Evaluation of Automatic Quantification Programs for Myocardial Perfusion SPECT using Monte Carlo Simulations

Abstract: Coronary artery disease (CAD) is a condition in which the blood supply to the myocardium is partially or completely blocked due to plaque build-up in the artery wall. Myocardial SPECT imaging is a commonly used technique to assess the blood flow to the myocardium and the function of the heart. The interpretation of the images can be difficult and subjective. Several automatic myocardial SPECT quantification programs have been developed to standardize the interpretation and analysis of myocardial images. These programs are based on different mathematical models for determination of the heart shape and different normal databases for detection of abnormality, which may lead to disparity in the results given by these different programs. The aims of the work were to validate two automatic quantification programs, i.e. AutoQUANT and 4D-MSPECT, concerning important parameters in the detection of CAD, such as extent, severity and semi-quantitative segment scores and to show that a combination of a patient-like digital phantom and the Monte Carlo technique is useful in simulating clinical patient images. Results: The development of a radioactive ink-based stack phantom was performed for accurate generation of SPECT images. The comparison between the images generated by the stack phantom and the corresponding Monte Carlo simulated images showed good agreement. Differences between the one-day and two-day protocols, as well as between the two quantification programs, were found. Interdependence between the measures of extent and severity were observed. The interdependence was identified as being due to the algorithm used by the quantification program in the assessment of extent and severity. A model was developed to account for this interdependence. The ability of myocardial SPECT to discriminate between transmural and subendocardial lesions was observed for the simulated cases, while in the clinical situations this difference is difficult to observe due to the interdependence of the extent and severity. Conclusions: The disparity between the two programs in assessing the same patient images was found. The combination of the NCAT digital phantom and the SIMIND program was found to be useful in simulating clinical myocardial SPECT study. This technique could be a promising tool since realistic patient studies can be performed in a controlled way and used as input in different automatic quantification programs for accurate comparative evaluations.