Development of attenuation correction methods using deep learning in brain-perfusion single-photon emission computed tomography.

PurposeAttenuation correction (AC) based on low-dose computed tomography (CT) could be more accurate in brain single-photon emission computed tomography (SPECT) than the widely used Chang method, and, therefore, has the potential to improve both semi-quantitative analysis and visual image interpretation. The present study evaluated CT-based AC for dopamine transporter SPECT with I-123-ioflupane.Materials and methodsSixty-two consecutive patients in whom I-123-ioflupane SPECT including low-dose CT had been performed were recruited retrospectively at 3 centres. For each patient, 3 different SPECT images were reconstructed: without AC, with Chang AC and with CT-based AC. Distribution volume ratio (DVR) images were obtained by scaling voxel intensities using the whole brain without striata as reference. For assessing the impact of AC on semi-quantitative analysis, specific-to-background ratios (SBR) in caudate and putamen were obtained by fully automated SPM8-based region of interest (ROI) analysis and tested for their diagnostic power using receiver-operator-characteristic (ROC) analysis. For assessing the impact of AC on visual image reading, screenshots of stereotactically normalized DVR images presented in randomized order were interpreted independently by two raters at each centre.ResultsCT-based AC resulted in intermediate SBRs about half way between no AC and Chang. Maximum area under the ROC curve was achieved by the putamen SBR, with negligible impact of AC (0.924, 0.935 and 0.938 for no, CT-based and Chang AC). Diagnostic accuracy of visual interpretation also did not depend on AC.ConclusionsThe impact of CT-based versus Chang AC on the interpretation of I-123-ioflupane SPECT is negligible. Therefore, CT-based AC cannot be recommended for routine use in clinical patient care, not least because of the additional radiation exposure.

Objectives:This study aimed to determine the most important perfusion score in patient selection for coronary angiography (CA) by quantitatively evaluating myocardial perfusion scintigraphy (MPS).Methods:Patients who underwent MPS single-photon emission computerized tomography/computed tomograph imaging in our clinic between December 2017 and January 2019, without coronary artery disease (CAD) history, followed by CA were included in the study. CA was considered positive when there is a stenosis of 70% or more in at least one coronary vessel. The summed stress score, rest score, and differential score; total perfusion deficit (TPD); and the defect’s extent obtained from non-attenuation-corrected (NC) and attenuation-corrected (AC) images of 80 patients were evaluated using the Mann-Whitney U test. A p value of <0.05 was considered significant. Receiver operating characteristic (ROC) analysis was performed.Results:The scores obtained from NC and AC images showed a significant difference between the two groups for all scores except for the extent and TPD scores at rest from AC images. The applied ROC curves’ highest diagnostic value was determined as the TPD score at stress (TPDS) obtained from NC images (area under the curve: 0.880, 95% confidence interval, 0.807-0.952, p<0.001). The cut-off value obtained for the TPDS from the ROC curve was found to be 5.5.Conclusion:The scores obtained from NC images have more power to detect CAD than those obtained from AC images. Patients with no prior CAD history with TPDS score higher than 5 in MPS should be referred for CA with priority.

Summary Aim: 18F-fluoride PET/CT is a promising tool for bone scanning. Recently, guidelines concerning the conduct of 18F-fluoride PET/CT have been published. One open question of the German guideline was the necessity of attenuation correction for 18F-fluoride PET/CT. We evaluated the need for a CT-based attenuation correction in 18F-fluoride PET/CT scans for the detection of bone lesions. Patients and methods: We retrospectively analyzed wholebody 18F-Fluoride PET/CT scans of 59 cancer patients. The lesions were categorized as malignant, benign or inconclusive. This assignment was performed for every lesion in both: attenuation corrected (AC) and non-attenuation- corrected (NAC) images. The maximum standardized uptake values (SUVmax) of the lesion in the AC images were also determined. Results: All bone lesions were detected in both image modalities. The AC images revealed 201 lesions categorized as malignant, 114 as benign and 35 as inconclusive. Without an AC, the results were 209, 116 and 25, respectively (p &gt; 0.05). 10/35 lesions categorized as inconclusive in the AC images were categorized as malignant in the NAC images, whereas 8 lesions were confirmed after comparison with other imaging modalities and follow-up data and 2 lesions were categorized as benign. The SUVmax for lesions identified as malignant showed a broad overlap with the SUV max of benign lesions and can consequently not be used for differentiation. Conclusion: An AC is not necessary for detecting bone lesions on 18F-fluoride PET/CT scans as the detection capability is identical for NAC imaging and lesion assignment was even better than with AC imaging. SUVmax seems not to improve the differentiation between malignant and benign bone lesions.

The effects of attenuation correction (AC) on tumor detection in SPECT oncology have not been studied as intensively as in PET oncology. The different attenuation mechanism in SPECT compared to PET imaging may lead to different effects of AC on tumor detection for the two modalities. Using the theory of filtered-backprojection reconstruction algorithm, we analyzed how the reconstructed images may appear when AC is not performed (non-AC) so that we can reveal the effects of AC on tumor detection. We then designed and performed a series of simulation studies to evaluate the results of the theoretical analysis. Finally, we analyzed the effects of Chang's post-reconstruction AC technique on tumor detection. Our results show that (1) for lung tumors, when the tumor-to-lung uptake ratio is above a threshold, tumors are visible in both AC and non-AC images; around this threshold, tumors disappear in non-AC images, and below this threshold, tumors appear photophenic in non-AC images. These effects are similar to those in PET oncology. (2) In mediastinum and abdomen areas, non-AC images have lower tumor-to-background ratio (TBR) than the true uptake ratio. This effect is opposite to that in PET oncology. (3) Chang's post-reconstruction AC technique does not improve TBR as compared with non-AC. As a conclusion, accurate AC can improve TBR as compared with non-AC, and thus can potentially improve tumor detectability. Chang's AC technique does not improve TBR as compared to non-AC, and is thus not likely to improve tumor detection in SPECT oncology

Somatostatin receptor scintigraphy (SRS) is well-established in neuroendocrine tumour (NET) imaging. This study evaluated the impact of attenuation correction (AC) on SRS SPECT data in patients examined by SPECT-CT. Planar scintigraphy and SPECT-CT of 17 patients (10 men, seven women; age, 40-74 years; mean, 62 years) suffering from NET were included. For the visual assessment of AC, the intensity and contrast of foci classified as pathological were rated in both the non-attenuation corrected (NAC) and the attenuation corrected (AC) SPECT images using a 5-point score. The change in signal intensity after AC was semiquantified two-fold for each focus in both SPECT(AC) and SPECT(NAC): firstly by using tumour-to-background (TB) ratios (defined as T(max)/B(mean)) for the determination of a TB(AC)/TB(NAC) ratio. Secondly, by a T(max,AC)/T(max,NAC) ratio. Both ratios were correlated to the focus depth. A total of 46 pathological foci were found. Focus contrast and intensity significantly increased in 14/46 foci (30%) after AC (mean, 3.7-4.0) in the visual analysis (P<0.001). While TB ratios increased only in 24/46 foci after AC and no correlation between the T(BAC)/T(BNAC) ratio and focus depth (r=0.027; P=0.856) was found, T(max) was higher after AC in all foci and the T(max,AC)/T(max,NAC) ratio showed the expected correlation to focus depth (r=0.650; P<0.01), indicating the superiority of the Tmax approach for the demonstration of the effects of attenuation correction on focal uptake. Attenuation correction of SRS SPECT data by SPECT-CT results in visually more clearly contrasted foci. Moreover, as focus intensity increases, especially in the more centrally localised foci, CT-based AC has a potential to further improve the sensitivity of SRS SPECT.

N-isopropyl-p-123I-iodoamphetamine brain perfusion SPECT has been used with various attenuation coefficients (μ-values); however, optimization is required. This study aimed to determine the optimal μ-value (μopt-value) for Chang attenuation correction (AC) using clinical data by comparing the Chang method and CT-based AC. Methods: We used 100 patients (reference group, 60; disease group, 40) who underwent N-isopropyl-p-123I-iodoamphetamine SPECT. SPECT images of the reference group were obtained to calculate the AC using the Chang method (μ-values, 0.07-0.20; 0.005 interval) and the CT-based method, both without scatter correction (SC) and with SC. The μopt-value with the smallest mean percentage error for the brain regions of the reference group was calculated. Agreement between the Chang and CT-based methods applying the μopt-value was evaluated using Bland-Altman analysis. Additionally, the percentage error in the region of hypoperfusion in the diseased group was compared with the percentage error in the same region in the reference group when the μopt-value was applied. Results: The μopt-values were 0.140 for Chang without SC and 0.160 for Chang with SC. In the Chang method, with the μopt-value applied, fixed and proportional biases were observed in the Bland-Altman analysis (both P < 0.05), and there was a tendency for the percentage error to be underestimated in the limbic regions and overestimated in the central brain regions. There was no significant difference between the disease group and the reference group in the region of hypoperfusion in either Chang without SC or Chang with SC. Conclusion: The present study revealed that the μopt-values of the Chang method are 0.140 without SC and 0.160 with SC.

Objectives:Myocardial perfusion scintigraphy (MPS) is an important diagnostic test for detecting of coronary artery stenosis (CAS); however, tissue attenuation can lead to a difference in accuracy. We evaluated the diagnostic accuracy of attenuation-corrected (AC) and non-attenuation-corrected (NC) MPS for the detection of CAS.Methods:We retrospectively recruited patients who underwent invasive coronary angiography within 10 months after Tc-99m sestamibi MPS. The AC and NC perfusion images were analyzed separately, and each myocardial segment was scored based on relative uptake from 0 to 4. The summed stress score (SSS), summed rest score (SRS), and summed difference score (SDS) were calculated. The diagnostic performances were analyzed using the area under the curve (AUC) of the receiver operating characteristic curve.Results:From 117 patients, significant coronary stenosis was present in 66 patients (56%). The SSS and SRS obtained from NC-images were higher than those from AC, supporting the presence of attenuation artifacts in NC images. The AUC of SSS and SDS were significantly higher than those of SRS in both AC- and NC-images, but no significant difference was found between the AUC of SSS, and those of SDS. The optimal cut-offs were >12 for AC-SSS, >15 for NC-SSS, >4 for AC-SDS and >3 for NC-SDS. There was no statistically significant difference in the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy among AC-SSS, NC-SSS, AC-SDS, and NC-SDS.Conclusion:NC-based Tc-99m-sestamibi MPS promised comparable accuracy to AC images by using different cut-off values for diagnosis.

For research and clinical PET brain studies performed on PET/CT systems, the CT image is often of little benefit beyond attenuation correction. The goal of this work is both to investigate the quantitative accuracy of CT-based attenuation correction (CTAC) for PET brain studies and to determine the lowest-dose protocol that performs adequately. Measurements were performed on a GE Discovery ST PET/CT scanner using the GE kVp-dependent CTAC algorithm. The effects of pitch on CT images were investigated by acquiring CT images of a Defrise disk phantom. The effects of pitch and x-ray tube voltage and current were investigated using a human skull phantom encased in plastic filled with F-18 radioactivity. This phantom was imaged and CT-based attenuation correction factors (CTACF) were created for several permutations of pitch (0.562:1, 0.938:1, 1.375:1, 1.75:1), voltage (80, 100, 120, 140 kVp) and current (10, 100mA). Emission images were acquired and reconstructed using the 3-D reprojection algorithm with the various CTACFs. The measured mean activity concentration is independent of pitch, kVp, and mA and accurate on an absolute scale to ~5%. Anatomically sized regional differences in the brain region indicate that tube voltages less than 100 kVp may not perform adequately (10% of values have a discrepancy greater than 5%). Results indicate that the higher pitch, lower current, and tube voltages down to 100 kVp perform equivalently to higher-dose configurations.

Chang's method, the most widely used attenuation correction (AC) in brain single-photon emission computed tomography (SPECT), requires delineation of the outer contour of the head. Manual and automatic threshold-based methods are prone to errors due to variability of tracer uptake in the scalp. The present study proposes a new method for fully automated delineation of the head based on stereotactical normalization. The method was validated for SPECT with I-123-ioflupane. The new method was compared to threshold-based delineation in 62 unselected patients who had received I-123-ioflupane SPECT at one of 3 centres. The impact on diagnostic power was tested for semi-quantitative analysis and visual reading of the SPECT images (six independent readers). The two delineation methods produced highly consistent semi-quantitative results. This was confirmed by receiver operating characteristic analyses in which the putamen specific-to-background ratio achieved highest area under the curve with negligible effect of the delineation method: 0.935 versus 0.938 for stereotactical normalization and threshold-based delineation, respectively. Visual interpretation of DVR images was also not affected by the delineation method. Delineation of the head contour by stereotactical normalization appears useful for Chang AC in I-123-ioflupane SPECT. It is robust and does not require user interaction. •Chang attenuation correction in brain SPECT requires delineation of the head contour. •Manual and threshold-based methods are prone to errors. •The study proposes a fully-automated method for delineation based on stereotactical normalization. •The method is shown to work reliably in I-123-ioflupane SPECT. •It might improve the workflow of I-123-ioflupane SPECT in everyday patient care.

We evaluated a novel normal database (NDB) generated using single photon emission computed tomography (SPECT) data obtained from healthy brains by using a SPECT/CT system, analyzed using a resolution recovery (RR) technique applied to the three-dimensional stereotactic surface projection (3D-SSP) technique. We used a three-dimensional ordered subset expectation maximization method (3D-OSEM) with applied scatter correction (SC), attenuation correction, and RR to reconstruct the data. We verified the accuracy of the novel NDB's values (Z, extent, and error scores), and compared the novel NDB to the 3D-SSP technique by using simulated misery perfusion-related patient data from a conventional NDB. In addition, Z, extent, and error scores at the precuneus, cuneus, and posterior cingulate were compared under different reconstruction conditions by using the patient data. In the simulation, Z scores decreased when using the novel NDB corrected using computed tomography-based attenuation correction (CTAC), SC, and RR. The extent scores of the posterior cingulate increased using the novel NDB, relative to the other NDBs. The error score with the novel NDB without RR decreased by 15% compared to that of the conventional NDB. Z scores generated from patient data decreased in the novel NDB with RR. The extent scores tended to decrease in the novel NDB with RR. The extent scores in the novel NDB with RR improved at the posterior cingulate, compared to the scores with the other NDBs. However, applying RR to the novel NDB conferred no advantage because the cut-off of the current Z score must be reconsidered when using the additive RR technique.

For PET/CT systems, PET image reconstruction requires corresponding CT images for anatomical localization and attenuation correction. In the case of PET respiratory gating, multiple gated CT scans can offer phase-matched attenuation and motion correction, at the expense of increased radiation dose. We aim to minimize the dose of the CT scan, while preserving adequate image quality for the purpose of PET attenuation correction by introducing sparse view CT data acquisition.We investigated sparse view CT acquisition protocols resulting in ultra-low dose CT scans designed for PET attenuation correction. We analyzed the tradeoffs between the number of views and the integrated tube current per view for a given dose using CT and PET simulations of a 3D NCAT phantom with lesions inserted into liver and lung. We simulated seven CT acquisition protocols with {984, 328, 123, 41, 24, 12, 8} views per rotation at a gantry speed of 0.35 s. One standard dose and four ultra-low dose levels, namely, 0.35 mAs, 0.175 mAs, 0.0875 mAs, and 0.043 75 mAs, were investigated. Both the analytical Feldkamp, Davis and Kress (FDK) algorithm and the Model Based Iterative Reconstruction (MBIR) algorithm were used for CT image reconstruction. We also evaluated the impact of sinogram interpolation to estimate the missing projection measurements due to sparse view data acquisition. For MBIR, we used a penalized weighted least squares (PWLS) cost function with an approximate total-variation (TV) regularizing penalty function. We compared a tube pulsing mode and a continuous exposure mode for sparse view data acquisition. Global PET ensemble root-mean-squares-error (RMSE) and local ensemble lesion activity error were used as quantitative evaluation metrics for PET image quality.With sparse view sampling, it is possible to greatly reduce the CT scan dose when it is primarily used for PET attenuation correction with little or no measureable effect on the PET image. For the four ultra-low dose levels simulated, sparse view protocols with 41 and 24 views best balanced the tradeoff between electronic noise and aliasing artifacts. In terms of lesion activity error and ensemble RMSE of the PET images, these two protocols, when combined with MBIR, are able to provide results that are comparable to the baseline full dose CT scan. View interpolation significantly improves the performance of FDK reconstruction but was not necessary for MBIR. With the more technically feasible continuous exposure data acquisition, the CT images show an increase in azimuthal blur compared to tube pulsing. However, this blurring generally does not have a measureable impact on PET reconstructed images.Our simulations demonstrated that ultra-low-dose CT-based attenuation correction can be achieved at dose levels on the order of 0.044 mAs with little impact on PET image quality. Highly sparse 41- or 24- view ultra-low dose CT scans are feasible for PET attenuation correction, providing the best tradeoff between electronic noise and view aliasing artifacts. The continuous exposure acquisition mode could potentially be implemented in current commercially available scanners, thus enabling sparse view data acquisition without requiring x-ray tubes capable of operating in a pulsing mode.

Attenuation correction (AC), as an essential component of quantitative SPECT imaging, is highly challenging in the absence of CT or transmission scanning in dedicated cardiac SPECT devices. This study investigated the potential of direct attenuation correction (AC) of <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">99m</sup> Tc-sestamibi SPECT myocardial perfusion imaging (MPI) in the image domain using residual (ResNet) and UNet deep convolutional neural networks. MPI-SPECT <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">99m</sup> Tc-ses-tamibi images of 99 subjects were retrospectively used, considering non-attenuation corrected SPECT images as input and CT-based attenuation corrected (CT-AC) SPECT images as reference. Chang’s AC technique was also performed as a baseline for evaluating the performance of deep learning-based models. The quantitative accuracy of the attenuation corrected SPECT images was validated using the proposed methods considering the corresponding CT-AC SPECT images of 19 external datasets as reference. Image-derived metrics, including the voxel-wise mean absolute error (MAE), relative error (RE), structural similarity index measure (SSIM), as well as clinically relevant indices, such as total perfusion deficit (TPD), were calculated. The quantitative results obtained from the external validation dataset of the deep learning models exhibited good agreement with the reference CT-based attenuation corrected MPI-SPECT images. The deep learning models showed superior performance over Chang’s AC method. The ResNet and UNet models resulted in MAE of 20.24±17.63 and 13.65±11.23 and SSIM of 0.99±0.04 and 0.98±0.05, respectively. In comparison, Chang’s approach led to MAE and SSIM of 80.27±47.56 and 0.93±0.09, respectively. Similarly, the clinical evaluation showed a mean TPD of 12.78±9.22% and 12.57±8.93% for ResNet and UNet models, respectively, compared to TPD of 12.84±8.63% for reference SPECT CT-AC images. In contrast, Chang’s approach led to a mean TPD of 16.68±11.24%. The evaluation of deep learning approaches implemented in this work showed the promising potential of deep learning-based techniques in producing attenuation corrected MPI-SPECT images in the image domain without anatomical imaging.

BackgroundIn integrated PET/MR hybrid imaging the evaluation of PET performance characteristics according to the NEMA standard NU 2–2007 is challenging because of incomplete MR-based attenuation correction (AC) for phantom imaging. In this study, a strategy for CT-based AC of the NEMA image quality (IQ) phantom is assessed. The method is systematically evaluated in NEMA IQ phantom measurements on an integrated PET/MR system.MethodsNEMA IQ measurements were performed on the integrated 3.0 Tesla PET/MR hybrid system (Biograph mMR, Siemens Healthcare). AC of the NEMA IQ phantom was realized by an MR-based and by a CT-based method. The suggested CT-based AC uses a template μ-map of the NEMA IQ phantom and a phantom holder for exact repositioning of the phantom on the systems patient table. The PET image quality parameters contrast recovery, background variability, and signal-to-noise ratio (SNR) were determined and compared for both phantom AC methods. Reconstruction parameters of an iterative 3D OP-OSEM reconstruction were optimized for highest lesion SNR in NEMA IQ phantom imaging.ResultsUsing a CT-based NEMA IQ phantom μ-map on the PET/MR system is straightforward and allowed performing accurate NEMA IQ measurements on the hybrid system. MR-based AC was determined to be insufficient for PET quantification in the tested NEMA IQ phantom because only photon attenuation caused by the MR-visible phantom filling but not the phantom housing is considered. Using the suggested CT-based AC, the highest SNR in this phantom experiment for small lesions (<= 13 mm) was obtained with 3 iterations, 21 subsets and 4 mm Gaussian filtering.ConclusionThis study suggests CT-based AC for the NEMA IQ phantom when performing PET NEMA IQ measurements on an integrated PET/MR hybrid system. The superiority of CT-based AC for this phantom is demonstrated by comparison to measurements using MR-based AC. Furthermore, optimized PET image reconstruction parameters are provided for the highest lesion SNR in NEMA IQ phantom measurements.

The purpose of this study was to evaluate the diagnostic value of attenuation-corrected single photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) on the inferior wall compared to uncorrected (NC) SPECT MPI between obese and nonobese patients. A total of 157 consecutive patients (122 males and 35 females, with median age: 57.4 ± 11 years) who underwent AC technetium 99m-methoxyisobutylisonitrile (AC Tc99m-sestamibi) SPECT MPI were included to the study. A hybrid SPECT and transmission computed tomography (CT) system was used for the diagnosis with 1-day protocol, and stress imaging was performed first. During attenuation correction (AC) processing on a Xeleris Workstation using Myovation cardiac software with ordered subset expectation maximization (OSEM), iterative reconstruction with attenuation correction (IRAC) and NC images filtered back projection (FBP) were used. For statistical purposes, P < 0.05 was considered significant. This study included 73 patients with body mass index (BMI) <30 and 84 patients with BMI ≥ 30. In patients with higher BMI, increased amount of both visual and semiquantitative attenuation of the inferior wall was detected. IRAC reconstruction corrects the diaphragm attenuation of the inferior wall better than FBP. AC with OSEM iterative reconstruction significantly improves the diagnostic value of stress-only SPECT MPI in patients with normal weight and those who are obese, but the improvements are significantly greater in obese patients. Stress-only SPECT imaging with AC provides shorter and lower radiation exposure.

Purpose: The main goal of this study was to determine the optimal collimator in the absence of medium energy collimators along with the impact of Attenuation Correction (AC) and different iterative reconstruction protocols on the quantitative evaluation of Gallium-67 (67Ga) SPECT/CT imaging.&#x0D; Materials and Methods: A GE Discovery 670 dual-head SPECT/CT scanner and a NEMA phantom filled with 67Ga solution were used to scan the patients. The projections were acquired with both Low Energy High Resolution (LEHR) and High Energy General Purpose (HEGP) collimators, and CT images were acquired to evaluate the effect of attenuation correction. SPECT data were reconstructed using the ordered subset expectation maximization (OSEM) method with various combinations of iterations and subsets. The performance was quantified, and a clinical study validated the phantom study.&#x0D; Results: Acquired images by the HEGP collimator yielded higher Contrast Recovery (CR) and Contrast to Noise Ratio (CNR) in images with AC than those without non-AC (41.6% and 74.2%, respectively). The CNR in all spheres after AC was increased by 80.4% (82.1%) for the HEGP collimator against the LEHR collimator. Also, an increase in iterations × subsets from 16 to 48 led to the Coefficient of Variation (COV) increasing by 17.2%, 16.67%, 15.50%, 14.4%, 14.2%, and 14.1% for 10 mm to 37 mm sphere diameter, respectively.&#x0D; Conclusion: CT-based AC and HEGP collimators can yield improved 67Ga SPECT quantification compared to Non-AC and LEHR collimators. The choice of the optimal collimator with the reconstruction protocol led to changes in the image quality and quantitative accuracy, emphasizing the need to carefully select the appropriate combination of data acquisition factors.&#x0D;