Quality assurance of late gadolinium enhancement cardiac magnetic resonance images: a deep learning classifier for confidence in the presence or absence of abnormality with potential to prompt real-time image optimization.

BACKGROUND: Late gadolinium enhancement (LGE) of the myocardium has significant diagnostic and prognostic implications, with even small areas of enhancement being important. Distinguishing between definitely normal and definitely abnormal LGE images is usually straightforward, but diagnostic uncertainty arises when reporters are not sure whether the observed LGE is genuine or not. This uncertainty might be resolved by repetition (to remove artifact) or further acquisition of intersecting images, but this must take place before the scan finishes. Real-time quality assurance by humans is a complex task requiring training and experience, so being able to identify which images have an intermediate likelihood of LGE while the scan is ongoing, without the presence of an expert is of high value. This decision-support could prompt immediate image optimization or acquisition of supplementary images to confirm or refute the presence of genuine LGE. This could reduce ambiguity in reports. METHODS: Short-axis, phase-sensitive inversion recovery late gadolinium images were extracted from our clinical cardiac magnetic resonance (CMR) database and shuffled. Two, independent, blinded experts scored each individual slice for "LGE likelihood" on a visual analog scale, from 0 (absolute certainty of no LGE) to 100 (absolute certainty of LGE), with 50 representing clinical equipoise. The scored images were split into two classes-either "high certainty" of whether LGE was present or not, or "low certainty." The dataset was split into training, validation, and test sets (70:15:15). A deep learning binary classifier based on the EfficientNetV2 convolutional neural network architecture was trained to distinguish between these categories. Classifier performance on the test set was evaluated by calculating the accuracy, precision, recall, F1-score, and area under the receiver operating characteristics curve (ROC AUC). Performance was also evaluated on an external test set of images from a different center. RESULTS: One thousand six hundred and forty-five images (from 272 patients) were labeled and split at the patient level into training (1151 images), validation (247 images), and test (247 images) sets for the deep learning binary classifier. Of these, 1208 images were "high certainty" (255 for LGE, 953 for no LGE), and 437 were "low certainty". An external test comprising 247 images from 41 patients from another center was also employed. After 100 epochs, the performance on the internal test set was accuracy = 0.94, recall = 0.80, precision = 0.97, F1-score = 0.87, and ROC AUC = 0.94. The classifier also performed robustly on the external test set (accuracy = 0.91, recall = 0.73, precision = 0.93, F1-score = 0.82, and ROC AUC = 0.91). These results were benchmarked against a reference inter-expert accuracy of 0.86. CONCLUSION: Deep learning shows potential to automate quality control of late gadolinium imaging in CMR. The ability to identify short-axis images with intermediate LGE likelihood in real-time may serve as a useful decision-support tool. This approach has the potential to guide immediate further imaging while the patient is still in the scanner, thereby reducing the frequency of recalls and inconclusive reports due to diagnostic indecision.

Sex- and age-specific normal values for automated quantitative pixel-wise myocardial perfusion cardiovascular magnetic resonance.

AIMS: Recently developed in-line automated cardiovascular magnetic resonance (CMR) myocardial perfusion mapping has been shown to be reproducible and comparable with positron emission tomography (PET), and can be easily integrated into clinical workflows. Bringing quantitative myocardial perfusion CMR into routine clinical care requires knowledge of sex- and age-specific normal values in order to define thresholds for disease detection. This study aimed to establish sex- and age-specific normal values for stress and rest CMR myocardial blood flow (MBF) in healthy volunteers. METHODS AND RESULTS: A total of 151 healthy volunteers recruited from two centres underwent adenosine stress and rest myocardial perfusion CMR. In-line automatic reconstruction and post processing of perfusion data were implemented within the Gadgetron software framework, creating pixel-wise perfusion maps. Rest and stress MBF were measured, deriving myocardial perfusion reserve (MPR) and were subdivided by sex and age. Mean MBF in all subjects was 0.62 ± 0.13 mL/g/min at rest and 2.24 ± 0.53 mL/g/min during stress. Mean MPR was 3.74 ± 1.00. Compared with males, females had higher rest (0.69 ± 0.13 vs. 0.58 ± 0.12 mL/g/min, P < 0.01) and stress MBF (2.41 ± 0.47 vs. 2.13 ± 0.54 mL/g/min, P = 0.001). Stress MBF and MPR showed significant negative correlations with increasing age (r = -0.43, P < 0.001 and r = -0.34, P < 0.001, respectively). CONCLUSION: Fully automated in-line CMR myocardial perfusion mapping produces similar normal values to the published CMR and PET literature. There is a significant increase in rest and stress MBF, but not MPR, in females and a reduction of stress MBF and MPR with advancing age, advocating the use of sex- and age-specific reference ranges for diagnostic use.

Cardiovascular magnetic resonance phenotyping of heart failure with mildly reduced ejection fraction.

AIMS: The 2016 European Society of Cardiology Heart Failure Guidelines defined a new category: heart failure with mid-range ejection fraction (HFmrEF) of 40-49%. This new category was highlighted as having limited evidence and research was advocated into underlying characteristics, pathophysiology, and diagnosis. We used multi-parametric cardiovascular magnetic resonance (CMR) to define the cardiac phenotype of presumed non-ischaemic HFmrEF. METHODS AND RESULTS: Patients (N = 300, 62.7 ± 13 years, 63% males) with a clinical diagnosis of heart failure with no angina symptoms, history of myocardial infarction, or coronary intervention were prospectively recruited. Patients underwent clinical assessment and CMR including T1 mapping, extracellular volume (ECV) mapping, late gadolinium enhancement, and measurement of myocardial blood flow at rest and maximal hyperaemia. Of 273 patients in the final analysis, 93 (34%) patients were categorized as HFmrEF, 46 (17%) as heart failure with preserved ejection fraction (HFpEF), and 134 (49%) as heart failure with reduced ejection fraction (HFrEF). Nineteen (20%) patients with HFmrEF had evidence of occult ischaemic heart disease. Diffuse fibrosis and hyperaemic myocardial blood flow were similar in HFmrEF and HFpEF, but HFmrEF showed significantly lower native T1 (1311 ± 32 vs. 1340 ± 45 ms, P < 0.001), ECV (24.6 ± 3.2 vs. 26.3 ± 3.1%, P < 0.001), and higher myocardial perfusion reserve (2.75 ± 0.84 vs. 2.28 ± 0.84, P < 0.001) compared with HFrEF. CONCLUSION: Patients with HFmrEF share most phenotypic characteristics with HFpEF, including the degree of microvascular impairment and fibrosis, but have a high prevalence of occult ischaemic heart disease similar to HFrEF. Further work is needed to confirm how the phenotype of HFmrEF responds to medical therapy.

Saturation-pulse prepared heart-rate independent inversion-recovery (SAPPHIRE) biventricular T1 mapping: inter-field strength, head-to-head comparison of diastolic, systolic and dark-blood measurements.

BACKGROUND: To assess the feasibility of biventricular SAPPHIRE T1 mapping in vivo across field strengths using diastolic, systolic and dark-blood (DB) approaches. METHODS: 10 healthy volunteers underwent same-day non-contrast cardiovascular magnetic resonance at 1.5 Tesla (T) and 3 T. Left and right ventricular (LV, RV) T1 mapping was performed in the basal, mid and apical short axis using 4-variants of SAPPHIRE: diastolic, systolic, 0th and 2nd order motion-sensitized DB and conventional modified Look-Locker inversion recovery (MOLLI). RESULTS: LV global myocardial T1 times (1.5 T then 3 T results) were significantly longer by diastolic SAPPHIRE (1283 ± 11|1600 ± 17 ms) than any of the other SAPPHIRE variants: systolic (1239 ± 9|1595 ± 13 ms), 0th order DB (1241 ± 10|1596 ± 12) and 2nd order DB (1251 ± 11|1560 ± 20 ms, all p < 0.05). In the mid septum MOLLI and diastolic SAPPHIRE exhibited significant T1 signal contamination (longer T1) at the blood-myocardial interface not seen with the other 3 SAPPHIRE variants (all p < 0.025). Additionally, systolic, 0th order and 2nd order DB SAPPHIRE showed narrower dispersion of myocardial T1 times across the mid septum when compared to diastolic SAPPHIRE (interquartile ranges respectively: 25 ms, 71 ms, 73 ms vs 143 ms, all p < 0.05). RV T1 mapping was achievable using systolic, 0th and 2nd order DB SAPPHIRE but not with MOLLI or diastolic SAPPHIRE. All 4 SAPPHIRE variants showed excellent re-read reproducibility (intraclass correlation coefficients 0.953 to 0.996). CONCLUSION: These small-scale preliminary healthy volunteer data suggest that DB SAPPHIRE has the potential to reduce partial volume effects at the blood-myocardial interface, and that systolic SAPPHIRE could be a feasible solution for right ventricular T1 mapping. Further work is needed to understand the robustness of these sequences and their potential clinical utility.

Quantitative Myocardial Perfusion Predicts Outcomes in Patients With Prior Surgical Revascularization.

BACKGROUND: Patients with previous coronary artery bypass graft (CABG) surgery typically have complex coronary disease and remain at high risk of adverse events. Quantitative myocardial perfusion indices predict outcomes in native vessel disease, but their prognostic performance in patients with prior CABG is unknown. OBJECTIVES: In this study, we sought to evaluate whether global stress myocardial blood flow (MBF) and perfusion reserve (MPR) derived from perfusion mapping cardiac magnetic resonance (CMR) independently predict adverse outcomes in patients with prior CABG. METHODS: This was a retrospective analysis of consecutive patients with prior CABG referred for adenosine stress perfusion CMR. Perfusion mapping was performed in-line with automated quantification of MBF. The primary outcome was a composite of all-cause mortality and major adverse cardiovascular events defined as nonfatal myocardial infarction and unplanned revascularization. Associations were evaluated with the use of Cox proportional hazards models after adjusting for comorbidities and CMR parameters. RESULTS: A total of 341 patients (median age 67 years, 86% male) were included. Over a median follow-up of 638 days (IQR: 367-976 days), 81 patients (24%) reached the primary outcome. Both stress MBF and MPR independently predicted outcomes after adjusting for known prognostic factors (regional ischemia, infarction). The adjusted hazard ratio (HR) for 1 mL/g/min of decrease in stress MBF was 2.56 (95% CI: 1.45-4.35) and for 1 unit of decrease in MPR was 1.61 (95% CI: 1.08-2.38). CONCLUSIONS: Global stress MBF and MPR derived from perfusion CMR independently predict adverse outcomes in patients with previous CABG. This effect is independent from the presence of regional ischemia on visual assessment and the extent of previous infarction.

Age matters: differences in exercise-induced cardiovascular remodelling in young and middle aged healthy sedentary individuals.

AIMS: Remodelling of the cardiovascular system (including heart and vasculature) is a dynamic process influenced by multiple physiological and pathological factors. We sought to understand whether remodelling in response to a stimulus, exercise training, altered with healthy ageing. METHODS: A total of 237 untrained healthy male and female subjects volunteering for their first time marathon were recruited. At baseline and after 6 months of unsupervised training, race completers underwent tests including 1.5T cardiac magnetic resonance, brachial and non-invasive central blood pressure assessment. For analysis, runners were divided by age into under or over 35 years (U35, O35). RESULTS: Injury and completion rates were similar among the groups; 138 runners (U35: n = 71, women 49%; O35: n = 67, women 51%) completed the race. On average, U35 were faster by 37 minutes (12%). Training induced a small increase in left ventricular mass in both groups (3 g/m2, P < 0.001), but U35 also increased ventricular cavity sizes (left ventricular end-diastolic volume (EDV)i +3%; left ventricular end-systolic volume (ESV)i +8%; right ventricular end-diastolic volume (EDV)i +4%; right ventricular end-systolic volume (ESV)i +5%; P < 0.01 for all). Systemic aortic compliance fell in the whole sample by 7% (P = 0.020) and, especially in O35, also systemic vascular resistance (-4% in the whole sample, P = 0.04) and blood pressure (systolic/diastolic, whole sample: brachial -4/-3 mmHg, central -4/-2 mmHg, all P < 0.001; O35: brachial -6/-3 mmHg, central -6/-4 mmHg, all P < 0.001). CONCLUSION: Medium-term, unsupervised physical training in healthy sedentary individuals induces measurable remodelling of both heart and vasculature. This amount is age dependent, with predominant cardiac remodelling when younger and predominantly vascular remodelling when older.

Use of quantitative cardiovascular magnetic resonance myocardial perfusion mapping for characterization of ischemia in patients with left internal mammary coronary artery bypass grafts.

BACKGROUND: Quantitative myocardial perfusion mapping using cardiovascular magnetic resonance (CMR) is validated for myocardial blood flow (MBF) estimation in native vessel coronary artery disease (CAD). Following coronary artery bypass graft (CABG) surgery, perfusion defects are often detected in territories supplied by the left internal mammary artery (LIMA) graft, but their interpretation and subsequent clinical management is variable. METHODS: We assessed myocardial perfusion using quantitative CMR perfusion mapping in 38 patients with prior CABG surgery, all with angiographically-proven patent LIMA grafts to the left anterior descending coronary artery (LAD) and no prior infarction in the LAD territory. Factors potentially determining MBF in the LIMA-LAD myocardial territory, including the impact of delayed contrast arrival through the LIMA graft were evaluated. RESULTS: Perfusion defects were reported on blinded visual analysis in the LIMA-LAD territory in 27 (71%) cases, despite LIMA graft patency and no LAD infarction. Native LAD chronic total occlusion (CTO) was a strong independent predictor of stress MBF (B = - 0.41, p = 0.014) and myocardial perfusion reserve (MPR) (B = - 0.56, p = 0.005), and was associated with reduced stress MBF in the basal (1.47 vs 2.07 ml/g/min; p = 0.002) but not the apical myocardial segments (1.52 vs 1.87 ml/g/min; p = 0.057). Extending the maximum arterial time delay incorporated in the quantitative perfusion algorithm, resulted only in a small increase (3.4%) of estimated stress MBF. CONCLUSIONS: Perfusion defects are frequently detected in LIMA-LAD subtended territories post CABG despite LIMA patency. Although delayed contrast arrival through LIMA grafts causes a small underestimation of MBF, perfusion defects are likely to reflect true reductions in myocardial blood flow, largely due to proximal native LAD disease.

Prognostic Value of Pulmonary Transit Time and Pulmonary Blood Volume Estimation Using Myocardial Perfusion CMR.

OBJECTIVES: The purpose of this study was to explore the prognostic significance of PTT and PBVi using an automated, inline method of estimation using CMR. BACKGROUND: Pulmonary transit time (PTT) and pulmonary blood volume index (PBVi) (the product of PTT and cardiac index), are quantitative biomarkers of cardiopulmonary status. The development of cardiovascular magnetic resonance (CMR) quantitative perfusion mapping permits their automated derivation, facilitating clinical adoption. METHODS: In this retrospective 2-center study of patients referred for clinical myocardial perfusion assessment using CMR, analysis of right and left ventricular cavity arterial input function curves from first pass perfusion was performed automatically (incorporating artificial intelligence techniques), allowing estimation of PTT and subsequent derivation of PBVi. Association with major adverse cardiovascular events (MACE) and all-cause mortality were evaluated using Cox proportional hazard models, after adjusting for comorbidities and CMR parameters. RESULTS: A total of 985 patients (67% men, median age 62 years [interquartile range (IQR): 52 to 71 years]) were included, with median left ventricular ejection fraction (LVEF) of 62% (IQR: 54% to 69%). PTT increased with age, male sex, atrial fibrillation, and left atrial area, and reduced with LVEF, heart rate, diabetes, and hypertension (model r2 = 0.57). Over a median follow-up period of 28.6 months (IQR: 22.6 to 35.7 months), MACE occurred in 61 (6.2%) patients. After adjusting for prognostic factors, both PTT and PBVi independently predicted MACE, but not all-cause mortality. There was no association between cardiac index and MACE. For every 1 × SD (2.39-s) increase in PTT, the adjusted hazard ratio for MACE was 1.43 (95% confidence interval [CI]: 1.10 to 1.85; p = 0.007). The adjusted hazard ratio for 1 × SD (118 ml/m2) increase in PBVi was 1.42 (95% CI: 1.13 to 1.78; p = 0.002). CONCLUSIONS: Pulmonary transit time (and its derived parameter pulmonary blood volume index), measured automatically without user interaction as part of CMR perfusion mapping, independently predicted adverse cardiovascular outcomes. These biomarkers may offer additional insights into cardiopulmonary function beyond conventional predictors including ejection fraction.

A comparison of standard and high dose adenosine protocols in routine vasodilator stress cardiovascular magnetic resonance: dosage affects hyperaemic myocardial blood flow in patients with severe left ventricular systolic impairment.

BACKGROUND: Adenosine stress perfusion cardiovascular magnetic resonance (CMR) is commonly used in the assessment of patients with suspected ischaemia. Accepted protocols recommend administration of adenosine at a dose of 140 µg/kg/min increased up to 210 µg/kg/min if required. Conventionally, adequate stress has been assessed using change in heart rate, however, recent studies have suggested that these peripheral measurements may not reflect hyperaemia and can be blunted, in particular, in patients with heart failure. This study looked to compare stress myocardial blood flow (MBF) and haemodynamic response with different dosing regimens of adenosine during stress perfusion CMR in patients and healthy controls. METHODS: 20 healthy adult subjects were recruited as controls to compare 3 adenosine perfusion protocols: standard dose (140 µg/kg/min for 4 min), high dose (210 µg/kg/min for 4 min) and long dose (140 µg/kg/min for 8 min). 60 patients with either known or suspected coronary artery disease (CAD) or with heart failure and different degrees of left ventricular (LV) dysfunction underwent adenosine stress with standard and high dose adenosine within the same scan. All studies were carried out on a 3 T CMR scanner. Quantitative global myocardial perfusion and haemodynamic response were compared between doses. RESULTS: In healthy controls, no significant difference was seen in stress MBF between the 3 protocols. In patients with known or suspected CAD, and those with heart failure and mild systolic impairment (LV ejection fraction (LVEF) ≥ 40%) no significant difference was seen in stress MBF between standard and high dose adenosine. In those with LVEF < 40%, there was a significantly higher stress MBF following high dose adenosine compared to standard dose (1.33 ± 0.46 vs 1.10 ± 0.47 ml/g/min, p = 0.004). Non-responders to standard dose adenosine (defined by an increase in heart rate (HR) < 10 bpm) had a significantly higher stress HR following high dose (75 ± 12 vs 70 ± 14 bpm, p = 0.034), but showed no significant difference in stress MBF. CONCLUSIONS: Increasing adenosine dose from 140 to 210 µg/kg/min leads to increased stress MBF in patients with significantly impaired LV systolic function. Adenosine dose in clinical perfusion assessment may need to be increased in these patients.

Diagnosis and risk stratification in hypertrophic cardiomyopathy using machine learning wall thickness measurement: a comparison with human test-retest performance.

BACKGROUND: Left ventricular maximum wall thickness (MWT) is central to diagnosis and risk stratification of hypertrophic cardiomyopathy, but human measurement is prone to variability. We developed an automated machine learning algorithm for MWT measurement and compared precision (reproducibility) with that of 11 international experts, using a dataset of patients with hypertrophic cardiomyopathy. METHODS: 60 adult patients with hypertrophic cardiomyopathy, including those carrying hypertrophic cardiomyopathy gene mutations, were recruited at three institutes in the UK from August, 2018, to September, 2019: Barts Heart Centre, University College London Hospital (The Heart Hospital), and Leeds Teaching Hospitals NHS Trust. Participants had two cardiovascular magnetic resonance scans (test and retest) on the same day, ensuring no biological variability, using four cardiac MRI scanner models represented across two manufacturers and two field strengths. End-diastolic short-axis MWT was measured in test and retest by 11 international experts (from nine centres in six countries) and an automated machine learning method, which was trained to segment endocardial and epicardial contours on an independent, multicentre, multidisease dataset of 1923 patients. Machine learning MWT measurement was done with a method based on solving Laplace's equation. To assess test-retest reproducibility, we estimated the absolute test-retest MWT difference (precision), the coefficient of variation (CoV) for duplicate measurements, and the number of patients reclassified between test and retest according to different thresholds (MWT >15 mm and >30 mm). We calculated the sample size required to detect a prespecified MWT change between pairs of scans for machine learning and each expert. FINDINGS: 1440 MWT measurements were analysed, corresponding to two scans from 60 participants by 12 observers (11 experts and machine learning). Experts differed in the MWT they measured, ranging from 14·9 mm (SD 4·2) to 19·0 mm (4·7; p<0·0001 for trend). Machine learning-measured mean MWT was 16·8 mm (4·1). Machine learning precision was superior, with a test-retest difference of 0·7 mm (0·6) compared with experts, who ranged from 1·1 mm (0·9) to 3·7 mm (2·0; p values for machine learning vs expert comparison ranging from <0·0001 to 0·0073) and a significantly lower CoV than for all experts (4·3% [95% CI 3·3-5·1] vs 5·7-12·1% across experts). On average, 38 (64%) patients were designated as having MWT greater than 15 mm by machine learning compared with 27 (45%) to 50 (83%) patients by experts; five (8%) patients were reclassified in test-retest by machine learning compared with four (7%) to 12 (20%) by experts. With a cutoff point of more than 30 mm for implantable cardioverter-defibrillator, three experts would have changed recommendations between tests a total of four times, but machine learning was consistent. Using machine learning, a clinical trial to detect a 2 mm MWT change would need 2·3 times (range 1·6-4·6) fewer patients. INTERPRETATION: In this preliminary study, machine learning MWT measurement in hypertrophic cardiomyopathy is superior to human experts with potential implications for diagnosis, risk stratification, and clinical trials. FUNDING: European Regional Development Fund and Barts Charity.

Measurement of T1 Mapping in Patients With Cardiac Devices: Off-Resonance Error Extends Beyond Visual Artifact but Can Be Quantified and Corrected.

Background: Measurement of myocardial T1 is increasingly incorporated into standard cardiovascular magnetic resonance (CMR) protocols, however accuracy may be reduced in patients with metallic cardiovascular implants. Measurement is feasible in segments free from visual artifact, but there may still be off-resonance induced error. Aim: To quantify off-resonance induced T1 error in patients with metallic cardiovascular implants, and validate a method for error correction for a conventional MOLLI pulse sequence. Methods: Twenty-four patients with cardiac implantable electronic devices (CIEDs: 46% permanent pacemakers, PPMs; 33% implantable loop recorders, ILRs; and 21% implantable cardioverter-defibrillators, ICDs); and 31 patients with aortic valve replacement (AVR) (45% metallic) were studied. Paired mid-myocardial short-axis MOLLI and single breath-hold off-resonance field maps were acquired at 1.5 T. T1 values were measured by AHA segment, and segments with visual artifact were excluded. T1 correction was applied using a published relationship between off-resonance and T1. The accuracy of the correction was assessed in 10 healthy volunteers by measuring T1 before and after external placement of an ICD generator next to the chest to generate off-resonance. Results: T1 values in healthy volunteers with an ICD were underestimated compared to without (967 ± 52 vs. 997 ± 26 ms respectively, p = 0.0001), but were similar after correction (p = 0.57, residual difference 2 ± 27 ms). Artifact was visible in 4 ± 12, 42 ± 31, and 53 ± 27% of AHA segments in patients with ILRs, PPMs, and ICDs, respectively. In segments without artifact, T1 was underestimated by 63 ms (interquartile range: 7-143) per patient. The greatest error for patients with ILRs, PPMs and ICDs were 79, 146, and 191 ms, respectively. The presence of an AVR did not generate T1 error. Conclusion: Even when there is no visual artifact, there is error in T1 in patients with CIEDs, but not AVRs. Off-resonance field map acquisition can detect error in measured T1, and a correction can be applied to quantify T1 MOLLI accurately.

The myocardial phenotype of Fabry disease pre-hypertrophy and pre-detectable storage.

AIMS: Cardiac involvement in Fabry disease (FD) occurs prior to left ventricular hypertrophy (LVH) and is characterized by low myocardial native T1 with sphingolipid storage reflected by cardiovascular magnetic resonance (CMR) and electrocardiogram (ECG) changes. We hypothesize that a pre-storage myocardial phenotype might occur even earlier, prior to T1 lowering. METHODS AND RESULTS: FD patients and age-, sex-, and heart rate-matched healthy controls underwent same-day ECG with advanced analysis and multiparametric CMR [cines, global longitudinal strain (GLS), T1 and T2 mapping, stress perfusion (myocardial blood flow, MBF), and late gadolinium enhancement (LGE)]. One hundred and fourteen Fabry patients (46 ± 13 years, 61% female) and 76 controls (49 ± 15 years, 50% female) were included. In pre-LVH FD (n = 72, 63%), a low T1 (n = 32/72, 44%) was associated with a constellation of ECG and functional abnormalities compared to normal T1 FD patients and controls. However, pre-LVH FD with normal T1 (n = 40/72, 56%) also had abnormalities compared to controls: reduced GLS (-18 ± 2 vs. -20 ± 2%, P < 0.001), microvascular changes (lower MBF 2.5 ± 0.7 vs. 3.0 ± 0.8 mL/g/min, P = 0.028), subtle T2 elevation (50 ± 4 vs. 48 ± 2 ms, P = 0.027), and limited LGE (%LGE 0.3 ± 1.1 vs. 0%, P = 0.004). ECG abnormalities included shorter P-wave duration (88 ± 12 vs. 94 ± 15 ms, P = 0.010) and T-wave peak time (Tonset - Tpeak; 104 ± 28 vs. 115 ± 20 ms, P = 0.015), resulting in a more symmetric T wave with lower T-wave time ratio (Tonset - Tpeak)/(Tpeak - Tend) (1.5 ± 0.4 vs. 1.8 ± 0.4, P < 0.001) compared to controls. CONCLUSION: FD has a measurable myocardial phenotype pre-LVH and pre-detectable myocyte storage with microvascular dysfunction, subtly impaired GLS and altered atrial depolarization and ventricular repolarization intervals.

Non-invasive Ischaemia Testing in Patients With Prior Coronary Artery Bypass Graft Surgery: Technical Challenges, Limitations, and Future Directions.

Coronary artery bypass graft (CABG) surgery effectively relieves symptoms and improves outcomes. However, patients undergoing CABG surgery typically have advanced coronary atherosclerotic disease and remain at high risk for symptom recurrence and adverse events. Functional non-invasive testing for ischaemia is commonly used as a gatekeeper for invasive coronary and graft angiography, and for guiding subsequent revascularisation decisions. However, performing and interpreting non-invasive ischaemia testing in patients post CABG is challenging, irrespective of the imaging modality used. Multiple factors including advanced multi-vessel native vessel disease, variability in coronary hemodynamics post-surgery, differences in graft lengths and vasomotor properties, and complex myocardial scar morphology are only some of the pathophysiological mechanisms that complicate ischaemia evaluation in this patient population. Systematic assessment of the impact of these challenges in relation to each imaging modality may help optimize diagnostic test selection by incorporating clinical information and individual patient characteristics. At the same time, recent technological advances in cardiac imaging including improvements in image quality, wider availability of quantitative techniques for measuring myocardial blood flow and the introduction of artificial intelligence-based approaches for image analysis offer the opportunity to re-evaluate the value of ischaemia testing, providing new insights into the pathophysiological processes that determine outcomes in this patient population.

Quantitative cardiovascular magnetic resonance myocardial perfusion mapping to assess hyperaemic response to adenosine stress.

AIMS: Assessment of hyperaemia during adenosine stress cardiovascular magnetic resonance (CMR) remains a clinical challenge with lack of a gold-standard non-invasive clinical marker to confirm hyperaemic response. This study aimed to validate maximum stress myocardial blood flow (SMBF) measured using quantitative perfusion mapping for assessment of hyperaemic response and compare this to current clinical markers of adenosine stress. METHODS AND RESULTS: Two hundred and eighteen subjects underwent adenosine stress CMR. A derivation cohort (22 volunteers) was used to identify a SMBF threshold value for hyperaemia. This was tested in a validation cohort (37 patients with suspected coronary artery disease) who underwent invasive coronary physiology assessment on the same day as CMR. A clinical cohort (159 patients) was used to compare SMBF to other physiological markers of hyperaemia [splenic switch-off (SSO), heart rate response (HRR), and blood pressure (BP) fall]. A minimum SMBF threshold of 1.43 mL/g/min was derived from volunteer scans. All patients in the coronary physiology cohort demonstrated regional maximum SMBF (SMBFmax) >1.43 mL/g/min and invasive evidence of hyperaemia. Of the clinical cohort, 93% had hyperaemia defined by perfusion mapping compared to 71% using SSO and 81% using HRR. There was no difference in SMBFmax in those with or without SSO (2.58 ± 0.89 vs. 2.54 ± 1.04 mL/g/min, P = 0.84) but those with HRR had significantly higher SMBFmax (2.66 1.86 mL/g/min, P < 0.001). HRR >15 bpm was superior to SSO in predicting adequate increase in SMBF (AUC 0.87 vs. 0.62, P < 0.001). CONCLUSION: Adenosine-induced increase in myocardial blood flow is accurate for confirmation of hyperaemia during stress CMR studies and is superior to traditional, clinically used markers of adequate stress such as SSO and BP response.

Automated Inline Analysis of Myocardial Perfusion MRI with Deep Learning.

PURPOSE: To develop a deep neural network-based computational workflow for inline myocardial perfusion analysis that automatically delineates the myocardium, which improves the clinical workflow and offers a "one-click" solution. MATERIALS AND METHODS: In this retrospective study, consecutive adenosine stress and rest perfusion scans were acquired from three hospitals between October 1, 2018 and February 27, 2019. The training and validation set included 1825 perfusion series from 1034 patients (mean age, 60.6 years ± 14.2 [standard deviation]). The independent test set included 200 scans from 105 patients (mean age, 59.1 years ± 12.5). A convolutional neural network (CNN) model was trained to segment the left ventricular cavity, myocardium, and right ventricle by processing an incoming time series of perfusion images. Model outputs were compared with manual ground truth for accuracy of segmentation and flow measures derived on a global and per-sector basis with t test performed for statistical significance. The trained models were integrated onto MR scanners for effective inference. RESULTS: The mean Dice ratio of automatic and manual segmentation was 0.93 ± 0.04. The CNN performed similarly to manual segmentation and flow measures for mean stress myocardial blood flow (MBF; 2.25 mL/min/g ± 0.59 vs 2.24 mL/min/g ± 0.59, P = .94) and mean rest MBF (1.08 mL/min/g ± 0.23 vs 1.07 mL/min/g ± 0.23, P = .83). The per-sector MBF values showed no difference between the CNN and manual assessment (P = .92). A central processing unit-based model inference on the MR scanner took less than 1 second for a typical perfusion scan of three slices. CONCLUSION: The described CNN was capable of cardiac perfusion mapping and integrated an automated inline implementation on the MR scanner, enabling one-click analysis and reporting in a manner comparable to manual assessment. Supplemental material is available for this article. © RSNA, 2020.

Assessment of Multivessel Coronary Artery Disease Using Cardiovascular Magnetic Resonance Pixelwise Quantitative Perfusion Mapping.

OBJECTIVES: The authors sought to compare the diagnostic accuracy of quantitative perfusion maps to visual assessment (VA) of first-pass perfusion images for the detection of multivessel coronary artery disease (MVCAD). BACKGROUND: VA of first-pass stress perfusion cardiac magnetic resonance (CMR) may underestimate ischemia in MVCAD. Pixelwise perfusion mapping allows quantitative measurement of regional myocardial blood flow, which may improve ischemia detection in MVCAD. METHODS: One hundred fifty-one subjects recruited at 2 centers underwent stress perfusion CMR with myocardial perfusion mapping, and invasive coronary angiography with coronary physiology assessment. Ischemic burden was assessed by VA of first-pass images and by quantitative measurement of stress myocardial blood flow using perfusion maps. RESULTS: In patients with MVCAD (2-vessel [2VD] or 3-vessel disease [3VD]; n = 95), perfusion mapping identified significantly more segments with perfusion defects (median segments per patient 12 [interquartile range (IQR): 9 to 16] by mapping vs. 8 [IQR: 5 to 9.5] by VA; p < 0.001). Ischemic burden (IB) measured using mapping was higher in MVCAD compared with IB measured using VA (3VD mapping 100 % (75% to 100%) vs. first-pass 56% (38% to 81%) ; 2VD mapping 63% (50% to 75%) vs. first-pass 41% (31% to 50%); both p < 0.001), but there was no difference in single-vessel disease (mapping 25% (13% to 44%) vs. 25% (13% to 31%). Perfusion mapping was superior to VA for the correct identification of extent of coronary disease (78% vs. 58%; p < 0.001) due to better identification of 3VD (87% vs. 40%) and 2VD (71% vs. 48%). CONCLUSIONS: VA of first-pass stress perfusion underestimates ischemic burden in MVCAD. Pixelwise quantitative perfusion mapping increases the accuracy of CMR in correctly identifying extent of coronary disease. This has important implications for assessment of ischemia and therapeutic decision-making.

Automated detection of left ventricle in arterial input function images for inline perfusion mapping using deep learning: A study of 15,000 patients.

PURPOSE: Quantification of myocardial perfusion has the potential to improve the detection of regional and global flow reduction. Significant effort has been made to automate the workflow, where one essential step is the arterial input function (AIF) extraction. Failure to accurately identify the left ventricle (LV) prevents AIF estimation required for quantification, therefore high detection accuracy is required. This study presents a robust LV detection method using the convolutional neural network (CNN). METHODS: CNN models were trained by assembling 25,027 scans (N = 12,984 patients) from three hospitals, seven scanners. Performance was evaluated using a hold-out test set of 5721 scans (N = 2805 patients). Model inputs were a time series of AIF images (2D+T). Two variations were investigated: (1) two classes (2CS) for background and foreground (LV mask), and (2) three classes (3CS) for background, LV, and RV. The final model was deployed on MRI scanners using the Gadgetron reconstruction software framework. RESULTS: Model loading on the MRI scanner took ~340 ms and applying the model took ~180 ms. The 3CS model successfully detected the LV in 99.98% of all test cases (1 failure out of 5721). The mean Dice ratio for 3CS was 0.87 ± 0.08 with 92.0% of all cases having Dice >0.75. The 2CS model gave a lower Dice ratio of 0.82 ± 0.22 (P < 1e-5). There was no significant difference in foot-time, peak-time, first-pass duration, peak value, and area-under-curve (P > .2) comparing automatically extracted AIF signals with signals from manually drawn contours. CONCLUSIONS: A CNN-based solution to detect the LV blood pool from the arterial input function image series was developed, validated, and deployed. A high LV detection accuracy of 99.98% was achieved.

The Prognostic Significance of Quantitative Myocardial Perfusion: An Artificial Intelligence-Based Approach Using Perfusion Mapping.

BACKGROUND: Myocardial perfusion reflects the macro- and microvascular coronary circulation. Recent quantitation developments using cardiovascular magnetic resonance perfusion permit automated measurement clinically. We explored the prognostic significance of stress myocardial blood flow (MBF) and myocardial perfusion reserve (MPR, the ratio of stress to rest MBF). METHODS: A 2-center study of patients with both suspected and known coronary artery disease referred clinically for perfusion assessment. Image analysis was performed automatically using a novel artificial intelligence approach deriving global and regional stress and rest MBF and MPR. Cox proportional hazard models adjusting for comorbidities and cardiovascular magnetic resonance parameters sought associations of stress MBF and MPR with death and major adverse cardiovascular events (MACE), including myocardial infarction, stroke, heart failure hospitalization, late (>90 day) revascularization, and death. RESULTS: A total of 1049 patients were included with a median follow-up of 605 (interquartile range, 464-814) days. There were 42 (4.0%) deaths and 188 MACE in 174 (16.6%) patients. Stress MBF and MPR were independently associated with both death and MACE. For each 1 mL·g-1·min-1 decrease in stress MBF, the adjusted hazard ratios for death and MACE were 1.93 (95% CI, 1.08-3.48, P=0.028) and 2.14 (95% CI, 1.58-2.90, P<0.0001), respectively, even after adjusting for age and comorbidity. For each 1 U decrease in MPR, the adjusted hazard ratios for death and MACE were 2.45 (95% CI, 1.42-4.24, P=0.001) and 1.74 (95% CI, 1.36-2.22, P<0.0001), respectively. In patients without regional perfusion defects on clinical read and no known macrovascular coronary artery disease (n=783), MPR remained independently associated with death and MACE, with stress MBF remaining associated with MACE only. CONCLUSIONS: In patients with known or suspected coronary artery disease, reduced MBF and MPR measured automatically inline using artificial intelligence quantification of cardiovascular magnetic resonance perfusion mapping provides a strong, independent predictor of adverse cardiovascular outcomes.