Automated quantitative assessment of myocardial perfusion in rodent PET/SPECT images

The aim of this study was to automatically assessment of the ⁶⁴Cu-PTSM, ^99mTc-MIBI and ²⁰¹Tl myocardial perfusion image using Multi Gaussian Mixture Model (MGMM). Myocardial infarction (MI) model was prepared through permanently ligated left circumflex artery of rats. PET/SPECT images were acquired using small animal PET/SPECT scanner (InveonTM, Siemens) with physiological monitoring system (BioVet, m2m Imag. Corp.). PET images were obtained at 10 min post injection of 37 MBq/0.2 mL of 64Cu-PTSM via tail vein as list-mode data. SPECT images were acquired at 60 min post IV injection of 110 MBq of Tc99m-MIBI and Tl-201/0.1 mL with 126-154 keV (Tc), 63-77 keV (Tl) energy window and 1.0 mm pinhole collimator. The acquired data was reconstructed by OSEM2D algorithm with 4 iterations. To automatically make the myocardial contour and generate polar map, we used Cedars-Sinai method. Myocardial infarct region of polar map were evaluated by using MGMM method for adaptive threshold calculation and predefined threshold method for absolute threshold calculation. Rat myocardium was stained by hematoxylin and eosin (H&E) for reference value of infarct size after the imaging. The infarct size of PET/SPECT image was defined by infarction area percentage of the total left myocardium. Comparative analysis was performed between MI size of histology and polar map. The size difference in ⁶⁴Cu-PTSM PET polar map was 6.96 3.53% from the absolute value 40% and 4.40 0.58% from MGMM4. The difference in ^99mTc-MIBI SPECT polar map was 8.87 1.16% from the absolute value 40%, and 3.47 1.99% from MGMM4. The difference in ²⁰¹Tl SPECT polar map was 11.73 1.45% from the absolute value 40%, and 2.97 1.46% from MGMM4. MGMM method showed the possibility of quantitative evaluation for my