A procedure for wall detection in [18F]FDG positron emission tomography heart studies

Quantitative positron emission tomography (PET) heart studies require the accurate localization of regions of interest (ROIs) on the myocardial wall (MW) and left ventricle (LV). The procedure is often inaccurate, especially when there is low tracer uptake. We implemented a data processing technique to improve the accuracy of the localization of ROIs on the MW and LV in fluorine-18 labelled deoxyglucose ([18F]FDG) PET heart studies. This technique combines transmission data, acquired before tracer administration and used for attenuation correction, and dynamic emission data (DY), acquired to obtain myocardial time-activity curves and used to calculate regional myocardial glucose utilization, to generate a new set of "transmission" images (TRDY) with enhanced contrast between MW and LV. These new transmission images identify the extravascular myocardial tissue and can be used for ROI placement. Validation of the method was performed in 25 patients, studied after an oral glucose load, by drawing irregular ROIs on three transaxial slices outlining the septum and anterior-apical and lateral wall on the last frame of the DY images (steady state) and then on the TRDY images. Two kinds of analysis were performed on a total of 225 myocardial segments: (1) mean counts per pixel in the DY images from ROIs independently drawn on DY and TRDY images were compared; (2) TRDY ROIs were copied onto DY images and repositioned in the event of mismatch between ROIs and myocardial tissue edge. Mean counts per pixel in the DY images from the original and the repositioned TRDY ROIs were compared. An excellent correlation was found in both cases (using TRDY and DY ROIs: y=0.908 x+0.068, r=0.97; using TRDY ROIs alone: y=0.975 x+0.006, r=0.99). This technique can be used for clinical applications in physiological and pathological conditions in which the myocardial [18F]FDG uptake is reduced or minimal, including diabetes and myocardial infarction.