Deep learning approach using SPECT-to-PET translation for attenuation correction in CT-less myocardial perfusion SPECT imaging

OBJECTIVE

Deep learning approaches have attracted attention for improving the scoring accuracy in computed tomography-less single photon emission computed tomography (SPECT). In this study, we proposed a novel deep learning approach referring to positron emission tomography (PET). The aims of this study were to analyze the agreement of representative voxel values and perfusion scores of SPECT-to-PET translation model-generated SPECT (SPECTSPT) against PET in 17 segments according to the American Heart Association (AHA).

METHODS

This retrospective study evaluated the patient-to-patient stress, resting SPECT, and PET datasets of 71 patients. The SPECTSPT generation model was trained (stress: 979 image pairs, rest: 987 image pairs) and validated (stress: 421 image pairs, rest: 425 image pairs) using 31 cases of SPECT and PET image pairs using an image-to-image translation network. Forty of 71 cases of left ventricular base-to-apex short-axis images were translated to SPECTSPT in the stress and resting state (stress: 1830 images, rest: 1856 images). Representative voxel values of SPECT and SPECTSPT in the 17 AHA segments against PET were compared. The stress, resting, and difference scores of 40 cases of SPECT and SPECTSPT were also compared in each of the 17 segments.

RESULTS

For AHA 17-segment-wise analysis, stressed SPECT but not SPECTSPT voxel values showed significant error from PET at basal anterior regions (segments #1, #6), and at mid inferoseptal regions (segments #8, #9, and #10). SPECT, but not SPECTSPT, voxel values at resting state showed significant error at basal anterior regions (segments #1, #2, and #6), and at mid inferior regions (segments #8, #9, and #11). Significant SPECT overscoring was observed against PET in basal-to-apical inferior regions (segments #4, #10, and #15) during stress. No significant overscoring was observed in SPECTSPT at stress, and only moderate over and underscoring in the basal inferior region (segment #4) was found in the resting and difference states.

CONCLUSIONS

Our PET-supervised deep learning model is a new approach to correct well-known inferior wall attenuation in SPECT myocardial perfusion imaging. As standalone SPECT systems are used worldwide, the SPECTSPT generation model may be applied as a low-cost and practical clinical tool that provides powerful auxiliary information for the diagnosis of myocardial blood flow.

Comparison of deep learning-based denoising methods in cardiac SPECT

BACKGROUND

Myocardial perfusion SPECT (MPS) images often suffer from artefacts caused by low-count statistics. Poor-quality images can lead to misinterpretations of perfusion defects. Deep learning (DL)-based methods have been proposed to overcome the noise artefacts. The aim of this study was to investigate the differences among several DL denoising models.

METHODS

Convolution neural network (CNN), residual neural network (RES), UNET and conditional generative adversarial neural network (cGAN) were generated and trained using ordered subsets expectation maximization (OSEM) reconstructed MPS studies acquired with full, half, three-eighths and quarter acquisition time. All DL methods were compared against each other and also against images without DL-based denoising. Comparisons were made using half and quarter time acquisition data. The methods were evaluated in terms of noise level (coefficient of variation of counts, CoV), structural similarity index measure (SSIM) in the myocardium of normal patients and receiver operating characteristic (ROC) analysis of realistic artificial perfusion defects inserted into normal MPS scans. Total perfusion deficit scores were used as observer rating for the presence of a perfusion defect.

RESULTS

All the DL denoising methods tested provided statistically significantly lower noise level than OSEM without DL-based denoising with the same acquisition time. CoV of the myocardium counts with the different DL noising methods was on average 7% (CNN), 8% (RES), 7% (UNET) and 14% (cGAN) lower than with OSEM. All DL methods also outperformed full time OSEM without DL-based denoising in terms of noise level with both half and quarter acquisition time, but this difference was not statistically significant. cGAN had the lowest CoV of the DL methods at all noise levels. Image quality and polar map uniformity of DL-denoised images were also better than reduced acquisition time OSEM's. SSIM of the reduced acquisition time OSEM was overall higher than with the DL methods. The defect detection performance of full time OSEM measured as area under the ROC curve (AUC) was on average 0.97. Half time OSEM, CNN, RES and UNET provided equal or nearly equal AUC. However, with quarter time data CNN, RES and UNET had an average AUC of 0.93, which was lower than full time OSEM's AUC, but equal to quarter acquisition time OSEM. cGAN did not achieve the defect detection performance of the other DL methods. Its average AUC with half time data was 0.94 and 0.91 with quarter time data.

CONCLUSIONS

DL-based denoising effectively improved noise level with slightly lower perfusion defect detection performance than full time reconstruction. cGAN achieved the lowest noise level, but at the same time the poorest defect detection performance among the studied DL methods.

Increasing angular sampling through deep learning for stationary cardiac SPECT image reconstruction

BACKGROUND

The GE Discovery NM (DNM) 530c/570c are dedicated cardiac SPECT scanners with 19 detector modules designed for stationary imaging. This study aims to incorporate additional projection angular sampling to improve reconstruction quality. A deep learning method is also proposed to generate synthetic dense-view image volumes from few-view counterparts.

METHODS

By moving the detector array, a total of four projection angle sets were acquired and combined for image reconstructions. A deep neural network is proposed to generate synthetic four-angle images with 76 ([Formula: see text]) projections from corresponding one-angle images with 19 projections. Simulated data, pig, physical phantom, and human studies were used for network training and evaluation. Reconstruction results were quantitatively evaluated using representative image metrics. The myocardial perfusion defect size of different subjects was quantified using an FDA-cleared clinical software.

RESULTS

Multi-angle reconstructions and network results have higher image resolution, improved uniformity on normal myocardium, more accurate defect quantification, and superior quantitative values on all the testing data. As validated against cardiac catheterization and diagnostic results, deep learning results showed improved image quality with better defect contrast on human studies.

CONCLUSION

Increasing angular sampling can substantially improve image quality on DNM, and deep learning can be implemented to improve reconstruction quality in case of stationary imaging.

A novel deep-learning-based approach for automatic reorientation of 3D cardiac SPECT images

PURPOSE

Reconstructed transaxial cardiac SPECT images need to be reoriented into standard short-axis slices for subsequent accurate processing and analysis. We proposed a novel deep-learning-based method for fully automatic reorientation of cardiac SPECT images and evaluated its performance on data from two clinical centers.

METHODS

We used a convolutional neural network to predict the 6 rigid-body transformation parameters and a spatial transformation network was then implemented to apply these parameters on the input images for image reorientation. A novel compound loss function which balanced the parametric similarity and penalized discrepancy of the prediction and training dataset was utilized in the training stage. Data from a set of 322 patients underwent data augmentation to 6440 groups of images for the network training, and a dataset of 52 patients from the same center and 23 patients from another center were used for evaluation. Similarity of the 6 parameters was analyzed between the proposed and the manual methods. Polar maps were generated from the output images and the averaged count values of the 17 segments were computed from polar maps to evaluate the quantitative accuracy of the proposed method.

RESULTS

All the testing patients achieved automatic reorientation successfully. Linear regression results showed the 6 predicted rigid parameters and the average count value of the 17 segments having good agreement with the reference manual method. No significant difference by paired t-test was noticed between the rigid parameters of our method and the manual method (p > 0.05). Average count values of the 17 segments show a smaller difference of the proposed and manual methods than those between the existing and manual methods.

CONCLUSION

The results strongly indicate the feasibility of our method in accurate automatic cardiac SPECT reorientation. This deep-learning-based reorientation method has great promise for clinical application and warrants further investigation.

Development of a new Python-based cardiac phantom for myocardial SPECT imaging

PURPOSE

The aim of this work was to develop a digital dynamic cardiac phantom able to mimic gated myocardial perfusion single photon emission computed tomography (SPECT) images.

METHODS

A software code package was written to construct a cardiac digital phantom based on mathematical ellipsoidal model utilizing powerful numerical and mathematic libraries of python programing language. An ellipsoidal mathematical model was adopted to create the left ventricle geometrical volume including myocardial boundaries, left ventricular cavity, with incorporation of myocardial wall thickening and motion. Realistic myocardial count density from true patient studies was used to simulate statistical intensity variation during myocardial contraction. A combination of different levels of defect extent and severity were precisely modeled taking into consideration defect size variation during cardiac contraction. Wall thickening was also modeled taking into account the effect of partial volume.

RESULTS

It has been successful to build a python-based software code that is able to model gated myocardial perfusion SPECT images with variable left ventricular volumes and ejection fraction. The recent flexibility of python programming enabled us to manipulate the shape and control the functional parameters in addition to creating variable sized-defects, extents and severities in different locations. Furthermore, the phantom code also provides different levels of image filtration mimicking those filters used in image reconstruction and their influence on image quality. Defect extent and severity were found to impact functional parameter estimation in consistence to clinical examinations.

CONCLUSION

A python-based gated myocardial perfusion SPECT phantom has been successfully developed. The phantom proved to be reliable to assess cardiac software analysis tools in terms of perfusion and functional parameters. The software code is under further development and refinement so that more functionalities and features can be added.

Comparison of coronary CT angiography versus functional imaging for CABG patients: A resource utilization analysis

AIMS

The impact of anatomical versus functional testing in patients with prior coronary artery bypass surgery (CABG) is poorly defined. We therefore sought to determine the rates of downstream investigations and the attendant healthcare costs in CABG patients undergoing CCTA versus SPECT.

METHODS AND RESULTS

2754 consecutive CABG patients were imaged by SPECT (2163) or CCTA (591). 425 patients (15.4%) underwent downstream testing which was more common in those imaged with CCTA versus SPECT (23.18% vs 13.31% respectively, p < 0.01). When a propensity score adjustment was made for differences in baseline characteristics, the findings in downstream testing persisted (p < 0.01). When patients who subsequently underwent repeat revascularization (arguably the highest risk patients) were removed from the analysis, downstream testing remained more frequent in CCTA (12.7%) versus SPECT imaged patients (8.8%) (p = 0.01). Costs of downstream tests per patient were two-fold greater in the CCTA group in comparison to the SPECT group ($366.79 ± 29.59 vs $167.35 ± 10.12 respectively, p < 0.01). Conversely, total costs which included the index costs were less in the CCTA group, $764.66 ± 29.59 versus $1396.73 ± 1012 for the SPECT cohort, p < 0.0001).

CONCLUSIONS

Index imaging with SPECT versus CCTA in CABG patients was associated with fewer downstream tests, less ICA, less repeat revascularization but greater expense. Cost however is only part of the decision making process that determines an optimal index test. Until CCTA demonstrates improved risk stratification over SPECT in CABG patients it is likely SPECT will remain the preferred first imaging test.

EANM procedural guidelines for myocardial perfusion scintigraphy using cardiac-centered gamma cameras

An increasing number of Nuclear Medicine sites in Europe are using cardiac-centered gamma cameras for myocardial perfusion scintigraphy (MPS). Three cardiac-centered gamma cameras are currently the most frequently used in Europe: the D-SPECT (Spectrum Dynamics), the Alcyone (Discovery NM 530c and Discovery NM/CT 570c; General Electric Medical Systems), and the IQ-SPECT (Siemens Healthcare). The increased myocardial count sensitivity of these three cardiac-centered systems has allowed for a decrease in the activities of radiopharmaceuticals injected to patients for myocardial perfusion imaging and, consequently, radiation exposure of patients. When setting up protocols for MPS, the overall objective should be to maintain high diagnostic accuracy of MPS, while injecting the lowest activities reasonably achievable to reduce the level of radiation exposure of patient and staff. These guidelines aim at providing recommendations for acquisition protocols and image interpretation using cardiac-centered cameras. As each imaging system has specific design and features for image acquisition and analysis, these guidelines have been separated into three sections for each gamma camera system. These recommendations have been written by the members of the Cardiovascular Committee of EANM and were based on their own experience with each of these systems and on the existing literature.

The Relationship Between Absence Coronary Artery Calcification and Myocardial Perfusion Single Photon Emission Computed Tomography

Background

Coronary artery calcification score (CACS) is well validated prognostic tool in coronary artery disease (CAD). The data on the prevalence of myocardial ischemia on myocardial perfusion single photon emission computed tomography (MPS) in symptomatic patients with zero CACS and low to intermediate risk probability is lacking and controversial. The aim of our study was to evaluate the capability of zero CACS to exclude myocardial ischemia on MPS.

Methods

A total of 157 patients ((mean age 53 ± 10 years), 88 (56%) female patients, 69 (44%) male patients) who were suspected to have CAD and having low to intermediate pretest likelihood for CAD underwent CACS on dedicated computed tomography (CT) scanners. CACS was reported as zero in all patients, subsequently all patients underwent MPS. Patients with abnormal MPS underwent additional imaging with coronary computed tomography angiography (CCTA).

Results

All patients had zero CACS, of which 122 (78%) had normal MPS, and 35 (22%) had abnormal MPS. Abnormal MPS included fixed defect in 22 (13%), equivocal in 10 (6%), and reversible defect in four (3%) patients. All patients with abnormal MPS had further imaging with CTCA. CTCA was normal in 30 (85%) patients, one patient had coronary artery stenosis more than 50%, one patient had coronary artery stenosis less than 50%, one patient had anomalous origin coronary artery, and two patients had myocardial bridging. Patients with abnormal MPS and normal coronary artery had dilated cardiomyopathy in 14 (40%), asymmetrical septal hypotrophy in one (3%), and mitral valve disease in three (9%).

Conclusions

Zero CACS in stable patients with low or intermediate risk indicated very low likelihood of obstructive CAD, less than 1%. Patients with zero CACS and normal MPS most likely will not benefit from further testing; however, patients with abnormal MPS will need further imaging with CCTA. CCTA is helpful in this group of patients for evaluation of coronary artery and cardiac morphology.

CT-based SPECT attenuation correction and assessment of infarct size: results from a cardiac phantom study

RATIONALE

Myocardial perfusion SPECT is a commonly performed, well established, clinically useful procedure for the management of patients with coronary artery disease. However, the attenuation of photons from myocardium impacts the quantification of infarct sizes. CT-Attenuation Correction (AC) potentially resolves this problem. This contention was investigated by analyzing various parameters for infarct size delineation in a cardiac phantom model.

METHODS

A thorax phantom with a left ventricle (LV), fillable defects, lungs, spine and liver was used. The defects were combined to simulate 6 infarct sizes (5-20% LV). The LV walls were filled with 100120 kBq/ml ^99mTc and the liver with 10-12 kBq/ml ^99mTc. The defects were filled with water of 50% LV activity to simulate transmural and non-transmural infarction, respectively. Imaging of the phantom was repeated for each configuration in a SPECT/CT system. The defects were positioned in the anterior as well as in the inferior wall. Data were acquired in two modes: 32 views, 30 s/view, 180° and 64 views, 15 s/view, 360° orbit. Images were reconstructed iteratively with scatter correction and resolution recovery. Polar maps were generated and defect sizes were calculated with variable thresholds (40-60%, in 5% steps). The threshold yielding the best correlation and the lowest mean deviation from the true extents was considered optimal.

RESULTS

AC data showed accurate estimation of transmural defect extents with an optimal threshold of 50% [non attenuation correction (NAC): 40%]. For the simulation of non-transmural defects, a threshold of 55% for AC was found to yield the best results (NAC: 45%). The variability in defect size due to the location (anterior versus inferior) of the defect was reduced by 50% when using AC data indicating the benefit from using AC. No difference in the optimal threshold was observed between the different orbits.

CONCLUSION

Cardiac SPECT/CT shows an improved capability for quantitative defect size assessment in phantom studies due to the positive effects of attenuation correction.

Advances in SPECT and PET Hardware

There have been significant recent advances in single photon emission computed tomography (SPECT) and positron emission tomography (PET) hardware. Novel collimator designs, such as multi-pinhole and locally focusing collimators arranged in geometries that are optimized for cardiac imaging have been implemented to reduce imaging time and radiation dose. These new collimators have been coupled with solid state photon detectors to further improve image quality and reduce scanner size. The new SPECT scanners demonstrate up to a 7-fold increase in photon sensitivity and up to 2 times improvement in image resolution. Although PET scanners are used primarily for oncological imaging, cardiac imaging can benefit from the improved PET sensitivity of 3D systems without inter-plane septa and implementation of the time-of-flight reconstruction. Additionally, resolution recovery techniques are now implemented by all major PET vendors. These new methods improve image contrast, image resolution, and reduce image noise. Simultaneous PET/magnetic resonance (MR) hybrid systems have been developed. Solid state detectors with avalanche photodiodes or digital silicon photomultipliers have also been utilized in PET. These new detectors allow improved image resolution, higher count rate, as well as a reduced sensitivity to electromagnetic MR fields.