Click here to


Are you sure ?

Yes, do it No, cancel

Partial Volume Correction (PVC) in Quantitative 18F-FDG PET/CT Imaging On Intratumoral Dose Response Assessment

S Chen*, S Chang, D Krauss, D Yan, William Beaumont Hospital, Royal Oak, MI


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Partial volume effect in PET imaging may affect tumor dose response assessment. This study investigated the impacts of PVC on tumor voxel dose response predicted using serial FDG-PET/CT imaging feedbacks.

Methods: FDG-PET/CT images were obtained at pretreatment and weekly during chemoradiotherapy of 30 HNC patients. PET images were reconstructed using the blob-ordered-subsets time-of-flight algorithm with a voxel 4x4x4mm³. Deconvolution-based PVC algorithms were applied on each PET image. Weekly PET/CT images of each patient were registered to the pretreatment PET/CT image to construct tumor dose response matrix (DRM). The DRM value of a tumor voxel represents the average change rate of its metabolic activity with respect to the pretreatment/baseline activity, SUV0. Resistant subvolumes (RV)s = { v|DRM(v) = a x SUV0(v)? + c} were calculated with and without PVC respectively and used to predict local tumor failure/control using ROC test. True positive/negative is defined as a tumor that has an RV larger/smaller than a cutoff would be a local failure/control tumor. The constant a, b and c were determined by maximizing the area under the curve (AUC). The mean and SD of tumor voxel SUV0 and DRM discrepancies induced by PVC were also calculated on individual tumors.

Results: Both the RVs calculated with and without PVC achieved the same AUC = 0.96. For all tumor voxels, the mean±SD of SUV0 discrepancy between with and without PVC was -0.7%±9.3% and 0.6%±8.8% for DRM. The SUV0 and DRM discrepancies were correlated with the FDG uptake heterogeneity of each tumor with Spearman correlation coefficients 0.71 and 0.52 respectively.

Conclusion: Local tumor failure/control could be equally well predicted with using FDG-PET/CT imaging feedbacks with/without PVC. However, discrepancies of 9.3% and 8.8% (one SD) for SUV0 and DRM were identified. This could imply the reliability of the prediction method needs more clinical followup to confirm.


FDG PET, Quantitative Imaging, Dose Response


TH- Response Assessment: Imaging-based: PET

Contact Email