Short-axis Slices Research Articles

Abstract 12818: Clinical Significance of Myocardial Perfusion in Dilated Cardiomyopathy and Hypokinetic Non-Dilated Cardiomyopathy

Introduction: Dilated cardiomyopathy (DCM) is characterized by left ventricular (LV) systolic dysfunction with LV dilatation. Hypokinetic non-dilated cardiomyopathy (HNDC) characterized by LV dysfunction without LV dilatation is proposed as a new category of DCM. Microvascular dysfunction in DCM is associated with the degree of LV dysfunction. However, the association between myocardial perfusion and clinical spectrum of DCM has not been elucidated. Hypothesis: We assessed the hypothesis that myocardial perfusion is correlated with the clinical spectrum of DCM. Methods: Thirty-six subjects (20 non-ischemic cardiomyopathy (NICM) patients and 16 control subjects, mean age: 59 ± 2 years) underwent CMR at 3T. Cine imaging, adenosine stress perfusion, T1 mapping, and late gadolinium enhancement were performed. NICM group was divided into DCM (n=10) and HNDC (n=10). Native T1 and extracellular volume fraction (ECV) were measured as an averaged values in the septum of 3 short-axis slices of base-to-apex LV myocardium. Myocardial perfusion reserve (MPR) was calculated as the ratio of stress/rest myocardial blood flow. Results: LV ejection fraction was significantly lower in NICM compared to control (29.2±1.6% vs. 56.6±1.5%, p<0.001). Native T1 and ECV were significantly higher in NICM compared to control (native T1: 1296±12ms vs. 1228±8ms, p<0.001, ECV: 32.9±1.7% vs. 27.3±0.62%, p<0.001), but they showed no significant differences between DCM and HNDC. MPR in HNDC was significantly higher than that in DCM (2.0±0.51 vs. 1.5±0.27, p=0.026), and comparable to that in control (1.9±0.11, p=0.717, Fig 1). MPR was negatively correlated with LV end diastolic volume index (r=-0.333, p=0.047, Fig 2), but MPR showed no significant correlations with native T1 and ECV. Conclusions: HNDC showed preserved MPR in contrast to impaired MPR in DCM. Myocardial perfusion may be a promising factor for the clinical spectrum of DCM.

Accuracy of free-breathing multi-parametric SASHA in identifying T1 and T2 elevations in pediatric orthotopic heart transplant patients.

T1/T2 parametric mapping may reveal patterns of elevation ("hotspots") in myocardial diseases, such as rejection in orthotopic heart transplant (OHT) patients. This study aimed to evaluate the diagnostic accuracy of free-breathing (FB) multi-parametric SAturation recovery single-SHot Acquisition (mSASHA) T1/T2 mapping in identifying hotspots present on conventional Breath-held Modified Look-Locker Inversion recovery (BH MOLLI) T1 and T2-prepared balanced steady-state free-precession (BH T2p-bSSFP) maps in pediatric OHT patients. Pediatric OHT patients underwent noncontrast 1.5T CMR with BH MOLLI T1 and T2p-bSSFP and prototype FB mSASHA T1/T2 mapping in 8 short-axis slices. FB and BH T1/T2 hotspots were segmented using semi-automated thresholding (ITK-SNAP) and their 3D coordinate locations were collected (3-Matic, Materialise, Leuven, Belgium). Receiver operator characteristic curve analysis and measures of central tendency were utilized. 40 imaging datasets from 23 pediatric OHT patients were obtained. FB mSASHA yielded a sensitivity of 82.8% for T1 and 80% for T2 maps when compared to the standard BH MOLLI, as well as 100% specificity for both T1 and T2 maps. When identified on both FB and BH maps, hotspots overlapped in all cases, with an average long axis offset between FB and BH hotspot centers of 5.8mm (IQR 3.5-8.2) on T1 and 5.9mm (IQR 3.5-8.2) on T2 maps. FB mSASHA T1/T2 maps can identify hotspots present on conventional BH T1/T2 maps in pediatric patients with OHT, with high sensitivity, specificity, and overlap in 3D space. Free-breathing mapping may improve patient comfort and facilitate OHT assessment in younger patient populations.

Automatic reorientation by deep learning to generate short-axis SPECT myocardial perfusion images

BackgroundSingle photon emission computed tomography (SPECT) myocardial perfusion images (MPI) can be displayed both in traditional short-axis (SA) cardiac planes and polar maps for interpretation and quantification. It is essential to reorient the reconstructed transaxial SPECT MPI into standard SA slices. This study is aimed to develop a deep-learning-based approach for automatic reorientation of MPI. MethodsA total of 254 patients were enrolled, including 226 stress SPECT MPIs and 247 rest SPECT MPIs. Fivefold cross-validation with 180 stress and 201 rest MPIs was used for training and internal validation; the remaining images were used for testing. The rigid transformation parameters (translation and rotation) from manual reorientation were annotated by an experienced nuclear cardiologist and used as the reference standard. A convolutional neural network (CNN) was designed to predict the transformation parameters. Then, the derived transform was applied to the grid generator and sampler in spatial transformer network (STN) to generate the reoriented image. A loss function containing mean absolute errors for translation and mean square errors for rotation was employed. A three-stage optimization strategy was adopted for model optimization: (1) optimize the translation parameters while fixing the rotation parameters; (2) optimize rotation parameters while fixing the translation parameters; (3) optimize both translation and rotation parameters together. ResultsIn the test set, the Spearman determination coefficients of the translation distances and rotation angles between the model prediction and the reference standard were 0.993 in X axis, 0.992 in Y axis, 0.994 in Z axis, 0.987 along X axis, 0.990 along Y axis and 0.996 along Z axis, respectively. For the 46 stress MPIs in the test set, the Spearman determination coefficients were 0.858 in percentage of profusion defect (PPD) and 0.858 in summed stress score (SSS); for the 46 rest MPIs in the test set, the Spearman determination coefficients were 0.9 in PPD and 0.9 in summed rest score (SRS). ConclusionsOur deep learning-based LV reorientation method is able to accurately generate the SA images. Technical validations and subsequent evaluations of measured clinical parameters show that it has great promise for clinical use.

Fractal analysis: Left ventricular trabecular complexity cardiac MRI adds independent risks for heart failure with preserved ejection fraction in participants with end-stage renal disease

PurposeTo measure left ventricular (LV) trabecular complexity by fractal dimension (FD) in patients with end-stage renal disease (ESRD), and assess whether FD was an independent risk factor for heart failure with preserved ejection fraction (HFpEF), or a significant predictor for adverse outcome in this population. MethodsThe study retrospectively enrolled 104 participants with ESRD who underwent 3.0 T cardiac magnetic resonance imaging (MRI) from June 2018 to November 2020. LV trabeculation was quantified with fractal analysis of short-axis cine slices to estimate the FD. Logistic regression analyses were used to evaluate FD and cardiac MRI parameters and to find independent risk predictors. Cox proportional hazard regression was used to investigate the association between FD and MACE. ResultsLV FD was higher in in the HFpEF group than those in the non-HFpEF group, with the greatest difference near the base of the ventricle. Age, minimum left atrial volume index, and LV mean basal FD were independent predictors for HFpEF in patients with ESRD. Combining the mean basal FD with typical predictive factors resulted in a C-index (0.902 vs 0.921), which was not significantly higher. Same improvements were found for net reclassification improvement [0.642; 95% confidence interval (CI), 0.254–1.029] and integrated discrimination index (0.026; 95% CI, 0.008–0.061). Participants with a LV global FD above the cutoff value (1.278) had higher risks of MACE in ESRD patients. ConclusionsLV trabecular complexity measured by FD was an independent risk factor for HFpEF, and a significant predictor for MACE among patients with ESRD.

Normal values for native T1 at 1.5 T in the pericardial fluid of healthy volunteers.

T1 mapping cardiovascular magnetic resonance (CMR) imaging has been used to characterize pericardial effusions. The aim of this study was to measure pericardial fluid native T1 values in healthy volunteers to establish normal values. Prospectively recruited volunteers (n = 30) underwent CMR at 1.5 T, and native T1 maps were acquired using a modified look-locker inversion recovery 5s(3s)3s acquisition scheme. A volume of pericardial fluid was imaged in a short-axis slice and in a slice perpendicular to the short-axis orientation. A reliable measurement had a region of interest (ROI) size > 10 mm2, coefficient of variation < 10%, and a relative difference < 5% between the two slice orientations. In 26/30 (87%) of volunteers, there was a sufficient amount of pericardial fluid to enable reliable measurement. Native T1 of pericardial fluid was 3262 ± 163 (95% normal limits 2943-3581 ms) and did not differ in the perpendicular slice orientation (3267 ± 173 ms, P = 0.75), due to sex (female 3311 ± 177 vs. male 3220 ± 142 ms, P = 0.17), age (R 2 = 0.03, P = 0.44), heart rate (R 2 = 0.005, P = 0.7), or size of the ROI (0.06, P = 0.23). This study shows that T1 values can be reliably measured in the pericardial fluid of healthy volunteers. It is the first to report normal reference ranges for T1 values at 1.5 T in the pericardial fluid of healthy volunteers.

Evaluation of convolutional neural networks for the detection of inter-breath-hold motion from a stack of cardiac short axis slice images

PurposeThis study aimed to develop and validate a deep learning-based method that detects inter-breath-hold motion from an estimated cardiac long axis image reconstructed from a stack of short axis cardiac cine images.MethodsCardiac cine magnetic resonance image data from all short axis slices and 2-/3-/4-chamber long axis slices were considered for the study. Data from 740 subjects were used for model development, and data from 491 subjects were used for testing. The method utilized the slice orientation information to calculate the intersection line of a short axis plane and a long axis plane. An estimated long axis image is shown along with a long axis image as a motion-free reference image, which enables visual assessment of the inter-breath-hold motion from the estimated long axis image. The estimated long axis image was labeled as either a motion-corrupted or a motion-free image. Deep convolutional neural network (CNN) models were developed and validated using the labeled data.ResultsThe method was fully automatic in obtaining long axis images reformatted from a 3D stack of short axis slices and predicting the presence/absence of inter-breath-hold motion. The deep CNN model with EfficientNet-B0 as a feature extractor was effective at motion detection with an area under the receiver operating characteristic (AUC) curve of 0.87 for the testing data.ConclusionThe proposed method can automatically assess inter-breath-hold motion in a stack of cardiac cine short axis slices. The method can help prospectively reacquire problematic short axis slices or retrospectively correct motion.

Optimizing Deep Learning for Cardiac MRI Segmentation: The Impact of Automated Slice Range Classification

Cardiac magnetic resonance imaging is crucial for diagnosing cardiovascular diseases, but lengthy postprocessing and manual segmentation can lead to observer bias. Deep learning (DL) has been proposed for automated cardiac segmentation; however, its effectiveness is limited by the slice range selection from base to apex. In this study, we integrated an automated slice range classification step to identify basal to apical short-axis slices before DL-based segmentation. We employed publicly available Multi-Disease, Multi-View & Multi-Center Right Ventricular Segmentation in Cardiac MRI data set with short-axis cine data from 160 training, 40 validation, and 160 testing cases. Three classification and seven segmentation DL models were studied. The top-performing segmentation model was assessed with and without the classification model. Model validation to compare automated and manual segmentation was performed using Dice score and Hausdorff distance and clinical indices (correlation score and Bland-Altman plots). The combined classification (CBAM-integrated 2D-CNN) and segmentation model (2D-UNet with dilated convolution block) demonstrated superior performance, achieving Dice scores of 0.952 for left ventricle (LV), 0.933 for right ventricle (RV), and 0.875 for myocardium, compared to the stand-alone segmentation model (0.949 for LV, 0.925 for RV, and 0.867 for myocardium). Combined classification and segmentation model showed high correlation (0.92-0.99) with manual segmentation for biventricular volumes, ejection fraction, and myocardial mass. The mean absolute difference (2.8-8.3mL) for clinical parameters between automated and manual segmentation was within the interobserver variability range, indicating comparable performance to manual annotation. Integrating an initial automated slice range classification step into the segmentation process improves the performance of DL-based cardiac chamber segmentation.

Extended-volume image-derived models of coronary microcirculation.

Recent advances in tissue clearing and high-throughput imaging have enabled the acquisition of extended-volume microvasculature images at a submicron resolution. The objective of this study was to extract information from this type of images by integrating a sequence of 3D image processing steps on Terabyte scale datasets. We acquired coronary microvasculature images throughout an entire short-axis slice of a 3-month-old Wistar-Kyoto rat heart. This dataset covered 13 × 10 × 0.6 mm at a resolution of 0.933 × 0.933 × 1.866 μm and occupied 700 Gigabytes of disk space. We used chunk-based image segmentation, combined with an efficient graph generation technique, to quantify the microvasculature in the large-scale images. Specifically, we focused on the microvasculature with a vessel diameter up to 15 μm. Morphological data for the complete short-axis ring were extracted within 16 h using this pipeline. From the analyses, we identified that microvessel lengths in the rat coronary microvasculature varied from 6 to 300 μm. However, their distribution was heavily skewed toward shorter lengths, with a mode of 16.5 μm. In contrast, vessel diameters ranged from 3 to 15 μm and had an approximately normal distribution of 6.5 ± 2 μm. The tools and techniques from this study will serve other investigations into the microcirculation, and the wealth of data from this study will enable the analysis of biophysical mechanisms using computer models.

Quantification of myocardial blood flow using stress cardiac magnetic resonance for the detection of coronary artery disease

Abstract Funding Acknowledgements Type of funding sources: Private company. Main funding source(s): GE Healthcare. Background Myocardial blood flow (MBF) analysis using stress cardiac magnetic resonance (CMR) has been shown to detect obstructive coronary artery disease (CAD); however, evaluation of its diagnostic performance has primarily been limited to single-center studies. AQUA-MBF (Assessment of QUAntitative MBF using stress CMR) is an international study with the goal of assessing the diagnostic performance of stress MBF for the detection of CAD. In this study, we aim to determine how stress MBF assessment compared against visual analysis (VA) of stress CMR images for the detection of CAD. Methods. 144 individuals (89 (62%) men, age 62±16 years, 97 (68%) hypertension, 54 (38%) diabetes, 92 (64%) hyperlipidemia) from 9 centers who underwent dual sequence stress CMR (1.5T or 3.0T GE Healthcare) and also had either a coronary computed tomography angiography (CTA, n=31), invasive coronary angiogram (ICA, n=95), or low pre-test probability for CAD (n=18) were included. CAD was defined as the presence of: (1) a stenosis ≥50% in the left main coronary artery or ≥70% in the 1 major vessel based on ICA or CTA or (2) an invasive fractional flow reserve (FFR) ≤ 0.8. Absence of obstructive coronary disease (NOCAD) was defined as a no history of myocardial infarction and stenosis &amp;lt;50% by ICA or CTA, 50–70% stenosis with FFR&amp;gt;0.8, or a young individual with no cardiac risk factors. Myocardial perfusion imaging was performed during first pass perfusion of a gadolinium-based contrast agent following the administration of adenosine or regadenoson with a low-resolution image acquired to assess the arterial input function and 2–3 short axis slices acquired to assess myocardial perfusion. VA was performed by 2 experienced cardiologists who assigned a grade of 1–5 based on the probability a study was abnormal. Stress MBF values were determined for each of the 16 myocardial segments using Fermi deconvolution (CircleCVI). The global stress MBF was calculated as the average value of two segments with the lowest values from each of the three coronary artery territories. Unpaired t-test was used to compare stress MBF values between CAD vs NOCAD. Receiver-operating characteristics curves were used to determine diagnostic performance. Results. 60 patients had CAD (20: 1-vessel, 26: 2-vessel, and 14: 3-vessel) (Figure 1A) while 84 had NOCAD (Figure 1B). The global stress MBF in CAD was lower than in NOCAD (1.63±0.52ml/g/min vs 2.41±0.68ml/g/min, p&amp;lt;0.0001) (Figure 2A). Stress MBF had a higher AUC than VA (reader 1 – 0.83 vs 0.73, p = 0.07: reader 2–0.83 vs 0.71, p = 0.02). The optimal stress MBF cut-off value for detecting CAD was 2.05ml/g/min (Figure 2B). Conclusions. In this multicenter study, we show that global stress MBF reported as a single value can identify patients with CAD at least as accurately as VA performed by physicians experienced in the interpretation of stress CMR images.

Comparison of the diagnostic accuracy of dynamic CT perfusion and cardiac CMR with quantitative perfusion: a subanalysis of ADVANTAGE II study

Abstract Funding Acknowledgements Type of funding sources: None. Background The ADVANTAGE II study has evaluated the performance of dynamic myocardial perfusion assessed by CT (CTP) in stented patient using ICA and in particular invasive coronary physiology (FFR and IMR) as gold-standard. Recently, quantitative myocardial blood flow (MBF) analysis using stress cardiac magnetic resonance (CMR) has been showed to detect obstructive coronary artery disease (CAD) and coronary microvascular dysfunction (CMD). In a substudy of ADVANTAGE II we aimed to evaluate the performance of quantitative stress CMR in comparison to CTP. Methods We present data from 39 subjects with clinically indicated ICA who underwent both vasodilator stress CMR using 1.5T MR scanner (General Electric) and dynamic stress myocardial CTP using a last generation scanner characterized by a 16-cm Z-axis coverage. In CMR Quantitative myocardial perfusion imaging was performed using the dual sequence technique (i.e. low-resolution slice for arterial input function and 2–3 short axis slices for myocardial perfusion). Resting MBF (rMBF), Stress MBF (sMBF), and myocardial perfusion reserve (MPR) were measured using Fermi deconvolution (Circle Cvi42). Quantitative perfusion results were displayed on color-encoded pixel-wise maps and on a 16-segment plot with segmental values listed. Results When FFR was used as gold standard the diagnostic accuracy of CTP was higher than those of CCTA (84.7% vs. 61.5%, p&amp;lt;0.001). When IMR was used as gold standard, diagnostic accuracy of CTP was higher than CCTA (82.9% vs. 59.3%, p&amp;lt;0.01, respectively). When the presence of at least one abnormal physiological parameter among FFR and IMR was used as gold standard, the diagnostic accuracy of CTP was higher than that of CCTA (90.8%vs 68%). Quantitative stress CMR using 2.2 ml/min/g as cut-off has a sensitivity of 83.3%, 77.7% and 100% vs QCA, FFR and IMR, respectively with higher performance of dynamic CTP in comparison to quantitative stress CMR. Conclusions In this substudy of ADVANTAGE II, we showed that global sMBF in MRI can accurately identify patients with CAD. Dynamic CTP seems to offer higher value of diagnostic performance in comparison to quantitative CMR that seems to be very promising, but more data are needed.

Cardiovascular magnetic resonance derived biventricular strain in healthy adolescents: reference values and comparison between myocardial tagging and feature-tracking

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – EU funding. Main funding source(s): Instituto de Salud Carlos III (ISCIII) European Regional Development Fun/European Social Fund (A way to make Europe/Investing in your future). Background Strain measurements have demonstrated to be earlier and more sensitive markers of contractile dysfunction than ejection fraction (EF) alone. Myocardial tagging (MT) is considered the reference standard for CMR strain measurements, but its clinical application is limited because acquisition and analysis are laborious and time consuming. Feature-tracking (FT) CMR allows strain quantification using steady-state in free-precession (SSFP) sequence, which is obtained in routine CMR protocols, being easily applicable in clinical practice. Literature regarding reference strain values in pediatric populations is scarce and, to the best of our knowledge, there are no studies comparing FT and MT in healthy adolescents. Purpose To provide CMR strain reference values in healthy adolescents and compare FT to MT results. Methods A CMR study was performed in healthy adolescents enrolled in the "Early ImaginG Markers of unhealthy lifestyles in Adolescents" (EnIGMA) project using a 3-Tesla CMR scanner between March and October 2021. The imaging protocol included a cine SSFP to provide images for chamber size and function, including FT; and MT acquisitions using spatial modulation of magnetization (SPAMM) technique. Cine acquisitions included the entire left ventricular (LV) short axis and long-axis 4 chamber (4c)- and 2 chamber (2c)-views, while MT acquisitions included three short axis slices (basal, mid and apical) and long-axis 4c-view. Endo- and epicardial borders were manually contoured in the end-diastolic (ED) phase. A non-rigid, elastic algorithm was used for segmentation and registration for strain quantification. If tracking was suboptimal, the software only allowed adaptations of the contours on the ED image. Longitudinal (LS) and circumferential (CS) strains were obtained for the LV and right ventricle (RV). Results CMR scans were performed in 123 adolescents (64 girls, 52.0%) with a mean age of 16.0 years (standard deviation [SD]=0.4), a mean LVEF of 62.5% (SD=4.1) and a mean RVEF of 56.2% (SD=4.6). Reference percentiles (P) 3, 10, 25, 50, 90 and 97 for strain values by sex are provided in Figure 1. FT-measured LVLS and LVCS were higher in girls than in boys (-19.8 vs -17.8%, p&amp;lt;0.01; and -22.2 vs -21.0%, p&amp;lt;0.01; respectively), and these differences were confirmed with MT analysis. There were no consistent between-sex differences for RVLS and RVCS. Bland-Altman analyses showed reasonable agreement between FT and MT for LVLS, LVCS, RVLS and RVCS (Figure 2 A, B, C and D). Conclusions This study provided CMR derived reference values for left and right ventricle strain in adolescents aged 15 to 18 years. This work suggests that FT is a reasonably valid technique to measure LV and RV strain in this population. This information may be useful for diagnostic and prognostic implications in adolescent patients.

Left Ventricular Vertical Run-Length Nonuniformity MRI Adds Prognostic Value to MACE in Patients with End-Stage Renal Disease.

Vertical run-length nonuniformity (VRLN) is a texture feature representing heterogeneity within native T1 images and reflects the extent of cardiac fibrosis. In uremic cardiomyopathy, interstitial fibrosis was the major histological alteration. The prognostic value of VRLN in patients with end-stage renal disease (ESRD) remains unclear. To evaluate the prognostic value of VRLN MRI in patients with ESRD. Prospective. A total of 127 ESRD patients (30 participants in the major adverse cardiac events, MACE group). 3.0 T/steady-state free precession sequence, modified Look-Locker imaging. MRI image qualities were assessed by three independent radiologists. VRLN values were measured in the myocardium on the mid-ventricular short-axis slice of T1 mapping. Left ventricular (LV) mass, LV end-diastolic and end-systolic volume, as well as LV global strain cardiac parameters were measured. The primary endpoint was the incident of MACE from enrollment time to January 2023. MACE is a composite endpoint consisting of all-cause mortality, acute myocardial infarction, stroke, heart failure hospitalization, and life-threatening arrhythmia. Cox proportional-hazards regression was performed to test whether VRLN independently correlated with MACE. The intraclass correlation coefficients of VRLN were calculated to evaluate intraobserver and interobserver reproducibility. The C-index was computed to examine the prognostic value of VRLN. P-value <0.05 were considered statistically significant. Participants were followed for a median of 26 months. VRLN, age, LV end-systolic volume index, and global longitudinal strain remained significantly associated with MACE in the multivariable model. Adding VRLN to a baseline model containing clinical and conventional cardiac MRI parameters significantly improved the accuracy of the predictive model (C-index of the baseline model: 0.781 vs. the model added VRLN: 0.814). VRLN is a novel marker for risk stratification toward MACE in patients with ESRD, superior to native T1 mapping and LV ejection fraction. 2 TECHNICAL EFFICACY STAGE: 2.

Myocardial Scar Detection Using High-Resolution Free-Breathing 3D Dark-Blood and Standard Breath-Holding 2D Bright-Blood Late Gadolinium Enhancement MRI: A Comparison of Observer Confidence.

To compare observer confidence for myocardial scar detection using 3 different late gadolinium enhancement (LGE) data sets by 2 observers with different levels of experience. Forty-one consecutive patients, who were referred for 3D dark-blood LGE MRI before implantable cardioverter-defibrillator implantation or ablation therapy and who underwent 2D bright-blood LGE MRI within a time frame of 3 months, were prospectively included. From all 3D dark-blood LGE data sets, a stack of 2D short-axis slices was reconstructed. All acquired LGE data sets were anonymized and randomized and evaluated by 2 independent observers with different levels of experience in cardiovascular imaging (beginner and expert). Confidence in detection of ischemic scar, nonischemic scar, papillary muscle scar, and right ventricular scar for each LGE data set was scored using a using a 3-point Likert scale (1 = low, 2 = medium, or 3 = high). Observer confidence scores were compared using the Friedman omnibus test and Wilcoxon signed-rank post hoc test. For the beginner observer, a significant difference in confidence regarding ischemic scar detection was observed in favor of reconstructed 2D dark-blood LGE compared with standard 2D bright-blood LGE (p = 0.030) while for the expert observer, no significant difference was found (p = 0.166). Similarly, for right ventricular scar detection, a significant difference in confidence was observed in favor of reconstructed 2D dark-blood LGE compared with standard 2D bright-blood LGE (p = 0.006) while for the expert observer, no significant difference was found (p = 0.662). Although not significantly different for other areas of interest, 3D dark-blood LGE and its derived 2D dark-blood LGE data set showed a tendency to score higher for all areas of interest at both experience levels. The combination of dark-blood LGE contrast and high isotropic voxels may contribute to increased observer confidence in myocardial scar detection, independent of observer's experience level but in particular for beginner observers.

125I-labeled 2-[4-(2-iodophenyl)piperidino]cyclopentanol (125I-OI5V) imaging visualized augmented sigma-1 receptor expression according to the severity of myocardial ischemia

BackgroundWe aimed to explore how the severity of myocardial ischemia affects myocardial sigma-1 receptor (Sig-1R) expression using 125I-labeled 2-[4-(2-iodophenyl)piperidino]cyclopentanol (125I-OI5V) imaging. Methods and resultsThe left coronary artery was occluded for 30, 20, and 10 minute, to vary the severity of myocardial ischemia, followed by reperfusion. Dual-tracer autoradiography of the left ventricular short-axis slices was performed 3 and 7 days after reperfusion. 125I-OI5V was injected 30 minute before sacrifice and the area at risk (AAR) was evaluated by 99mTc-MIBI. Intense 125I-OI5V uptake was observed in the AAR and was significantly increased with increasing ischemia duration. To evaluate salvaged and nonsalvaged areas (preserved and decreased perfusion areas), triple-tracer autoradiography was performed 3 days after reperfusion. After dual-tracer autoradiography, 201Tl was injected 20 minute post 125I-OI5V injection. On triple-tracer autoradiography, the AAR/normally perfused area 125I-OI5V uptake ratio was positively correlated with the nonsalvaged area/whole left ventricular (LV) area ratio (P < .05). The AAR/normally perfused area 125I-OI5V uptake ratio was negatively correlated with the 201Tl uptake ratio of the AAR to normally perfused areas (P < .05). The comparison of the immunostaining distribution of 125I-OI5V and the macrophage marker CD68 revealed that 125I-OI5V was present mainly in, and immediately adjacent to the macrophage infiltration area. ConclusionsSignificant 125I-OI5V uptake in the AAR depends on the duration of ischemia and reduced 201Tl uptake; furthermore, 125I-OI5V was found in and around the macrophage infiltrate area. These results indicate that iodine-labeled OI5V is a promising tool for visualizing Sig-1R expression according to the ischemic burden.

Efficient labelling for efficient deep learning: the benefit of a multiple-image-ranking method to generate high volume training data applied to ventricular slice level classification in cardiac MRI.

Getting the most value from expert clinicians' limited labelling time is a major challenge for artificial intelligence (AI) development in clinical imaging. We present a novel method for ground-truth labelling of cardiac magnetic resonance imaging (CMR) image data by leveraging multiple clinician experts ranking multiple images on a single ordinal axis, rather than manual labelling of one image at a time. We apply this strategy to train a deep learning (DL) model to classify the anatomical position of CMR images. This allows the automated removal of slices that do not contain the left ventricular (LV) myocardium. Anonymised LV short-axis slices from 300 random scans (3,552 individual images) were extracted. Each image's anatomical position relative to the LV was labelled using two different strategies performed for 5 hours each: (I) 'one-image-at-a-time': each image labelled according to its position: 'too basal', 'LV', or 'too apical' individually by one of three experts; and (II) 'multiple-image-ranking': three independent experts ordered slices according to their relative position from 'most-basal' to 'most apical' in batches of eight until each image had been viewed at least 3 times. Two convolutional neural networks were trained for a three-way classification task (each model using data from one labelling strategy). The models' performance was evaluated by accuracy, F1-score, and area under the receiver operating characteristics curve (ROC AUC). After excluding images with artefact, 3,323 images were labelled by both strategies. The model trained using labels from the 'multiple-image-ranking strategy' performed better than the model using the 'one-image-at-a-time' labelling strategy (accuracy 86% vs. 72%, P=0.02; F1-score 0.86 vs. 0.75; ROC AUC 0.95 vs. 0.86). For expert clinicians performing this task manually the intra-observer variability was low (Cohen's κ=0.90), but the inter-observer variability was higher (Cohen's κ=0.77). We present proof of concept that, given the same clinician labelling effort, comparing multiple images side-by-side using a 'multiple-image-ranking' strategy achieves ground truth labels for DL more accurately than by classifying images individually. We demonstrate a potential clinical application: the automatic removal of unrequired CMR images. This leads to increased efficiency by focussing human and machine attention on images which are needed to answer clinical questions.

Deep-active-learning approach towards accurate right ventricular segmentation using a two-level uncertainty estimation.

The Right Ventricle (RV) is currently recognised to be a significant and important prognostic factor for various pathologies. Its assessment is made possible using Magnetic Resonance Imaging (CMRI) short-axis slices. Yet, due to the challenging issues of this cavity, radiologists still perform its delineation manually, which is tedious, laborious, and time-consuming. Therefore, to automatically tackle the RV segmentation issues, Deep-Learning (DL) techniques seem to be the axis of the most recent promising approaches. Along with its potential at dealing with shape variations, DL techniques highly requires a large number of labelled images to avoid over-fitting. Subsequently, with the produced large amounts of data in the medical industry, preparing annotated datasets manually is still time-consuming, and requires high skills to be accomplished. To benefit from a significant number of labelled and unlabelled CMRI images, a Deep-Active-Learning (DAL) approach is proposed in this paper to segment the RV. Thus, three main steps are distinguished. First, a personalised labelled dataset is gathered and augmented to allow initial learning. Secondly, a U-Net based architecture is modified towards efficient initial accuracy. Finally, a two-level uncertainty estimation technique is settled to enable the selection of complementary unlabelled data. The proposed pipeline is enhanced with a customised postprocessing step, in which epistemic uncertainty and Dense Conditional Random Fields are used. The proposed approach is tested on 791 images gathered from 32 public patients and 1230 images of 50 custom subjects. The obtained results show an increased dice coefficient from 0.86 to 0.91 with a decreased Hausdorff distance from 7.55 to 7.45.

Automated identification of myocardial perfusion defects in dynamic cardiac computed tomography using deep learning

PurposeThe purpose of this study was to develop and evaluate deep convolutional neural network (CNN) models for quantifying myocardial blood flow (MBF) as well as for identifying myocardial perfusion defects in dynamic cardiac computed tomography (CT) images. MethodsAdenosine stress cardiac CT perfusion data acquired from 156 patients having or being suspected with coronary artery disease were considered for model development and validation. U-net-based deep CNN models were developed to segment the aorta and myocardium and to localize anatomical landmarks. Color-coded MBF maps were obtained in short-axis slices from the apex to the base level and were used to train a deep CNN classifier. Three binary classification models were built for the detection of perfusion defect in the left anterior descending artery (LAD), the right coronary artery (RCA), and the left circumflex artery (LCX) territories. ResultsMean Dice scores were 0.94 (±0.07) and 0.86 (±0.06) for the aorta and myocardial deep learning-based segmentations, respectively. With the localization U-net, mean distance errors were 3.5 (±3.5) mm and 3.8 (±2.4) mm for the basal and apical center points, respectively. The classification models identified perfusion defects with the accuracy of mean area under the receiver operating curve (AUROC) values of 0.959 (±0.023) for LAD, 0.949 (±0.016) for RCA, and 0.957 (±0.021) for LCX. ConclusionThe presented method has the potential to fully automate the quantification of MBF and subsequently identify the main coronary artery territories with myocardial perfusion defects in dynamic cardiac CT perfusion.

Accuracy of Peguero-Lo Presti ECG Criteria for Left Ventricular Hypertrophy Diagnosis in Elderly Thai Patients Using Cardiovascular Magnetic Resonance as a Gold Standard

Background: The conventional electrocardiogram (ECG) criteria have shown low sensitivity for LVH detection in the elderly. Objective: To evaluate the diagnosis accuracy of the new ECG criteria proposed by Peguero-Lo Presti in this population using cardiovascular magnetic resonance (CMR) as a gold standard. Materials and Methods: Patients older than 60 years that underwent CMR (1.5 T and 3 T, Philips) between 2012 and 2019 were included. Exclusion criteria were abnormal CMR findings and abnormal baseline ECG. Left ventricular mass index (LVMI) was calculated from short-axis slices covering the entire left ventricle. LVH was defined as LVMI of 63 g/m² or more in male patients and 48 g/m² or more in female patients. LVH by new ECG criteria was defined by the deepest S + SV4 of 2.8 or more in males and 2.3 or more in females. Sensitivity, specificity, and accuracy between the new criteria and the Cornell voltage criteria were compared. Results: Five hundred fifty-nine patients were included with a mean age of 71±7 years old, and with 63.69% of female, 71.56% had hypertension, and 16.46% were octogenarian. Eighty patients (14.3%) had LVH detected by CMR. The new ECG criteria showed higher sensitivity compared with the Cornell voltage criteria especially in the octogenarian with a sensitivity of 15% versus 10% in overall and 31.25% versus 18.75% in octogenarian, but the AUC did not show significant difference at AUC of 0.56 versus 0.54 (p=0.67). Conclusion: Using CMR as the gold standard, the new ECG criteria showed higher sensitivity than the Cornell voltage criteria, especially in the octogenarian, without significant different accuracy. Keywords: Electrocardiography; Elderly population; Left ventricular hypertrophy; Cardiac magnetic resonance; Peguero-Lo Presti Criteria

Regional extracellular volume within late gadolinium enhancement-positive myocardium to differentiate cardiac sarcoidosis from myocarditis of other etiology: a cardiovascular magnetic resonance study

BackgroundCardiovascular magnetic resonance (CMR) plays a pivotal role in diagnosing myocardial inflammation. In addition to late gadolinium enhancement (LGE), native T1 and T2 mapping as well as extracellular volume (ECV) are essential tools for tissue characterization. However, the differentiation of cardiac sarcoidosis (CS) from myocarditis of other etiology can be challenging. Positron-emission tomography-computed tomography (PET-CT) regularly shows the highest Fluordesoxyglucose (FDG) uptake in LGE positive regions. It was therefore the aim of this study to investigate, whether native T1, T2, and ECV measurements within LGE regions can improve the differentiation of CS and myocarditis compared with using global native T1, T2, and ECV values alone. MethodsPET/CT confirmed CS patients and myocarditis patients (both acute and chronic) from a prospective registry were compared with respect to regional native T1, T2, and ECV. Acute and chronic myocarditis were defined based on the 2013 European Society of Cardiology position paper on myocarditis. All parametric measures and ECV were acquired in standard fashion on three short-axis slices according to the ConSept study for global values and within PET-CT positive regions of LGE. ResultsBetween 2017 and 2020, 33 patients with CS and 73 chronic and 35 acute myocarditis patients were identified. The mean ECV (± SD) in LGE regions of CS patients was higher than in myocarditis patients (CS vs. acute and chronic, respectively: 0.65 ± 0.12 vs. 0.45 ± 0.13 and 0.47 ± 0.1; p < 0.001). Acute and chronic myocarditis patients had higher global native T1 values (1157 ± 54 ms vs. 1196 ± 63 ms vs. 1215 ± 74 ms; p = 0.001). There was no difference in global T2 and ECV values between CS and acute or chronic myocarditis patients. ConclusionThis is the first study to show that the calculation of regional ECV within LGE-positive regions may help to differentiate CS from myocarditis. Further studies are warranted to corroborate these findings.

Res-DUnet: A small-region attentioned model for cardiac MRI-based right ventricular segmentation

Right ventricular function has been associated with a variety of cardiovascular diseases. In the clinical study of right ventricular function, an important step is segmentation of the right ventricle so that functional indicators of the heart can be quantified and evaluated based on the segmented region. Compared to left ventricle, right ventricle (RV) is more difficult to segment due to its irregular shape variations and blurred borders. To improve segmentation accuracy on slices with very small regions of RV, a two-stage deep learning model called Res-DUnet is developed to segment the right ventricle on short-axis slices of cardiac magnetic resonance imaging. The model is divided into two modules, the localization module for extracting the regions of interest (ROIs) and the segmentation module for segmenting the right ventricle. The two-stage model achieves a mean dice score (DSC) of 90.55, a mean Hausdorff distance (HD) of 6.67 and a mean recall of 92.68, outperforming the state-of-the-art models. The clinical indicators derived from the model are analyzed for consistency with the ground truth. The results showed that the model’s performance is comparable to that of the radiologists.