Quantification of serial tumor glucose metabolism.

Abstract

We developed a method to improve the quantitative precision of FDG-PET scans in cancer patients. The total-lesion evaluation method generates a correlation coefficient (r) constrained Patlak parametric image of the lesion together with three calculated glucose metabolic indices: (a) the total-lesion metabolic index ("KT-tle", ml/min/lesion); (b) the total-lesion voxel index ("VT-tle", voxels/lesion); and (c) the global average metabolic index ("KV-tle", ml/min/voxel). The glucose metabolic indices obtained from conventional region of interest (ROI) and multiplane evaluation were used as standards to evaluate the accuracy of the total-lesion evaluation method. Computer simulations and four patients with metastatic melanoma before and after chemotherapy were studied. Computer simulations showed that the total-lesion evaluation method has improved precision (% s.d. < 0.6%) and accuracy (approximately 10% error) compared with the conventional ROI method (% s.d. approximately 5%; approximately 25% error). The KT-tle and VT-tle indices from human FDG-PET studies using the total-lesion evaluation method showed excellent correlations with the corresponding values obtained from the conventional ROI methods and multiplane evaluation (r approximately 1.0) and CT lesion volume measurements. This method is a simple but reliable way to quantitatively monitor tumor FDG uptake. The method has several advantages over the conventional ROI method: (a) less sensitive to the ROI definition, (b) no need for image registration of serial scan data and (c) includes tumor volume changes in the global tumor metabolism.