Regularized ML reconstruction for time/dose reduction in18F-fluoride PET/CT studies

Optimization of a Bayesian penalized likelihood algorithm (Q.Clear) for 18F-NaF bone PET/CT images acquired over shorter durations using a custom-designed phantom

Background

The Bayesian penalized likelihood (BPL) algorithm Q.Clear (GE Healthcare) allows fully convergent iterative reconstruction that results in better image quality and quantitative accuracy, while limiting image noise. The present study aimed to optimize BPL reconstruction parameters for ¹⁸F-NaF PET/CT images and to determine the feasibility of ¹⁸F-NaF PET/CT image acquisition over shorter durations in clinical practice.

Methods

A custom-designed thoracic spine phantom consisting of several inserts, soft tissue, normal spine, and metastatic bone tumor, was scanned using a Discovery MI PET/CT scanner (GE Healthcare). The phantom allows optional adjustment of activity distribution, tumor size, and attenuation. We reconstructed PET images using OSEM + PSF + TOF (2 iterations, 17 subsets, and a 4-mm Gaussian filter), BPL + TOF (β = 200 to 700), and scan durations of 30–120 s. Signal-to-noise ratios (SNR), contrast, and coefficients of variance (CV) as image quality indicators were calculated, whereas the quantitative measures were recovery coefficients (RC) and RC linearity over a range of activity. We retrospectively analyzed images from five persons without bone metastases (male, n = 1; female, n = 4), then standardized uptake values (SUV), CV, and SNR at the 4th, 5th, and 6th thoracic vertebra were calculated in BPL + TOF (β = 400) images.

Results

The optimal reconstruction parameter of the BPL was β = 400 when images were acquired at 120 s/bed. At 90 s/bed, the BPL with a β value of 400 yielded 24% and 18% higher SNR and contrast, respectively, than OSEM (2 iterations; 120 s acquisitions). The BPL was superior to OSEM in terms of RC and the RC linearity over a range of activity, regardless of scan duration. The SUV_max were lower in BPL, than in OSEM. The CV and vertebral SNR in BPL were superior to those in OSEM.

Conclusions

The optimal reconstruction parameters of ¹⁸F-NaF PET/CT images acquired over different durations were determined. The BPL can reduce PET acquisition to 90 s/bed in ¹⁸F-NaF PET/CT imaging. Our results suggest that BPL (β = 400) on SiPM-based TOF PET/CT scanner maintained high image quality and quantitative accuracy even for shorter acquisition durations.

Evaluation of block-sequential regularized expectation maximization reconstruction of 68Ga-DOTATOC, 18F-fluoride, and 11C-acetate whole-body examinations acquired on a digital time-of-flight PET/CT scanner

Background

Block-sequential regularized expectation maximization (BSREM) is a fully convergent iterative image reconstruction algorithm. We hypothesize that tracers with different distribution patterns will result in different optimal settings for the BSREM algorithm. The aim of this study was to evaluate the image quality with variations in the applied β-value and acquisition time for three positron emission tomography (PET) tracers. NEMA image quality phantom measurements and clinical whole-body digital time-of-flight (TOF) PET/computed tomography (CT) examinations with ⁶⁸Ga-DOTATOC (n = 13), ¹⁸F-fluoride (n = 10), and ¹¹C-acetate (n = 13) were included. Each scan was reconstructed using BSREM with β-values of 133, 267, 400, and 533, and ordered subsets expectation maximization (OSEM; 3 iterations, 16 subsets, and 5-mm Gaussian post-processing filter). Both reconstruction methods included TOF and point spread function (PSF) recovery. Quantitative measures of noise, signal-to-noise ratio (SNR), and signal-to-background ratio (SBR) were analysed for various acquisition times per bed position (bp).

Results

The highest β-value resulted in the lowest level of noise, which in turn resulted in the highest SNR and lowest SBR. Noise levels equal to or lower than those of OSEM were found with β-values equal to or higher than 400, 533, and 267 for ⁶⁸Ga-DOTATOC, ¹⁸F-fluoride, and ¹¹C-acetate, respectively. The specified β-ranges resulted in increased SNR at a minimum of 25% (P < 0.0001) and SBR at a maximum of 23% (P < 0.0001) as compared to OSEM. At a reduced acquisition time by 25% for ⁶⁸Ga-DOTATOC and ¹⁸F-fluoride, and 67% for ¹¹C-acetate, BSREM with β-values equal to or higher than 533 resulted in noise equal to or lower than that of OSEM at full acquisition duration (2 min/bp for ⁶⁸Ga-DOTATOC and ¹⁸F-fluoride, 3 min/bp for ¹¹C-acetate). The reduced acquisition time with β 533 resulted in increased SNR (16–26%, P < 0.003) and SBR (6–18%, P < 0.0001 (P = 0.07 for ¹¹C-acetate)) compared to the full acquisition OSEM.

Conclusions

Within tracer-specific ranges of β-values, BSREM reconstruction resulted in increased SNR and SBR with respect to conventional OSEM reconstruction. Similar SNR, SBR, and noise levels could be attained with BSREM at relatively shorter acquisition times or, alternatively, lower administered dosages, compared to those attained with OSEM.

Background based Gaussian mixture model lesion segmentation in PET

PURPOSE

Quantitative (18)F-fluorodeoxyglucose positron emission tomography is limited by the uncertainty in lesion delineation due to poor SNR, low resolution, and partial volume effects, subsequently impacting oncological assessment, treatment planning, and follow-up. The present work develops and validates a segmentation algorithm based on statistical clustering. The introduction of constraints based on background features and contiguity priors is expected to improve robustness vs clinical image characteristics such as lesion dimension, noise, and contrast level.

METHODS

An eight-class Gaussian mixture model (GMM) clustering algorithm was modified by constraining the mean and variance parameters of four background classes according to the previous analysis of a lesion-free background volume of interest (background modeling). Hence, expectation maximization operated only on the four classes dedicated to lesion detection. To favor the segmentation of connected objects, a further variant was introduced by inserting priors relevant to the classification of neighbors. The algorithm was applied to simulated datasets and acquired phantom data. Feasibility and robustness toward initialization were assessed on a clinical dataset manually contoured by two expert clinicians. Comparisons were performed with respect to a standard eight-class GMM algorithm and to four different state-of-the-art methods in terms of volume error (VE), Dice index, classification error (CE), and Hausdorff distance (HD).

RESULTS

The proposed GMM segmentation with background modeling outperformed standard GMM and all the other tested methods. Medians of accuracy indexes were VE <3%, Dice >0.88, CE <0.25, and HD <1.2 in simulations; VE <23%, Dice >0.74, CE <0.43, and HD <1.77 in phantom data. Robustness toward image statistic changes (±15%) was shown by the low index changes: <26% for VE, <17% for Dice, and <15% for CE. Finally, robustness toward the user-dependent volume initialization was demonstrated. The inclusion of the spatial prior improved segmentation accuracy only for lesions surrounded by heterogeneous background: in the relevant simulation subset, the median VE significantly decreased from 13% to 7%. Results on clinical data were found in accordance with simulations, with absolute VE <7%, Dice >0.85, CE <0.30, and HD <0.81.

CONCLUSIONS

The sole introduction of constraints based on background modeling outperformed standard GMM and the other tested algorithms. Insertion of a spatial prior improved the accuracy for realistic cases of objects in heterogeneous backgrounds. Moreover, robustness against initialization supports the applicability in a clinical setting. In conclusion, application-driven constraints can generally improve the capabilities of GMM and statistical clustering algorithms.

Quantitative comparison of OSEM and penalized likelihood image reconstruction using relative difference penalties for clinical PET

Ordered subset expectation maximization (OSEM) is the most widely used algorithm for clinical PET image reconstruction. OSEM is usually stopped early and post-filtered to control image noise and does not necessarily achieve optimal quantitation accuracy. As an alternative to OSEM, we have recently implemented a penalized likelihood (PL) image reconstruction algorithm for clinical PET using the relative difference penalty with the aim of improving quantitation accuracy without compromising visual image quality. Preliminary clinical studies have demonstrated visual image quality including lesion conspicuity in images reconstructed by the PL algorithm is better than or at least as good as that in OSEM images. In this paper we evaluate lesion quantitation accuracy of the PL algorithm with the relative difference penalty compared to OSEM by using various data sets including phantom data acquired with an anthropomorphic torso phantom, an extended oval phantom and the NEMA image quality phantom; clinical data; and hybrid clinical data generated by adding simulated lesion data to clinical data. We focus on mean standardized uptake values and compare them for PL and OSEM using both time-of-flight (TOF) and non-TOF data. The results demonstrate improvements of PL in lesion quantitation accuracy compared to OSEM with a particular improvement in cold background regions such as lungs.