Determination of optimal regularization factor in Bayesian penalized likelihood reconstruction of brain PET images using [¹⁸F]FDG and [¹¹C]PiB

Abstract

Purpose The Bayesian penalized likelihood (BPL) reconstruction algorithm, Q.Clear, can achieve a higher signal‐to‐noise ratio on images and more accurate quantitation than ordered subset‐expectation maximization (OSEM). The reconstruction parameter (β) in BPL requires optimization according to the radiopharmaceutical tracer. The present study aimed to define the optimal β value in BPL required to diagnose Alzheimer disease from brain positron emission tomography (PET) images acquired using 18F‐fluoro‐2‐deoxy‐D‐glucose ([18F]FDG) and 11C‐labeled Pittsburg compound B ([11C]PiB). Methods Images generated from Hoffman 3D brain and cylindrical phantoms were acquired using a Discovery PET/computed tomography (CT) 710 and reconstructed using OSEM + time‐of‐flight (TOF) under clinical conditions and BPL + TOF (β = 20–1000). Contrast was calculated from images generated by the Hoffman 3D brain phantom, and noise and uniformity were calculated from those generated by the cylindrical phantom. Five cognitively healthy controls and five patients with Alzheimer disease were assessed using [18F]FDG and [11C]PiB PET to validate the findings from the phantom study. The β values were restricted by the findings of the phantom study, then one certified nuclear medicine physician and two certified nuclear medicine technologists visually determined optimal β values by scoring the quality parameters of image contrast, image noise, cerebellar stability, and overall image quality of PET images from 1 (poor) to 5 (excellent). Results The contrast in BPL satisfied the Japanese Society of Nuclear Medicine (JSNM) criterion of ≥55% and exceeded that of OSEM at ranges of β = 20–450 and 20–600 for [18F]FDG and [11C]PiB, respectively. The image noise in BPL satisfied the JSNM criterion of ≤15% and was below that in OSEM when β = 150–1000 and 400–1000 for [18F]FDG and [11C]PiB, respectively. The phantom study restricted the ranges of β values to 100–300 and 300–500 for [18F]FDG and [11C]PiB, respectively. The BPL scores for gray–white matter contrast and image noise, exceeded those of OSEM in [18F]FDG and [11C]PiB images regardless of β values. Visual evaluation confirmed that the optimal β values were 200 and 450 for [18F]FDG and [11C]PiB, respectively. Conclusions The BPL achieved better image contrast and less image noise than OSEM, while maintaining quantitative standardized uptake value ratios (SUVR) due to full convergence, more rigorous noise control, and edge preservation. The optimal β values for [18F]FDG and [11C]PiB brain PET were apparently 200 and 450, respectively. The present study provides useful information about how to determine optimal β values in BPL for brain PET imaging.

Towards standardization of tau PET imaging corresponding to various tau PET tracer: multi-center phantom study

“…The convergence of PET images using an iterative reconstruction algorithm depends on the target activity, the acquired PET counts, and the target size or shape. We previously found that the convergence rate of contrast was independent of target activity because the conditions of the iterative reconstruction algorithm were the same regardless of phantom activity [26,27]. The high levels of whole-brain activity in tau PET images in rst generation tau PET tracers due to non-speci c brain uptake were decreased in second generation tau PET tracers.…”

“…Thus, the Japanese Society of Nuclear Medicine (JSNM) proposed phantom test procedures and criteria to standardize brain [ 18 F]FDG and amyloid PET imaging [25]. Several studies then determined optimal reconstruction conditions for brain [ 18 F]FDG and amyloid PET imaging using an iterative reconstruction method based on the JSNM phantom test criteria [26,27]. The FDA approved [ 18 F] ortaucipir as the rst tau PET ligand in 2020 [28].…”

Phantom and clinical evaluation of the Bayesian penalised likelihood reconstruction algorithm Q.Clear without PSF correction in amyloid PET images

“…The Japanese Society of Nuclear Medicine (JSNM) has proposed a phantom test procedure to determine optimal reconstruction conditions and standardise amyloid PET imaging [ 5 ]. The determination process of reconstruction conditions for iterative reconstruction methods have been established for brain 2-deoxy-2-[ 18 F]fluoro- d -glucose ([ 18 F]FDG) and amyloid PET [ 6 , 7 ].…”

“…Although optimal β values for whole-body imaging using [ 18 F]FDG have been reported [ 10 , 11 , 13 , 16 , 17 ], little is known about optimal β values for brain PET imaging. The optimal β values were 200 for neuro-oncological [ 18 F]FDG imaging, β = 350–500 for pediatric brain [ 18 F]FDG imaging, 150 or 300 for [ 18 F]flutemetamol imaging, and β = 450 for carbon-11-labeled Pittsburgh compound B ([ 11 C]PiB) [ 7 , 17 – 19 ].…”

Q.Clear Reconstruction for Reducing the Scanning Time for 68Gallium-DOTA-FAPI-04 PET/MR Imaging

“…The Q.Clear algorithm uses a customizable penalization factor (β) for noise suppression, and can achieve multiple iterations while suppressing background noise [7][8]. Recently, several studies have reported that, compared with traditional OSEM reconstruction, Q.Clear improved the quality of PET images by increasing SUVs within lesions, signal-to-noise ratio (SNR), and spatial resolution [9][10][11]. However, large SUVs from Q.Clear PET reconstruction appear problematic when interpreted according to current criteria for quantitative evaluation, because they may lead to overestimation of the disease burden and thus limit mainstream application in the clinic [12].…”