Correction of vascular component in 18F-FDG brain PET: Least squared method with a weighted function using blood time-activity curve

Objectives For brain ¹⁸F- FDG dynamic PET, we developed a new method to correct vascular component in calculating the cerebral metabolic rate for glucose (CMRGlc) without ¹⁵O-CO imaging.

Methods Seven normal volunteers were employed. ¹⁵O-CO PET and 80-minute dynamic FDG PET was scanned. Serial arterial blood and plasma counts were measured during the study. We set 28 ROIs in each brain. Cerebral blood volume (CBV) was estimated with ¹⁵O-CO data. Three-compartment model analysis was performed with dynamic FDG data. Conventional vascular activity correction was performed with four parameters (K1 – k4) fitting analysis and ¹⁵O-CO CBV data, standard CMRGlc was estimated. Five parameters (K1 – k4 and V (vascular component)) fitting analysis was performed with non-linear least squared method with uniform fitting and with weighted fitting. The blood count curve in dynamic FDG study was adopted for the weighted function. Error between standard CMRGlc and CMRGlc with uniform fitting (Error-U), and error between standard CMRGlc and CMRGlc with weighted fitting (Error-W) were evaluated.

Results In the total of 196 ROIs data, Error-W (6.6±6.8%) yielded less values than Error-U (14.2±14.1%) (p<0.001). Weighted non-linear fitting dedicated more corrected CMRGlc estimates than conventional uniform fitting analysis.

Conclusions For brain ¹⁸F- FDG dynamic PET, we developed a non-linear weighted fitting method to correct vascular component in calculating CMRGlc without ¹⁵O-CO imaging.