaging software. The mean value of the volumes of the CT bone erosions detected at two readings was used to calculate inter-rater agreement. RESULTS The overall sensitivity, specificity and accuracy of radiography for detecting erosions were 25.5%, 98.3% and 70.1%, respectively. Using computer-assisted manual segmentation (outlining) technique, erosion volume on CT measurements per subject was ranged from 0.001 cm³ to 2.01 cm³. Spearman's RAMRIS score of each wrist bones in all subjects (n = 25) were correlated with the total erosion volume on CT (p < 0.0001), with the ratio between erosion volume and the corresponding bone volume on a percentage basis (p < 0.0001). The total Sharp/van der Heijde erosion score of the all wrist bones was correlate with the RAMRIS score (p = 0.008). The intraclass correlation coefficients (ICC) for manual segmentation showed high agreement (ICC = 0.901). CONCLUSIONS Considering CT as the reference method, CR showed very low sensitivity. A close correlation with CT erosion volumes supports the OMERACT RAMRIS erosion score as a semiquantitative measure of joint damage in RA. Although the computer-assisted manual segmentation can be beneficial for diagnostic decision in cross-sectional CT examinations of the wrist in RA, this technique will require further evaluation in terms of responsiveness.