Artificial Intelligence Enabled Interpretation of ECG Images to Predict Hematopoietic Cell Transplantation Toxicity.

Artificial intelligence enabled interpretation of electrocardiogram waveform images (AI-ECG) can identify patterns predictive of future adverse cardiac events. We hypothesized such an approach, which is well described in general medical and surgical patients, would provide prognostic information with respect to the risk of cardiac complications and overall mortality in patients undergoing hematopoietic cell transplantation (HCT) for blood malignancy. We retrospectively subjected ECGs obtained pre-HCT to an externally trained, deep learning model designed to predict risk of atrial fibrillation (AF). Included were 1,377 patients (849 autologous HCT and 528 allogeneic HCT recipients). Median follow-up was 2.9 years. The three-year cumulative incidence of AF was 9% (95% CI: 7-12%) in autologous HCT patients and 13% (10-16%) in allogeneic HCT patients. In the entire cohort, pre-HCT AI-ECG estimate of AF risk correlated highly with development of clinical AF (Hazard Ratio (HR) 7.37, 3.53-15.4, p <0.001), inferior overall survival (HR: 2.4; 1.3-4.5, p = 0.004), and greater risk of non-relapse mortality (HR 3.36, 1.39-8.13, p = 0.007), without increased risk of relapse. Significant associations with mortality were only noted in allo HCT recipients, where the risk of non-relapse mortality was greater. Compared to calcineurin inhibitor-based graft versus host disease prophylaxis, the use of post-transplantation cyclophosphamide resulted in greater 90-day incidence of AF (13% versus 5%, p = 0.01), corresponding to temporal changes in AI-ECG AF prediction post HCT. In summary, AI-ECG can inform risk of post-transplant cardiac outcomes and survival in HCT patients and represents a novel strategy for personalized risk assessment after HCT.

Performance of AI-Enabled Electrocardiogram in the Prediction of Metabolic Dysfunction-Associated Steatotic Liver Disease.

BACKGROUND AND AIMS: Accessible noninvasive screening tools for metabolic dysfunction-associated steatotic liver disease (MASLD) are needed. We aim to explore the performance of a deep-learning based artificial intelligence (AI) model in distinguishing the presence of MASLD using 12-lead electrocardiogram (ECG). METHODS: This is a retrospective study of adults diagnosed with MASLD in Olmsted County, Minnesota, between 1996 and 2019. Both cases and controls had ECGs performed within 6 years before and 1 year after study entry. An AI-based ECG model using a convolutional neural network was trained, validated, and tested in 70%, 10% and 20% of the cohort, respectively. External validation was performed in an independent cohort from Mayo Clinic Enterprise. The primary outcome was the performance of ECG to identify MASLD, alone or when added to clinical parameters. RESULTS: 3,468 MASLD cases and 25,407 controls were identified. The AI-ECG model predicted the presence of MASLD with an area under the curve (AUC) of 0.69 (original cohort) and 0.62 (validation cohort). The performance was similar or superior to age- and sex-adjusted models using body mass index (BMI) (AUC=0.71), presence of diabetes, hypertension or hyperlipidemia (AUC=0.68) or diabetes alone (AUC=0.66). The model combining ECG, BMI, diabetes, and alanine aminotransferase had the highest AUC (0.76 (original); 0.72 (validation)). CONCLUSION: This is a proof-of-concept study that an AI-based ECG model can detect MASLD with a comparable or superior performance as compared to the models using a single clinical parameter but not superior to the combination of clinical parameters. ECG can serve as another screening tool for MASLD in the non-hepatology space.

Pediatric sex estimation using AI-enabled ECG analysis: influence of pubertal development.

AI-enabled ECGs have previously been shown to accurately predict patient sex in adults and correlate with sex hormone levels. We aimed to test the ability of AI-enabled ECGs to predict sex in the pediatric population and study the influence of pubertal development. AI-enabled ECG models were created using a convolutional neural network trained on pediatric 10-second, 12-lead ECGs. The first model was trained de novo using pediatric data. The second model used transfer learning from a previously validated adult data-derived algorithm. We analyzed the first ECG from 90,133 unique pediatric patients (aged ≤18 years) recorded between 1987-2022, and divided the cohort into training, validation, and testing datasets. Subgroup analysis was performed on prepubertal (0-7 years), peripubertal (8-14 years), and postpubertal (15-18 years) patients. The cohort was 46.7% male, with 21,678 prepubertal, 26,740 peripubertal, and 41,715 postpubertal children. The de novo pediatric model demonstrated 81% accuracy and an area under the curve (AUC) of 0.91. Model sensitivity was 0.79, specificity was 0.83, positive predicted value was 0.84, and the negative predicted value was 0.78, for the entire test cohort. The model's discriminatory ability was highest in postpubertal (AUC = 0.98), lower in the peripubertal age group (AUC = 0.91), and poor in the prepubertal age group (AUC = 0.67). There was no significant performance difference observed between the transfer learning and de novo models. AI-enabled interpretation of ECG can estimate sex in peripubertal and postpubertal children with high accuracy.

Machine learning for prediction of ventricular arrhythmia episodes from intracardiac electrograms of automatic implantable cardioverter-defibrillators.

BACKGROUND: Despite effectiveness of the implantable cardioverter-defibrillator (ICD) in saving patients with life-threatening ventricular arrhythmias (VAs), the temporal occurrence of VA after ICD implantation is unpredictable. OBJECTIVE: The study aimed to apply machine learning (ML) to intracardiac electrograms (IEGMs) recorded by ICDs as a unique biomarker for predicting impending VAs. METHODS: The study included 13,516 patients who received Biotronik ICDs and enrolled in the CERTITUDE registry between January 1, 2010, and December 31, 2020. Database extraction included IEGMs from standard quarterly transmissions and VA event episodes. The processed IEGM data were pulled from device transmissions stored in a centralized Home Monitoring Service Center and reformatted into an analyzable format. Long-range (baseline or first scheduled remote recording), mid-range (scheduled remote recording every 90 days), or short-range predictions (IEGM within 5 seconds before the VA onset) were used to determine whether ML-processed IEGMs predicted impending VA events. Convolutional neural network classifiers using ResNet architecture were employed. RESULTS: Of 13,516 patients (male, 72%; age, 67.5 ± 11.9 years), 301,647 IEGM recordings were collected; 27,845 episodes of sustained ventricular tachycardia or ventricular fibrillation were observed in 4467 patients (33.0%). Neural networks based on convolutional neural networks using ResNet-like architectures on far-field IEGMs yielded an area under the curve of 0.83 with a 95% confidence interval of 0.79-0.87 in the short term, whereas the long-range and mid-range analyses had minimal predictive value for VA events. CONCLUSION: In this study, applying ML to ICD-acquired IEGMs predicted impending ventricular tachycardia or ventricular fibrillation events seconds before they occurred, whereas midterm to long-term predictions were not successful. This could have important implications for future device therapies.

Artificial Intelligence of Arterial Doppler Waveforms to Predict Major Adverse Outcomes Among Patients Evaluated for Peripheral Artery Disease.

BACKGROUND: Patients with peripheral artery disease are at increased risk for major adverse cardiac events, major adverse limb events, and all-cause death. Developing tools capable of identifying those patients with peripheral artery disease at greatest risk for major adverse events is the first step for outcome prevention. This study aimed to determine whether computer-assisted analysis of a resting Doppler waveform using deep neural networks can accurately identify patients with peripheral artery disease at greatest risk for adverse outcome events. METHODS AND RESULTS: Consecutive patients (April 1, 2015, to December 31, 2020) undergoing ankle-brachial index testing were included. Patients were randomly allocated to training, validation, and testing subsets (60%/20%/20%). Deep neural networks were trained on resting posterior tibial arterial Doppler waveforms to predict major adverse cardiac events, major adverse limb events, and all-cause death at 5 years. Patients were then analyzed in groups based on the quartiles of each prediction score in the training set. Among 11 384 total patients, 10 437 patients met study inclusion criteria (mean age, 65.8±14.8 years; 40.6% women). The test subset included 2084 patients. During 5 years of follow-up, there were 447 deaths, 585 major adverse cardiac events, and 161 MALE events. After adjusting for age, sex, and Charlson comorbidity index, deep neural network analysis of the posterior tibial artery waveform provided independent prediction of death (hazard ratio [HR], 2.44 [95% CI, 1.78-3.34]), major adverse cardiac events (HR, 1.97 [95% CI, 1.49-2.61]), and major adverse limb events (HR, 11.03 [95% CI, 5.43-22.39]) at 5 years. CONCLUSIONS: An artificial intelligence-enabled analysis of Doppler arterial waveforms enables identification of major adverse outcomes among patients with peripheral artery disease, which may promote early adoption and adherence of risk factor modification.

Clinician Adoption of an Artificial Intelligence Algorithm to Detect Left Ventricular Systolic Dysfunction in Primary Care.

OBJECTIVE: To compare the clinicians' characteristics of "high adopters" and "low adopters" of an artificial intelligence (AI)-enabled electrocardiogram (ECG) algorithm that alerted for possible low left ventricular ejection fraction (EF) and the subsequent effectiveness of detecting patients with low EF. METHODS: Clinicians in 48 practice sites of a US Midwest health system were cluster-randomized by the care team to usual care or to receive a notification that suggested ordering an echocardiogram in patients flagged as potentially having low EF based on an AI-ECG algorithm. Enrollment was between June 26, 2019, and July 30, 2019; participation concluded on March 31, 2020. This report is focused on those clinicians randomized to receive the notification of the AI-ECG algorithm. At the patient level, data were analyzed for the proportion of patients with positive AI-ECG results. Adoption was defined as the clinician order of an echocardiogram after prompted by the alert. RESULTS: A total of 165 clinicians and 11,573 patients were included in this analysis. Among patients with positive AI-ECG, high adopters (n=41) were twice as likely to diagnose patients with low EF (33.9%) vs low adopters, n=124, (16.9%); odds ratio, 1.62; 95% CI, 1.21 to 2.17). High adopters were more often advanced practice providers (eg, nurse practitioners and physician assistants) vs physicians, Family Medicine vs Internal Medicine specialty, and tended to have less complex patients. CONCLUSION: Clinicians who most frequently followed the recommendations of an AI tool were twice as likely to diagnose low EF. Those clinicians with less complex patients were more likely to be high adopters. TRIAL REGISTRATION: Clinicaltrials.gov Identifier: NCT04000087.

Artificial intelligence for the evaluation of peripheral artery disease using arterial Doppler waveforms to predict abnormal ankle-brachial index.

BACKGROUND: Patients with peripheral artery disease (PAD) are at increased risk for major adverse limb and cardiac events including mortality. Developing screening tools capable of accurate PAD identification is a necessary first step for strategies of adverse outcome prevention. This study aimed to determine whether machine analysis of a resting Doppler waveform using deep neural networks can accurately identify patients with PAD. METHODS: Consecutive patients (4/8/2015 - 12/31/2020) undergoing rest and postexercise ankle-brachial index (ABI) testing were included. Patients were randomly allocated to training, validation, and testing subsets (70%/15%/15%). Deep neural networks were trained on resting posterior tibial arterial Doppler waveforms to predict normal (> 0.9) or PAD (⩽ 0.9) using rest and postexercise ABI. A separate dataset of 151 patients who underwent testing during a period after the model had been created and validated (1/1/2021 - 3/31/2021) was used for secondary validation. Area under the receiver operating characteristic curves (AUC) were constructed to evaluate test performance. RESULTS: Among 11,748 total patients, 3432 patients met study criteria: 1941 with PAD (mean age 69 ± 12 years) and 1491 without PAD (64 ± 14 years). The predictive model with highest performance identified PAD with an AUC 0.94 (CI = 0.92-0.96), sensitivity 0.83, specificity 0.88, accuracy 0.85, and positive predictive value (PPV) 0.90. Results were similar for the validation dataset: AUC 0.94 (CI = 0.91-0.98), sensitivity 0.91, specificity 0.85, accuracy 0.89, and PPV 0.89 (postexercise ABI comparison). CONCLUSION: An artificial intelligence-enabled analysis of a resting Doppler arterial waveform permits identification of PAD at a clinically relevant performance level.

Artificial Intelligence-Enabled ECG to Identify Silent Atrial Fibrillation in Embolic Stroke of Unknown Source.

OBJECTIVES: Embolic strokes of unknown source (ESUS) are common and often suspected to be caused by unrecognized paroxysmal atrial fibrillation (AF). An AI-enabled ECG (AI-ECG) during sinus rhythm has been shown to identify patients with unrecognized AF. We pursued this study to determine if the AI-ECG model differentiates between patients with ESUS and those with known causes of stroke, and to evaluate whether the AF prediction by AI-ECG among patients with ESUS was associated with the results of prolonged ambulatory cardiac rhythm monitoring. MATERIALS AND METHODS: We reviewed consecutive patients admitted with acute ischemic stroke to a comprehensive stroke center between January 2018 and August 2019 and employed the TOAST classification to categorize the mechanisms of ischemia. Use and results of ambulatory cardiac rhythm monitoring after discharge were gathered. We ran the AI-ECG model to obtain AF probabilities from all ECGs acquired during the hospitalization and compared those probabilities in patients with ESUS versus those with known stroke causes (apart from AF), and between patients with and without AF detected by ambulatory cardiac rhythm monitoring. RESULTS: The study cohort had 930 patients, including 263 patients (28.3%) with known AF or AF diagnosed during the index hospitalization and 265 cases (28.5%) categorized as ESUS. Ambulatory cardiac rhythm monitoring was performed in 226 (85.3%) patients with ESUS. AF probability by AI-ECG was not associated with ESUS. However, among patients with ESUS, the probability of AF by AI-ECG was associated with a higher likelihood of AF detection by ambulatory monitoring (P = 0.004). A probability of AF by AI-ECG greater than 0.20 was associated with AF detection by ambulatory cardiac rhythm monitoring with an OR of 5.47 (95% CI 1.51-22.51). CONCLUSIONS: AI-ECG may help guide the use of prolonged ambulatory cardiac rhythm monitoring in patients with ESUS to identify those who might benefit from anticoagulation.

Artificial Intelligence-Augmented Electrocardiogram Detection of Left Ventricular Systolic Dysfunction in the General Population.

OBJECTIVE: To validate an artificial intelligence-augmented electrocardiogram (AI-ECG) algorithm for the detection of preclinical left ventricular systolic dysfunction (LVSD) in a large community-based cohort. METHODS: We identified a randomly selected community-based cohort of 2041 subjects age 45 years or older in Olmsted County, Minnesota. All participants underwent a study echocardiogram and ECG. We first assessed the performance of the AI-ECG to identify LVSD (ejection fraction ≤40%). After excluding participants with clinical heart failure, we further assessed the AI-ECG to detect preclinical LVSD among all patients (n=1996) and in a high-risk subgroup (n=1348). Next we modelled an imputed screening program for preclinical LVSD detection where a positive AI-ECG triggered an echocardiogram. Finally, we assessed the ability of the AI-ECG to predict future LVSD. Participants were enrolled between January 1, 1997, and September 30, 2000; and LVSD surveillance was performed for 10 years after enrollment. RESULTS: For detection of LVSD in the total population (prevalence, 2.0%), the area under the receiver operating curve for AI-ECG was 0.97 (sensitivity, 90%; specificity, 92%); in the high-risk subgroup (prevalence 2.7%), the area under the curve was 0.97 (sensitivity, 92%; specificity, 93%). In an imputed screening program, identification of one preclinical LSVD case would require 88.3 AI-ECGs and 8.7 echocardiograms in the total population and 65.7 AI-ECGs and 5.5 echocardiograms in the high-risk subgroup. The unadjusted hazard ratio for a positive AI-ECG for incident LVSD over 10 years was 2.31 (95% CI, 1.32 to 4.05; P=.004). CONCLUSION: Artificial intelligence-augmented ECG can identify preclinical LVSD in the community and warrants further study as a screening tool for preclinical LVSD.

Cost Effectiveness of an Electrocardiographic Deep Learning Algorithm to Detect Asymptomatic Left Ventricular Dysfunction.

OBJECTIVE: To evaluate the cost-effectiveness of an artificial intelligence electrocardiogram (AI-ECG) algorithm under various clinical and cost scenarios when used for universal screening at age 65. PATIENTS AND METHODS: We used decision analytic modeling to perform a cost-effectiveness analysis of the use of AI-ECG to screen for asymptomatic left ventricular dysfunction (ALVD) once at age 65 compared with no screening. This screening consisted of an initial screening decision tree and subsequent construction of a Markov model. One-way sensitivity analysis on various disease and cost parameters to evaluate cost-effectiveness at both $50,000 per quality-adjusted life year (QALY) and $100,000 per QALY willingness-to-pay threshold. RESULTS: We found that for universal screening at age 65, the novel AI-ECG algorithm would cost $43,351 per QALY gained, test performance, disease characteristics, and testing cost parameters significantly affect cost-effectiveness, and screening at ages 55 and 75 would cost $48,649 and $52,072 per QALY gained, respectively. Overall, under most of the clinical scenarios modeled, coupled with its robust test performance in both testing and validation cohorts, screening with the novel AI-ECG algorithm appears to be cost-effective at a willingness-to-pay threshold of $50,000. CONCLUSION: Universal screening for ALVD with the novel AI-ECG appears to be cost-effective under most clinical scenarios with a cost of <$50,000 per QALY. Cost-effectiveness is particularly sensitive to both the probability of disease progression and the cost of screening and downstream testing. To improve cost-effectiveness modeling, further study of the natural progression and treatment of ALVD and external validation of AI-ECG should be undertaken.

Artificial Intelligence (AI)-Empowered Echocardiography Interpretation: A State-of-the-Art Review.

Echocardiography (Echo), a widely available, noninvasive, and portable bedside imaging tool, is the most frequently used imaging modality in assessing cardiac anatomy and function in clinical practice. On the other hand, its operator dependability introduces variability in image acquisition, measurements, and interpretation. To reduce these variabilities, there is an increasing demand for an operator- and interpreter-independent Echo system empowered with artificial intelligence (AI), which has been incorporated into diverse areas of clinical medicine. Recent advances in AI applications in computer vision have enabled us to identify conceptual and complex imaging features with the self-learning ability of AI models and efficient parallel computing power. This has resulted in vast opportunities such as providing AI models that are robust to variations with generalizability for instantaneous image quality control, aiding in the acquisition of optimal images and diagnosis of complex diseases, and improving the clinical workflow of cardiac ultrasound. In this review, we provide a state-of-the art overview of AI-empowered Echo applications in cardiology and future trends for AI-powered Echo technology that standardize measurements, aid physicians in diagnosing cardiac diseases, optimize Echo workflow in clinics, and ultimately, reduce healthcare costs.

External validation of a deep learning electrocardiogram algorithm to detect ventricular dysfunction.

OBJECTIVE: To validate a novel artificial-intelligence electrocardiogram algorithm (AI-ECG) to detect left ventricular systolic dysfunction (LVSD) in an external population. BACKGROUND: LVSD, even when asymptomatic, confers increased morbidity and mortality. We recently derived AI-ECG to detect LVSD using ECGs based on a large sample of patients treated at the Mayo Clinic. METHODS: We performed an external validation study with subjects from the Know Your Heart Study, a cross-sectional study of adults aged 35-69 years residing in two cities in Russia, who had undergone both ECG and transthoracic echocardiography. LVSD was defined as left ventricular ejection fraction ≤ 35%. We assessed the performance of the AI-ECG to identify LVSD in this distinct patient population. RESULTS: Among 4277 subjects in this external population-based validation study, 0.6% had LVSD (compared to 7.8% of the original clinical derivation study). The overall performance of the AI-ECG to detect LVSD was robust with an area under the receiver operating curve of 0.82. When using the LVSD probability cut-off of 0.256 from the original derivation study, the sensitivity, specificity, and accuracy in this population were 26.9%, 97.4%, 97.0%, respectively. Other probability cut-offs were analysed for different sensitivity values. CONCLUSIONS: The AI-ECG detected LVSD with robust test performance in a population that was very different from that used to develop the algorithm. Population-specific cut-offs may be necessary for clinical implementation. Differences in population characteristics, ECG and echocardiographic data quality may affect test performance.

Spectrum bias in algorithms derived by artificial intelligence: a case study in detecting aortic stenosis using electrocardiograms.

Aims: Spectrum bias can arise when a diagnostic test is derived from study populations with different disease spectra than the target population, resulting in poor generalizability. We used a real-world artificial intelligence (AI)-derived algorithm to detect severe aortic stenosis (AS) to experimentally assess the effect of spectrum bias on test performance. Methods and results: All adult patients at the Mayo Clinic between 1 January 1989 and 30 September 2019 with transthoracic echocardiograms within 180 days after electrocardiogram (ECG) were identified. Two models were developed from two distinct patient cohorts: a whole-spectrum cohort comparing severe AS to any non-severe AS and an extreme-spectrum cohort comparing severe AS to no AS at all. Model performance was assessed. Overall, 258 607 patients had valid ECG and echocardiograms pairs. The area under the receiver operator curve was 0.87 and 0.91 for the whole-spectrum and extreme-spectrum models, respectively. Sensitivity and specificity for the whole-spectrum model was 80% and 81%, respectively, while for the extreme-spectrum model it was 84% and 84%, respectively. When applying the AI-ECG derived from the extreme-spectrum cohort to patients in the whole-spectrum cohort, the sensitivity, specificity, and area under the curve dropped to 83%, 73%, and 0.86, respectively. Conclusion: While the algorithm performed robustly in identifying severe AS, this study shows that limiting datasets to clearly positive or negative labels leads to overestimation of test performance when testing an AI algorithm in the setting of classifying severe AS using ECG data. While the effect of the bias may be modest in this example, clinicians should be aware of the existence of such a bias in AI-derived algorithms.

Detecting cardiomyopathies in pregnancy and the postpartum period with an electrocardiogram-based deep learning model.

AIMS: Cardiovascular disease is a major threat to maternal health, with cardiomyopathy being among the most common acquired cardiovascular diseases during pregnancy and the postpartum period. The aim of our study was to evaluate the effectiveness of an electrocardiogram (ECG)-based deep learning model in identifying cardiomyopathy during pregnancy and the postpartum period. METHODS AND RESULTS: We used an ECG-based deep learning model to detect cardiomyopathy in a cohort of women who were pregnant or in the postpartum period seen at Mayo Clinic. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. We compared the diagnostic probabilities of the deep learning model with natriuretic peptides and a multivariable model consisting of demographic and clinical parameters. The study cohort included 1807 women; 7%, 10%, and 13% had left ventricular ejection fraction (LVEF) of 35% or less, <45%, and <50%, respectively. The ECG-based deep learning model identified cardiomyopathy with AUCs of 0.92 (LVEF ≤ 35%), 0.89 (LVEF < 45%), and 0.87 (LVEF < 50%). For LVEF of 35% or less, AUC was higher in Black (0.95) and Hispanic (0.98) women compared to White (0.91). Natriuretic peptides and the multivariable model had AUCs of 0.85 to 0.86 and 0.72, respectively. CONCLUSIONS: An ECG-based deep learning model effectively identifies cardiomyopathy during pregnancy and the postpartum period and outperforms natriuretic peptides and traditional clinical parameters with the potential to become a powerful initial screening tool for cardiomyopathy in the obstetric care setting.