Interpretation of SPECT wall motion with deep learning.

OBJECTIVES: We sought to develop a novel deep learning (DL) workflow to interpret single-photon emission computed tomography (SPECT) wall motion. BACKGROUND: Wall motion assessment with SPECT is limited by image temporal and spatial resolution. Visual interpretation of wall motion can be subjective and prone to error. Artificial intelligence (AI) may improve accuracy of wall motion assessment. METHODS: A total of 1038 patients undergoing rest electrocardiogram (ECG)-gated SPECT and echocardiography were included. Using echocardiography as truth, a DL-model (DL-model 1) was trained to predict the probability of abnormal wall motion. Of the 1038 patients, 317 were used to train a DL-model (DL-model 2) to assess regional wall motion. A 10-fold cross-validation was adopted. Diagnostic performance of DL was compared with human readers and quantitative parameters. RESULTS: The area under the receiver operating characteristic curve (AUC) and accuracy (ACC) of DL model (AUC: .82 [95% CI: .79-.85]; ACC: .88) were higher than human (AUC: .77 [95% CI: .73-.81]; ACC: .82; P < .001) and quantitative parameter (AUC: .74 [95% CI: .66-.81]; ACC: .78; P < .05). The net reclassification index (NRI) was 7.7%. The AUC and accuracy of DL model for per-segment and per-vessel territory diagnosis were also higher than human reader. The DL model generated results within 30 seconds with operable guided user interface (GUI) and therefore could provide preliminary interpretation. CONCLUSIONS: DL can be used to improve interpretation of rest SPECT wall motion as compared with current human readers and quantitative parameter diagnosis.

Myocardial blood flow quantification with SPECT.

INTRODUCTION: The addition of absolute myocardial blood flow (MBF) data improves the diagnostic and prognostic accuracy of relative perfusion imaging with nuclear medicine. Cardiac-specific gamma cameras allow measurement of MBF with SPECT. METHODS: This paper reviews the evidence supporting the use of SPECT to measure myocardial blood flow (MBF). Studies have evaluated SPECT MBF in large animal models and compared it in humans with invasive angiographic measurements and against the clinical standard of PET MBF. The repeatability of SPECT MBF has been determined in both single-site and multi-center trials. RESULTS: SPECT MBF has excellent correlation with microspheres in an animal model, with the number of stenoses and fractional flow reserve, and with PET-derived MBF. The inter-user coefficient of variability is ∼20% while the COV of test-retest MBF is ∼30%. SPECT MBF improves the sensitivity and specificity of the detection of multi-vessel disease over relative perfusion imaging and provides incremental value in predicting adverse cardiac events. CONCLUSION: SPECT MBF is a promising technique for providing clinically valuable information in the assessment of coronary artery disease.

Multicenter Evaluation of the Feasibility of Clinical Implementation of SPECT Myocardial Blood Flow Measurement: Intersite Variability and Imaging Time.

BACKGROUND: Single-center studies have shown that single photon emission computed tomography myocardial blood flow (MBF) measurement is accurate compared with MBF measured with microspheres in a porcine model, positron emission tomography, and angiography. Clinical implementation requires consistency across multiple sites. The study goal is to determine the intersite processing repeatability of single photon emission computed tomography MBF and the additional camera time required. METHODS: Five sites (Canada, Italy, Japan, Germany, and Singapore) each acquired 25 to 35 MBF studies at rest and with pharmacological stress using technetium-99m-tetrofosmin on a pinhole-collimated cadmium-zinc-telluride-based cardiac single photon emission computed tomography camera with standardized list-mode imaging and processing protocols. Patients had intermediate to high pretest probability of coronary artery disease. MBF was measured locally and at a core laboratory using commercially available software. The time a room was occupied for an MBF study was compared with that for a standard rest/stress myocardial perfusion study. RESULTS: With motion correction, the overall correlation in MBF between core laboratory and local site was 0.93 (range, 0.87-0.97) at rest, 0.90 (range, 0.84-0.96) at stress, and 0.84 (range, 0.70-0.92) for myocardial flow reserve. The local-to-core difference in global MBF (bias-MBF) was 5.4% (-3.8% to 14.8%; median [interquartile range]) at rest and 5.4% (-6.2% to 19.4%) at stress. Between the 5 sites, bias-MBF ranged from -1.6% to 11.0% at rest and from -1.9% to 16.3% at stress; the interquartile range in bias-MBF was between 9.3% (4.8%-14.0%) and 22.3% (-10.3% to 12.0%) at rest and between 17.0% (-11.3% to 5.6%) and 33.3% (-10.4% to 22.9%) at stress and was not significantly different between most sites. Both bias and interquartile range were like previously reported interobserver variability and less than the SD of the test-retest difference of 30%. The overall difference in myocardial flow reserve was 1.52% (-10.6% to 11.3%). There were no significant differences between with and without motion correction. The average additional acquisition time varied between sites from 44 to 79 minutes. CONCLUSIONS: The average bias-MBF and bias-MFR values were small with standard deviations substantially less than the test-retest variability. This demonstrates that MBF can be measured consistently across multiple sites and further supports that this technique can be reliably implemented. REGISTRATION: URL: https://www. CLINICALTRIALS: gov; Unique identifier: NCT03427749.

Exponential dosing to standardize myocardial perfusion image quality with rubidium-82 PET.

BACKGROUND: 82Rb PET is commonly performed using the same injected activity in all patients, resulting in lower image quality in larger patients. This study compared 82Rb dosing with exponential vs proportional functions of body weight on the standardization of myocardial perfusion image (MPI) quality. METHODS: Two sequential cohorts of N = 60 patients were matched by patient weight. Rest and dipyridamole stress 82Rb PET was performed using 0.1 MBq·kg-2 exponential and 9 MBq·kg-1 proportional dosing. MPI scans were compared qualitatively with visual image quality scoring (IQS) and quantitatively using the myocardium-to-blood contrast-to-noise ratio (CNR) and blood background signal-to-noise ratio (SNR) as a function of body weight. RESULTS: Average (min-max) patient body weight was 81 ± 18 kg (46-137 kg). Proportional dosing resulted in decreasing CNR, SNR, and visual IQS with increasing body weight (P < 0.05). Exponential dosing eliminated the weight-dependent decreases in these image quality metrics that were observed in the proportional dosing group. CONCLUSION: 82Rb PET dosing as an exponential (squared) function of body weight produced consistent stress perfusion image quality over a wide range of patient weights. Dramatically lower doses can be used in lighter patients, with the equivalent population dose shifted toward the heavier patients to standardize diagnostic image quality.

Improved precision of SPECT myocardial blood flow using a net tracer retention model.

BACKGROUND: Noninvasive quantification of absolute myocardial blood flow (MBF) and myocardial flow reserve (MFR) provides incremental benefit to relative myocardial perfusion imaging (MPI) to diagnose and manage heart disease. MBF can be measured with single-photon emission computed tomography (SPECT) but the uncertainty in the measured values is high. Standardization and optimization of protocols for SPECT MBF measurements will improve the consistency of this technique. One element of the processing protocol is the choice of kinetic model used to analyze the dynamic image series. PURPOSE: This study evaluates if a net tracer retention model (RET) will provide a better fit to the acquired data and greater test-retest precision than a one-compartment model (1CM) for SPECT MBF, with (+MC) and without (-MC) manual motion correction. METHODS: Data from previously acquired rest-stress MBF studies (31 SPECT-PET and 30 SPECT-SPECT) were reprocessed ± MC. Rate constants (K1) were extracted using 1CM and RET, +/-MC, and compared pairwise with standard PET MBF measurements using cross-validation to obtain calibration parameters for converting SPECT rate constants to MBF and to assess the goodness-of-fit of the calibration curves. Precision (coefficient of variation of test re-test relative differences, COV) of flow measurements was computed for 1CM and RET ± MC using data from the repeated SPECT MBF studies. RESULTS: Both the RET model and MC improved the goodness-of-fit of the SPECT MBF calibration curves to PET. All models produced minimal bias compared with PET (mean bias < 0.6%). The SPECT-SPECT MBF COV significantly improved from 34% (1CM+MC) to 28% (RET+MC, P = 0.008). CONCLUSION: The RET+MC model provides a better calibration of SPECT to PET and blood flow measurements with better precision than the 1CM, without loss of accuracy.

Acquisition, Processing, and Interpretation of PET 18F-FDG Viability and Inflammation Studies.

PURPOSE OF REVIEW: This article reviews the acquisition protocols and image interpretation for 18F-fluorodeoxyglucose (18F-FDG) imaging with positron emission tomography (PET) applied to the evaluation of myocardial viability and inflammation. RECENT FINDINGS: Cardiac PET with 18F-FDG provides essential information for the assessment of myocardial viability and inflammation and is usually combined with PET perfusion imaging using 82Rb or 13N-ammonia. Viable myocardium maintains glucose metabolism which can be detected via the uptake of 18F-FDG by PET imaging. The patient is prepared for viability imaging by shifting the metabolism of the heart to maximize the uptake of glucose and hence of 18F-FDG. Comparison of the 18F-FDG and myocardial perfusion images allows distinction between regions of the myocardium that are hibernating and thus may recover function with intervention, from those that are infarcted. Increased glucose utilization in the inflammatory cells also makes 18F-FDG a useful imaging technique in conditions such as cardiac sarcoidosis. Here, suppression of normal myocardial uptake is essential for accurate image interpretation. 18F-FDG PET broadens the scope of information potentially available through a cardiac PET study. With careful patient preparation, it provides valuable insights into myocardial viability and inflammatory processes such as sarcoidosis.

Feasibility of attenuation map alignment in pinhole cardiac SPECT using exponential data consistency conditions.

PURPOSE: Dedicated cardiac SPECT systems do not typically include an integrated CT scanner and thus attenuation correction requires registration of separately acquired transmission scans. Data consistency conditions are equations that express the redundancy between projections while taking into account the attenuation effects. This study assessed the feasibility of applying exponential data consistency conditions to rebinned pinhole projections for attenuation-map registration in pinhole cardiac SPECT. METHODS: Simulations of an anthropomorphic computer phantom with three different tracer activity distributions were performed with and without clinical levels of noise in the projection data. The first activity distribution contained activity only within the myocardium which satisfied the assumptions of the data consistency conditions. The other two distributions violated these assumptions by adding background activity and uptake in the liver. Simulations included acquisitions with 360, 31, and 9 pinhole projections and detector pixel sizes of 0.75 and 2.5 mm. A metric based on the average difference between pairs of exponential projections was used to evaluate registration accuracy. RESULTS: When activity is restricted to the myocardium, the registration error was 3.0 mm for 31 noisy pinhole projections with a detector size of 2.5 mm. When activity is added to the background and the liver, a correction for the extra-cardiac activity is needed but when applied, a registration error of 6.0 mm was achieved. CONCLUSION: These results suggest that it may be feasible to use exponential data consistency conditions to register pinhole cardiac SPECT and CT transmission data. Taxonomy: 8-6 (IM-SPECT/Registration).

Guidelines on Setting Up Stations for Remote Viewing of Nuclear Medicine and Molecular Imaging Studies During COVID-19.

The current pandemic has created a situation where nuclear medicine practitioners and medical physicists read or process nuclear medicine images remotely from their home office. This article presents recommendations on the components and specifications when setting up a remote viewing station for nuclear medicine imaging.

Measuring SPECT myocardial blood flow at the University of Ottawa Heart Institute.

The introduction of new cardiac SPECT cameras has made it practical to do dynamic SPECT imaging and opened the door to performing myocardial blood flow (MBF) imaging with SPECT. In this paper, we describe in detail our approach to dynamic SPECT MBF imaging using a multi-pinhole cardiac SPECT camera and commercially available kinetic analysis software. We use a 1-day rest/stress protocol with 370 MBq injected at rest and 1,000 MBq at stress with a 1- to 2-hour interval between rest and stress imaging. The tracer is injected mechanically over 30 seconds using a syringe pump. Projection data are acquired in listmode for a duration of 11 minutes and then reframed into a dynamic series. Each image is reconstructed independently using vendor-supplied software. The dynamic images are corrected for residual activity and manually corrected for motion using rigid-body translation. The uptake rate constant, K1, is calculated using a 1-tissue-compartment kinetic model and converted to MBF using a previously determined extraction fraction correction.

Reduced acquisition times for measurement of myocardial blood flow with 99mTc-tetrofosmin and solid-state detector SPECT.

BACKGROUND: Measurement of myocardial blood flow (MBF) is feasible using SPECT imaging but the acquisition requires more time than usual. Our study assessed the impact of reducing acquisition times on the accuracy and repeatability of the uptake rate constant (K1). METHODS: Twenty-nine patients underwent two rest/stress studies with Tc-99m-tetrofosmin 18 ± 13 days apart, using a one-day rest/stress dynamic SPECT imaging protocol with a solid-state cardiac camera. A 5-minute static image was acquired prior to tracer injection for subtraction of residual activity, followed immediately by 11-minute of list-mode data collection. Static image acquisition times of 0.5, 1, and 3 minutes and dynamic imaging times of 5, 7, and 9 minutes were simulated by truncating list-mode data. Images were reconstructed with/without attenuation correction and with/without motion correction. Kinetic parameters were calculated using a 1-tissue-compartment model. RESULTS: K1 increased with reduced dynamic but not static imaging time (P < 0.001). The increase in K1 for a 9-minute scan was small (4.7 ± 5.3%) compared with full-length studies. The repeatability of K1 did not change significantly (13 ± 12%, P > 0.17). CONCLUSIONS: A shortened imaging protocol of 3-minute (rest) or 30-second (stress) static image acquisition and 9 minutes of dynamic image acquisition altered K1 by less than 5% compared to a previously validated 11-minute acquisition.

Patient-specific SPECT imaging protocols to standardize image noise.

BACKGROUND: In addition to acquired photon counts, image noise depends on the image reconstruction algorithm. This work develops patient-specific activity or acquisition time protocols to standardize the average noise in a reconstructed image for different patients, cameras, and reconstruction algorithms. METHODS: Image noise was calculated for images from 43 patients acquired on both a conventional and a multiple-pinhole cardiac SPECT camera. Functions were found to relate image noise to radiotracer activity, scan time, and body mass and were validated by normalizing the image noise in a test set of 58 patients. RESULTS: There was a 3.6-fold difference in photon sensitivity between the two cameras but a 16-fold difference in activity-scan time was necessary to match the noise levels. Image noise doubled from 45 to 128 kg for the conventional camera (12.8 minutes) and tripled for the multiple-pinhole camera (5 minutes) for 350 MBq (9.5 mCi) 99mTc-tetrofosmin. It was 16.3% and 6.1% respectively for an average sized patient. CONCLUSIONS: A linear scaling of activity with respect to the patient weight normalizes image noise but the scaling factors depend on the choice of camera and image reconstruction parameters. Therefore, equivalent numbers of acquired photon counts are not sufficient to guarantee equivalent image noise.

Noise heterogeneity in attenuation-corrected cardiac SPECT images increases perfusion value uncertainty near the base of the heart.

BACKGROUND: Dedicated cardiac SPECT cameras which employ multi-pinhole detectors have variable photon sensitivity within the camera's field-of-view such that a lower number of photon counts is typically detected from the base of the heart than from the apex. Consequently, the noise in a reconstructed image is expected to be higher at the base than at the apex of the heart. METHODS: Patient emission images were resampled to create statistical replicates which were reconstructed with and without attenuation correction. Noise images were computed using one standard deviation of the replicated images. These were evaluated for 93 patients with normal study results, each imaged with both a dual-headed parallel-hole camera and a multi-pinhole camera. Statistics for a normal database (NDB) of images from the 93 patients were also calculated. RESULTS: Image noise (1.7-fold) and NDB uncertainty (1.3-fold) increase significantly from the apex-to-the base of the heart in attenuation-corrected multi-pinhole SPECT images. The differences for non-attenuation-corrected images or those acquired with a parallel-hole camera were not significant. CONCLUSIONS: For best interpretation of attenuation-corrected images acquired with multi-pinhole cameras, knowledge of NDB uncertainty gradients should be taken into consideration.

A Clinical Tool to Identify Candidates for Stress-First Myocardial Perfusion Imaging.

OBJECTIVES: This study sought to develop a clinical model that identifies a lower-risk population for coronary artery disease that could benefit from stress-first myocardial perfusion imaging (MPI) protocols and that can be used at point of care to risk stratify patients. BACKGROUND: There is an increasing interest in stress-first and stress-only imaging to reduce patient radiation exposure and improve patient workflow and experience. METHODS: A secondary analysis was conducted on a single-center cohort of patients undergoing single-photon emission computed tomography (SPECT) and positron emission tomography (PET) studies. Normal MPI was defined by the absence of perfusion abnormalities and other ischemic markers and the presence of normal left ventricular wall motion and left ventricular ejection fraction. A model was derived using a cohort of 18,389 consecutive patients who underwent SPECT and was validated in a separate cohort of patients who underwent SPECT (n = 5,819), 1 internal cohort of patients who underwent PET (n=4,631), and 1 external PET cohort (n = 7,028). RESULTS: Final models were made for men and women and consisted of 9 variables including age, smoking, hypertension, diabetes, dyslipidemia, typical angina, prior percutaneous coronary intervention, prior coronary artery bypass graft, and prior myocardial infarction. Patients with a score ≤1 were stratified as low risk. The model was robust with areas under the curve of 0.684 (95% confidence interval [CI]: 0.674 to 0.694) and 0.681 (95% CI: 0.666 to 0.696) in the derivation cohort, 0.745 (95% CI: 0.728 to 0.762) and 0.701 (95% CI: 0.673 to 0.728) in the SPECT validation cohort, 0.672 (95% CI: 0.649 to 0.696) and 0.686 (95% CI: 0.663 to 0.710) in the internal PET validation cohort, and 0.756 (95% CI: 0.740 to 0.772) and 0.737 (95% CI: 0.716 to 0.757) in the external PET validation cohort in men and women, respectively. Men and women who scored ≤1 had negative likelihood ratios of 0.48 and 0.52, respectively. CONCLUSIONS: A novel model, based on easily obtained clinical variables, is proposed to identify patients with low probability of having abnormal MPI results. This point-of-care tool may be used to identify a population that might qualify for stress-first MPI protocols.

Test-Retest Precision of Myocardial Blood Flow Measurements With 99mTc-Tetrofosmin and Solid-State Detector Single Photon Emission Computed Tomography.

BACKGROUND: Measurement of myocardial blood flow (MBF) with single photon emission computed tomography (SPECT) is feasible using cardiac cameras with solid-state detectors. SPECT MBF has been shown to be accurate when compared with positron emission tomography MBF measured in the same patients. However, the value of a test result applied to an individual patient depends strongly on the precision or repeatability of the test. The purpose of our study is to measure the precision of SPECT MBF measurements using 99mTc-tetrofosmin and a solid-state cardiac camera. METHODS: SPECT MBF was measured in 30 patients and repeated at a mean interval of 18 days. MBF was evaluated from images with and without attenuation correction based on a separately acquired CT scan. The dynamic images were processed independently by 2 operators using in-house kinetic analysis software that applied a 1-tissue-compartment model. The K1 rate constant was converted to MBF using previously determined extraction fraction corrections. Correction for patient body motion was applied manually. RESULTS: The average coefficient of variation (COV) in the differences between the 2 MBF measurements was between 28% and 31%. The interobserver COV was between 11% and 15%. Myocardial flow reserve is the ratio of MBF measured at stress and rest, and the COV is correspondingly higher. The COV for the difference in repeated myocardial flow reserve was 33% to 38%, whereas the interobserver COV was 13% to 22%. CONCLUSIONS: The COV for the difference in SPECT MBF measurements obtained on separate days is 28% to 31%. The corresponding COV for myocardial flow reserve is 33% to 38%.

Correction to: Patient-specific SPECT imaging protocols to standardize image noise.

The originally published version of this article contained typographical errors in the units of photon sensitivity. The units of counts · MBq-1 · min-1 and kcounts · mCi-1 · min-1 were mistakenly recorded as counts · MBq · min and kcounts · mCi · min respectively. The original article has been corrected.