Machine learning-based prediction of conversion coefficients for I-123 metaiodobenzylguanidine heart-to-mediastinum ratio.

PURPOSE: We developed a method of standardizing the heart-to-mediastinal ratio in 123I-labeled meta-iodobenzylguanidine (MIBG) images using a conversion coefficient derived from a dedicated phantom. This study aimed to create a machine-learning (ML) model to estimate conversion coefficients without using a phantom. METHODS: 210 Monte Carlo (MC) simulations of 123I-MIBG images to obtain conversion coefficients using collimators that differed in terms of hole diameter, septal thickness, and length. Simulated conversion coefficients and collimator parameters were prepared as training datasets, then a gradient-boosting ML was trained to estimate conversion coefficients from collimator parameters. Conversion coefficients derived by ML were compared with those that were MC simulated and experimentally derived from 613 phantom images. RESULTS: Conversion coefficients were superior when estimated by ML compared with the classical multiple linear regression model (root mean square deviations: 0.021 and 0.059, respectively). The experimental, MC simulated, and ML-estimated conversion coefficients agreed, being, respectively, 0.54, 0.55, and 0.55 for the low-; 0.74, 0.70, and 0.72 for the low-middle; and 0.88, 0.88, and 0.88 for the medium-energy collimators. CONCLUSIONS: The ML model estimated conversion coefficients without the need for phantom experiments. This means that conversion coefficients were comparable when estimated based on collimator parameters and on experiments.

The utility of heart-to-mediastinum ratio using a planar image created from IQ-SPECT with Iodine-123 meta-iodobenzylguanidine.

AIMS: 123I-labeled meta-iodobenzylguanidine (MIBG) has used a planar image to measure the heart-to-mediastinum ratio (HMR). However, planar images are not available from IQ-SPECT with SMARTZOOM collimator due to its multi-focal collimation. Since we created the planar-equivalent (IQ-planar) images by adding all slices of the IQ-SPECT coronal image. The aim of this study was to demonstrate the utility of the new method for calculating HMR. METHODS: The planar image and transverse images of IQ-SPECT with attenuation and scatter corrections (ACSC) and without ACSC (NC) were obtained. Multi-planar reconstruction and ray-summation processing were applied to create IQ-planar images with NC and ACSC. Linear regression between the measured HMR from the planar image and the mathematically calculated HMR was used to calibrate HMR to standardized values. RESULTS: Scatterplots and linear regression lines between planar and IQ-planar HMRs before and after cross-calibration showed systematic differences in both NC and ACSC conditions. The IQ-planar HMR with NC and ACSC was significantly higher compared with that of the conventional planar image. However, the IQ-planar HMR with NC and ACSC after cross-calibration was similar to the standardized HMR calculated by planar image. CONCLUSION: The IQ-planar HMR using the new ray-summation processing method could be used along with the conventional planar HMR.

Experimental evaluation of the GE NM/CT 870 CZT clinical SPECT system equipped with WEHR and MEHRS collimator.

PURPOSE: A high-energy-resolution whole-body SPECT-CT device (NM/CT 870 CZT; C-SPECT) equipped with a CZT detector has been developed and is being used clinically. A MEHRS collimator has also been developed recently, with an expected improvement in imaging accuracy using medium-energy radionuclides. The objective of this study was to compare and analyze the accuracies of the following devices: a WEHR collimator and the MEHRS collimator installed on a C-SPECT, and a NaI scintillation detector-equipped Anger-type SPECT (A-SPECT) scanner, with a LEHR and LMEGP. METHODS: A line phantom was used to measure the energy resolutions including collimator characteristics in the planar acquisition of each device using 99m Tc and 123 I. We also measured the system's sensitivity and high-contrast resolution using a lead bar phantom. We evaluated SPECT spatial resolution, high-contrast resolution, radioactivity concentration linearity, and homogeneity, using a basic performance evaluation phantom. In addition, the effect of scatter correction was evaluated by varying the sub window (SW) employed for scattering correction. RESULTS: The energy resolution with 99m Tc was 5.6% in C-SPECT with WEHR and 9.9% in A-SPECT with LEHR. Using 123I, the results were 9.1% in C-SPECT with WEHR, 5.5% in C-SPECT with MEHRS, and 10.4% in A-SPECT with LMEGP. The planar spatial resolution was similar under all conditions, but C-SPECT performed better in SPECT acquisition. High-contrast resolution was improved in C-SPECT under planar condition and SPECT. The sensitivity and homogeneity were improved by setting the SW for scattering correction to 3% of the main peak in C-SPECT. CONCLUSION: C-SPECT demonstrates excellent energy resolution and improved high-contrast resolution for each radionuclide. In addition, when using 123I, careful attention should be paid to SW for scatter correction. By setting the appropriate SW, C-SPECT with MEHRS has an excellent scattered ray removal effect, and highly homogenous imaging is possible while maintaining the high-contrast resolution.

[Validation of Simulation Codes for Nuclear Imaging Using Digital Phantoms].

Validation study of simulation codes was performed based on the measurement of a sphere phantom and the National Electrical Manufacturers Association (NEMA) body phantoms. SIMIND and Prominence Processor were used for the simulation. Both source and density maps were generated using the characteristics of 99mTc energy. A full width at half maximum (FWHM) of the sphere phantom was measured and simulated. Simulated recovery coefficient and the background count coefficient of variation were also compared with the measured values in the body phantom study. When the two simulation codes were compared with actual measurements, maximum relative errors of FWHM values were 3.6% for Prominence Processor and -10.0% for SIMIND. The maximum relative errors of relative recovery coefficients exhibited 11.8% for Prominence Processor and -2.0% for SIMIND in the body phantom study. The coefficients of variation of the SPECT count in the background were significantly different among the measurement and two simulation codes. The simulated FWHM values and recovery coefficients paralleled measured results. However, the noise characteristic differed among actual measurements and two simulation codes in the background count statistics.

[A Nationwide Survey on Additional Scan in Nuclear Medicine Imaging].

The aim of the present study was to clarify the routine protocols and the frequency of added or omitted imaging on nuclear medicine imaging in Japan. A nationwide survey on routine protocols and current state of added or omitted imaging in major nuclear medicine imaging were performed for Japanese nuclear medicine technologist. The survey showed that the routine protocols were almost 100% fixed, some of the routine protocols were found to be useful and percentage of imaging techniques such as single photon emission computed tomography/computed tomography that increased patient burden and reduced through put were low. Furthermore, the survey showed that additional or omission imaging were frequently performed on bone scintigraphy and positron emission tomography and added or omitted judgements were often depend upon the rule of thumb by nuclear medicine technologist. In this study, we have concluded that the quality of examination and the diagnosis might depend on the knowledge of nuclear medicine technologist, performed added or omitted imaging.

Creation and characterization of normal myocardial perfusion imaging databases using the IQ·SPECT system.

BACKGROUND: Image acquisition by short-time single-photon emission-computed tomography (SPECT) has been made feasible by IQ·SPECT. The aim of this study was to generate normal databases (NDBs) of thallium-201 (201Tl) myocardial perfusion imaging for IQ·SPECT, and characterize myocardial perfusion distribution. METHODS AND RESULTS: We retrospectively enrolled 159 patients with a low likelihood of cardiac diseases from four hospitals in Japan. All patients underwent short-time 201Tl myocardial perfusion IQ·SPECT with or without attenuation and scatter correction (ACSC) in either supine or prone position. The mean myocardial counts were calculated using 17-segment polar maps. Three NDBs were derived from supine and prone images as well as supine images with ACSC. Differences between the supine and prone positions were observed in the uncorrected sex-segregated NDBs in the mid-inferolateral counts (p ≤ 0.016 for males and p ≤ 0.002 for females). Differences between IQ·SPECT and conventional SPECT were also observed in the mid-anterior, inferolateral, and apical lateral counts (p ≤ 0.009 for males and p ≤ 0.003 for females). Apical low counts attributed to myocardial thinning were observed in the apical anterior and apex segments in the supine IQ·SPECT NDB with ACSC. CONCLUSIONS: There were significant differences between uncorrected supine and prone NDBs, between uncorrected supine NDB and supine NDB with ACSC, and between uncorrected supine NDB and conventional SPECT NDB. Understanding the pattern of normal distribution in IQ-SPECT short-time acquisitions with and without ACSC will be helpful for interpretation of imaging findings in patients with coronary artery disease (CAD) or low likelihood of CAD and the NDBs will aid in quantitative analysis.

Influence of ROI definition on the heart-to-mediastinum ratio in planar 123I-MIBG imaging.

BACKGROUND: Iodine-123-metaiodobenzylguanidine (123I-MIBG) imaging with estimation of the heart-to-mediastinum ratio (HMR) has been established for risk assessment in patients with chronic heart failure. Our aim was to evaluate the effect of different methods of ROI definition on the renderability of HMR to normal or decreased sympathetic innervation. METHODS AND RESULTS: The results of three different methods of ROI definition (clinical routine (CLI), simple standardization (STA), and semi-automated (AUT) were compared. Ranges of 95% limits of agreement (LoA) of inter-observer variabilities were 0.28 and 0.13 for STA and AUT, respectively. Considering a HMR of 1.60 as the lower limit of normal, 13 of 32 (41%) for method STA and 5 of 32 (16%) for method AUT of all HMR measurements could not be classified to normal or pathologic. Ranges of 95% LoA of inter-method variabilities were 0.72 for CLI vs AUT, 0.65 for CLI vs STA, and 0.31 for STA vs AUT. CONCLUSION: Different methods of ROI definition result in different ranges of the LoA of the measured HMR with relevance for rendering the results to normal or pathological innervation. We could demonstrate that standardized protocols can help keep methodological variabilities limited, narrowing the gray zone of renderability.

A European myocardial 123I-mIBG cross-calibration phantom study.

AIM: Planar myocardial 123I-meta-iodobenzylguanidine (123I-mIBG) scintigraphy is a highly reproducible technique. However, differences in collimator use are one of the most important factors that cause variation among institutions and studies in heart-to-mediastinum (H/M) ratio. Therefore, standardization among various gamma camera-collimator combinations is needed. Previously, a phantom has been developed to cross-calibrate different acquisition conditions in Japan. For further cross-calibration of European myocardial 123I-mIBG imaging, the aim of this study was to collect 123I-mIBG data for H/M ratios from common European gamma camera vendors. METHODS: 210 experiments were performed in 27 European institutions. Based on these experiments, conversion coefficients for each gamma camera-collimator combination were calculated. An averaged conversion coefficient of 0.88 was used to calculate a standardized H/M ratio. RESULTS: On average, LE-collimator-derived H/M ratios were significantly lower compared to ME-collimator-derived H/M ratios. The mean conversion coefficients ranged from 0.553 to 0.605 for the LE-collimator group and from 0.824 to 0.895 for the ME-collimator group. CONCLUSION: Clinically established H/M ratios can be converted into standardized H/M ratios using cross-calibrated conversion coefficients. This standardization is important for identifying appropriate thresholds for adequate risk stratification. In addition, this cross-calibration enables comparison between different national and international data.

Diagnostic accuracy of an artificial neural network compared with statistical quantitation of myocardial perfusion images: a Japanese multicenter study.

PURPOSE: Artificial neural networks (ANN) might help to diagnose coronary artery disease. This study aimed to determine whether the diagnostic accuracy of an ANN-based diagnostic system and conventional quantitation are comparable. METHODS: The ANN was trained to classify potentially abnormal areas as true or false based on the nuclear cardiology expert interpretation of 1001 gated stress/rest 99mTc-MIBI images at 12 hospitals. The diagnostic accuracy of the ANN was compared with 364 expert interpretations that served as the gold standard of abnormality for the validation study. Conventional summed stress/rest/difference scores (SSS/SRS/SDS) were calculated and compared with receiver operating characteristics (ROC) analysis. RESULTS: The ANN generated a better area under the ROC curves (AUC) than SSS (0.92 vs. 0.82, p < 0.0001), indicating better identification of stress defects. The ANN also generated a better AUC than SDS (0.90 vs. 0.75, p < 0.0001) for stress-induced ischemia. The AUC for patients with old myocardial infarction based on rest defects was 0.97 (0.91 for SRS, p = 0.0061), and that for patients with and without a history of revascularization based on stress defects was 0.94 and 0.90 (p = 0.0055 and p < 0.0001 vs. SSS, respectively). The SSS/SRS/SDS steeply increased when ANN values (probability of abnormality) were >0.80. CONCLUSION: The ANN was diagnostically accurate in various clinical settings, including that of patients with previous myocardial infarction and coronary revascularization. The ANN could help to diagnose coronary artery disease.

Comparison of phase dyssynchrony analysis using gated myocardial perfusion imaging with four software programs: Based on the Japanese Society of Nuclear Medicine working group normal database.

PURPOSE: Left ventricular (LV) phase dyssynchrony parameters based on gated myocardial perfusion imaging varied among software programs. The aim of this study was to determine normal ranges and factors affecting phase parameters. METHODS: Normal databases were derived from the Japanese Society of Nuclear Medicine working group (n = 69). The programs were Emory Cardiac Toolbox with SyncTool (ECTb), Quantitative Gated SPECT (QGS), Heart Function View (HFV), and cardioREPO (cREPO); parameters of phase standard deviation (PSD), 95% bandwidth, and entropy were compared with parameters with ECTb as a reference. RESULTS: PSD (degree) was 5.3 ± 3.3 for QGS (P < .0001), 5.4 ± 2.5 for HFV (P < .0001), and 10.3 ± 3.2 for cREPO (P = n. s.) compared with 11.5 ± 5.5 for ECTb. Phase bandwidth with three programs differed significantly from ECTb. Gender differences were significant for all programs, indicating larger variation in males. After adjustment of LV volumes between genders, the difference disappeared except for QGS. The phase parameters showed wider variations in patients with the lower ejection fraction (EF) and larger LV volumes, depending on software types. CONCLUSION: Based on normal ranges of phase dyssynchrony parameters in four software programs, dependency on genders, LV volume, and EF should be considered, indicating the need for careful comparison among different software programs.

Reducing the small-heart effect in pediatric gated myocardial perfusion single-photon emission computed tomography.

BACKGROUND: We compared two reconstruction algorisms and two cardiac functional evaluation software programs in terms of their accuracy for estimating ejection fraction (EF) of small hearts (SH). METHODS: The study group consisted of 66 pediatric patients. Data were reconstructed using a filtered back projection (FBP) method without the resolution correction (RC) and an iterative method based on an ordered subset expectation maximization (OSEM) algorithm with the RC. EF was evaluated using two software programs of quantitative gated single-photon emission computed tomography (SPECT) (QGS) and cardioREPO. We compared the EF of gated myocardial perfusion SPECT to echocardiographic measurement (Echo). RESULTS: Forty-eight of 66 patients had an end-systolic volume < 20 mL which was used as the criterion for being included in the SH group, and the SH effect led to an overestimation of EF. While significant differences were observed between Echo (66.9 ± 5.0%) and QGS-FBP without RC (76.9 ± 8.4%, P < .0001), QGS-OSEM with RC (76.6 ± 8.6%, P < .0001), and cardioREPO-FBP without RC (72.1 ± 10.0%, P = .0011), no significant difference was observed between Echo and cardioREPO-OSEM with RC (67.4 ± 6.1%) in SH group. CONCLUSIONS: In pediatric gated myocardial perfusion SPECT, the SH effect can be significantly reduced when an OSEM algorithm is used with RC in combination with the specific cardioREPO algorithm.

Development and validation of a direct-comparison method for cardiac (123)I-metaiodobenzylguanidine washout rates derived from late 3-hour and 4-hour imaging.

PURPOSE: The washout rate (WR) has been used in (123)I-metaiodobenzylguanidine (MIBG) imaging to evaluate cardiac sympathetic innervation. However, WR varies depending on the time between the early and late MIBG scans. Late scans are performed at either 3 or 4 hours after injection of MIBG. The aim of this study was to directly compare the WR at 3 hours (WR3h) with the WR at 4 hours (WR4h). METHODS: We hypothesized that the cardiac count would reduce linearly between the 3-hour and 4-hour scans. A linear regression model for cardiac counts at two time-points was generated. We enrolled a total of 96 patients who underwent planar (123)I-MIBG scintigraphy early (15 min) and during the late phase at both 3 and 4 hours. Patients were randomly divided into two groups: a model-creation group (group 1) and a clinical validation group (group 2). Cardiac counts at 15 minutes (countearly), 3 hours (count3h) and 4 hours (count4h) were measured. Cardiac count4h was mathematically estimated using the linear regression model from countearly and count3h. RESULTS: In group 1, the actual cardiac count4h/countearly was highly significantly correlated with count3h/countearly (r = 0.979). In group 2, the average estimated count4h was 92.8 ± 31.9, and there was no significant difference between this value and the actual count4h (91.9 ± 31.9). Bland-Altman analysis revealed a small bias of -0.9 with 95 % limits of agreement of -6.2 and +4.3. WR4h calculated using the estimated cardiac count4h was comparable to the actual WR4h (24.3 ± 9.6 % vs. 25.1 ± 9.7 %, p = ns). Bland-Altman analysis and the intraclass correlation coefficient showed that there was excellent agreement between the estimated and actual WR4h. CONCLUSION: The linear regression model that we used accurately estimated cardiac count4h using countearly and count3h. Moreover, WR4h that was mathematically calculated using the estimated count4h was comparable to the actual WR4h.

Diagnostic Performance of Artificial Neural Network for Detecting Ischemia in Myocardial Perfusion Imaging.

BACKGROUND: The purpose of this study was to apply an artificial neural network (ANN) in patients with coronary artery disease (CAD) and to characterize its diagnostic ability compared with conventional visual and quantitative methods in myocardial perfusion imaging (MPI). METHODS AND RESULTS: A total of 106 patients with CAD were studied with MPI, including multiple vessel disease (49%), history of myocardial infarction (27%) and coronary intervention (30%). The ANN detected abnormal areas with a probability of stress defect and ischemia. The consensus diagnosis based on expert interpretation and coronary stenosis was used as the gold standard. The left ventricular ANN value was higher in the stress-defect group than in the no-defect group (0.92±0.11 vs. 0.25±0.32, P<0.0001) and higher in the ischemia group than in the no-ischemia group (0.70±0.40 vs. 0.004±0.032, P<0.0001). Receiver-operating characteristics curve analysis showed comparable diagnostic accuracy between ANN and the scoring methods (0.971 vs. 0.980 for stress defect, and 0.882 vs. 0.937 for ischemia, both P=NS). The relationship between the ANN and defect scores was non-linear, with the ANN rapidly increased in ranges of summed stress score of 2-7 and summed defect score of 2-4. CONCLUSIONS: Although the diagnostic ability of ANN was similar to that of conventional scoring methods, the ANN could provide a different viewpoint for judging abnormality, and thus is a promising method for evaluating abnormality in MPI.

Multicenter cross-calibration of I-123 metaiodobenzylguanidine heart-to-mediastinum ratios to overcome camera-collimator variations.

BACKGROUND: The heart-to-mediastinum ratio (HMR) of (123)I-metaiodobenzylguanidine (MIBG) showed variations among institutions and needs to be standardized among various scinticamera-collimator combinations. METHODS: A total of 225 phantom experiments were performed in 84 institutions to calculate cross-calibration coefficients of HMR. Based on phantom studies, a conversion coefficient for each camera-collimator system was created, including low-energy (LE, n = 125) and a medium-energy (ME, n = 100) collimators. An average conversion coefficient from the most common ME group was used to calculate the standard HMR. In clinical MIBG studies (n = 52) from three institutions, HMRs were standardized from both LE- and ME-type collimators and classified into risk groups of <1.60, 1.60-2.19, and ≥2.20. RESULTS: The average conversion coefficients from the individual camera-collimator condition to the mathematically calculated reference HMR ranged from 0.55 to 0.75 for LE groups and from 0.83 to 0.95 for ME groups. The conversion coefficient of 0.88 was used to unify HMRs from all acquisition conditions. Using the standardized HMR, clinical studies (n = 52) showed good agreement between LE and ME types regarding three risk groups (κ = 0.83, P < .0001, complete agreement in 90%, 42% of the patients reclassified into the same risk group). CONCLUSION: By using the reference HMR and conversion coefficients for the system, HMRs with various conditions can be converted to the standard HMRs in a range of normal to low HMRs.