BSREM for Brain Metastasis Detection with 18F-FDG-PET/CT in Lung Cancer Patients

Abstract

The aim of the study was to analyze the use of block sequential regularized expectation maximization (BSREM) with different β-values for the detection of brain metastases in digital fluorine-18 labeled 2-deoxy-2-fluoro-D-glucose (18F-FDG) PET/CT in lung cancer patients. We retrospectively analyzed staging/restaging 18F-FDG PET/CT scans of 40 consecutive lung cancer patients with new brain metastases, confirmed by MRI. PET images were reconstructed using BSREM (β-values of 100, 200, 300, 400, 500, 600, 700) and OSEM. Two independent blinded readers (R1 and R2) evaluated each reconstruction using a 4-point scale for general image quality, noise, and lesion detectability. SUVmax of metastases, brain background, target-to-background ratio (TBR), and contrast recovery (CR) ratio were recorded for each reconstruction. Among all reconstruction techniques, differences in qualitative parameters were analyzed using non-parametric Friedman test, while differences in quantitative parameters were compared using analysis of variances for repeated measures. Cohen’s kappa (k) was used to measure inter-reader agreement. The overall detectability of brain metastases was highest for BSREM200 (R1: 2.83 ± 1.17; R2: 2.68 ± 1.32) and BSREM300 (R1: 2.78 ± 1.23; R2: 2.68 ± 1.36), followed by BSREM100, which had lower accuracy owing to noise. The highest median TBR was found for BSREM100 (R1: 2.19 ± 1.05; R2: 2.42 ± 1.08), followed by BSREM200 and BSREM300. Image quality ratings were significantly different among reconstructions (p < 0.001). The median quality score was higher for BSREM100-300, and both noise and metastases’ SUVmax decreased with increasing β-value. Inter-reader agreement was particularly high for the detectability of photopenic metastases and blurring (all k > 0.65). BSREM200 and BSREM300 yielded the best results for the detection of brain metastases, surpassing both BSREM400 and OSEM, typically used in clinical practice.

Optimization of Bayesian penalized likelihood reconstruction for 68Ga-prostate-specific membrane antigen-11 PET/computed tomography

“…Different radiotracer characteristics likely influence the optimal β value. For example, β 400 for whole body 18 F-FDG [10], β 200-300 for brain metastasis [7], β 300 for whole body 18 F-fluciclovine [15]. This is likely due to the inherent imaging characteristics of 68 Ga tracers which result in increased imaged compared to fluorinated tracers.…”

“…Our patient population was also heterogenous, including patients at different stages of their diagnosis, with different clinical indications for PSMA PET/CTs and with a wide variety of tumor-node-metastasis stages. Most of the studies to date have looked into the optimal β values for fluorinated PET tracers, with a few publications looking into the optimal β values in 18 F-FDG PET/CT imaging for pulmonary nodules [23,24] and brain metastases in lung cancer [7]; however, these do not translate well in real-world clinical practice, where patients come for a multitude of scan indications, and it would not be feasible to select different β values for different patients or even different regions of the body of the same patient. Blinding all readers to the β values and to one another's interpretation reduced the impact of bias.…”

“…If these γ-rays have a scatter interaction in the patient, lower energy γ-ray in the energy window of the PET system is generated. These cannot be differentiated from the true coincidence of positron annihilation [7], which in turn increases image noise and reduces contrast and resolution.…”

HYPR4D kernel method on TOF PET data with validations including image-derived input function

“…We compared its performance in terms of contrast recovery and noise suppression as well as quality of time-activity curves from relatively small structures to the performance achieved with all the available TOF reconstructions with point-spread-function (PSF) resolution modeling on the system, namely PSF-TOFOSEM with and without standard (3.5mm FWHM transaxial and 1-4-1 axial) post filter and PSF-TOFBSREM (i.e. TOF Q.Clear which is the newly introduced stateof-the-art clinical reconstruction method) [9,10] with various beta values (regularization strengths) using data acquired from contrast phantom and human subjects. In addition, the impact of the reconstruction method on the derivation of an image-derived input function (IDIF) obtained from a novel voxel search technique was investigated.…”

Analysis of image quality by regulating beta function of BSREM reconstruction algorithm and comparison with conventional reconstructions in carcinoma breast studies of PET CT with BGO detector

“…For example, the optimum β factor in 18F-fluorodeoxyglucose ( 18 F-FDG) PET has been reported between 300 and 500 for the whole-body scan [13,14] and between 100 to 200 for brain scans [14], 400-600 for pulmonary nodules [15,16], in Ga68 PSMA it is reported as 400-550 [17], for 18F-fluciclovine it is written as 300 [18] and 4000 for low count 90Y [19]. Further, most of the previous studies were performed in the time of flight (TOF)-based lutetium-yttrium oxyorthosilicate PET CT system [8,[14][15][16][18][19][20][21][22], and very few studies are available bismuth germanate oxide (BGO)-based fivering detector PET CT systems [13,23], and some recent studies are also conducted in PET MR [17,24,25]. Due to the recent use of BSREM reconstruction in scanners, its impact on clinical use and research applications is still being evaluated [21][22][23][24].…”