Background/Purpose: To describe the methodology of a novel computer-based technique for outlining erosions on plain radiographs in rheumatoid arthritis (RA) patients, and assess the responsiveness and correlation with the Sharp erosion score. 

Methods: A sample of 112 patients with mild to severe rheumatoid arthritis, for whom serial (baseline and one follow-up) bilateral hand radiographs were available, was obtained from the National Databank for Rheumatic Diseases. A radiologist, blinded to time point, used the software to outline the erosions at each MCP and PIP joint. Severely diseased joints were excluded.  The computer-based method partially automates the process of outlining erosions on hand radiographs, and calculates erosion area (see figure). For 99 subjects, the total Sharp erosion score for the MCP and PIP joints of digits 2 to 5 was available. Because all radiographs were acquired prior to the advent of DMARDS, we expected radiographic damage to increase with time. Therefore, the chronological order of serial radiographs was used as the standard to assess performance. The percent change and standardized response mean (SRM) were used to quantify responsiveness, and linear regression established the correlation with the Sharp score. 

Results: 78% of subjects showed an increase in erosion area over time, and 5% showed no change. Changes in the computer-based and Sharp methods were highly correlated (r = 0.75, p < 0.01). The SRM was 0.46. The reader time was approximately 0.6 minutes per hand radiograph.

Conclusion: The computer-based method is fast and can accurately assess progression of disease in patients with RA. The high correlation of the software measure with the Sharp score suggests it can be used as a surrogate for this measure.