Defining the real-world reproducibility of visual grading of left ventricular function and visual estimation of left ventricular ejection fraction: impact of image quality, experience and accreditation

Self-supervised learning for label-free segmentation in cardiac ultrasound

Segmentation and measurement of cardiac chambers from ultrasound is critical, but laborious and poorly reproducible. Neural networks can assist, but supervised approaches require the same problematic manual annotations. We build a pipeline for self-supervised segmentation combining computer vision, clinical knowledge, and deep learning. We train on 450 echocardiograms and test on 18,423 echocardiograms (including external data), using the resulting segmentations to calculate measurements. Coefficient of determination (r2) between clinically measured and pipeline-predicted measurements (0.55-0.84) are comparable to inter-clinician variation and to supervised learning. Average accuracy for detecting abnormal chambers is 0.85 (0.71-0.97). A subset of test echocardiograms (n = 553) have corresponding cardiac MRIs (the gold standard). Correlation between pipeline and MRI measurements is similar to that of clinical echocardiogram. Finally, the pipeline segments the left ventricle with an average Dice score of 0.89 (95% CI [0.89]). Our results demonstrate a manual-label free, clinically valid, and scalable method for segmentation from ultrasound.

The value of handheld ultrasound in point-of-care or at home EF prediction

In this paper, AI-enabled handheld ultrasound is used in point-of-care or at home, and evaluate the accuracy of it for left ventricular ejection fraction (LVEF) evaluation. It provides a simple, convenient, and practical tool for the patients with heart disease, especially those with heart failure. The AI model used for this AI-enabled handheld ultrasound is a machine learning model trained with tens of thousands of ultrasound four-chamber cardiograms. The LVEF evaluation accuracy of the AI model was compared by the experts performing ultrasound four-chamber cardiogram detection in 100 patients on high-end ultrasound in the hospital. In the 100 clinical trials, the sensitivity, specificity, and accuracy of the AI model were 91%, 95%, and 98%, respectively. Then 10 cases were used to compare the LVEF results of hospital tests with the predicted results of the AI model. The difference between the two is less than 10%. Finally, over the course of one month, the AI-enabled handheld ultrasound was employed to conduct regular evaluations of left LVEF for point-of-care purposes on a group of 10 patients diagnosed with heart failure. The LVEF evaluation accuracy of AI-enabled handheld ultrasound is more than 96%, which was higher than that of experts in high-end ultrasound in hospitals. The easy-to-use AI-enabled handheld ultrasound can evaluate the LVEF in the point of care or at home and get the same accuracy as the high-end ultrasound equipment in the hospital. It may play an important role in monitoring cardiac function at home for the ambulatory heart failure patients.

Assessment of the Left Ventricular Dysfunction in Patients With Acromegaly Using Global Longitudinal Strain by Two-Dimensional Speckle Tracking Echocardiography and Tissue Doppler Imaging

Introduction Acromegaly is a rare disease resulting from excess growth hormone (GH) and insulin-like growth factor-1 (IGF1), with cardiovascular complications being frequently encountered, leading to increased morbidity and mortality. The aim of the study was to determine the frequency of left ventricular systolic dysfunction in patients with acromegaly using global longitudinal strain (GLS) by 2D speckle tracking echocardiography and also the frequency of left ventricular diastolic dysfunction by tissue Doppler imaging (TDI). Materials and method A cross-sectional study involving 20 acromegaly patients with normal left ventricular (LV) systolic function as measured by ejection fraction and 20 controls with age, sex, and comorbidities matched were included in the study from 2021 to 2023. All these patients underwent 2D speckle tracking echocardiography to assess GLS and TDI with conventional 2D Echocardiography to assess diastolic function. Results GLS was significantly lower in the acromegaly group, which was -15.79±2.54 (mean±SD), than in the control group, which was -17.47±0.98 (mean±SD) with p < 0.05, indicating significant LV systolic dysfunction in the acromegaly group. The majority of the acromegaly group had abnormal GLS (n=11; 55%). The majority of the acromegaly patients with increased left atrial volume index (LAVi) had abnormal GLS (n=8/11; 72.7%). Also, the majority of the acromegaly patients with increased LVMi had abnormal GLS (n=8/12; 66.66%). TDI study for diastolic dysfunction showed no significant difference between the acromegaly group and the control group (p > 0.05). LAVi in the acromegaly group was 31.35±6.22 (mean±SD), and in the control group was 27.00±4.81(mean±SD) with p < 0.05, which was statistically significant. LAVi was more in the active acromegaly group (4 males and 2 females) than inactive acromegaly group (3 males and 2 females). LVMi in the acromegaly group was 100.32±24.335 (mean±SD), and in the control group, it was 85.85±19.63 (mean±SD) with p < 0.05, indicating more LV hypertrophy. LVMi was more in the active (5 females and 4 males) than the inactive acromegaly group (2 females and 1 male). Statistical significance was observed in LVID between the acromegaly group, which was 4.75±0.52 (mean±SD), and the control group, which was 4.41±0.49 (mean±SD) with a p < 0.05. Septal medial early diastolic velocity (e` med) in the acromegaly group was 0.08±0.03 (mean±SD), and in the control group was 0.10±0.02 (mean±SD) with p < 0.05, which was statistically significant. The multiple linear regression analysis revealed that acromegaly, hypertension, and higher body surface were the most important predictors of abnormal GLS. Conclusions Abnormal GLS indicating subclinical LV systolic dysfunction in patients with acromegaly can be evaluated by 2D speckle tracking echocardiography. Active acromegaly patients had more abnormal GLS, increased LAVi, and increased LVMi than inactive acromegaly patients. LV diastolic dysfunction was not remarkable when the acromegaly group and control group were assessed as the comorbidities were matched. The presence of acromegaly, hypertension, and higher body surface area had a significant negative effect on GLS.

AI in Echocardiography: State-of-the-art Automated Measurement Techniques and Clinical Applications

The artificial intelligence (AI) technology in automated measurements has seen remarkable advancements across various vendors, thereby offering new opportunities in echocardiography. Fully automated software particularly has the potential to elevate the analysis and the interpretation of medical images to a new level compared to previous algorithms. Tasks that traditionally required significant time, such as ventricular and atrial volume measurements and Doppler tracing, can now be performed swiftly through AI's automated phase setting and waveform tracing capabilities. The benefits of AI-driven systems include high-precision and reliable measurements, significant time savings, and enhanced workflow efficiency. By automating routine tasks, AI can reduce the burden on clinicians, allowing them to gather additional information, perform additional tests, and improve patient care. While many studies confirm the accuracy and the reproducibility of AI-driven techniques, it is crucial for clinicians to verify AI-generated measurements and ensure high-quality imaging and Doppler waveforms to fully take advantage of the benefits from these technologies. This review discusses the current state of AI-driven automated measurements in echocardiography, their impact on clinical practice, and the strategies required for the effective integration of AI into clinical workflows.

Impact of experience on visual and Simpson's biplane echocardiographic assessment of left ventricular ejection fraction

BACKGROUND

In clinical routine, health care professionals with various levels of experience assess left ventricular ejection fraction (LVEF) by echocardiography. The aim was to investigate to what extent visual and Simpson's biplane assessment of LVEF, using two-dimensional (2D) transthoracic echocardiography (TTE), is affected by the evaluator's experience.

METHODS

Ultrasound images of 140 patients were assessed, visually and with Simpson's biplane method, by six evaluators divided into three groups based on echocardiographic experience level (beginner, intermediate and expert). The evaluators were blinded to each other's LVEF assessments. Bland-Altman analyses (bias±SD) were performed to assess agreement. P-values < 0.05 with the performed paired t-test were considered statistically significant.

RESULTS

Level of agreement in LVEF was good between evaluators within the expert group: visual = LVEFexpert 1 vs LVEFexpert 2: -0.4 ± 6.4 (p = 0.46); Simpson's biplane = LVEFexpert 1 vs LVEFexpert 2: 0.96 ± 7.0 (p = 0.11), somewhat lower within the intermediate group: visual = LVEFintermediate 1 vs LVEFintermediate 2: -1.2 ± 4.4 (p = 0.004); Simpson's biplane = LVEFintermediate 1 vs LVEF intermediate 2: -3.3 ± 5.0 (p < 0.001) and lowest for beginners: visual = LVEFbeginner 1 vs LVEFbeginner 2: 2.3 ± 9.8 (p = 0.007), Simpson's biplane = LVEFbeginner 1 vs LVEF beginner 2: -1.8 ± 8.7 (p = 0.02). The agreement between LVEFexpert and LVEFs by the two other groups was: visual = LVEFexpert vs LVEFbeginner: 1.5 ± 6.0 (p = 0.005); LVEFintermediate: -3.0 ± 4.4 (p < 0.001) and Simpson's biplane = LVEFexpert vs LVEFbeginner: 3.2 ± 6.3 (p < 0.001); LVEFintermediate: -2.2 ± 4.7 (p < 0.001).

CONCLUSIONS

The evaluator's level of experience affects visual and Simpson's biplane assessment of LVEF by 2D-TTE, with highest variability being among beginners. Furthermore, a second opinion is recommended when assessing reduced LVEF even for evaluators with intermediate and expert experience.

SimLVSeg: Simplifying Left Ventricular Segmentation in 2-D+Time Echocardiograms With Self- and Weakly Supervised Learning

OBJECTIVE

Achieving reliable automatic left ventricle (LV) segmentation from echocardiograms is challenging due to the inherent sparsity of annotations in the dataset, as clinicians typically only annotate two specific frames for diagnostic purposes. Here we aim to address this challenge by introducing simplified LV segmentation (SimLVSeg), a novel paradigm that enables video-based networks for consistent LV segmentation from sparsely annotated echocardiogram videos.

METHODS

SimLVSeg consists of two training stages: (i) self-supervised pre-training with temporal masking, which involves pre-training a video segmentation network by capturing the cyclic patterns of echocardiograms from largely unannotated echocardiogram frames, and (ii) weakly supervised learning tailored for LV segmentation from sparse annotations.

RESULTS

We extensively evaluated SimLVSeg using EchoNet-Dynamic, the largest echocardiography dataset. SimLVSeg outperformed state-of-the-art solutions by achieving a 93.32% (95% confidence interval: 93.21-93.43%) dice score while being more efficient. We further conducted an out-of-distribution test to showcase SimLVSeg's generalizability on distribution shifts (CAM US dataset).

CONCLUSION

Our findings show that SimLVSeg exhibits excellent performance on LV segmentation with a relatively cheaper computational cost. This suggests that adopting video-based networks for LV segmentation is a promising research direction to achieve reliable LV segmentation. Our code is publicly available at https://github.com/BioMedIA-MBZUAI/SimLVSeg.

Active Inference and Deep Generative Modeling for Cognitive Ultrasound

Ultrasound (US) has the unique potential to offer access to medical imaging to anyone, everywhere. Devices have become ultraportable and cost-effective, akin to the stethoscope. Nevertheless, and despite many advances, US image quality and diagnostic efficacy are still highly operator- and patient-dependent. In difficult-to-image patients, image quality is often insufficient for reliable diagnosis. In this article, we put forth the idea that US imaging systems can be recast as information-seeking agents that engage in reciprocal interactions with their anatomical environment. Such agents autonomously adapt their transmit-receive sequences to fully personalize imaging and actively maximize information gain in situ. To that end, we will show that the sequence of pulse-echo experiments that a US system performs can be interpreted as a perception-action loop: the action is the data acquisition, probing tissue with acoustic waves and recording reflections at the detection array, and perception is the inference of the anatomical and or functional state, potentially including associated diagnostic quantities. We then equip systems with a mechanism to actively reduce uncertainty and maximize diagnostic value across a sequence of experiments, treating action and perception jointly using Bayesian inference given generative models of the environment and action-conditional pulse-echo observations. Since the representation capacity of the generative models dictates both the quality of inferred anatomical states and the effectiveness of inferred sequences of future imaging actions, we will be greatly leveraging the enormous advances in deep generative modeling (generative AI), which are currently disrupting many fields and society at large. Finally, we show some examples of cognitive, closed-loop, US systems that perform active beamsteering and adaptive scanline selection based on deep generative models that track anatomical belief states.

Magnetocardiography in the Evaluation of Sudden Cardiac Death Risk: A Systematic Review

Sudden cardiac death (SCD) is responsible for 15%-20% of deaths globally/year, predominantly due to ventricular arrhythmias (VA) caused by vulnerable cardiac substrate. Identifying those at risk has proved difficult with several limitations of current methods. We evaluated the evidence for magnetocardiography (MCG) in predicting SCD events. We searched Embase/Medline databases for English language papers evaluating MCG in patients at risk of VA. A total of 119 papers were screened with 27 papers included for analysis (23 case-controlled, four cohort studies); study sizes varied (n = 12 to 158). Etiology was ischemic cardiomyopathy (ICM) in 22, dilated cardiomyopathy in 2, arrhythmogenic cardiomyopathy in 1 and mixed in 2. In patients with ICM there were consistent discriminatory features seen using time-based and signal-complexity measures that persisted when evaluating the independence of these parameters. Current flow analysis demonstrated promising discriminatory results in other etiologies. The features studied support the role of MCG in identifying substrate for VA, particularly in ICM.

Active learning for left ventricle segmentation in echocardiography

BACKGROUND AND OBJECTIVE

Training deep learning models for medical image segmentation require large annotated datasets, which can be expensive and time-consuming to create. Active learning is a promising approach to reduce this burden by strategically selecting the most informative samples for segmentation. This study investigates the use of active learning for efficient left ventricle segmentation in echocardiography with sparse expert annotations.

METHODS

We adapt and evaluate various sampling techniques, demonstrating their effectiveness in judiciously selecting samples for segmentation. Additionally, we introduce a novel strategy, Optimised Representativeness Sampling, which combines feature-based outliers with the most representative samples to enhance annotation efficiency.

RESULTS

Our findings demonstrate a substantial reduction in annotation costs, achieving a remarkable 99% upper bound performance while utilising only 20% of the labelled data. This equates to a reduction of 1680 images needing annotation within our dataset. When applied to a publicly available dataset, our approach yielded a remarkable 70% reduction in required annotation efforts, representing a significant advancement compared to baseline active learning strategies, which achieved only a 50% reduction. Our experiments highlight the nuanced performance of diverse sampling strategies across datasets within the same domain.

CONCLUSIONS

The study provides a cost-effective approach to tackle the challenges of limited expert annotations in echocardiography. By introducing a distinct dataset, made publicly available for research purposes, our work contributes to the field's understanding of efficient annotation strategies in medical image segmentation.

Vision-language foundation model for echocardiogram interpretation

The development of robust artificial intelligence models for echocardiography has been limited by the availability of annotated clinical data. Here, to address this challenge and improve the performance of cardiac imaging models, we developed EchoCLIP, a vision-language foundation model for echocardiography, that learns the relationship between cardiac ultrasound images and the interpretations of expert cardiologists across a wide range of patients and indications for imaging. After training on 1,032,975 cardiac ultrasound videos and corresponding expert text, EchoCLIP performs well on a diverse range of benchmarks for cardiac image interpretation, despite not having been explicitly trained for individual interpretation tasks. EchoCLIP can assess cardiac function (mean absolute error of 7.1% when predicting left ventricular ejection fraction in an external validation dataset) and identify implanted intracardiac devices (area under the curve (AUC) of 0.84, 0.92 and 0.97 for pacemakers, percutaneous mitral valve repair and artificial aortic valves, respectively). We also developed a long-context variant (EchoCLIP-R) using a custom tokenizer based on common echocardiography concepts. EchoCLIP-R accurately identified unique patients across multiple videos (AUC of 0.86), identified clinical transitions such as heart transplants (AUC of 0.79) and cardiac surgery (AUC 0.77) and enabled robust image-to-text search (mean cross-modal retrieval rank in the top 1% of candidate text reports). These capabilities represent a substantial step toward understanding and applying foundation models in cardiovascular imaging for preliminary interpretation of echocardiographic findings.

Usefulness of Heart Failure Categories Based on Left Ventricular Ejection Fraction

BACKGROUND

Heart failure guidelines have recently introduced a narrow category with mildly reduced left ventricular ejection fraction (LVEF) (heart failure with mildly reduced ejection fraction; LVEF 41%-49%) between the previous categories of reduced (heart failure with reduced ejection fraction; LVEF ≤40%) and preserved (heart failure with preserved ejection fraction; LVEF ≥50%) ejection fraction. Grouping of continuous measurements into narrow categories can be questioned if their variability is high.

METHODS AND RESULTS

We constructed a cohort of all 9716 new cases of chronic heart failure with an available LVEF in Stockholm, Sweden, from January 1, 2015, until December 31, 2020. All values of LVEF were collected over time, and patients were followed up until death, moving out of Stockholm, or end of study. Mixed models were used to quantify within-person variance in LVEF, and multistate Markov models, with death as an absorbing state, to quantify the stability of LVEF categories. LVEF values followed a normal distribution. The SD of the within-person variance in LVEF over time was 7.4%. The mean time spent in any LVEF category before transition to another category was on average <1 year for heart failure with mildly reduced ejection fraction. Probabilities of transitioning between categories during the first year were substantial; patients with heart failure with mildly reduced ejection fraction had a probability of <25% of remaining in that category 1 year later.

CONCLUSIONS

LVEF follows a normal distribution and has considerable variability over time, which may impose a risk for underuse of efficient treatment. The heart failure with mildly reduced ejection fraction category is especially inconstant. Assumptions of a patient's current LVEF should take this variability and the normal distribution of LVEF into account.

Automatic assessment of left ventricular function for hemodynamic monitoring using artificial intelligence and transesophageal echocardiography

We have developed a method to automatically assess LV function by measuring mitral annular plane systolic excursion (MAPSE) using artificial intelligence and transesophageal echocardiography (autoMAPSE). Our aim was to evaluate autoMAPSE as an automatic tool for rapid and quantitative assessment of LV function in critical care patients. In this retrospective study, we studied 40 critical care patients immediately after cardiac surgery. First, we recorded a set of echocardiographic data, consisting of three consecutive beats of midesophageal two- and four-chamber views. We then altered the patient's hemodynamics by positioning them in anti-Trendelenburg and repeated the recordings. We measured MAPSE manually and used autoMAPSE in all available heartbeats and in four LV walls. To assess the agreement with manual measurements, we used a modified Bland-Altman analysis. To assess the precision of each method, we calculated the least significant change (LSC). Finally, to assess trending ability, we calculated the concordance rates using a four-quadrant plot. We found that autoMAPSE measured MAPSE in almost every set of two- and four-chamber views (feasibility 95%). It took less than a second to measure and average MAPSE over three heartbeats. AutoMAPSE had a low bias (0.4 mm) and acceptable limits of agreement (- 3.7 to 4.5 mm). AutoMAPSE was more precise than manual measurements if it averaged more heartbeats. AutoMAPSE had acceptable trending ability (concordance rate 81%) during hemodynamic alterations. In conclusion, autoMAPSE is feasible as an automatic tool for rapid and quantitative assessment of LV function, indicating its potential for hemodynamic monitoring.

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.

Harnessing the Power of AI: A Comprehensive Review of Left Ventricular Ejection Fraction Assessment With Echocardiography

The quantification of left ventricular ejection fraction (LVEF) has important clinical utility in the assessment of cardiac function and is vital for the diagnosis of cardiovascular diseases. A transthoracic echocardiogram serves as the most commonly used tool for LVEF assessment for several reasons, including, its noninvasive nature, great safety profile, real-time image processing ability, portability, and cost-effectiveness. However, transthoracic echocardiogram is highly dependent on the clinical skill of the sonographer and interpreting physician. Moreover, even amongst well-trained clinicians, significant interobserver variability exists in the quantification of LVEF. In search of possible solutions, the usage of artificial intelligence (AI) has been increasingly tested in the clinical setting. While AI-derived ejection fraction is in the preliminary stages of development, it has shown promise in its ability to rapidly quantify LVEF, decrease variability, increase accuracy, and utilize higher-order processing capabilities. This review will delineate the latest advancements of AI in evaluating LVEF through echocardiography and explore the challenges and future trajectory of this emerging domain.

Automatic measurements of left ventricular volumes and ejection fraction by artificial intelligence: clinical validation in real time and large databases

AIMS

Echocardiography is a cornerstone in cardiac imaging, and left ventricular (LV) ejection fraction (EF) is a key parameter for patient management. Recent advances in artificial intelligence (AI) have enabled fully automatic measurements of LV volumes and EF both during scanning and in stored recordings. The aim of this study was to evaluate the impact of implementing AI measurements on acquisition and processing time and test-retest reproducibility compared with standard clinical workflow, as well as to study the agreement with reference in large internal and external databases.

METHODS AND RESULTS

Fully automatic measurements of LV volumes and EF by a novel AI software were compared with manual measurements in the following clinical scenarios: (i) in real time use during scanning of 50 consecutive patients, (ii) in 40 subjects with repeated echocardiographic examinations and manual measurements by 4 readers, and (iii) in large internal and external research databases of 1881 and 849 subjects, respectively. Real-time AI measurements significantly reduced the total acquisition and processing time by 77% (median 5.3 min, P < 0.001) compared with standard clinical workflow. Test-retest reproducibility of AI measurements was superior in inter-observer scenarios and non-inferior in intra-observer scenarios. AI measurements showed good agreement with reference measurements both in real time and in large research databases.

CONCLUSION

The software reduced the time taken to perform and volumetrically analyse routine echocardiograms without a decrease in accuracy compared with experts.

Prognostic Value of Cardiac Magnetic Resonance Feature Tracking Strain in Aortic Stenosis

BACKGROUND

Recent data have suggested that global longitudinal strain (GLS) could be useful for risk stratification of patients with severe aortic stenosis (AS). In this study, we aimed to investigate the prognostic role of GLS in patients with AS and also its incremental value in relation to left ventricular ejection fraction (LVEF) and late gadolinium enhancement (LGE).

METHODS

We analysed all consecutive patients with AS and LGE-CMR in our institution. Survival data were obtained from office of national statistics, a national body where all deaths in England are registered by law. Death certificates were obtained from the general register office.

RESULTS

Some 194 consecutive patients with aortic stenosis were investigated with CMR at baseline and followed up for 7.3 ± 4 years. On multivariate Cox regression analysis, only increasing age remained significant for both all-cause and cardiac mortality, while LGE (any pattern) retained significance for all-cause mortality and had a trend to significance for cardiac mortality. Kaplan-Meier survival analysis demonstrated that patients in the best and middle GLS tertiles had significantly better mortality compared to patients in the worst GLS tertiles. Importantly though, sequential Cox proportional-hazard analysis demonstrated that GLS did not have significant incremental prognostic value for all-cause mortality or cardiac mortality in addition to LVEF and LGE.

CONCLUSIONS

Our study has demonstrated that age and LGE but not GLS are significant poor prognostic indicators in patients with moderate and severe AS.

ChatGPT Helped Me Write This Talk Title, but Can It Read an Echocardiogram?

While multidisciplinary collaboration in echocardiography is not new, machine learning has the potential to further improve it. In this transcript of the ASE 2023 Annual Feigenbaum lecture, advancements in foundation models are discussed, including their advantages, current disadvantages, and future potential for echocardiography.

Development and Challenges of Pre-Heart Failure with Preserved Ejection Fraction

Pre-heart failure with preserved ejection fraction (Pre-HFpEF) is a critical link to the development of heart failure with preserved ejection fraction (HFpEF). Early recognition and early intervention of pre-HFpEF will halt the progression of HFpEF. This article addresses the concept proposal, development, and evolution of pre-HFpEF, the mechanisms and risks of pre-HFpEF, the screening methods to recognize pre-HFpEF, and the treatment of pre-HFpEF. Despite the challenges, we believe more focus on the topic will resolve more problems.

Growth of targeted neonatal echocardiography in Chinese neonatal intensive care units: gaps in practice and training

To evaluate clinical practice, neonatologists' attitudes, and the extent of training and accreditation regarding targeted neonatal echocardiography (TnEcho) among Chinese neonatologists. A web-based questionnaire was emailed to 331 neonatologists across China who completed training in subspecialty neonatology. The survey covered various aspects of TnEcho, including the characteristics of clinical practice, attitudes towards its usefulness, and perceived barriers to implementation and training methods. Survey response rate was 68.0% (225/331). Seventy-nine (35.1%) respondents stated that TnEcho was utilized in their NICUs. Most respondents reported the use of echocardiography to evaluate hemodynamic significance of the patent ductus arteriosus (PDA, 94.9%). The eyeballing technique was most used to evaluate left (82.3%) and right (77.2%) ventricular function. Most respondents (87.3-96.2%) positively valued the role of TnEcho in providing timely and longitudinal hemodynamic information to guide cardiovascular care. Access to TnEcho was more likely in centers with on-site pediatric cardiology service (p = .003), larger bed capacity (p = .004), or level IV status (p = .003). Lack of experienced practitioners with echocardiography expertise (88.9%) and accredited training programs (85.8%) was perceived to be the major barrier to implementation. Of concern, most practitioners with TnEcho skills received training in an informal manner through workshops (60.8%) or self-directed learning (54.4%). Conclusions: The use of TnEcho for longitudinal evaluation of infants with hemodynamic instability is growing within Chinese NICUs. There is an urgent need to develop standardized training programs and accreditation for TnEcho which are adapted to the Chinese context.

Cyclical Self-Supervision for Semi-Supervised Ejection Fraction Prediction From Echocardiogram Videos

Left-ventricular ejection fraction (LVEF) is an important indicator of heart failure. Existing methods for LVEF estimation from video require large amounts of annotated data to achieve high performance, e.g. using 10,030 labeled echocardiogram videos to achieve mean absolute error (MAE) of 4.10. Labeling these videos is time-consuming however and limits potential downstream applications to other heart diseases. This paper presents the first semi-supervised approach for LVEF prediction. Unlike general video prediction tasks, LVEF prediction is specifically related to changes in the left ventricle (LV) in echocardiogram videos. By incorporating knowledge learned from predicting LV segmentations into LVEF regression, we can provide additional context to the model for better predictions. To this end, we propose a novel Cyclical Self-Supervision (CSS) method for learning video-based LV segmentation, which is motivated by the observation that the heartbeat is a cyclical process with temporal repetition. Prediction masks from our segmentation model can then be used as additional input for LVEF regression to provide spatial context for the LV region. We also introduce teacher-student distillation to distill the information from LV segmentation masks into an end-to-end LVEF regression model that only requires video inputs. Results show our method outperforms alternative semi-supervised methods and can achieve MAE of 4.17, which is competitive with state-of-the-art supervised performance, using half the number of labels. Validation on an external dataset also shows improved generalization ability from using our method.

Comparison of visual estimation and quantitative measurement of left ventricular ejection fraction in untrained perioperative echocardiographers

BACKGROUND

Perioperative evaluation of the left ventricular systolic function is essential information to help diagnose and manage life-threatening perioperative emergencies. Although quantifying the left ventricular ejection fraction (LVEF) is recommended to determine the left ventricular function, it may not always be feasible in emergency perioperative settings. This study compared the visual estimation of LVEF (eyeballing) by noncardiac anesthesiologists with the quantitative LVEF measured using a modified Simpson's biplane method.

METHODS

Transesophageal echocardiographic (TEE) studies of 35 patients were selected and 3 different echocardiographic views (the mid-esophageal four chamber view, the mid-esophageal two chamber view, and the transgastric mid-papillary short axis view) were recovered from each study and displayed in random order. Two cardiac anesthesiologists certified in perioperative echocardiography independently measured LVEF using the modified Simpson method and categorized LVEF into five grades: hyperdynamic LVEF, normal, mildly reduced LVEF, moderately reduced LVEF and severely reduced LVEF. Seven noncardiac anesthesiologists with limited experience in echocardiography also reviewed the same TEE studies and estimated the LVEF and graded LV function. The precision of the LV function classification and the correlation between visual estimation of LVEF and quantitative LVEF were calculated. The agreement of measurements between the two methods was also assessed.

RESULTS

Pearson's correlation between the LVEF estimated by the participants and the quantitative LVEF using the modified Simpson method was 0.818 (p < 0.001). Of a total of 245 responses, 120 (49.0%) responses were correct grading of the LV function. Participants were able to classify the LV function more accurately in the LV function grades 1 and 5 (65.3%). The 95% level of agreement of the Bland-Altman method was - 11.3-24.5. -21.9-22.6, - 23.1-26.5, - 20.5-22.0 and - 26.6-11.1 for LV grade 1 to 5, respectively.

CONCLUSION

Visual estimation of LVEF in perioperative TEE has acceptable accuracy in untrained echocardiographers and can be used for rescue TEE.

Blinded, randomized trial of sonographer versus AI cardiac function assessment

Artificial intelligence (AI) has been developed for echocardiography1-3, although it has not yet been tested with blinding and randomization. Here we designed a blinded, randomized non-inferiority clinical trial (ClinicalTrials.gov ID: NCT05140642; no outside funding) of AI versus sonographer initial assessment of left ventricular ejection fraction (LVEF) to evaluate the impact of AI in the interpretation workflow. The primary end point was the change in the LVEF between initial AI or sonographer assessment and final cardiologist assessment, evaluated by the proportion of studies with substantial change (more than 5% change). From 3,769 echocardiographic studies screened, 274 studies were excluded owing to poor image quality. The proportion of studies substantially changed was 16.8% in the AI group and 27.2% in the sonographer group (difference of -10.4%, 95% confidence interval: -13.2% to -7.7%, P < 0.001 for non-inferiority, P < 0.001 for superiority). The mean absolute difference between final cardiologist assessment and independent previous cardiologist assessment was 6.29% in the AI group and 7.23% in the sonographer group (difference of -0.96%, 95% confidence interval: -1.34% to -0.54%, P < 0.001 for superiority). The AI-guided workflow saved time for both sonographers and cardiologists, and cardiologists were not able to distinguish between the initial assessments by AI versus the sonographer (blinding index of 0.088). For patients undergoing echocardiographic quantification of cardiac function, initial assessment of LVEF by AI was non-inferior to assessment by sonographers.

Reliability of estimating left ventricular ejection fraction in clinical routine: a validation study of the SWEDEHEART registry

OBJECTIVE

Patients hospitalized with acute coronary syndrome (ACS) in Sweden routinely undergo an echocardiographic examination with assessment of left ventricular ejection fraction (LVEF). LVEF is a measurement widely used for outcome prediction and treatment guidance. The obtained LVEF is categorized as normal (> 50%) or mildly, moderately, or severely impaired (40-49, 30-39, and < 30%, respectively) and reported to the nationwide registry for ACS (SWEDEHEART). The purpose of this study was to determine the reliability of the reported LVEF values by validating them against an independent re-evaluation of LVEF.

METHODS

A random sample of 130 patients from three hospitals were included. LVEF re-evaluation was performed by two independent reviewers using the modified biplane Simpson method and their mean LVEF was compared to the LVEF reported to SWEDEHEART. Agreement between reported and re-evaluated LVEF was assessed using Gwet's AC2 statistics.

RESULTS

Analysis showed good agreement between reported and re-evaluated LVEF (AC2: 0.76 [95% CI 0.69-0.84]). The LVEF re-evaluations were in agreement with the registry reported LVEF categorization in 86 (66.0%) of the cases. In 33 (25.4%) of the cases the SWEDEHEART-reported LVEF was lower than re-evaluated LVEF. The opposite relation was found in 11 (8.5%) of the cases (p < 0.005).

CONCLUSION

Independent validation of SWEDEHEART-reported LVEF shows an overall good agreement with the re-evaluated LVEF. However, a tendency towards underestimation of LVEF was observed, with the largest discrepancy between re-evaluated LVEF and registry LVEF in subjects with subnormal LV-function in whom the reported assessment of LVEF should be interpreted more cautiously.

A Stress Test of Artificial Intelligence: Can Deep Learning Models Trained From Formal Echocardiography Accurately Interpret Point-of-Care Ultrasound?

OBJECTIVES

To test if a deep learning (DL) model trained on echocardiography images could accurately segment the left ventricle (LV) and predict ejection fraction on apical 4-chamber images acquired by point-of-care ultrasound (POCUS).

METHODS

We created a dataset of 333 videos from cardiac POCUS exams acquired in the emergency department. For each video we derived two ground-truth labels. First, we segmented the LV from one image frame and second, we classified the EF as normal, reduced, or severely reduced. We then classified the media's quality as optimal, adequate, or inadequate. With this dataset we tested the accuracy of automated LV segmentation and EF classification by the best-in-class echocardiography trained DL model EchoNet-Dynamic.

RESULTS

The mean Dice similarity coefficient for LV segmentation was 0.72 (N = 333; 95% CI 0.70-0.74). Cohen's kappa coefficient for agreement between predicted and ground-truth EF classification was 0.16 (N = 333). The area under the receiver-operating curve for the diagnosis of heart failure was 0.74 (N = 333). Model performance improved with video quality for the tasks of LV segmentation and diagnosis of heart failure, but was unchanged with EF classification. For all tasks the model was less accurate than the published benchmarks for EchoNet-Dynamic.

CONCLUSIONS

Performance of a DL model trained on formal echocardiography worsened when challenged with images captured during resuscitations. DL models intended for assessing bedside ultrasound should be trained on datasets composed of POCUS images. Such datasets have yet to be made publicly available.

Utility of Handheld Ultrasound Performed by Cardiology Fellows in Patients Presenting with Suspected ST-Elevation Myocardial Infarction

Background

In academic hospitals, cardiology fellows may be the first point of contact for patients presenting with suspected ST-elevation myocardial infarction (STEMI) or acute coronary syndrome (ACS). In this study, we sought to determine the role of handheld ultrasound (HHU) in patients with suspected acute myocardial injury (AMI) when used by fellows in training, its association with the year of training in cardiology fellowship, and its influence on clinical care.

Methods

This prospective study's sample population comprised patients who presented to the Loma Linda University Medical Center Emergency Department with suspected acute STEMI. On-call cardiology fellows performed bedside cardiac HHU at the time of AMI activation. All patients subsequently underwent standard transthoracic echocardiography (TTE). The impact of the detection of wall motion abnormalities (WMAs) on HHU in regard to clinical decision-making, including whether the patient would undergo urgent invasive angiography, was also evaluated.

Results

Eighty-two patients (mean age: 65 years, 70% male) were included. The use of HHU by cardiology fellows resulted in a concordance correlation coefficient of 0.71 (95% confidence interval: 0.58-0.81) between HHU and TTE for left ventricular ejection fraction (LVEF), and a concordance correlation coefficient of 0.76 (0.65-0.84) for wall motion score index. Patients with WMA on HHU were more likely to undergo invasive angiogram during hospitalization (96% vs. 75%, P < 0.01). The time interval between the performance of HHU to initiation of cardiac catheterization (time-to-cath) was shorter in patients with abnormal versus normal HHU examinations (58 ± 32 min vs. 218 ± 388 min, P = 0.06). Finally, among patients who underwent angiography, those with WMA were more likely to undergo angiography within 90 min of presentation (96% vs. 66%, P < 0.001).

Conclusion

HHU can be reliably used by cardiology fellows in training for measurement of LVEF and assessment of wall motion abnormalities, with good correlation to findings obtained via standard TTE. HHU-identified WMA at first contact was associated with higher rates of angiography as well as sooner angiography compared to patients without WMA.

Heart failure in the general population and impact of the 2021 European Society of Cardiology Heart Failure Guidelines

AIM

The diagnosis of heart failure (HF) has been refined in several steps in recent years, reflecting evolving diagnostic and therapeutic approaches. The European Society of Cardiology (ESC) recently published a modified definition of HF in the 2021 heart failure (HF) guidelines. The impact of this new diagnostic algorithm on the prevalence of HF is not known. The aim of this study was to describe the contemporary prevalence of HF in a representative, completely phenotyped sample from the general population.

METHODS AND RESULTS

This analysis was conducted among 7074 participants (aged 45-78 years, 51.5% women) from the population-based Hamburg City Health Study. Compared with the 2016 version, HF prevalence increased with the 2021 HF guidelines from 4.31% to 4.83% (12% increase). This increase was driven by a higher number of subjects with HF with reduced/mildly-reduced ejection fraction (0.47% to 0.52%; 1.37% to 2.12%), while the number of subjects with HF with preserved ejection fraction decreased from 2.46% to 2.19%. Importantly, this did not impact the known risk factor profiles of the phenotypes. Although four drugs are recommended for all subjects with HFrEF in the new guidelines, several adjunctive therapies are recommended for dedicated cases/scenarios (e.g. <1% eligibility for ivabradine/vericiguat/devices).

CONCLUSION

Heart failure remains common in a contemporary general population sample. The number of patients with HF will increase when the current diagnostic criteria are applied. This offers opportunities to initiate preventive therapies, especially in patients with HFmrEF and HFrEF.

Real-world performance, long-term efficacy, and absence of bias in the artificial intelligence enhanced electrocardiogram to detect left ventricular systolic dysfunction

Aims

Some artificial intelligence models applied in medical practice require ongoing retraining, introduce unintended racial bias, or have variable performance among different subgroups of patients. We assessed the real-world performance of the artificial intelligence-enhanced electrocardiogram to detect left ventricular systolic dysfunction with respect to multiple patient and electrocardiogram variables to determine the algorithm’s long-term efficacy and potential bias in the absence of retraining.

Methods and results

Electrocardiograms acquired in 2019 at Mayo Clinic in Minnesota, Arizona, and Florida with an echocardiogram performed within 14 days were analyzed (n = 44 986 unique patients). The area under the curve (AUC) was calculated to evaluate performance of the algorithm among age groups, racial and ethnic groups, patient encounter location, electrocardiogram features, and over time. The artificial intelligence-enhanced electrocardiogram to detect left ventricular systolic dysfunction had an AUC of 0.903 for the total cohort. Time series analysis of the model validated its temporal stability. Areas under the curve were similar for all racial and ethnic groups (0.90–0.92) with minimal performance difference between sexes. Patients with a ‘normal sinus rhythm’ electrocardiogram (n = 37 047) exhibited an AUC of 0.91. All other electrocardiogram features had areas under the curve between 0.79 and 0.91, with the lowest performance occurring in the left bundle branch block group (0.79).

Conclusion

The artificial intelligence-enhanced electrocardiogram to detect left ventricular systolic dysfunction is stable over time in the absence of retraining and robust with respect to multiple variables including time, patient race, and electrocardiogram features.

Maternal left ventricular function and adverse neonatal outcomes in women with cardiac disease

Purpose

To evaluate the relationship between maternal left ventricular systolic function, utero-placental circulation, and risk of adverse neonatal outcomes in women with cardiac disease.

Methods

119 women managed in the pregnancy heart clinic (2019–2021) were identified. Women were classified by their primary cardiac condition. Adverse neonatal outcomes were: low birth weight (< 2500 g), small-for-gestational-age (< 10th birth-weight centile), pre-term delivery (< 37 weeks’ gestation), and fetal demise (> 20 weeks’ gestation). Parameters of left ventricular systolic function (global longitudinal strain, radial strain, ejection fraction, average S’, and cardiac output) were calculated and pulsatility index was recorded from last growth scan.

Results

Adverse neonatal outcomes occurred in 28 neonates (24%); most frequently in valvular heart disease (n = 8) and cardiomyopathy (n = 7). Small-for-gestational-age neonates were most common in women with cardiomyopathy (p = 0.016). Early pregnancy average S’ (p = 0.03), late pregnancy average S’ (p = 0.02), and late pregnancy cardiac output (p = 0.008) were significantly lower in women with adverse neonatal outcomes than in those with healthy neonates. There was a significant association between neonatal birth-weight centile and global longitudinal strain (p = 0.04) and cardiac output (p = 0.0002) in late pregnancy. Pulsatility index was highest in women with cardiomyopathy (p = 0.007), and correlated with average S’ (p < 0.0001) and global longitudinal strain (p = 0.03) in late pregnancy.

Conclusion

Women with cardiac disease may not tolerate cardiovascular adaptations required during pregnancy to support fetal growth. Adverse neonatal outcomes were associated with reduced left ventricular systolic function and higher pulsatility index. The association between impaired systolic function and reduced fetal growth is supported by insufficient utero-placental circulation.

Automated Endocardial Border Detection and Left Ventricular Functional Assessment in Echocardiography Using Deep Learning

Endocardial border detection is a key step in assessing left ventricular systolic function in echocardiography. However, this process is still not sufficiently accurate, and manual retracing is often required, causing time-consuming and intra-/inter-observer variability in clinical practice. To address these clinical issues, more accurate and normalized automatic endocardial border detection would be valuable. Here, we develop a deep learning-based method for automated endocardial border detection and left ventricular functional assessment in two-dimensional echocardiographic videos. First, segmentation of the left ventricular cavity was performed in the six representative projections for a cardiac cycle. We employed four segmentation methods: U-Net, UNet++, UNet3+, and Deep Residual U-Net. UNet++ and UNet3+ showed a sufficiently high performance in the mean value of intersection over union and Dice coefficient. The accuracy of the four segmentation methods was then evaluated by calculating the mean value for the estimation error of the echocardiographic indexes. UNet++ was superior to the other segmentation methods, with the acceptable mean estimation error of the left ventricular ejection fraction of 10.8%, global longitudinal strain of 8.5%, and global circumferential strain of 5.8%, respectively. Our method using UNet++ demonstrated the best performance. This method may potentially support examiners and improve the workflow in echocardiography.

Echo-SyncNet: Self-Supervised Cardiac View Synchronization in Echocardiography

In echocardiography (echo), an electrocardiogram (ECG) is conventionally used to temporally align different cardiac views for assessing critical measurements. However, in emergencies or point-of-care situations, acquiring an ECG is often not an option, hence motivating the need for alternative temporal synchronization methods. Here, we propose Echo-SyncNet, a self-supervised learning framework to synchronize various cross-sectional 2D echo series without any human supervision or external inputs. The proposed framework takes advantage of two types of supervisory signals derived from the input data: spatiotemporal patterns found between the frames of a single cine (intra-view self-supervision) and interdependencies between multiple cines (inter-view self-supervision). The combined supervisory signals are used to learn a feature-rich and low dimensional embedding space where multiple echo cines can be temporally synchronized. Two intra-view self-supervisions are used, the first is based on the information encoded by the temporal ordering of a cine (temporal intra-view) and the second on the spatial similarities between nearby frames (spatial intra-view). The inter-view self-supervision is used to promote the learning of similar embeddings for frames captured from the same cardiac phase in different echo views. We evaluate the framework with multiple experiments: 1) Using data from 998 patients, Echo-SyncNet shows promising results for synchronizing Apical 2 chamber and Apical 4 chamber cardiac views, which are acquired spatially perpendicular to each other; 2) Using data from 3070 patients, our experiments reveal that the learned representations of Echo-SyncNet outperform a supervised deep learning method that is optimized for automatic detection of fine-grained cardiac cycle phase; 3) We go one step further and show the usefulness of the learned representations in a one-shot learning scenario of cardiac key-frame detection. Without any fine-tuning, key frames in 1188 validation patient studies are identified by synchronizing them with only one labeled reference cine. We do not make any prior assumption about what specific cardiac views are used for training, and hence we show that Echo-SyncNet can accurately generalize to views not present in its training set. Project repository: github.com/fatemehtd/Echo-SyncNet>.

The impact of a perceptual learning module on novices' ability to visually estimate left ventricular ejection fraction by transesophageal echocardiography: a randomized controlled study

PURPOSE

Echocardiography is a difficult tool to master. Competency requires the supervised interpretation of hundreds of exams. Perceptual learning modules (PLMs) are novel learning tools that aim to speed up this learning process by enabling learners to go online and interpret numerous clinical images, followed systematically by expert feedback. We developed and tested a PLM aimed at improving novices' ability to quickly visually estimate left ventricular ejection fraction (LVEF) on transesophageal echocardiography images, a critical skill in acute care. We hypothesized that using the PLM would improve the accuracy and the speed of learners' estimations.

METHODS

Learners without echocardiography experience were randomly assigned to a group that used the 96-case PLM (n = 26) or a control group (n = 26) that did not. Both groups took a pre-test and an immediate post-test that measured the accuracy of their visual estimations during a first session. At six months, participants also completed a delayed post-test.

RESULTS

In the immediate post-test, the PLM group showed significantly better accuracy than the control group (median absolute estimation error 6.1 vs 9.0; difference 95% CI, 1.0 to 4.6; P < 0.001). Nevertheless, at six months, estimation errors were similar in both groups (median absolute estimation error 10.0 vs 10.0; difference 95% CI, -1.3 to 2.1; P = 0.27).

CONCLUSIONS

Participation in a short online PLM significantly improved novices' short-term ability to accurately estimate LVEF visually, compared with controls. The effect was not sustained at six months.

TRIAL REGISTRATION

www.clinicaltrials.gov (NCT03245567); registered 7 August 2017.

The relation between cardiac 123I-mIBG scintigraphy and functional response 1 year after CRT implantation

Aims 

Cardiac resynchronization therapy (CRT) is a disease-modifying therapy in patients with chronic heart failure (CHF). Current guidelines ascribe CRT eligibility on three parameters only: left ventricular ejection fraction (LVEF), QRS duration, and New York Heart Association (NYHA) functional class. However, one-third of CHF patients does not benefit from CRT. This study evaluated whether ¹²³I-meta-iodobenzylguanidine (¹²³I-mIBG) assessed cardiac sympathetic activity could optimize CRT patient selection.

Methods and results 

A total of 78 stable CHF subjects (age 66.8 ± 9.6 years, 73% male, LVEF 25.2 ± 6.7%, QRS duration 153 ± 23 ms, NYHA 2.2 ± 0.7) referred for CRT implantation were enrolled. Subjects underwent ¹²³I-mIBG scintigraphy prior to implantation. Early and late heart-to-mediastinum (H/M) ratio and ¹²³I-mIBG washout were calculated. CRT response was defined as either an increase of LVEF to >35%, any improvement in LVEF of >10%, QRS shortening to <150 ms, or improvement in NYHA class of >1 class. In 33 patients LVEF increased to >35%, QRS decreased <150 ms in 36 patients, and NYHA class decreased in 33 patients. Late H/M ratio and hypertension were independent predictors of LVEF improvement to >35% (P = 0.0014 and P = 0.0149, respectively). In addition, early H/M ratio, LVEF, and absence of diabetes mellitus (DM) were independent predictors for LVEF improvement by >10%. No independent predictors were found for QRS shortening to <150 ms or improvement in NYHA class.

Conclusion 

Early and late H/M ratio were independent predictors of CRT response when improvement of LVEF was used as measure of response. Therefore, cardiac ¹²³I-mIBG scintigraphy may be used as a tool to optimize selection of subjects that might benefit from CRT.

Impact of catheter ablation for atrial fibrillation in patients with heart failure and left ventricular systolic dysfunction

INTRODUCTION AND AIMS

Catheter ablation has been shown to improve left ventricular (LV) ejection fraction (LVEF) in patients with atrial fibrillation (AF) and heart failure (HF). Our aim was to assess the impact of AF ablation on the outcome of patients with HF and LV systolic dysfunction.

METHODS

We performed a retrospective observational cohort study of all patients with HF and LVEF <50% and with no apparent cause for systolic dysfunction other than AF who underwent catheter ablation in a tertiary referral center between July 2016 and November 2018. The primary endpoint was a ≥5% improvement in LVEF. Secondary endpoints included improvement in New York Heart Association (NYHA) class and reduction in LV end-diastolic diameter (LVEDD) and left atrial diameter (LAD).

RESULTS

Of 153 patients who underwent AF ablation in this period, 22 (77% male, median age 61 [IQR 54-64] years) fulfilled the inclusion criteria. Median follow-up was 11.1 months (IQR 6.1-19.0). After ablation, median LVEF increased from 40% (IQR 33-41) to 58% (IQR 55-62) (p<0.01), mean NYHA class improved from 2.35±0.49 to 1.3±0.47 (p<0.001), and median LAD and LVEDD decreased from 48.0 (IQR 43.5-51.5) mm to 44 (IQR 40-49) mm (p<0.01) and from 61.0 (IQR 54.0-64.8) mm to 55.0 (52.2-58.0) mm (p<0.01), respectively.

CONCLUSION

In patients with HF and LV systolic dysfunction, AF ablation is associated not only with improved functional status but also with favorable structural remodeling, including improvement in LVEF and decreases in LAD and LVEDD.

A machine learning algorithm supports ultrasound-naïve novices in the acquisition of diagnostic echocardiography loops and provides accurate estimation of LVEF

Left ventricular ejection fraction (LVEF) is the most important parameter in the assessment of cardiac function. A machine-learning algorithm was trained to guide ultrasound-novices to acquire diagnostic echocardiography images. The artificial intelligence (AI) algorithm then estimates LVEF from the captured apical-4-chamber (AP4), apical-2-chamber (AP2), and parasternal-long-axis (PLAX) loops. We sought to test this algorithm by having first-year medical students without previous ultrasound knowledge scan real patients. Nineteen echo-naïve first-year medical students were trained in the basics of echocardiography by a 2.5 h online video tutorial. Each student then scanned three patients with the help of the AI. Image quality was graded according to the American College of Emergency Physicians scale. If rated as diagnostic quality, the AI calculated LVEF from the acquired loops (monoplane and also a "best-LVEF" considering all views acquired in the particular patient). These LVEF calculations were compared to images of the same patients captured and read by three experts (ground-truth LVEF [GT-EF]). The novices acquired diagnostic-quality images in 33/57 (58%), 49/57 (86%), and 39/57 (68%) patients in the PLAX, AP4, and AP2, respectively. At least one of the three views was obtained in 91% of the attempts. We found an excellent agreement between the machine's LVEF calculations from images acquired by the novices with the GT-EF (bias of 3.5% ± 5.6 and r = 0.92, p < 0.001 in the "best-LVEF" algorithm). This pilot study shows first evidence that a machine-learning algorithm can guide ultrasound-novices to acquire diagnostic echo loops and provide an automated LVEF calculation that is in agreement with a human expert.

Point-of-Care Echocardiography and Hemodynamic Monitoring in Cirrhosis and Acute-on-Chronic Liver Failure in the COVID-19 Era

Point-of-Care (POC) transthoracic echocardiography (TTE) is transforming the management of patients with cirrhosis presenting with septic shock, acute kidney injury, hepatorenal syndrome and acute-on-chronic liver failure (ACLF) by correctly assessing the hemodynamic and volume status at the bedside using combined echocardiography and POC ultrasound (POCUS). When POC TTE is performed by the hepatologist or intensivist in the intensive care unit (ICU), and interpreted remotely by a cardiologist, it can rule out cardiovascular conditions that may be contributing to undifferentiated shock, such as diastolic dysfunction, myocardial infarction, myocarditis, regional wall motion abnormalities and pulmonary embolism. The COVID-19 pandemic has led to a delay in seeking medical treatment, reduced invasive interventions and deferment in referrals leading to "collateral damage" in critically ill patients with liver disease. Thus, the use of telemedicine in the ICU (Tele-ICU) has integrated cardiology, intensive care, and hepatology practices across the spectrum of ICU, operating room, and transplant healthcare. Telecardiology tools have improved bedside diagnosis when introduced as part of COVID-19 care by remote supervision and interpretation of POCUS and echocardiographic data. In this review, we present the contemporary approach of using POC echocardiography and offer a practical guide for primary care hepatologists and gastroenterologists for cardiac assessment in critically ill patients with cirrhosis and ACLF. Evidenced based use of Tele-ICU can prevent delay in cardiac diagnosis, optimize safe use of expert resources and ensure timely care in the setting of critically ill cirrhosis, ACLF and liver transplantation in the COVID-19 era.

Artificial Intelligence, Data Sensors and Interconnectivity: Future Opportunities for Heart Failure

A higher proportion of patients with heart failure have benefitted from a wide and expanding variety of sensor-enabled implantable devices than any other patient group. These patients can now also take advantage of the ever-increasing availability and affordability of consumer electronics. Wearable, on- and near-body sensor technologies, much like implantable devices, generate massive amounts of data. The connectivity of all these devices has created opportunities for pooling data from multiple sensors – so-called interconnectivity – and for artificial intelligence to provide new diagnostic, triage, risk-stratification and disease management insights for the delivery of better, more personalised and cost-effective healthcare. Artificial intelligence is also bringing important and previously inaccessible insights from our conventional cardiac investigations. The aim of this article is to review the convergence of artificial intelligence, sensor technologies and interconnectivity and the way in which this combination is set to change the care of patients with heart failure.

Video-based AI for beat-to-beat assessment of cardiac function

Accurate assessment of cardiac function is crucial for the diagnosis of cardiovascular disease1, screening for cardiotoxicity2 and decisions regarding the clinical management of patients with a critical illness3. However, human assessment of cardiac function focuses on a limited sampling of cardiac cycles and has considerable inter-observer variability despite years of training4,5. Here, to overcome this challenge, we present a video-based deep learning algorithm-EchoNet-Dynamic-that surpasses the performance of human experts in the critical tasks of segmenting the left ventricle, estimating ejection fraction and assessing cardiomyopathy. Trained on echocardiogram videos, our model accurately segments the left ventricle with a Dice similarity coefficient of 0.92, predicts ejection fraction with a mean absolute error of 4.1% and reliably classifies heart failure with reduced ejection fraction (area under the curve of 0.97). In an external dataset from another healthcare system, EchoNet-Dynamic predicts the ejection fraction with a mean absolute error of 6.0% and classifies heart failure with reduced ejection fraction with an area under the curve of 0.96. Prospective evaluation with repeated human measurements confirms that the model has variance that is comparable to or less than that of human experts. By leveraging information across multiple cardiac cycles, our model can rapidly identify subtle changes in ejection fraction, is more reproducible than human evaluation and lays the foundation for precise diagnosis of cardiovascular disease in real time. As a resource to promote further innovation, we also make publicly available a large dataset of 10,030 annotated echocardiogram videos.

Tool to improve qualitative assessment of left ventricular systolic function

Interactive online learning tools have revolutionized graduate medical education and can impart echocardiographic image interpretive skills. We created self-paced, interactive online training modules using a repository of echocardiography videos of normal and various degrees of abnormal left ventricles. In this study, we tested the feasibility of this learning tool. Thirteen anesthesia interns took a pre-test and then had 3 weeks to complete the training modules on their own time before taking a post-test. The average score on the post-test (74.6% ± 11.08%) was higher than the average score on the pre-test (57.7% ± 9.27%) (P < 0.001). Scores did not differ between extreme function (severe dysfunction or hyperdynamic function) and non-extreme function (normal function or mild or moderate dysfunction) questions on both the pre-test (P = 0.278) and post-test (P = 0.093). The interns scored higher on the post-test than the pre-test on both extreme (P = 0.0062) and non-extreme (P = 0.0083) questions. After using an online educational tool that allowed learning on their own time and pace, trainees improved their ability to correctly categorize left ventricular systolic function. Left ventricular systolic function is often a key echocardiographic question that can be difficult to master. The promising performance of this educational resource may lead to more time- and cost-effective methods for improving diagnostic accuracy among learners.

A Teaching Intervention Increases the Performance of Handheld Ultrasound Devices for Assessment of Left Ventricular Ejection Fraction

Background

Few studies have demonstrated the utility of a teaching program for evaluation of left ventricular ejection fraction (LVEF) of echocardiographic images acquired with high-end machines. No study to date explored the value of similar programs when a handheld ultrasound device is used. The aim of this study was to determine whether a teaching intervention could improve the accuracy and the reliability of LVEF visual assessment of echocardiographic images acquired with HUD.

Materials and Methods

Twenty echocardiograms acquired with a hand-held ultrasound device with a spectrum of LVEF were presented to 26 participants with varying experience in echocardiography (range 2-12 years) for single-point LVEF visual estimates. After this baseline assessment, participants underwent three training sessions which included analysis of the individual baseline results and review and interpretation of additional 60 cases from the same platform. After 2 months, 20 new echocardiograms were presented to the same 26 participants for visual LVEF assessment. For each participant, the visual LVEF for each case was compared with the reference LVEF (quantitative measurements by experts), and a difference of > ±5% was considered a misclassification.

Results

The misclassification rate was 61% preintervention and decreased to 41% after intervention (P < 0.0001). The mean absolute differences in LVEF between visual estimates and reference before and after intervention for all readers were -7.9 ± 9.6 and -1.2 ± 7.8, respectively (P < 0.0001). Inter-rater repeatability analysis was performed using the intraclass correlation coefficient. The intraclass correlation coefficient for inter-rater reliability was fair preintervention (0.65, 95% confidence interval [CI] 0.59 0.71) and good after intervention (0.80, 95% CI 0.73 0.87), and there were no differences when categorized according to the level of experience.

Conclusions

A teaching intervention can improve the accuracy and the reliability in the visual LVEF assessment of images acquired with handheld ultrasound device.

Use of the Valve Visualization on Echocardiography Grade Tool Improves Sensitivity and Negative Predictive Value of Transthoracic Echocardiogram for Exclusion of Native Valvular Vegetation

BACKGROUND

Transesophageal echocardiography (TEE) remains the preferred test to rule out infective endocarditis (IE) but is resource intensive and carries risk. Multiple studies report low sensitivity of transthoracic echocardiography (TTE) for detection of IE; however, these studies did not account for TTE quality. We test the validity of a simple valve visualization grading tool to stratify TTEs by quality and determine whether a high-quality TTE may be used to exclude valvular vegetation and forgo the need for TEE.

METHODS

The Valve Visualization on Echocardiography Grade (VEG) tool scores the TTE from 0 to 10 based on leaflet visualization and valve leaflet clarity. The tool was retrospectively applied to 309 sequential patients who underwent both TTE and TEE at an academic teaching hospital between 2011 and 2015. The TEE report was the gold standard for presence or absence of vegetation. Patients with prosthetic valves and pacemaker wires were excluded. Sensitivity of TTE for detecting vegetation was calculated at each VEG score, and the optimal cutoff was identified.

RESULTS

A total of 309 patients were included in the analysis. Among the 216 negative TTEs, 19 (9%) had a positive TEE. The median VEG score was 4. A VEG score cutoff >6 provided optimal sensitivity and was used as the cutoff. Overall, 75 (25%) patients had a VEG score >6, and 234 (75%) had a score ≤6. Sensitivity and negative predictive value for IE were higher in the VEG >6 versus VEG ≤6 group (sensitivity 96% vs 66%, negative predictive value 97.5% vs 90%; P < .05). The false-negative rate was lower (2.5% vs 10%; P = .04) in VEG > 6 versus VEG ≤ 6 groups, respectively.

CONCLUSIONS

Leaflet visualization and valve leaflet clarity are important components in the TTE evaluation of patients with suspected IE. This study demonstrates that the better the valve leaflets are visualized on TTE (as represented in this population by a score >6), the higher the confidence one can have that the TTE will not be falsely negative for vegetation(s) when vegetation(s) are not noted on these TTEs. If validated in future prospective studies, this may reduce the need to perform an invasive TEE in selected patients undergoing evaluation for native valve IE.