The implementation of a software system for performing quality assurance in radiology

Purpose: To compare three processing methods for quantitative analysis on Temporomandibular joint (TMJ) divided function. Methods: SPECT images from 20 patients were reconstructed using HOSEM Iterative reconstruction and quantitative analysis was performed using three different methods: (1) Summation of a circular ROI over adjacent transaxial slices, (2) 3 cm spherical VOI placed on the transaxial slice of highest activity and (3) Individually constrained VOI, where (1) is the current method being used in the department. Each SPECT scan was analysed with all three techniques by six physicists in the department. Coefficient of variation (COV) was used to assess inter-observer variability for each method. Results: The mean COV was found to be: for method (1) 0.0208, (2) 0.0179 and (3) 0.0433. A single factor ANOVA was performed on the data and they were found to be significantly different from alpha = 0.05 (F-value of 12.182 and F-crit. of 3.159). Post hoc analysis showed that there was no significant difference betweenmethods 1 and 2 (p-value = 0.1107), strong evidence of difference between methods 1 and 3 (p-value = 0.0029) and very strong evidence of difference between methods 2 and 3 (p-value = 0.0004). Conclusion: The method that shows the least variation between users was the fixed 3 cm spherical VOI. This method will now be implemented at Glasgow Royal Infirmary.