Precision measurement of cardiac structure and function in cardiovascular magnetic resonance using machine learning

BACKGROUND

Measurement of cardiac structure and function from images (e.g. volumes, mass and derived parameters such as left ventricular (LV) ejection fraction [LVEF]) guides care for millions. This is best assessed using cardiovascular magnetic resonance (CMR), but image analysis is currently performed by individual clinicians, which introduces error. We sought to develop a machine learning algorithm for volumetric analysis of CMR images with demonstrably better precision than human analysis.

METHODS

A fully automated machine learning algorithm was trained on 1923 scans (10 scanner models, 13 institutions, 9 clinical conditions, 60,000 contours) and used to segment the LV blood volume and myocardium. Performance was quantified by measuring precision on an independent multi-site validation dataset with multiple pathologies with n = 109 patients, scanned twice. This dataset was augmented with a further 1277 patients scanned as part of routine clinical care to allow qualitative assessment of generalization ability by identifying mis-segmentations. Machine learning algorithm ('machine') performance was compared to three clinicians ('human') and a commercial tool (cvi42, Circle Cardiovascular Imaging).

FINDINGS

Machine analysis was quicker (20 s per patient) than human (13 min). Overall machine mis-segmentation rate was 1 in 479 images for the combined dataset, occurring mostly in rare pathologies not encountered in training. Without correcting these mis-segmentations, machine analysis had superior precision to three clinicians (e.g. scan-rescan coefficients of variation of human vs machine: LVEF 6.0% vs 4.2%, LV mass 4.8% vs. 3.6%; both P < 0.05), translating to a 46% reduction in required trial sample size using an LVEF endpoint.

CONCLUSION

We present a fully automated algorithm for measuring LV structure and global systolic function that betters human performance for speed and precision.

Improving cardiovascular magnetic resonance access in low- and middle-income countries for cardiomyopathy assessment: rapid cardiovascular magnetic resonance

AIMS

To evaluate the impact of a simplified, rapid cardiovascular magnetic resonance (CMR) protocol embedded in care and supported by a partner education programme on the management of cardiomyopathy (CMP) in low- and middle-income countries (LMICs).

METHODS AND RESULTS

Rapid CMR focused particularly on CMP was implemented in 11 centres, 7 cities, 5 countries, and 3 continents linked to training courses for local professionals. Patients were followed up for 24 months to assess impact. The rate of subsequent adoption was tracked. Five CMR conferences were delivered (920 attendees-potential referrers, radiographers, reporting cardiologists, or radiologists) and five new centres starting CMR. Six hundred and one patients were scanned. Cardiovascular magnetic resonance indications were 24% non-contrast T2* scans [myocardial iron overload (MIO)] and 72% suspected/known cardiomyopathies (including ischaemic and viability). Ninety-eighty per cent of studies were of diagnostic quality. The average scan time was 22 ± 6 min (contrast) and 12 ± 4 min (non-contrast), a potential cost/throughput reduction of between 30 and 60%. Cardiovascular magnetic resonance findings impacted management in 62%, including a new diagnosis in 22% and MIO detected in 30% of non-contrast scans. Nine centres continued using rapid CMR 2 years later (typically 1-2 days per week, 30 min slots).

CONCLUSIONS

Rapid CMR of diagnostic quality can be delivered using available technology in LMICs. When embedded in care and a training programme, costs are lower, care is improved, and services can be sustained over time.

28Impact of non-invasive rapid cardiac magnetic resonance for the assessment of cardiomyopathies in developing countries

Background

Cardiovascular mortality is higher in developing countries. Part of that is suboptimal testing. Cardiac magnetic resonance (CMR) is the gold standard for measuring structure, function of the heart and adds incremental value by imaging scarring and to assess iron level. Despite the existence of MRI units, CMR is identified as a complex test, with poor training and availability in developing countries.

Purpose

To assess the potential impact of a faster CMR protocol at a multicentre level in developing countries; implementing it with an education program, for the assessment cardiomyopathies.

Methods

An international partnership. A rapid CMR protocol for the evaluation of cardiac volumes, function and tissue characterization (Cardiac Iron T2* and LGE for scar) Figure 1a. We deployed the protocol as a multicentre study: Argentina, Peru, India and South Africa accompanied by a program of education. Pre-scan clinical information, scanning data: complications, image quality and post-scan follow-up of participants for the assessment on impact, between 3 to 24 months.

Results

510 scans (4 countries, 6 cities, 12 centres) were performed with the rapid CMR protocol. Contrast studies in 378 (74%). There were no scan-related complications. Quality of the studies was maintained in a high level as an average of 89%. 97% of studies responded referral's question. All patients with contrast CMR scan have had at least one 2D echocardiogram before CMR. Average scan duration was 21±6 mins for contrast studies and 12±3 for non-contrast T2* protocol. The most common underlying diagnoses were non-ischaemic cardiomyopathy in 73% of participants (including cardiac iron level assessment in 26%, HCM in 17%, DCM in 15%), 27% for ischaemic cardiomyopathy and 15% for other pathologies. 4 of the 12 participant centres started to incorporate CMR for the first time. Findings impacted management in 60% of patients, including new diagnosis in 21% of participants. See table 1, figure 1b. For just cardiac iron assessment: 1/3 of participants had iron deposited in the heart with 14% of patients in severe levels.

Conclusions

CMR can be delivered faster and easier. When this abbreviated protocol is enabled with education, it can be implemented in developing countries with existing technology. This protocol shows high quality exam, with an important impact on patient's management.

Characteristics and impact on management Contrast studies Non-contrast studies All patients (%) 378 (74) 132 (36) Age, mean (range) years 54 (16–93) 24 (13–41) Male (%) 151 (39) 64 (48) Pre-echocardiography exam (%) 370 (98) 42 (32) Scanning duration mean (SD) 21 (6) 12 (3) Good quality exam (%) 329 (87) 120 (91) Impact on management Total All patients (%) 510 (100) Completely new diagnosis (%) 105 (21) Change/Addition of Medication (%) 128 (25) Intervention/ Surgery (%) 31 (6) Invasive angiography/biopsy (%) 25 (5) Hospital discharge/admission (%) 15 (3) TOTAL 306 (60%) SD: Standard Deviation.

Acknowledgement/Funding

Global Engagement UCL, UK Foreign & Commonwealth Office and The Peruvian Scientific, Technological Development and Technological Innovation (FONDECYT)

A randomized, multicenter, open-label, blinded end point trial comparing the effects of spironolactone to chlorthalidone on left ventricular mass in patients with early-stage chronic kidney disease: Rationale and design of the SPIRO-CKD trial

BACKGROUND

Chronic kidney disease (CKD) is associated with increased left ventricular (LV) mass and arterial stiffness. In a previous trial, spironolactone improved these end points compared with placebo in subjects with early-stage CKD, but it is not known whether these effects were specific to the drug or secondary to blood pressure lowering.

AIM

The aim was to investigate the hypothesis that spironolactone is superior to chlorthalidone in the reduction of LV mass while exerting similar effects on blood pressure.

DESIGN

This is a multicenter, prospective, randomized, open-label, blinded end point clinical trial initially designed to compare the effects of 40weeks of treatment with spironolactone 25mg once daily to chlorthalidone 25mg once daily on the co-primary end points of change in pulse wave velocity and change in LV mass in 350 patients with stages 2 and 3 CKD on established treatment with an angiotensin-converting enzyme inhibitor or an angiotensin receptor blocker. Because of slow recruitment rates, it became apparent that it would not be possible to recruit this sample size within the funded time period. The study design was therefore changed to one with a single primary end point of LV mass requiring 150 patients. Recruitment was completed on 31 December 2016, at which time 154 patients had been recruited. Investigations included cardiac magnetic resonance imaging, applanation tonometry, 24-hour ambulatory blood pressure monitoring, and laboratory tests. Subjects are assessed before and after 40weeks of randomly allocated drug therapy and at 46weeks after discontinuation of the study drug.

T1 mapping and T2 mapping at 3T for quantifying the area-at-risk in reperfused STEMI patients

BACKGROUND

Whether T1-mapping cardiovascular magnetic resonance (CMR) can accurately quantify the area-at-risk (AAR) as delineated by T2 mapping and assess myocardial salvage at 3T in reperfused ST-segment elevation myocardial infarction (STEMI) patients is not known and was investigated in this study.

METHODS

18 STEMI patients underwent CMR at 3T (Siemens Bio-graph mMR) at a median of 5 (4-6) days post primary percutaneous coronary intervention using native T1 (MOLLI) and T2 mapping (WIP #699; Siemens Healthcare, UK). Matching short-axis T1 and T2 maps covering the entire left ventricle (LV) were assessed by two independent observers using manual, Otsu and 2 standard deviation thresholds. Inter- and intra-observer variability, correlation and agreement between the T1 and T2 mapping techniques on a per-slice and per patient basis were assessed.

RESULTS

A total of 125 matching T1 and T2 mapping short-axis slices were available for analysis from 18 patients. The acquisition times were identical for the T1 maps and T2 maps. 18 slices were excluded due to suboptimal image quality. Both mapping sequences were equally prone to susceptibility artifacts in the lateral wall and were equally likely to be affected by microvascular obstruction requiring manual correction. The Otsu thresholding technique performed best in terms of inter- and intra-observer variability for both T1 and T2 mapping CMR. The mean myocardial infarct size was 18.8 ± 9.4 % of the LV. There was no difference in either the mean AAR (32.3 ± 11.5 % of the LV versus 31.6 ± 11.2 % of the LV, P = 0.25) or myocardial salvage index (0.40 ± 0.26 versus 0.39 ± 0.27, P = 0.20) between the T1 and T2 mapping techniques. On a per-slice analysis, there was an excellent correlation between T1 mapping and T2 mapping in the quantification of the AAR with an R(2) of 0.95 (P < 0.001), with no bias (mean ± 2SD: bias 0.0 ± 9.6 %). On a per-patient analysis, the correlation and agreement remained excellent with no bias (R(2) 0.95, P < 0.0001, bias 0.7 ± 5.1 %).

CONCLUSIONS

T1 mapping CMR at 3T performed as well as T2 mapping in quantifying the AAR and assessing myocardial salvage in reperfused STEMI patients, thereby providing an alternative CMR measure of the the AAR.

Diagnosis of apical hypertrophic cardiomyopathy: T-wave inversion and relative but not absolute apical left ventricular hypertrophy

BACKGROUND

Diagnosis of apical HCM utilizes conventional wall thickness criteria. The normal left ventricular wall thins towards the apex such that normal values are lower in the apical versus the basal segments. The impact of this on the diagnosis of apical hypertrophic cardiomyopathy has not been evaluated.

METHODS

We performed a retrospective review of 2662 consecutive CMR referrals, of which 75 patients were identified in whom there was abnormal T-wave inversion on ECG and a clinical suspicion of hypertrophic cardiomyopathy. These were retrospectively analyzed for imaging features consistent with cardiomyopathy, specifically: relative apical hypertrophy, left atrial dilatation, scar, apical cavity obliteration or apical aneurysm. For comparison, the same evaluation was performed in 60 healthy volunteers and 50 hypertensive patients.

RESULTS

Of the 75 patients, 48 met conventional HCM diagnostic criteria and went on to act as another comparator group. Twenty-seven did not meet criteria for HCM and of these 5 had no relative apical hypertrophy and were not analyzed further. The remaining 22 patients had relative apical thickening with an apical:basal wall thickness ratio >1 and a higher prevalence of features consistent with a cardiomyopathy than in the control groups with 54% having 2 or more of the 4 features. No individual in the healthy volunteer group had more than one feature and no hypertension patient had more than 2.

CONCLUSION

A cohort of individuals exist with T wave inversion, relative apical hypertrophy and additional imaging features of HCM suggesting an apical HCM phenotype not captured by existing diagnostic criteria.

Noncontact mapping guided ablation of right ventricular outflow tract ectopy in a patient with interruption of the inferior vena cava and azygos continuation

A 58-year-old woman with symptomatic multiple monomorphic premature ventricular beats of a right ventricular outflow tract origin was referred for ablation. An inferior vena cava interruption with azygos continuation was discovered during catheter placement. This case describes positioning of the noncontact mapping array and successful radiofrequency ablation in this challenging anatomy.