Comparison of SSS and SRS calculated from normal databases provided by QPS and 4D-MSPECT manufacturers and from identical institutional normals

Purpose There is proven evidence for the importance of myocardial perfusion-single-photon emission computed tomography (SPECT) with computerised determination of summed stress and rest scores (SSS/SRS) for the diagnosis of coronary artery disease (CAD). SSS and SRS can thereby be calculated semi-quantitatively using a 20-segment model by comparing tracer-uptake with values from normal databases (NDB). Four severity-degrees for SSS and SRS are normally used: < 4, 4-8, 9-13, and >= 14. Manufacturers' NDBs (M-NDBs) often do not fit the institutional (I) settings. Therefore, this study compared SSS and SRS obtained with the algorithms Quantitative Perfusion SPECT (QPS) and 4D-MSPECT using M-NDB and I-NDB. Methods I-NDBs were obtained using QPS and 4D-MSPECT from exercise stress data (450 MBq Tc-99-tetrofosmin, triple-head-camera, 30 s/view, 20 views/head) from 36 men with a low post-stress test CAD probability and visually normal SPECT findings. Patient group was 60 men showing the entire CAD-spectrum referred for routine perfusion-SPECT. Stress/rest results of automatic quantification of the 60 patients were compared to M-NDB and I-NDB. After reclassifying SSS/SRS into the four severity degrees, kappa (k) values were calculated to objectify agreement. Results Mean values (vs M-NDB) were 9.4 +/- 10.3 (SSS) and 5.8 +/- 9.7 (SRS) for QPS and 8.2 +/- 8.7 (SSS) and 6.2 +/- 7.8 (SRS) for 4D-MSPECT. Thirty seven of sixty SSS classifications (k=0.462) and 40/60 SRS classifications (k=0.457) agreed. Compared to I-NDB, mean values were 10.2 +/- 11.6 (SSS) and 6.5 +/- 10.4 (SRS) for QPS and 9.2 +/- 9.3 (SSS) and 7.2 +/- 8.6 (SRS) for 4D-MSPECT. Forty four of sixty patients agreed in SSS and SRS (kappa=0.621 resp. 0.58). Conclusion Considerable differences between SSS/SRS obtained with QPS and 4D-MSPECT were found when using M-NDB. Even using identical patients and identical I-NDB, the algorithms still gave substantial different results.