Deep learning-based partial volume correction in standard and low-dose positron emission tomography-computed tomography imaging.

Background: Positron emission tomography (PET) imaging encounters the obstacle of partial volume effects, arising from its limited intrinsic resolution, giving rise to (I) considerable bias, particularly for structures comparable in size to the point spread function (PSF) of the system; and (II) blurred image edges and blending of textures along the borders. We set out to build a deep learning-based framework for predicting partial volume corrected full-dose (FD + PVC) images from either standard or low-dose (LD) PET images without requiring any anatomical data in order to provide a joint solution for partial volume correction and de-noise LD PET images. Methods: We trained a modified encoder-decoder U-Net network with standard of care or LD PET images as the input and FD + PVC images by six different PVC methods as the target. These six PVC approaches include geometric transfer matrix (GTM), multi-target correction (MTC), region-based voxel-wise correction (RBV), iterative Yang (IY), reblurred Van-Cittert (RVC), and Richardson-Lucy (RL). The proposed models were evaluated using standard criteria, such as peak signal-to-noise ratio (PSNR), root mean squared error (RMSE), structural similarity index (SSIM), relative bias, and absolute relative bias. Results: Different levels of error were observed for these partial volume correction methods, which were relatively smaller for GTM with a SSIM of 0.63 for LD and 0.29 for FD, IY with an SSIM of 0.63 for LD and 0.67 for FD, RBV with an SSIM of 0.57 for LD and 0.65 for FD, and RVC with an SSIM of 0.89 for LD and 0.94 for FD PVC approaches. However, large quantitative errors were observed for multi-target MTC with an RMSE of 2.71 for LD and 2.45 for FD and RL with an RMSE of 5 for LD and 3.27 for FD PVC approaches. Conclusions: We found that the proposed framework could effectively perform joint de-noising and partial volume correction for PET images with LD and FD input PET data (LD vs. FD). When no magnetic resonance imaging (MRI) images are available, the developed deep learning models could be used for partial volume correction on LD or standard PET-computed tomography (PET-CT) scans as an image quality enhancement technique.

Attention-based deep neural network for partial volume correction in brain 18F-FDG PET imaging.

PURPOSE: This work set out to propose an attention-based deep neural network to predict partial volume corrected images from PET data not utilizing anatomical information. METHODS: An attention-based convolutional neural network (ATB-Net) is developed to predict PVE-corrected images in brain PET imaging by concentrating on anatomical areas of the brain. The performance of the deep neural network for performing PVC without using anatomical images was evaluated for two PVC methods, including iterative Yang (IY) and reblurred Van-Cittert (RVC) approaches. The RVC and IY PVC approaches were applied to PET images to generate the reference images. The training of the U-Net network for the partial volume correction was trained twice, once without using the attention module and once with the attention module concentrating on the anatomical brain regions. RESULTS: Regarding the peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and root mean square error (RMSE) metrics, the proposed ATB-Net outperformed the standard U-Net model (without attention compartment). For the RVC technique, the ATB-Net performed just marginally better than the U-Net; however, for the IY method, which is a region-wise method, the attention-based approach resulted in a substantial improvement. The mean absolute relative SUV difference and mean absolute relative bias improved by 38.02 % and 91.60 % for the RVC method and 77.47 % and 79.68 % for the IY method when using the ATB-Net model, respectively. CONCLUSIONS: Our results propose that without using anatomical data, the attention-based DL model could perform PVC on PET images, which could be employed for PVC in PET imaging.

The effects of various penalty parameter values in Q.Clear algorithm for rectal cancer detection on 18F-FDG images using a BGO-based PET/CT scanner: a phantom and clinical study.

BACKGROUND: The Q.Clear algorithm is a fully convergent iterative image reconstruction technique. We hypothesize that different PET/CT scanners with distinct crystal properties will require different optimal settings for the Q.Clear algorithm. Many studies have investigated the improvement of the Q.Clear reconstruction algorithm on PET/CT scanner with LYSO crystals and SiPM detectors. We propose an optimum penalization factor (ß) for the detection of rectal cancer and its metastases using a BGO-based detector PET/CT system which obtained via accurate and comprehensive phantom and clinical studies. METHODS: 18F-FDG PET-CT scans were acquired from NEMA phantom with lesion-to-background ratio (LBR) of 2:1, 4:1, 8:1, and 15 patients with rectal cancer. Clinical lesions were classified into two size groups. OSEM and Q.Clear (ß value of 100-500) reconstruction was applied. In Q.Clear, background variability (BV), contrast recovery (CR), signal-to-noise ratio (SNR), SUVmax, and signal-to-background ratio (SBR) were evaluated and compared to OSEM. RESULTS: OSEM had 11.5-18.6% higher BV than Q.Clear using ß value of 500. Conversely, RC from OSEM to Q.Clear using ß value of 500 decreased by 3.3-7.7% for a sphere with a diameter of 10 mm and 2.5-5.1% for a sphere with a diameter of 37 mm. Furthermore, the increment of contrast using a ß value of 500 was 5.2-8.1% in the smallest spheres compared to OSEM. When the ß value was increased from 100 to 500, the SNR increased by 49.1% and 30.8% in the smallest and largest spheres at LBR 2:1, respectively. At LBR of 8:1, the relative difference of SNR between ß value of 100 and 500 was 43.7% and 44.0% in the smallest and largest spheres, respectively. In the clinical study, as ß increased from 100 to 500, the SUVmax decreased by 47.7% in small and 31.1% in large lesions. OSEM demonstrated the least SUVmax, SBR, and contrast. The decrement of SBR and contrast using OSEM were 13.6% and 12.9% in small and 4.2% and 3.4%, respectively, in large lesions. CONCLUSIONS: Implementing Q.Clear enhances quantitative accuracies through a fully convergent voxel-based image approach, employing a penalization factor. In the BGO-based scanner, the optimal ß value for small lesions ranges from 200 for LBR 2:1 to 300 for LBR 8:1. For large lesions, the optimal ß value is between 400 for LBR 2:1 and 500 for LBR 8:1. We recommended ß value of 300 for small lesions and ß value of 500 for large lesions in clinical study.

Phantom and clinical evaluation of Block Sequential Regularized Expectation Maximization (BSREM) reconstruction algorithm in 68Ga-PSMA PET-CT studies.

In this study, we aimed to examine the effect of varying ß-values in the block sequential regularized expectation maximization (BSREM) algorithm under differing lesion sizes to determine an optimal penalty factor for clinical application. The National Electrical Manufacturers Association phantom and 15 prostate cancer patients were injected with 68Ga-PSMA and scanned using a GE Discovery IQ PET/CT scanner. Images were reconstructed using ordered subset expectation maximization (OSEM) and BSREM with different ß-values. Then, the background variability (BV), contrast recovery, signal-to-noise ratio, and lung residual error were measured from the phantom data, and the signal-to-background ratio (SBR) and contrast from the clinical data. The increment of BV using a ß-value of 100 was 120.0%, and the decrement of BV using a ß-value of 1000 was 40.5% compared to OSEM. As ß decreased from 1000 to 100, the [Formula: see text] increased by 59.0% for a sphere with a diameter of 10 mm and 26.4% for a sphere with a diameter of 37 mm. Conversely, [Formula: see text] increased by 140.5% and 29.0% in the smallest and largest spheres, respectively. Furthermore, the Δ[Formula: see text] and Δ[Formula: see text] were - 41.1% and - 36.7%, respectively. In the clinical study, OSEM exhibited the lowest SBR and contrast. When the ß-value was reduced from 500 to 100, the SBR and contrast increased by 69.7% and 71.8% in small and 35.6% and 33.0%, respectively, in large lesions. Moreover, the optimal ß-value decreased as lesion size decreased. In conclusion, a ß-value of 400 is optimal for small lesion reconstruction, while ß-values of 600 and 500 are optimal for large lesions in phantom and clinical studies, respectively.

Endorectal ultrasound radiomics in locally advanced rectal cancer patients: despeckling and radiotherapy response prediction using machine learning.

PURPOSE: The current study aimed to evaluate the association of endorectal ultrasound (EUS) radiomics features at different denoising filters based on machine learning algorithms and to predict radiotherapy response in locally advanced rectal cancer (LARC) patients. METHODS: The EUS images of forty-three LARC patients, as a predictive biomarker for predicting the treatment response of neoadjuvant chemoradiotherapy (NCRT), were investigated. For despeckling, the EUS images were preprocessed by traditional filters (bilateral, wiener, lee, frost, median, and wavelet filters). The rectal tumors were delineated by two readers separately, and radiomics features were extracted. The least absolute shrinkage and selection operator were used for feature selection. Classifiers including logistic regression (LR), K-nearest neighbor (KNN), support vector machine (SVM), random forest, naive Bayes, and decision tree were trained using stratified fivefold cross-validation for model development. The area under the curve (AUC) of the receiver operating characteristic curve followed by accuracy, precision, sensitivity, and specificity were obtained for model performance assessment. RESULTS: The wavelet filter had the best results with means of AUC: 0.83, accuracy: 77.41%, precision: 82.15%, and sensitivity: 79.41%. LR and SVM by having AUC: 0.71 and 0.76; accuracy: 70.0% and 71.5%; precision: 75.0% and 73.0%; sensitivity: 69.8% and 80.2%; and specificity: 70.0% and 60.9% had the highest model's performance, respectively. CONCLUSION: This study demonstrated that the EUS-based radiomics model could serve as pretreatment biomarkers in predicting pathologic features of rectal cancer. The wavelet filter and machine learning methods (LR and SVM) had good results on the EUS images of rectal cancer.

Statistical analysis of COVID-19 infection severity in lung lobes from chest CT.

Detection of the COVID 19 virus is possible through the reverse transcription-polymerase chain reaction (RT-PCR) kits and computed tomography (CT) images of the lungs. Diagnosis via CT images provides a faster diagnosis than the RT-PCR method does. In addition to low false-negative rate, CT is also used for prognosis in determining the severity of the disease and the proposed treatment method. In this study, we estimated a probability density function (PDF) to examine the infections caused by the virus. We collected 232 chest CT of suspected patients and had them labeled by two radiologists in 6 classes, including a healthy class and 5 classes of different infection severity. To segment the lung lobes, we used a pre-trained U-Net model with an average Dice similarity coefficient (DSC) greater than 0.96. First, we extracted the PDF to grade the infection of each lobe and selected five specific thresholds as feature vectors. We then assigned this feature vector to a support vector machine (SVM) model and made the final prediction of the infection severity. Using the T-Test statistics, we calculated the p-value at different pixel thresholds and reported the significant differences in the pixel values. In most cases, the p-value was less than 0.05. Our developed model was developed on roughly labeled data without any manual segmentation, which estimated lung infection involvements with the area under the curve (AUC) in the range of [0.64, 0.87]. The introduced model can be used to generate a systematic automated report for individual patients infected by COVID-19.

Temporal and extratemporal atrophic manifestation of temporal lobe epilepsy using voxel-based morphometry and corticometry: clinical application in lateralization of epileptogenic zone.

BACKGROUND: Advances in MRI acquisition and data processing have become important for revealing brain structural changes. Previous studies have reported widespread structural brain abnormalities and cortical thinning in patients with temporal lobe epilepsy (TLE), as the most common form of focal epilepsy. METHODS: In this research, healthy control cases (n = 20) and patients with left TLE (n = 19) and right TLE (n = 14) were recruited, all underwent 3.0 T MRI with magnetization-prepared rapid gradient echo sequence to acquire T1-weighted images. Morphometric alterations in gray matter were identified using voxel-based morphometry (VBM). Volumetric alterations in subcortical structures and cortical thinning were also determined. RESULTS: Patients with left TLE demonstrated more prevailing and widespread changes in subcortical volumes and cortical thickness than right TLE, mainly in the left hemisphere, compared to the healthy group. Both VBM analysis and subcortical volumetry detected significant hippocampal atrophy in ipsilateral compared to contralateral side in TLE group. In addition to hippocampus, subcortical volumetry found the thalamus and pallidum bilaterally vulnerable to the TLE. Furthermore, the TLE patients underwent cortical thinning beyond the temporal lobe, affecting gray matter cortices in frontal, parietal, and occipital lobes in the majority of patients, more prevalently for left TLE cases. Exploiting volume changes in individual patients in the hippocampus alone led to 63.6% sensitivity and 100% specificity for lateralization of TLE. CONCLUSION: Alteration of gray matter volumes in subcortical regions and neocortical temporal structures and also cortical gray matter thickness were evidenced as common effects of epileptogenicity, as manifested by the majority of cases in this study.

Compton scatter tomography in TOF-PET.

Scatter coincidences contain hidden information about the activity distribution on the positron emission tomography (PET) imaging system. However, in conventional reconstruction, the scattered data cause the blurring of images and thus are estimated and subtracted from detected coincidences. List mode format provides a new aspect to use time of flight (TOF) and energy information of each coincidence in the reconstruction process. In this study, a novel approach is proposed to reconstruct activity distribution using the scattered data in the PET system. For each single scattering coincidence, a scattering angle can be determined by the recorded energy of the detected photons, and then possible locations of scattering can be calculated based on the scattering angle. Geometry equations show that these sites lie on two arcs in 2D mode or the surface of a prolate spheroid in 3D mode, passing through the pair of detector elements. The proposed method uses a novel and flexible technique to estimate source origin locations from the possible scattering locations, using the TOF information. Evaluations were based on a Monte-Carlo simulation of uniform and non-uniform phantoms at different resolutions of time and detector energy. The results show that although the energy uncertainties deteriorate the image spatial resolution in the proposed method, the time resolution has more impact on image quality than the energy resolution. With progress of the TOF system, the reconstruction using the scattered data can be used in a complementary manner, or to improve image quality in the next generation of PET systems.

EBT GAFCHROMIC(TM) film dosimetry in compensator-based intensity modulated radiation therapy.

The electron benefit transfer (EBT) GAFCHROMIC films possess a number of features making them appropriate for high-quality dosimetry in intensity-modulated radiation therapy (IMRT). Compensators to deliver IMRT are known to change the beam-energy spectrum as well as to produce scattered photons and to contaminate electrons; therefore, the accuracy and validity of EBT-film dosimetry in compensator-based IMRT should be investigated. Percentage-depth doses and lateral-beam profiles were measured using EBT films in perpendicular orientation with respect to 6 and 18 MV photon beam energies for: (1) different thicknesses of cerrobend slab (open, 1.0, 2.0, 4.0, and 6.0 cm), field sizes (5×5, 10×10, and 20×20 cm(2)), and measurement depths (Dmax, 5.0 and 10.0 cm); and (2) step-wedged compensator in a solid phantom. To verify results, same measurements were implemented using a 0.125 cm(3) ionization chamber in a water phantom and also in Monte Carlo simulations using the Monte Carlo N-particle radiation transport computer code. The mean energy of photons was increased due to beam hardening in comparison with open fields at both 6 and 18 MV energies. For a 20×20 cm(2) field size of a 6 MV photon beam and a 6.0 cm thick block, the surface dose decreased by about 12% and percentage-depth doses increased up to 3% at 30.0 cm depth, due to the beam-hardening effect induced by the block. In contrast, at 18 MV, the surface dose increased by about 8% and depth dose reduced by 3% at 30.0 cm depth. The penumbral widths (80% to 20%) increase with block thickness, field size, and beam energy. The EBT film results were in good agreement with the ionization chamber dose profiles and Monte Carlo N-particle radiation transport computer code simulation behind the step-wedged compensator. Also, there was a good agreement between the EBT-film and the treatment-planning results on the anthropomorphic phantom. The EBT films can be accurately used as a 2D dosimeter for dose verification and quality assurance of compensator-based C-IMRT.

Differential dose volume histograms of Gamma knife in the presence of inhomogeneities using MRI-polymer gel dosimetry and MC simulation.

Polymer gel dosimeters offer a practical solution to 3D dose verification for conventional radiotherapy as well as intensity-modulated and stereotactic radiotherapy. In this study, EGSnrc calculated and PAGAT polymer gel dosimeter measured dose volume histograms (DVHs) for single-shot irradiations of the Gamma Knife (GK) unit were used to investigate the effects of the presence of inhomogeneities on 3D dose distribution. The head phantom was a custom-built 16 cm diameter Plexiglas sphere. Inside the phantom, there is a cubic cutout for inserting the gel vials and another cutout for inserting the inhomogeneities. Following irradiation with the GK unit, the polymer gel phantoms were scanned with a 1.5 T MRI scanner. Comparing the results of measurement in homogeneous and heterogeneous phantoms revealed that inserting inhomogeneities inside the homogeneous phantom did not cause considerable disturbances on dose distribution in irradiation with 8 mm collimator within low isodose levels (< 50%), which is essential for the dose sparing of sensitive structures. The results of simulation for homogeneous and inhomogeneous phantoms in irradiation with 18 mm collimator of the GK unit showed 23.24% difference in DVH within 90%-100% relative isodose level and also revealed that a significant part of the target (28.56%) received relative doses higher than the maximum dose, which exceeds the acceptance criterion (5%). Based on these results it is concluded that the presence of inhomogeneities inside the phantom can cause considerable errors in dose calculation within high isodose levels with respect to LGP prediction which assumes that the target is a homogeneous material. Moreover, it is demonstrated that the applied MC code is an accurate and stand-alone tool for 3D evaluation of dose distribution in irradiation with the GK unit, which can provide important, 3D plan evaluation criteria used in clinical practice.

Performance evaluation of MRI-based PAGAT polymer gel dosimeter in an inhomogeneous phantom using EGSnrc code on a Co-60 machine.

Relative isodose curves were obtained using PAGAT gel dosimeter on homogeneous and inhomogeneous phantoms. Distance-to-agreement (DTA) was calculated between simulated and measured values for both the homogeneous and inhomogeneous phantoms. All DTAs except one passed the acceptance criterion (+/-5 dose variation for selected isodose levels). Results of this study also showed the ability of the Monte Carlo modeling to provide accurate dosimetry, and revealed that the dose response of PAGAT polymer gel is dependent on the method of fabrication.