Image-derived input functions for determination of MRGlu in cardiac (18)F-FDG PET scans

Image-derived input functions (IDIF) are frequently used in cardiac (18)F-FDG PET studies for determination of the myocardial metabolic rate of glucose (MRGlu). The purpose of this study was to assess which vascular structure is most suited for defining the IDIF, using online arterial blood sampling (AS) as the gold standard.

In 18 patients with ischemic heart disease, 370 MBq FDG were injected during a hyperinsulinemic euglycemic clamp. Studies were performed with a Siemens/CTI HR+ PET scanner using a dynamic scanning protocol. A fully automated blood-sampling device was used for continuous AS. IDIF were obtained using regions of interest (ROIs) of 3 different sizes defined on the left ventricle (LV), left atrium (LA), ascending aorta (AA), and descending aorta (DA). MRGlu was calculated with all input functions. Ratios between MRGlu obtained with IDIF and AS were calculated for each patient.

Time-activity curves from smaller ROIs suffered more from statistical noise with only a modest reduction of spillover effects, which led to more variation in calculated MRGlu. Mean ratios of MRGlu obtained with IDIF and AS were close to 1 when AA and DA (0.97 +/- 0.07 and 1.00 +/- 0.11, respectively) were used to define the input function. However, when LA and LV were used, mean ratios were 0.81 +/- 0.06 and 0.79 +/- 0.08, respectively, reflecting a significant underestimation of MRGlu. The use of AA for defining the input function resulted in the best agreement with AS and the smallest interobserver variation.

The ascending aorta is the structure of choice for defining IDIF and a large ROI (diameter, approximately 15 mm) should be used to minimize the effects of statistical noise.