Impact of Various Image Reconstruction Methods on Joint Compensation of Respiratory Motion and Partial Volume Effects in Whole-Body18f-Fdg Pet/Ct Imaging: Patients With Non-Small Cell Lung Cancer Publisher



Ghafarian P¹ ; Rezaei S^{2, 3} ; Gharepapagh E^{2, 3} ; Sarkar S^{4, 5} ; Ay MR^{4, 5} Show Affiliations
Source: Iranian South Medical Journal Published:2022

Abstract

Background: The present study aims to assess the impact of various image reconstruction methods in18F-FDG PET/CT imaging on the quantification performance of the proposed technique for joint compensation of respiratory motion and partial volume effects (PVEs) in patients with non-small cell lung cancer. Materials and Methods: An image-based deconvolution technique was proposed, incorporating wavelet-based denoising within the Lucy-Richardson algorithm to jointly compensate for PVEs and respiratory motion. The method was evaluated using data from 15 patients with 60 non-small cell lung cancer. In these patients, the lesions were classified by size, location and Signal-to-Background Ratios (SBR). In each study, PET images were reconstructed using four different methods: OSEM with time-of-flight (TOF) information, OSEM with point spread function modelling (PSF), OSEM with both TOF and PSF (TOFPSF), and OSEM without PSF or TOF (OSEM). The Contrast to Noise Ratio (CNR), Coefficient of Variation (COV) and Standardized Uptake Values (SUV) were measured within the lesions and compared to images that were not processed using the joint-compensation technique. Furthermore, variabilities arising due to the choice of the reconstruction methods were assessed. Results: Processing the images using the proposed technique yielded significantly higher CNR and SUV, particularly in small spheres, for all the reconstruction methods and all the SBRs (P<0.05). Overall, the incorporation of wavelet-based denoising within the Lucy Richardson algorithm improved COV and CNR in all the cases (P<0.05(. In the patient data, the median values of the relative difference (%) of CNR for the compensated images in comparison to the uncompensated images were 40.9%, 41.2%, 45.3% and 40.8% for OSEM, PSF, TOF, and TOFPSF, respectively, in the small lesions (equivalent diameter <15 mm), 31.0%, 25.9%, 34.1% and 28.2% in the average-sized lesions (equivalent diameter<30 mm), 35.7%, 33.7%, 37.8% and 33.2% in the lesions in the lower lung lobes, 33.5%, 31.0%, 35.7% and 30.6% in the lesions in the upper lung lobes, 39.7%, 37.9%, 45.1% and 39.0% in the low-SBR lesions and 28.8%, 27.8%, 34.8% and 25.7% in the high-SBR lesions. Changes in motion amplitude, target size and SBRs in the patient data resulted in significant inter-method differences in the images reconstructed using different methods. Specifically, in a small target size, quantitative accuracy was highly dependent on the choice of the reconstruction method. Conclusion: Our results showed that joint compensation, and incorporation of wavelet-based denoising, yielded improved quantification from PET images. Quantitative accuracy is greatly affected by SBR, lesion size, breathing motion amplitude, as well as the choice of the reconstruction protocols. Overall, the choice of reconstruction algorithm combined with compensation method needs to be determined carefully. © 2022, Bushehr University of Medical Sciences. All rights reserved.