2.3.7. PET Image Analysis

PET image analysis was done with Carimas 2.9 software (Turku PET Centre, Turku, Finland) using the analysis protocol described in [24]. After visualization and definition of the heart long axis, the myocardial volumes of interest (VOIs) in the [15O]H2O images were semi-automatically defined. The VOIs contained the whole left ventricle and a blood pool as separate volumes. These contours were then imported to overlapping [68Ga]NODAGA-RGD PET images from non-gated, RPM-gated and MMS-gated reconstructions followed by visual confirmation of the overlap of contours and appropriate anatomical detail. For the RPM and the MMS, the gated images from bin number 5 (end-expiratory bin) were used in the analysis as they contained the best image quality. After visual confirmation, the contour locations needed further manual adjustment for subjects 4 and 6 to match the contours for the myocardial region in non-gated, MMS-gated and RPM-gated reconstructed PET. From both the myocardium and blood pool, the mean and standard deviation in PET uptake (Bq/mL) was then defined. For assessment of image quality, contrast ratio (CR) [26,27], signal-to-noise ratio (SNR) [28], coefficient of variation (CV) [28] and contrast-to-noise ratio (CNR) [26,27] were used. The uptake values were used to calculate the image quality metrics from each PET image series as follows:

where VOIsd is the standard deviation and VOImean is the average uptake in the corresponding volume of interest.

In addition to the image quality metrics, to evaluate the image quality of the PET images with RPM and MMS gating, manually selected profiles over a hot spot for gated images were extracted.