Background: Our segmentation algorithm for single-photon emission computed tomographic perfusion studies was tested in 244 patients treated by thrombolysis within 5 hours after onset of symptoms. This algorithm uses radial slices to approximate true three-dimensional gradients, determines the apex and basal plane, and creates a perfusion and volume polar map.
Methods and results: Perfusion defect size was compared with enzymatic infarct size and global and regional function. All patients underwent rest planar and tomographic 99mTc-labeled sestamibi scanning, contrast coronary angiography, and ventriculography 10 to 14 days after the start of treatment. Manual correction had to be performed in only 10% of the cases and presented no problems. The correlation coefficients (r) between planar and relative tomographic perfusion defects versus enzymatic infarct size were 0.71 and 0.73. A negative correlation was found with left ventricular ejection fraction: r = -0.65 and r = -0.60. A comparable correlation was also found between regional wall motion and perfusion defect size. Most correlations were higher in the case of anterior infarction. An excellent correlation was found between planar and tomographic defect size (r = 0.83).
Conclusions: In most cases, our segmentation algorithm delineates myocardial edges and basal plane automatically. A good correlation was found between perfusion defect size, enzymatic infarct size, and global and regional ventricular function.