Local large language models for privacy-preserving accelerated review of historic echocardiogram reports.

OBJECTIVES: The study developed framework that leverages an open-source Large Language Model (LLM) to enable clinicians to ask plain-language questions about a patient's entire echocardiogram report history. This approach is intended to streamline the extraction of clinical insights from multiple echocardiogram reports, particularly in patients with complex cardiac diseases, thereby enhancing both patient care and research efficiency. MATERIALS AND METHODS: Data from over 10 years were collected, comprising echocardiogram reports from patients with more than 10 echocardiograms on file at the Mount Sinai Health System. These reports were converted into a single document per patient for analysis, broken down into snippets and relevant snippets were retrieved using text similarity measures. The LLaMA-2 70B model was employed for analyzing the text using a specially crafted prompt. The model's performance was evaluated against ground-truth answers created by faculty cardiologists. RESULTS: The study analyzed 432 reports from 37 patients for a total of 100 question-answer pairs. The LLM correctly answered 90% questions, with accuracies of 83% for temporality, 93% for severity assessment, 84% for intervention identification, and 100% for diagnosis retrieval. Errors mainly stemmed from the LLM's inherent limitations, such as misinterpreting numbers or hallucinations. CONCLUSION: The study demonstrates the feasibility and effectiveness of using a local, open-source LLM for querying and interpreting echocardiogram report data. This approach offers a significant improvement over traditional keyword-based searches, enabling more contextually relevant and semantically accurate responses; in turn showing promise in enhancing clinical decision-making and research by facilitating more efficient access to complex patient data.

Clinical outcomes of intravenous iron therapy in patients with heart failure and iron deficiency: Meta-analysis and trial sequential analysis of randomized clinical trials.

BACKGROUND: Iron deficiency in patients with heart failure (HF) is underdiagnosed and undertreated. The role of intravenous (IV) iron is well-established to improve quality of life measures. Emerging evidence also supports its role in preventing cardiovascular events in patients with HF. METHODOLOGY: We conducted a literature search of multiple electronic databases. Randomized controlled trials that compared IV iron to usual care among patients with HF and reported cardiovascular (CV) outcomes were included. Primary outcome was the composite of first heart failure hospitalization (HFH) or CV death. Secondary outcomes included HFH (first or recurrent), CV death, all-cause mortality, hospitalization for any cause, gastrointestinal (GI) side effects, or any infection. We performed trial sequential and cumulative meta-analyses to evaluate the effect of IV iron on the primary endpoint, and on HFH. RESULTS: Nine trials enrolling 3337 patients were included. Adding IV iron to usual care significantly reduced the risk of first HFH or CV death [risk ratio (RR) 0.84; 95 % confidence interval (CI) 0.75-0.93; I2 = 0 %; number needed to treat (NNT) 18], which was primarily driven by a reduction in the risk of HFH of 25 %. IV iron also reduced the risk of the composite of hospitalization for any cause or death (RR 0.92; 95 % CI 0.85-0.99; I2 = 0 %; NNT 19). There was no significant difference in the risk of CV death, all-cause mortality, adverse GI events, or any infection among patients receiving IV iron compared to usual care. The observed benefits of IV iron were directionally consistent across trials and crossed both the statistical and trial sequential boundaries of benefit. CONCLUSION: In patients with HF and iron deficiency, the addition of IV iron to usual care reduces the risk of HFH without affecting the risk of CV or all-cause mortality.

Tricuspid Regurgitation: From Imaging to Clinical Trials to Resolving the Unmet Need for Treatment.

Tricuspid regurgitation (TR) is a highly prevalent and heterogeneous valvular disease, independently associated with excess mortality and high morbidity in all clinical contexts. TR is profoundly undertreated by surgery and is often discovered late in patients presenting with right-sided heart failure. To address the issue of undertreatment and poor clinical outcomes without intervention, numerous structural tricuspid interventional devices have been and are in development, a challenging process due to the unique anatomic and physiological characteristics of the tricuspid valve, and warranting well-designed clinical trials. The path from routine practice TR detection to appropriate TR evaluation, to conduction of clinical trials, to enriched therapeutic possibilities for improving TR access to treatment and outcomes in routine practice is complex. Therefore, this paper summarizes the key points and methods crucial to TR detection, quantitation, categorization, risk-scoring, intervention-monitoring, and outcomes evaluation, particularly of right-sided function, and to clinical trial development and conduct, for both interventional and surgical groups.

Implications of the Use of Artificial Intelligence Predictive Models in Health Care Settings : A Simulation Study.

BACKGROUND: Substantial effort has been directed toward demonstrating uses of predictive models in health care. However, implementation of these models into clinical practice may influence patient outcomes, which in turn are captured in electronic health record data. As a result, deployed models may affect the predictive ability of current and future models. OBJECTIVE: To estimate changes in predictive model performance with use through 3 common scenarios: model retraining, sequentially implementing 1 model after another, and intervening in response to a model when 2 are simultaneously implemented. DESIGN: Simulation of model implementation and use in critical care settings at various levels of intervention effectiveness and clinician adherence. Models were either trained or retrained after simulated implementation. SETTING: Admissions to the intensive care unit (ICU) at Mount Sinai Health System (New York, New York) and Beth Israel Deaconess Medical Center (Boston, Massachusetts). PATIENTS: 130 000 critical care admissions across both health systems. INTERVENTION: Across 3 scenarios, interventions were simulated at varying levels of clinician adherence and effectiveness. MEASUREMENTS: Statistical measures of performance, including threshold-independent (area under the curve) and threshold-dependent measures. RESULTS: At fixed 90% sensitivity, in scenario 1 a mortality prediction model lost 9% to 39% specificity after retraining once and in scenario 2 a mortality prediction model lost 8% to 15% specificity when created after the implementation of an acute kidney injury (AKI) prediction model; in scenario 3, models for AKI and mortality prediction implemented simultaneously, each led to reduced effective accuracy of the other by 1% to 28%. LIMITATIONS: In real-world practice, the effectiveness of and adherence to model-based recommendations are rarely known in advance. Only binary classifiers for tabular ICU admissions data were simulated. CONCLUSION: In simulated ICU settings, a universally effective model-updating approach for maintaining model performance does not seem to exist. Model use may have to be recorded to maintain viability of predictive modeling. PRIMARY FUNDING SOURCE: National Center for Advancing Translational Sciences.

A foundational vision transformer improves diagnostic performance for electrocardiograms.

The electrocardiogram (ECG) is a ubiquitous diagnostic modality. Convolutional neural networks (CNNs) applied towards ECG analysis require large sample sizes, and transfer learning approaches for biomedical problems may result in suboptimal performance when pre-training is done on natural images. We leveraged masked image modeling to create a vision-based transformer model, HeartBEiT, for electrocardiogram waveform analysis. We pre-trained this model on 8.5 million ECGs and then compared performance vs. standard CNN architectures for diagnosis of hypertrophic cardiomyopathy, low left ventricular ejection fraction and ST elevation myocardial infarction using differing training sample sizes and independent validation datasets. We find that HeartBEiT has significantly higher performance at lower sample sizes compared to other models. We also find that HeartBEiT improves explainability of diagnosis by highlighting biologically relevant regions of the EKG vs. standard CNNs. Domain specific pre-trained transformer models may exceed the classification performance of models trained on natural images especially in very low data regimes. The combination of the architecture and such pre-training allows for more accurate, granular explainability of model predictions.

Part 2: The Clinical Applications of Left Ventricular Myocardial Strain.

There is an increasing prevalence of cardiovascular disease and heart failure. Indices of left ventricular (LV) systolic function such as LV ejection fraction used to identify those at risk of adverse cardiac events such as heart failure may not be truly representative of LV systolic function in certain cardiac diseases. Given that LV ejection fraction reduction may represent more advanced irreversible stages of disease, measures of myocardial strain have emerged as a feasible and robust instrument for the early identification of heart disease and subtle LV systolic dysfunction. The purpose of this review was to provide an overview of emerging clinical applications of LV global longitudinal strain in valvular and cardiomyopathic diseases and coronavirus disease 2019.

Part 1: The Clinical Applications of Left Ventricular Myocardial Strain.

There is an increasing prevalence of cardiovascular disease and heart failure. Indices of left ventricular (LV) systolic function such as LV ejection fraction used to identify those at risk of adverse cardiac events such as heart failure may not be truly representative of LV systolic function in certain cardiac diseases. Given that LV ejection fraction reduction may represent more advanced irreversible stages of disease, measures of myocardial strain have emerged as a feasible and robust instrument for the early identification of heart disease and subtle LV systolic dysfunction. The purpose of this review was to provide an overview of myocardial strain concepts and emerging clinical applications of global longitudinal strain in cardio-oncology.

Uncommon spectral Doppler findings in patients with heart failure.

Spectral Doppler examination is necessary for full hemodynamic assessment of patients with systolic heart failure. It is fully incorporated into comprehensive echocardiographic examination. In this manuscript, we describe two uncommon findings in patients with established severe left ventricular systolic dysfunction: notched aortic regurgitation and merged mitral regurgitation.

Prognostic significance of incorporating exercise tissue doppler mitral annular early diastolic velocity in exercise diastolic dysfunction assessment.

BACKGROUND: The existing algorithm for defining exercise-induced diastolic dysfunction incorporates resting e' velocity as a surrogate of myocardial relaxation. The additive prognostic value of incorporating post-exercise e' velocity in definition of exercise-induced diastolic dysfunction is poorly studied. AIM: To define the additive prognostic value of post-exercise e' septal velocity in the assessment of exercise-induced diastolic dysfunction compared to the traditional approach. METHODS: This retrospective study included 1409 patients undergoing exercise treadmill echocardiography with available full set of diastolic variables. Doppler measures of diastolic function included resting septal e' velocity, post-exercise septal e' velocity, post-exercise E/e' ratio, and post-exercise tricuspid regurgitant jet velocity. Approaches incorporating resting septal e' velocity and post-exercise septal e' velocity were compared in defining exercise-induced diastolic dysfunction, and for association with adverse cardiovascular outcomes. RESULTS: The mean age of study subjects was 56.3 ± 16.5 years and 791 (56%) patients were women. A total of 524 patients had disagreement between resting and post exercise septal e' velocities, and these values showed only weak agreement (kappa statistics: .28, P = .02). All categories of the traditional exercise-induced DD approach incorporating resting septal e' velocity witnessed reclassification when exercise septal e' velocity was used. When both approaches were compared, increased event rates were only evident when both approaches agreed on exercise-induced diastolic dysfunction (HR: 1.92, P < .001, 95% CI: 1.37-2.69). This association persisted after multivariable adjustment and propensity score matching for covariates. CONCLUSION: Incorporation of post-exercise e' velocity into the set of variables defining exercise-induced diastolic dysfunction can improve the prognostic utility of diastolic function assessment.

Deep Learning for Echocardiography: Introduction for Clinicians and Future Vision: State-of-the-Art Review.

Exponential growth in data storage and computational power is rapidly narrowing the gap between translating findings from advanced clinical informatics into cardiovascular clinical practice. Specifically, cardiovascular imaging has the distinct advantage in providing a great quantity of data for potentially rich insights, but nuanced interpretation requires a high-level skillset that few individuals possess. A subset of machine learning, deep learning (DL), is a modality that has shown promise, particularly in the areas of image recognition, computer vision, and video classification. Due to a low signal-to-noise ratio, echocardiographic data tend to be challenging to classify; however, utilization of robust DL architectures may help clinicians and researchers automate conventional human tasks and catalyze the extraction of clinically useful data from the petabytes of collected imaging data. The promise is extending far and beyond towards a contactless echocardiographic exam-a dream that is much needed in this time of uncertainty and social distancing brought on by a stunning pandemic culture. In the current review, we discuss state-of-the-art DL techniques and architectures that can be used for image and video classification, and future directions in echocardiographic research in the current era.

Multi-center retrospective cohort study applying deep learning to electrocardiograms to identify left heart valvular dysfunction.

BACKGROUND: Aortic Stenosis and Mitral Regurgitation are common valvular conditions representing a hidden burden of disease within the population. The aim of this study was to develop and validate deep learning-based screening and diagnostic tools that can help guide clinical decision making. METHODS: In this multi-center retrospective cohort study, we acquired Transthoracic Echocardiogram reports from five Mount Sinai hospitals within New York City representing a demographically diverse cohort of patients. We developed a Natural Language Processing pipeline to extract ground-truth labels about valvular status and paired these to Electrocardiograms (ECGs). We developed and externally validated deep learning models capable of detecting valvular disease, in addition to considering scenarios of clinical deployment. RESULTS: We use 617,338 ECGs paired to transthoracic echocardiograms from 123,096 patients to develop a deep learning model for detection of Mitral Regurgitation. Area Under Receiver Operating Characteristic curve (AUROC) is 0.88 (95% CI:0.88-0.89) in internal testing, and 0.81 (95% CI:0.80-0.82) in external validation. To develop a model for detection of Aortic Stenosis, we use 617,338 Echo-ECG pairs for 128,628 patients. AUROC is 0.89 (95% CI: 0.88-0.89) in internal testing, going to 0.86 (95% CI: 0.85-0.87) in external validation. The model's performance increases leading up to the time of the diagnostic echo, and it performs well in validation against requirement of Transcatheter Aortic Valve Replacement procedures. CONCLUSIONS: Deep learning based tools can increase the amount of information extracted from ubiquitous investigations such as the ECG. Such tools are inexpensive, can help in earlier disease detection, and potentially improve prognosis.

The valves of the heart have flaps that open and close when the heart beats to maintain the flow of blood in the correct direction. Valvular disease, such as backflow or narrowing, puts additional strain upon heart muscles which can lead to heart failure. Usually, these conditions are diagnosed by doing an echocardiogram, an ultrasound scan of the heart and nearby blood vessels. The electrocardiogram (ECG) records the electrical signal generated by the heart and can be obtained more easily. We used deep learning neural networks, self-learning computer algorithms which excel at finding patterns within complex data. This enabled us to develop computer software able to diagnose valvular disease from ECGs. Earlier detection of such disease can help in improving overall outcome, while also reducing costs related to treatment.

The contribution of amyloid deposition in the aortic valve to calcification and aortic stenosis.

Calcific aortic valve disease (CAVD) and stenosis have a complex pathogenesis, and no therapies are available that can halt or slow their progression. Several studies have shown the presence of apolipoprotein-related amyloid deposits in close proximity to calcified areas in diseased aortic valves. In this Perspective, we explore a possible relationship between amyloid deposits, calcification and the development of aortic valve stenosis. These amyloid deposits might contribute to the amplification of the inflammatory cycle in the aortic valve, including extracellular matrix remodelling and myofibroblast and osteoblast-like cell proliferation. Further investigation in this area is needed to characterize the amyloid deposits associated with CAVD, which could allow the use of antisense oligonucleotides and/or isotype gene therapies for the prevention and/or treatment of CAVD.

Acute effect of edge-to-edge repair of mitral regurgitation on left heart mechanics and health status.

AIMS: Examine the impact of acute changes in left heart strain and volumes with percutaneous edge-to-edge MitraClip repair on improvement in health status assessed using Kansas City Cardiomyopathy Questionnaire-12 (KCCQ-12) score. METHODS: Changes in left atrial strain, left ventricular (LV) global longitudinal strain (LVGLS), LV end-systolic volume (LVESV), and end-diastolic volume (LVEDV) were evaluated in 50 patients undergoing MitraClip repair for symptomatic primary mitral regurgitation (PMR) and secondary mitral regurgitation (SMR) on transthoracic echocardiography before and 1 month after MitraClip. Multivariable regression was used to evaluate changes in left heart strain and volumes as predictors of change in KCCQ-12 scores, adjusting for baseline clinical and echocardiographic characteristics. RESULTS: Both PMR and SMR patients had significant increase in LVGLS and reduction in LVEDV and LVESV ( P  < 0.05) after MitraClip, reduction trend in left atrial conduit strain (PMR P  = 0.053; SMR P  = 0.12) but no significant change in LV ejection fraction. KCCQ-12 score improved significantly in both PMR ( P  < 0.001) and SMR cohorts ( P  < 0.001). Higher delta KCCQ-12 tertiles were associated with greater reduction in LVEDV ( P  = 0.022) after MitraClip. On multiple regression analysis, lower preprocedural Society of Thoracic Surgeons for Mitral Valve Replacement and KCCQ-12 score, and greater reduction in LVESV and left atrial strain conduit phase were associated with KCCQ-12 score improvement ( P  < 0.001). CONCLUSION: There is a significant increase in LVGLS and reduction in LVEDV, LVESV and left atrial strain conduit after edge-to-edge MitraClip repair in both PMR and SMR. Lower preprocedural Society of Thoracic Surgeons for Mitral Valve Replacement and KCCQ-12 score, and greater reduction in LVESV and left atrial conduit strain were associated with KCCQ-12 score improvement after MitraClip. Further studies are warranted to understand the mechanism and significance of our findings.

Using Shapes of COVID-19 Positive Patient-Specific Trajectories for Mortality Prediction.

Machine learning can be used to identify relevant trajectory shape features for improved predictive risk modeling, which can help inform decisions for individualized patient management in intensive care during COVID-19 outbreaks. We present explainable random forests to dynamically predict next day mortality risk in COVID -19 positive and negative patients admitted to the Mount Sinai Health System between March 1st and June 8th, 2020 using patient time-series data of vitals, blood and other laboratory measurements from the previous 7 days. Three different models were assessed by using time series with: 1) most recent patient measurements, 2) summary statistics of trajectories (min/max/median/first/last/count), and 3) coefficients of fitted cubic splines to trajectories. AUROC and AUPRC with cross-validation were used to compare models. We found that the second and third models performed statistically significantly better than the first model. Model interpretations are provided at patient-specific level to inform resource allocation and patient care.

BEHRTDAY: Dynamic Mortality Risk Prediction using Time-Variant COVID-19 Patient Specific Trajectories.

Incorporating repeated measurements of vitals and laboratory measurements can improve mortality risk-prediction and identify key risk factors in individualized treatment of COVID-19 hospitalized patients. In this observational study, demographic and laboratory data of all admitted patients to 5 hospitals of Mount Sinai Health System, New York, with COVID-19 positive tests between March 1st and June 8th, 2020, were extracted from electronic medical records and compared between survivors and non-survivors. Next day mortality risk of patients was assessed using a transformer-based model BEHRTDAY fitted to patient time series data of vital signs, blood and other laboratory measurements given the entire patients' hospital stay. The study population includes 3699 COVID-19 positive (57% male, median age: 67) patients. This model had a very high average precision score (0.96) and area under receiver operator curve (0.92) for next-day mortality prediction given entire patients' trajectories, and through masking, it learnt each variable's context.