Hui, Ng Suk (2024) Qualitative and quantitative accuracy evaluation of 18f-fdg PET/CT with TOF and NON-TOF system on beta value in BPL reconstruction. Project Report. Universiti Sains Malaysia. (Submitted)
|
PDF
- Submitted Version
Download (455kB) | Preview |
Abstract
The resolution and quantitative accuracy of Positron Emission Tomography (PET) greatly depend on the reconstruction algorithm. Ordered Subset Expectation Maximisation (OSEM) is unable to achieve full convergence as image noise grows with each iteration, potentially compromising quantitative accuracy. Penalised likelihood estimation algorithms, on the other hand, allow for an effective convergence that improves image quality by enhancing contrast and reducing noise. Q.Clear, a Bayesian Penalised Likelihood (BPL) algorithm, has demonstrated notable advancements in clinical image quality and quantification, particularly in detecting subtle abnormalities. This study compares the performance of BPL with OSEM in both Time-of-Flight (TOF) and non-TOF Positron Emission Tomography/ Computed Tomography (PET/CT) acquisitions. By varying the noise penalisation factor (beta value), which balances noise reduction and resolution, the study aims to determine optimal beta value for BPL, particularly in enhancing diagnostic accuracy and image quality for small lesions. Methods: A National Electrical Manufacturers Association (NEMA) image quality phantom filled with Fluorine-18 fluoro-2-deoxy-D-glucose (18F-FDG) at a 5:1 tumour-to-background ratio (TBR) was scanned on a lutetium-based PET/CT scanner. The images were reconstructed using the OSEM (16 subsets, 3 iterations) and Q.Clear algorithms, both of which include Point Spread Function (PSF) modelling. Q.Clear was investigated for beta values ranging from 100 to 2000. Both BPL and OSEM reconstructions were acquired with and without TOF information for comparison. These phantom images were evaluated quantitatively and qualitatively. The Recovery Coefficient (RC), Coefficient of Variance (COV), and Contrast-to-Noise Ratio (CNR) were measured to evaluate image quality and lesion detectability. Subjective image quality was performed by three experienced physicists. Results: Increasing beta values led to reduced RCmax and COV, with TOF acquisitions consistently exhibiting higher RCmax and COV values compared to non-TOF acquisitions. BPL achieved more accurate quantitative recovery and lower noise (except BPL 100) compared to OSEM reconstruction. Both CNRmean and CNRmax generally increased with sphere diameter. In TOF acquisitions, CNRmean and CNRmax peaked around BPL 700, declining thereafter, while in non-TOF acquisitions, both plateaued around BPL 1000 after an initial drop at BPL 100 and BPL 200. While OSEM generally maintained comparable CNR values across various beta values, BPL reconstruction exhibited improvements, particularly in reducing noise and enhancing contrast. In subjective evaluation, OSEM and BPL 450 were rated higher for sphere detectability in TOF acquisition, whereas BPL 2000 was identified as having the highest noise score in both TOF and non-TOF acquisitions. Conclusion: Our study findings suggested that a beta value of 450 was optimal for TOF acquisition, while a beta value of 300 was recommended for non-TOF acquisition. These results were consistent with the standard practice at IKN, where a BPL 450 with TOF acquisition is used.
| Item Type: | Monograph (Project Report) |
|---|---|
| Uncontrolled Keywords: | BPL reconstruction |
| Subjects: | R Medicine R Medicine > R Medicine (General) > R735-854 Medical education. Medical schools. Research R Medicine > R Medicine (General) > R856-857 Biomedical engineering. Electronics. Instrumentation |
| Divisions: | Kampus Kesihatan (Health Campus) > Pusat Pengajian Sains Kesihatan (School of Health Sciences) > Monograph |
| Depositing User: | Mr Abdul Hadi Mohammad |
| Date Deposited: | 02 Dec 2024 07:57 |
| Last Modified: | 06 Jan 2025 07:41 |
| URI: | http://eprints.usm.my/id/eprint/61360 |
Actions (login required)
 |
View Item |
