Abstract-- In assessing the size and severity of myocardial perfusion defects, either a count threshold is applied to the images, or they are compared to a database of healthy hearts. This study aims to determine the dependence of these databases and thresholds on patient, acquisition and reconstruction variables, by measuring myocardial wall thickness. Analysis was performed on myocardial perfusion studies from 38 normal patients and a series of phantom experiments. The variables investigated included patient gender, test type, liver interference, myocardium to background activity ratio, acquisition zoom factor, matrix size and reconstruction type. When attenuation correction (AC) and detector resolution compensation (DRC) was applied during reconstruction, no significant difference was found in myocardial wall thickness between males and females, rest and stress studies, the presence and absence of liver interference, or clinically relevant myocardium to background activity ratios. A significant difference was found between standard and zoomed acquisitions, and between simple reconstruction techniques and those containing SPECT corrections. Results suggest that when AC and DRC are applied during reconstruction, patient variables do not influence quantitative accuracy and therefore analysis does not require individual databases or thresholds. As reconstruction methods improve in accuracy and in their ability to reconstruct large matrices, new databases and thresholds will be needed, bringing us closer to perfect absolute quantitative accuracy. I. INTRODUCTION INGLE Photon Emission Computed Tomography (SPECT) is a well-established, non-invasive imaging method routinely used for the evaluation of patients with suspected coronary artery disease (CAD). Currently these studies are analyzed primarily by means of a qualitative visual inspection of images, however quantitative analysis is becoming more significant, particularly in the setting of clinical research. This work was supported by an Izaak Walton Killam Memorial Pre- Doctoral Fellowship and by VHHSC/UBC interim funding. K. Dixon is with the Physics and Astronomy Department of the University of British Columbia, Vancouver, BC Canada (telephone: 604-875-5252, e- mail: katdixon@physics.ubc.ca). A. Celler and A. Fung are with the Radiology Department of the University of British Columbia, Vancouver, BC Canada (telephone: 604-875-5252, e- mail: aceller@physics.ubc.ca and funga@interchange.ubc.ca, respectively). An example of quantitative analysis is the assessment of the size and severity of myocardial perfusion defects. Here either a predetermined count threshold is applied to the images or the images are compared to a database of healthy hearts. However, it is understood that these databases and thresholds are dependent on patient, acquisition and reconstruction variables [1]. Therefore, in order to produce accurate and reliable results, each analysis method should use the database that is specific to the camera, software and patient population which are being investigated. Generation of these numerous databases is obviously a cumbersome task. The goal of this investigation was to determine the dependence of absolute quantitative accuracy on these variables and hence to identify the situations in which separate databases or thresholds are really required. Myocardial wall thickness of the left ventricle (LV) was used as the measure of absolute quantitative accuracy, and the database variables which were investigated include patient gender, test type, stress type, liver interference, myocardium to background activity ratio, acquisition zoom factor, matrix size and reconstruction type. Analysis was carried out on a database of rest and stress studies from 37 patients with a less than 5% likelihood of CAD. This supplied a uniquely clinical perspective with enough data to provide comparisons with a statistical power of greater than 80%. Where possible the results of the analysis were verified by a set of phantom experiments. Phantom studies were also used to compare variables that could not be altered in a standard clinical myocardial perfusion study. II. METHODS A. Data acquisition - Clinical Data A database of myocardial perfusion studies was created from all consenting patients scanned with the transmission source on the Siemens E-cam camera at Vancouver General Hospital. Using categorization based on medical history and the results of stress electrocardiography (ECG) [2], 38 of these patients were determined to have a less than 5% likelihood of CAD. These 38 rest and 38 stress studies were therefore The effect of patient, acquisition and reconstruction variables on myocardial wall thickness as measured from myocardial perfusion SPECT studies Katherine L. Dixon, Student Member, IEEE, Anthony Fung, Anna Celler, Member, IEEE. S 0-7803-8701-5/04/$20.00 (C) 2004 IEEE