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.

Validation of American Society of Echocardiography Guideline-Recommended Parameters of Right Ventricular Dysfunction Using Artificial Intelligence Compared With Cardiac Magnetic Resonance Imaging.

BACKGROUND: Right ventricular (RV) function is important in the evaluation of cardiac function, but its assessment using standard transthoracic echocardiography (TTE) remains challenging. Cardiac magnetic resonance imaging (CMR) is considered the gold standard. The American Society of Echocardiography recommends surrogate measures of RV function and RV ejection fraction (RVEF) by TTE, including fractional area change (FAC), free wall strain (FWS), and tricuspid annular planar systolic excursion (TAPSE), but they require technical expertise in acquisition and quantification. METHODS: The aim of this study was to evaluate the sensitivity, specificity, and positive and negative predictive values of FAC, FWS, and TAPSE derived using a rapid, novel artificial intelligence (AI) software (LVivoRV) from a single-plane transthoracic echocardiographic apical four-chamber, RV-focused view without ultrasound-enhancing agents for detecting abnormal RV function compared with CMR-derived RVEF. RV dysfunction was defined as RVEF < 50% and RVEF < 40% on CMR. RESULTS: TTE and CMR were performed within a median of 10 days (interquartile range, 2-32 days) of each other in 225 consecutive patients without interval procedural or pharmacologic intervention. The sensitivity and negative predictive value to detect CMR-defined RV dysfunction when all three AI-derived parameters (FAC, FWS, and TAPSE) were abnormal were 91% and 96%, while those of expert physician reads were 91% and 97%. Specificity and positive predictive value were lower (50% and 32%) compared with expert physician-read echocardiograms (82% and 56%). CONCLUSIONS: AI-derived measurements of FAC, FWS, and TAPSE had excellent sensitivity and negative predictive value for ruling out significant RV dysfunction (CMR RVEF < 40%), comparable with that of expert physician readers, but lower specificity. Thus AI, using American Society of Echocardiography guidelines, may serve as a useful screening tool for rapid bedside assessment to exclude significant RV dysfunction.

Obstructive sleep apnoea and left ventricular diastolic dysfunction among first responders to the 9/11 World Trade Center terrorist attack: a cross-sectional study.

OBJECTIVES: Obstructive sleep apnoea (OSA) is often linked to cardiovascular disease. A limited number of studies have reported an association between OSA and left ventricular diastolic dysfunction (LVDD). However, prior studies were performed on small patient populations. Studies have shown a high prevalence of OSA among first responders to the 9/11 World Trade Center (WTC) terrorist attack. We investigated the relationship between OSA and LVDD in a large population of WTC responders. DESIGN: Cross-sectional study. SETTING: One-time screening programme as part of the WTC-CHEST Study (NCT10466218), performed at a quaternary medical centre in New York City, from November 2011 to June 2014. PARTICIPANTS: A total of 1007 participants with mean age of 51 years of mostly non-Hispanic white men were evaluated. Patients from the WTC Health Program-Clinical Center of Excellence, who were over the age of 39 years, were eligible to participate. RESULTS: Evaluation of those without OSA diagnosis showed no significant association with LVDD when comparing those screened (Berlin Questionnaire) as OSA high risk versus OSA low risk (p=0.101). Among those diagnosed with LVDD, there was a significant association when comparing those with and without patient-reported OSA (OR 1.50, 95% CI 1.13 to 2.00, p=0.005), but the significance was not maintained after adjusting for pertinent variables (OR 1.3, 0.94 to 1.75, p=0.119). Notably, comparing those with OSA diagnosis and those low risk of OSA, the OR for LVDD was significant (1.69, 1.24 to 2.31, p=0.001), and after adjusting for waist-hip ratio, diabetes and coronary artery calcium score percentile, the relationship remained significant (OR 1.45, 1.03 to 2.04, p=0.032). CONCLUSION: The strong association of OSA with LVDD in this population may inform future guidelines to recommend screening for LVDD in high-risk asymptomatic patients with OSA.

Assessment and validation of a novel fast fully automated artificial intelligence left ventricular ejection fraction quantification software.

BACKGROUND: Quantification of left ventricular ejection fraction (LVEF) by transthoracic echocardiography (TTE) is operator-dependent, time-consuming, and error-prone. LVivoEF by DIA is a new artificial intelligence (AI) software, which displays the tracking of endocardial borders and rapidly quantifies LVEF. We sought to assess the accuracy of LVivoEF compared to cardiac magnetic resonance imaging (cMRI) as the reference standard and to compare LVivoEF to the standard-of-care physician-measured LVEF (MD-EF) including studies with ultrasound enhancing agents (UEAs). METHODS: In 273 consecutive patients, we compared MD-EF and AI-derived LVEF to cMRI. AI-derived LVEF was obtained from a non-UEA four-chamber view without manual correction. Thirty-one patients were excluded: 25 had interval interventions or incomplete TTE or cMRI studies and six had uninterpretable non-UEA apical views. RESULTS: In the 242 subjects, the correlation between AI and cMRI was r = .890, similar to MD-EF and cMRI with r = .891 (p = 0.48). Of the 126 studies performed with UEAs, the correlation of AI using the unenhanced four-chamber view was r = .89, similar to MD-EF with r = .90. In the 116 unenhanced studies, AI correlation was r = .87, similar to MD-EF with r = .84. From Bland-Altman analysis, LVivoEF underreported the LVEF with a bias of 3.63 ± 7.40% EF points compared to cMRI while MD-EF to cMRI had a bias of .33 ± 7.52% (p = 0.80). CONCLUSIONS: Compared to cMRI, LVivoEF can accurately quantify LVEF from a standard apical four-chamber view without manual correction. Thus, LVivoEF has the ability to improve and expedite LVEF quantification.

Functional Effects of Cardiomyocyte Injury in COVID-19.

COVID-19 affects multiple organs. Clinical data from the Mount Sinai Health System show that substantial numbers of COVID-19 patients without prior heart disease develop cardiac dysfunction. How COVID-19 patients develop cardiac disease is not known. We integrated cell biological and physiological analyses of human cardiomyocytes differentiated from human induced pluripotent stem cells (hiPSCs) infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in the presence of interleukins (ILs) with clinical findings related to laboratory values in COVID-19 patients to identify plausible mechanisms of cardiac disease in COVID-19 patients. We infected hiPSC-derived cardiomyocytes from healthy human subjects with SARS-CoV-2 in the absence and presence of IL-6 and IL-1ß. Infection resulted in increased numbers of multinucleated cells. Interleukin treatment and infection resulted in disorganization of myofibrils, extracellular release of troponin I, and reduced and erratic beating. Infection resulted in decreased expression of mRNA encoding key proteins of the cardiomyocyte contractile apparatus. Although interleukins did not increase the extent of infection, they increased the contractile dysfunction associated with viral infection of cardiomyocytes, resulting in cessation of beating. Clinical data from hospitalized patients from the Mount Sinai Health System show that a significant portion of COVID-19 patients without history of heart disease have elevated troponin and interleukin levels. A substantial subset of these patients showed reduced left ventricular function by echocardiography. Our laboratory observations, combined with the clinical data, indicate that direct effects on cardiomyocytes by interleukins and SARS-CoV-2 infection might underlie heart disease in COVID-19 patients. IMPORTANCE SARS-CoV-2 infects multiple organs, including the heart. Analyses of hospitalized patients show that a substantial number without prior indication of heart disease or comorbidities show significant injury to heart tissue, assessed by increased levels of troponin in blood. We studied the cell biological and physiological effects of virus infection of healthy human iPSC-derived cardiomyocytes in culture. Virus infection with interleukins disorganizes myofibrils, increases cell size and the numbers of multinucleated cells, and suppresses the expression of proteins of the contractile apparatus. Viral infection of cardiomyocytes in culture triggers release of troponin similar to elevation in levels of COVID-19 patients with heart disease. Viral infection in the presence of interleukins slows down and desynchronizes the beating of cardiomyocytes in culture. The cell-level physiological changes are similar to decreases in left ventricular ejection seen in imaging of patients' hearts. These observations suggest that direct injury to heart tissue by virus can be one underlying cause of heart disease in COVID-19.

Echocardiography in the time of Covid-19: Ultrasound enhancing agents save time and augment diagnostic information.

BACKGROUND: There are currently no clear guidelines regarding the use of ultrasound enhancing agents (UEAs) with transthoracic echocardiography (TTE) for patients hospitalized with Covid-19. We investigated whether the performance of TTE with UEAs provides more diagnostic information and allows for shorter acquisition time compared to unenhanced TTE imaging in this patient population. METHODS: We analyzed the TTEs of 107 hospitalized Covid-19 patients between April and June 2020 who were administered UEAs (Definity®, Lantheus). The time to acquire images with and without UEAs was calculated. A level III echocardiographer determined if new, clinically significant findings were visualized with the addition of UEAs. RESULTS: There was a mean of 11.84±3.59 UEA cineloops/study vs 20.74±8.10 non-UEA cineloops/study (p < 0.0001). Mean time to acquire UEA cineloop images was 72.28±28.18 s/study compared to 188.07±86.04 s/study for non-UEA cineloop images (p < 0.0001). Forty-eight patients (45%) had at least one new finding on UEA imaging, with a total of 62 new findings seen. New information gained with UEAs was more likely to be found in patients with acute respiratory distress syndrome (21 vs 9, p < 0.001) and in those on mechanical ventilation (21 vs 15, p = 0.046). CONCLUSIONS: TTE with UEAs required less time and fewer cineloop images compared to non-UEA imaging in patients hospitalized with Covid-19. Additionally, Covid-19 patients with severe respiratory disease benefited most with regard to new diagnostic information. Health care personnel should consider early use of UEAs in select hospitalized Covid-19 patients in order to reduce exposure and optimize diagnostic yield.

Using Clinical and Echocardiographic Characteristics to Characterize the Risk of Ischemic Stroke in Patients with COVID-19.

BACKGROUND: COVID-19 has been associated with an increased incidence of ischemic stroke. The use echocardiography to characterize the risk of ischemic stroke in patients hospitalized with COVID-19 has not been explored. METHODS: We conducted a retrospective study of 368 patients hospitalized between 3/1/2020 and 5/31/2020 who had laboratory-confirmed infection with SARS-CoV-2 and underwent transthoracic echocardiography during hospitalization. Patients were categorized according to the presence of ischemic stroke on cerebrovascular imaging following echocardiography. Ischemic stroke was identified in 49 patients (13.3%). We characterized the risk of ischemic stroke using a novel composite risk score of clinical and echocardiographic variables: age <55, systolic blood pressure >140 mmHg, anticoagulation prior to admission, left atrial dilation and left ventricular thrombus. RESULTS: Patients with ischemic stroke had no difference in biomarkers of inflammation and hypercoagulability compared to those without ischemic stroke. Patients with ischemic stroke had significantly more left atrial dilation and left ventricular thrombus (48.3% vs 27.9%, p = 0.04; 4.2% vs 0.7%, p = 0.03). The unadjusted odds ratio of the composite novel COVID-19 Ischemic Stroke Risk Score for the likelihood of ischemic stroke was 4.1 (95% confidence interval 1.4-16.1). The AUC for the risk score was 0.70. CONCLUSIONS: The COVID-19 Ischemic Stroke Risk Score utilizes clinical and echocardiographic parameters to robustly estimate the risk of ischemic stroke in patients hospitalized with COVID-19 and supports the use of echocardiography to characterize the risk of ischemic stroke in patients hospitalized with COVID-19.

Multimodality cardiac imaging for management of a cardiac mass in the era of COVID-19: a case report.

BACKGROUND: Atrial myxomas are the most common benign cardiac tumours. This case highlights an unusual presentation and complex management of a patient who was incidentally found to have a left atrial tumour concerning for a myxoma. CASE SUMMARY: A 54-year-old-woman presented with symptoms of nausea and vomiting and was found to have a left atrial mass incidentally in addition to a renal infarct. She was also found to have COVID-19 and the mass was initially thought to be a thrombus. With the help of multimodality imaging, it was determined that the mass was an atrial myxoma and she was started on short-term anticoagulation to prevent recurrent embolization. After 6 weeks of anticoagulation, she successfully underwent elective resection of the mass which was confirmed to be myxoma with superimposed thrombus. DISCUSSION: It is difficult to differentiate cardiac tumours from intracardiac thrombus and multimodality cardiac imaging is crucial to make an accurate diagnosis. While the treatment of atrial myxomas involves early surgical resection, it becomes more complicated with concurrent COVID-19 infection.

Transcatheter embolic coils to treat peridevice leaks after left atrial appendage closure.

BACKGROUND: Left atrial appendage closure (LAAC) has proven to be an effective alternative to long-term oral anticoagulation in the prevention of thromboembolic events in patients with atrial fibrillation. In a minority of patients, inadequate seal may result in persistent peridevice flow and inability of the appendage to fully thrombose, thereby representing a potential source for thromboembolism. OBJECTIVE: The purpose of this study was to study the use of endovascular coiling of the appendage to address persistent peridevice leak in patients undergoing LAAC with the Watchman device. METHODS: This is a retrospective single-center analysis involving patients who underwent placement of a LAAC device and returned for endovascular coiling to address persistent device leak between 2018 and 2020. Baseline characteristics, procedural outcomes, and follow-up echocardiograms were analyzed to demonstrate the feasibility and safety of this technique. RESULTS: Patients (N = 20) were identified with a mean leak size of 3.8 ± 1.3 mm (range 2.5-7 mm), all of whom had a non-thrombosed appendage. Acute procedural success was achieved in 95% of patients. Complete or significant reduction in flow beyond the LAAC device was achieved in 61% and 33% of patients, respectively. The 1 procedure-related adverse event was a pericardial effusion before coil deployment, requiring percutaneous drainage. CONCLUSION: The clinical impact of residual peridevice leak post-Watchman implantation is a matter of continuing investigation. However, appendage coiling represents a new therapeutic tool to address this potential source for thromboembolism. Further studies should address the clinical impact of this technique, including the safety of discontinuing anticoagulation after successful coiling.

Abnormal left ventricular global longitudinal strain by speckle tracking echocardiography in COVID-19 patients.

COVID-19 infection can affect the cardiovascular system. We sought to determine if left ventricular global longitudinal strain (LVGLS) is affected by COVID-19 and if this has prognostic implications. Materials & methods: Retrospective study, with LVGLS was measured in 58 COVID-19 patients. Patients discharged were compared with those who died. Results: The mean LV ejection fraction (LVEF) and LVGLS for the cohort was 52.1 and -12.9 ± 4.0%, respectively. Among 30 patients with preserved LVEF (>50%), LVGLS was -15.7 ± 2.8%, which is lower than the reference mean LVGLS for a normal, healthy population. There was no significant difference in LVGLS or LVEF when comparing patients who survived to discharge or died. Conclusion: LVGLS was reduced in COVID-19 patients, although not significantly lower in those who died compared with survivors.

Biventricular strain by speckle tracking echocardiography in COVID-19: findings and possible prognostic implications.

The COVID-19 infection adversely affects the cardiovascular system. Transthoracic echocardiography has demonstrated diagnostic, prognostic and therapeutic utility. We report biventricular myocardial strain in COVID-19. Methods: Biventricular strain measurements were performed for 12 patients. Patients who were discharged were compared with those who needed intubation and/or died. Results: Seven patients were discharged and five died or needed intubation. Right ventricular strain parameters were decreased in patients with poor outcomes compared with those discharged. Left ventricular strain was decreased in both groups but was not statistically significant. Conclusion: Right ventricular strain was decreased in patients with poor outcomes and left ventricular strain was decreased regardless of outcome. Right ventricular strain measurements may be important for risk stratification and prognosis. Further studies are needed to confirm these findings.

Physiology of cardiomyocyte injury in COVID-19.

COVID-19 affects multiple organs. Clinical data from the Mount Sinai Health System shows that substantial numbers of COVID-19 patients without prior heart disease develop cardiac dysfunction. How COVID-19 patients develop cardiac disease is not known. We integrate cell biological and physiological analyses of human cardiomyocytes differentiated from human induced pluripotent stem cells (hiPSCs) infected with SARS-CoV-2 in the presence of interleukins, with clinical findings, to investigate plausible mechanisms of cardiac disease in COVID-19 patients. We infected hiPSC-derived cardiomyocytes, from healthy human subjects, with SARS-CoV-2 in the absence and presence of interleukins. We find that interleukin treatment and infection results in disorganization of myofibrils, extracellular release of troponin-I, and reduced and erratic beating. Although interleukins do not increase the extent, they increase the severity of viral infection of cardiomyocytes resulting in cessation of beating. Clinical data from hospitalized patients from the Mount Sinai Health system show that a significant portion of COVID-19 patients without prior history of heart disease, have elevated troponin and interleukin levels. A substantial subset of these patients showed reduced left ventricular function by echocardiography. Our laboratory observations, combined with the clinical data, indicate that direct effects on cardiomyocytes by interleukins and SARS-CoV-2 infection can underlie the heart disease in COVID-19 patients.

Characterization of Myocardial Injury in Patients With COVID-19.

BACKGROUND: Myocardial injury is frequent among patients hospitalized with coronavirus disease-2019 (COVID-19) and is associated with a poor prognosis. However, the mechanisms of myocardial injury remain unclear and prior studies have not reported cardiovascular imaging data. OBJECTIVES: This study sought to characterize the echocardiographic abnormalities associated with myocardial injury and their prognostic impact in patients with COVID-19. METHODS: We conducted an international, multicenter cohort study including 7 hospitals in New York City and Milan of hospitalized patients with laboratory-confirmed COVID-19 who had undergone transthoracic echocardiographic (TTE) and electrocardiographic evaluation during their index hospitalization. Myocardial injury was defined as any elevation in cardiac troponin at the time of clinical presentation or during the hospitalization. RESULTS: A total of 305 patients were included. Mean age was 63 years and 205 patients (67.2%) were male. Overall, myocardial injury was observed in 190 patients (62.3%). Compared with patients without myocardial injury, those with myocardial injury had more electrocardiographic abnormalities, higher inflammatory biomarkers and an increased prevalence of major echocardiographic abnormalities that included left ventricular wall motion abnormalities, global left ventricular dysfunction, left ventricular diastolic dysfunction grade II or III, right ventricular dysfunction and pericardial effusions. Rates of in-hospital mortality were 5.2%, 18.6%, and 31.7% in patients without myocardial injury, with myocardial injury without TTE abnormalities, and with myocardial injury and TTE abnormalities. Following multivariable adjustment, myocardial injury with TTE abnormalities was associated with higher risk of death but not myocardial injury without TTE abnormalities. CONCLUSIONS: Among patients with COVID-19 who underwent TTE, cardiac structural abnormalities were present in nearly two-thirds of patients with myocardial injury. Myocardial injury was associated with increased in-hospital mortality particularly if echocardiographic abnormalities were present.

Improved scoring system for the electrocardiographic diagnosis of left ventricular hypertrophy.

BACKGROUND: Left ventricular hypertrophy (LVH) is a common manifestation of cardiovascular disease and a risk factor for cardiovascular morbidity and mortality, but available methods for its electrocardiographic (ECG) diagnosis have limited accuracy. AIM: To investigate findings associated with LVH on ECG and developed an improved system for the diagnosis of LVH. METHODS: A cohort study comparing ECG data acquired within 30 days of transthoracic echocardiography (TTE) was performed. Multivariate regression analysis identified ECG findings associated with increased LV mass and mass index. A scoring system was derived and performance compared to established criteria for LVH. RESULTS: Data from 5486 outpatients with TTEs and corresponding ECGs were included in the derivation cohort, 333 (6.1%) of whom had LVH by TTE. In the primary regression analysis, findings associated with LVH were amplitudes of Q in V3, R in V6, S in V3, T in V6, P' in V1, P in V6, as well as R and T-axis discordance, R peak time in V6, QRS duration, weight, height, sex, and age. From this we derived a score consisting of 5 criteria, and validated it in an independent cohort of 910 patients. With a threshold of 1.5 points, sensitivity and specificity were 67.9% and 81.4%, and 62.5% and 83.2% in the derivation and validation cohorts, respectively. With a threshold of 2 points, sensitivity and specificity were 42.3% and 93.0%, and 37.5% and 93.4% in these cohorts. CONCLUSIONS: This score had superior sensitivity for detection of LVH by ECG while making a modest sacrifice in specificity compared to conventional criteria.

Multicenter evaluation of stress-first myocardial perfusion image triage by nuclear technologists and automated quantification.

BACKGROUND: A stress-first myocardial perfusion imaging (MPI) protocol saves time, is cost effective, and decreases radiation exposure. A limitation of this protocol is the requirement for physician review of the stress images to determine the need for rest images. This hurdle could be eliminated if an experienced technologist and/or automated computer quantification could make this determination. METHODS: Images from consecutive patients who were undergoing a stress-first MPI with attenuation correction at two tertiary care medical centers were prospectively reviewed independently by a technologist and cardiologist blinded to clinical and stress test data. Their decision on the need for rest imaging along with automated computer quantification of perfusion results was compared with the clinical reference standard of an assessment of perfusion images by a board-certified nuclear cardiologist that included clinical and stress test data. RESULTS: A total of 250 patients (mean age 61 years and 55% female) who underwent a stress-first MPI were studied. According to the clinical reference standard, 42 (16.8%) and 208 (83.2%) stress-first images were interpreted as "needing" and "not needing" rest images, respectively. The technologists correctly classified 229 (91.6%) stress-first images as either "needing" (n = 28) or "not needing" (n = 201) rest images. Their sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were 66.7%, 96.6%, 80.0%, and 93.5%, respectively. An automated stress TPD score ≥1.2 was associated with optimal sensitivity and specificity and correctly classified 179 (71.6%) stress-first images as either "needing" (n = 31) or "not needing" (n = 148) rest images. Its sensitivity, specificity, PPV, and NPV were 73.8%, 71.2%, 34.1%, and 93.1%, respectively. In a model whereby the computer or technologist could correct for the other's incorrect classification, 242 (96.8%) stress-first images were correctly classified. The composite sensitivity, specificity, PPV, and NPV were 83.3%, 99.5%, 97.2%, and 96.7%, respectively. CONCLUSION: Technologists and automated quantification software had a high degree of agreement with the clinical reference standard for determining the need for rest images in a stress-first imaging protocol. Utilizing an experienced technologist and automated systems to screen stress-first images could expand the use of stress-first MPI to sites where the cardiologist is not immediately available for interpretation.