Postreconstruction filtering of 3D PET images by using weighted higher-order singular value decomposition

Deep learning generation of preclinical positron emission tomography (PET) images from low-count PET with task-based performance assessment

BACKGROUND

Preclinical low-count positron emission tomography (LC-PET) imaging offers numerous advantages such as facilitating imaging logistics, enabling longitudinal studies of long- and short-lived isotopes as well as increasing scanner throughput. However, LC-PET is characterized by reduced photon-count levels resulting in low signal-to-noise ratio (SNR), segmentation difficulties, and quantification uncertainties.

PURPOSE

We developed and evaluated a novel deep-learning (DL) architecture-Attention based Residual-Dilated Net (ARD-Net)-to generate standard-count PET (SC-PET) images from LC-PET images. The performance of the ARD-Net framework was evaluated for numerous low count realizations using fidelity-based qualitative metrics, task-based segmentation, and quantitative metrics.

METHOD

Patient Derived tumor Xenograft (PDX) with tumors implanted in the mammary fat-pad were subjected to preclinical [18F]-Fluorodeoxyglucose (FDG)-PET/CT imaging. SC-PET images were derived from a 10 min static FDG-PET acquisition, 50 min post administration of FDG, and were resampled to generate four distinct LC-PET realizations corresponding to 10%, 5%, 1.6%, and 0.8% of SC-PET count-level. ARD-Net was trained and optimized using 48 preclinical FDG-PET datasets, while 16 datasets were utilized to assess performance. Further, the performance of ARD-Net was benchmarked against two leading DL-based methods (Residual UNet, RU-Net; and Dilated Network, D-Net) and non-DL methods (Non-Local Means, NLM; and Block Matching 3D Filtering, BM3D). The performance of the framework was evaluated using traditional fidelity-based image quality metrics such as Structural Similarity Index Metric (SSIM) and Normalized Root Mean Square Error (NRMSE), as well as human observer-based tumor segmentation performance (Dice Score and volume bias) and quantitative analysis of Standardized Uptake Value (SUV) measurements. Additionally, radiomics-derived features were utilized as a measure of quality assurance (QA) in comparison to true SC-PET. Finally, a performance ensemble score (EPS) was developed by integrating fidelity-based and task-based metrics. Concordance Correlation Coefficient (CCC) was utilized to determine concordance between measures. The non-parametric Friedman Test with Bonferroni correction was used to compare the performance of ARD-Net against benchmarked methods with significance at adjusted p-value ≤0.01.

RESULTS

ARD-Net-generated SC-PET images exhibited significantly better (p ≤ 0.01 post Bonferroni correction) overall image fidelity scores in terms of SSIM and NRMSE at majority of photon-count levels compared to benchmarked DL and non-DL methods. In terms of task-based quantitative accuracy evaluated by SUVMean and SUVPeak, ARD-Net exhibited less than 5% median absolute bias for SUVMean compared to true SC-PET and lower degree of variability compared to benchmarked DL and non-DL based methods in generating SC-PET. Additionally, ARD-Net-generated SC-PET images displayed higher degree of concordance to SC-PET images in terms of radiomics features compared to non-DL and other DL approaches. Finally, the ensemble score suggested that ARD-Net exhibited significantly superior performance compared to benchmarked algorithms (p ≤ 0.01 post Bonferroni correction).

CONCLUSION

ARD-Net provides a robust framework to generate SC-PET from LC-PET images. ARD-Net generated SC-PET images exhibited superior performance compared other DL and non-DL approaches in terms of image-fidelity based metrics, task-based segmentation metrics, and minimal bias in terms of task-based quantification performance for preclinical PET imaging.

Improving detection accuracy of perfusion defect in standard dose SPECT-myocardial perfusion imaging by deep-learning denoising

BACKGROUND

We previously developed a deep-learning (DL) network for image denoising in SPECT-myocardial perfusion imaging (MPI). Here we investigate whether this DL network can be utilized for improving detection of perfusion defects in standard-dose clinical acquisitions.

METHODS

To quantify perfusion-defect detection accuracy, we conducted a receiver-operating characteristic (ROC) analysis on reconstructed images with and without processing by the DL network using a set of clinical SPECT-MPI data from 190 subjects. For perfusion-defect detection hybrid studies were used as ground truth, which were created from clinically normal studies with simulated realistic lesions inserted. We considered ordered-subset expectation-maximization (OSEM) reconstruction with corrections for attenuation, resolution, and scatter and with 3D Gaussian post-filtering. Total perfusion deficit (TPD) scores, computed by Quantitative Perfusion SPECT (QPS) software, were used to evaluate the reconstructed images.

RESULTS

Compared to reconstruction with optimal Gaussian post-filtering (sigma = 1.2 voxels), further DL denoising increased the area under the ROC curve (AUC) from 0.80 to 0.88 (P-value < 10-4). For reconstruction with less Gaussian post-filtering (sigma = 0.8 voxels), thus better spatial resolution, DL denoising increased the AUC value from 0.78 to 0.86 (P-value < 10-4) and achieved better spatial resolution in reconstruction.

CONCLUSIONS

DL denoising can effectively improve the detection of abnormal defects in standard-dose SPECT-MPI images over conventional reconstruction.

Deep Learning-Based Diffusion-Weighted Magnetic Resonance Imaging in the Diagnosis of Ischemic Penumbra in Early Cerebral Infarction

The prefiltered image was imported into the local higher-order singular value decomposition (HOSVD) denoising algorithm (GL-HOSVD)-optimized diffusion-weighted imaging (DWI) image, which was compared with the deviation correction nonlocal mean (NL mean) and low-level edge algorithm (LR + edge). Regarding the peak signal-to-noise ratio (PSNR), root mean square error (RMSE), sensitivity, specificity, accuracy, and consistency, the application effect of the GL-HOSVD algorithm in DWI was investigated, and its adoption effect in the examination of ischemic penumbra (IP) of early acute cerebral infarction (ACI) patients was evaluated. A total of 210 patients with ACI were selected as the research subjects, who were randomly rolled into two groups. Those who were checked by conventional DWI were set as the control group, and those who used DWI based on the GL-HOSVD denoising algorithm were set as the observation group, with 105 people in each. Positron emission tomography (PET) test results were set as the gold standard to evaluate the application value of the two examination methods. It was found that under different noise levels, the peak signal-to-noise ratio (PSNR) of the GL-HOSVD algorithm and the root mean square error (RMSE) of the FA parameter were better than those of the nonlocal means (NL-means) of deviation correction and low-rank edge algorithm (LR + edge). The sensitivity, specificity, accuracy, and consistency (8.76%, 81.25%, 87.62%, and 0.52) of the observation group were higher than those of the control group (57.78%, 53.33%, 57.14%, and 0.35) (P < 0.05). Moreover, the apparent diffusion coefficient (ADC) of the DWI images of the observation group was basically consistent with that of the PET images, while the control group had a poor display effect and low definition. In summary, under different noise levels, the GL-HOSVD algorithm had a good denoising effect and greatly reduced fringe artifacts. DWI after denoising had high sensitivity, specificity, accuracy, and consistency in the detection of IP, which was worthy of clinical application and promotion.

Leveraging deep neural networks to improve numerical and perceptual image quality in low-dose preclinical PET imaging

The amount of radiotracer injected into laboratory animals is still the most daunting challenge facing translational PET studies. Since low-dose imaging is characterized by a higher level of noise, the quality of the reconstructed images leaves much to be desired. Being the most ubiquitous techniques in denoising applications, edge-aware denoising filters, and reconstruction-based techniques have drawn significant attention in low-count applications. However, for the last few years, much of the credit has gone to deep-learning (DL) methods, which provide more robust solutions to handle various conditions. Albeit being extensively explored in clinical studies, to the best of our knowledge, there is a lack of studies exploring the feasibility of DL-based image denoising in low-count small animal PET imaging. Therefore, herein, we investigated different DL frameworks to map low-dose small animal PET images to their full-dose equivalent with quality and visual similarity on a par with those of standard acquisition. The performance of the DL model was also compared to other well-established filters, including Gaussian smoothing, nonlocal means, and anisotropic diffusion. Visual inspection and quantitative assessment based on quality metrics proved the superior performance of the DL methods in low-count small animal PET studies, paving the way for a more detailed exploration of DL-assisted algorithms in this domain.

Performance evaluation of the Q.Clear reconstruction framework versus conventional reconstruction algorithms for quantitative brain PET-MR studies

Background

Q.Clear is a Bayesian penalized likelihood (BPL) reconstruction algorithm that presents improvements in signal-to-noise ratio (SNR) in clinical positron emission tomography (PET) scans. Brain studies in research require a reconstruction that provides a good spatial resolution and accentuates contrast features however, filtered back-projection (FBP) reconstruction is not available on GE SIGNA PET-Magnetic Resonance (PET-MR) and studies have been reconstructed with an ordered subset expectation maximization (OSEM) algorithm. This study aims to propose a strategy to approximate brain PET quantitative outcomes obtained from images reconstructed with Q.Clear versus traditional FBP and OSEM.

Methods

Contrast recovery and background variability were investigated with the National Electrical Manufacturers Association (NEMA) Image Quality (IQ) phantom. Resolution, axial uniformity and SNR were investigated using the Hoffman phantom. Both phantoms were scanned on a Siemens Biograph 6 TruePoint PET-Computed Tomography (CT) and a General Electric SIGNA PET-MR, for FBP, OSEM and Q.Clear. Differences between the metrics obtained with Q.Clear with different β values and FBP obtained on the PET-CT were determined.

Results

For in plane and axial resolution, Q.Clear with low β values presented the best results, whereas for SNR Q.Clear with higher β gave the best results. The uniformity results are greatly impacted by the β value, where β < 600 can yield worse uniformity results compared with the FBP reconstruction.

Conclusion

This study shows that Q.Clear improves contrast recovery and provides better resolution and SNR, in comparison to OSEM, on the PET-MR. When using low β values, Q.Clear can provide similar results to the ones obtained with traditional FBP reconstruction, suggesting it can be used for quantitative brain PET kinetic modelling studies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40658-021-00386-3.

Deep learning with noise-to-noise training for denoising in SPECT myocardial perfusion imaging

PURPOSE

Post-reconstruction filtering is often applied for noise suppression due to limited data counts in myocardial perfusion imaging (MPI) with single-photon emission computed tomography (SPECT). We study a deep learning (DL) approach for denoising in conventional SPECT-MPI acquisitions, and investigate whether it can be more effective for improving the detectability of perfusion defects compared to traditional postfiltering.

METHODS

Owing to the lack of ground truth in clinical studies, we adopt a noise-to-noise (N2N) training approach for denoising in SPECT-MPI images. We consider a coupled U-Net (CU-Net) structure which is designed to improve learning efficiency through feature map reuse. For network training we employ a bootstrap procedure to generate multiple noise realizations from list-mode clinical acquisitions. In the experiments we demonstrated the proposed approach on a set of 895 clinical studies, where the iterative OSEM algorithm with three-dimensional (3D) Gaussian postfiltering was used to reconstruct the images. We investigated the detection performance of perfusion defects in the reconstructed images using the non-prewhitening matched filter (NPWMF), evaluated the uniformity of left ventricular (LV) wall in terms of image intensity, and quantified the effect of smoothing on the spatial resolution of the reconstructed LV wall by using its full-width at half-maximum (FWHM).

RESULTS

Compared to OSEM with Gaussian postfiltering, the DL denoised images with CU-Net significantly improved the detection performance of perfusion defects at all contrast levels (65%, 50%, 35%, and 20%). The signal-to-noise ratio (SNR_D ) in the NPWMF output was increased on average by 8% over optimal Gaussian smoothing (P < 10^-4 , paired t-test), while the inter-subject variability was greatly reduced. The CU-Net also outperformed a 3D nonlocal means (NLM) filter and a convolutional autoencoder (CAE) denoising network in terms of SNR_D . In addition, the FWHM of the LV wall in the reconstructed images was varied by less than 1%. Furthermore, CU-Net also improved the detection performance when the images were processed with less post-reconstruction smoothing (a trade-off of increased noise for better LV resolution), with SNR_D improved on average by 23%.

CONCLUSIONS

The proposed DL with N2N training approach can yield additional noise suppression in SPECT-MPI images over conventional postfiltering. For perfusion defect detection, DL with CU-Net could outperform conventional 3D Gaussian filtering with optimal setting as well as NLM and CAE.

Liver segmentation from CT images using a sparse priori statistical shape model (SP-SSM)

This study proposes a new liver segmentation method based on a sparse a priori statistical shape model (SP-SSM). First, mark points are selected in the liver a priori model and the original image. Then, the a priori shape and its mark points are used to obtain a dictionary for the liver boundary information. Second, the sparse coefficient is calculated based on the correspondence between mark points in the original image and those in the a priori model, and then the sparse statistical model is established by combining the sparse coefficients and the dictionary. Finally, the intensity energy and boundary energy models are built based on the intensity information and the specific boundary information of the original image. Then, the sparse matching constraint model is established based on the sparse coding theory. These models jointly drive the iterative deformation of the sparse statistical model to approximate and accurately extract the liver boundaries. This method can solve the problems of deformation model initialization and a priori method accuracy using the sparse dictionary. The SP-SSM can achieve a mean overlap error of 4.8% and a mean volume difference of 1.8%, whereas the average symmetric surface distance and the root mean square symmetric surface distance can reach 0.8 mm and 1.4 mm, respectively.