Comparison of four methods for quantitative assessment of ^99mTc-MDP SPECT in patients with suspected condylar...

To compare variability, reproducibility and repeatability of four quantitative evaluation methods to interpret the ^99mTc-MDP SPECT reports in patients with clinically suspected unilateral condylar hyperplasia.

This was a descriptive observational study carried out with SPECT images of 38 patients with clinical and radiographic signs of unilateral condylar hyperplasia, and interpreted using four quantitative methods: 1) one image, variable-size region of interest (ROI); 2) one image, fixed-size ROI (1.76cm²), 3) five image variable-size ROI; 4) five image, fixed-size ROI (1.76cm²). Each of the images was reported simultaneously (but in an independent way) by two nuclear medicine experts, in both total radioactive counts as well as normalized counts to evaluate the reproducibility (inter-operator variability) and the repeatability (intra-operator variability).

Higher reproducibility and repeatability were obtained in 5-image fixed-size ROI method (intra-class correlation coefficient: 0.979 [0.959; 0.989]). A high grade of diagnostic agreement (97.4%) was also attained in fixed methods (Kappa 0.940, p value: .000) from either total or normalized counts. There was no difference between fixed-size 1 vs 5 image methods. The methods based on variable-size ROI had a low grade of agreement (Kappa<0.20). More positive cases were identified using one image, ROI variable total counts (27 cases), but when the counts were normalized, they presented a lower number (5 cases).

Five-image fixed-size ROI provides the best intra-operator and inter-operator reliability for the diagnosis of unilateral condylar hyperplasia. In the four methods using normalized counts fewer positive cases were detected (≥10%), unlike with total counts when more positive cases were found.