Artificial intelligence-enabled prediction of chemotherapy-induced cardiotoxicity from baseline electrocardiograms.

Anthracyclines can cause cancer therapy-related cardiac dysfunction (CTRCD) that adversely affects prognosis. Despite guideline recommendations, only half of the patients undergo surveillance echocardiograms. An AI model detecting reduced left ventricular ejection fraction from 12-lead electrocardiograms (ECG) (AI-EF model) suggests ECG features reflect left ventricular pathophysiology. We hypothesized that AI could predict CTRCD from baseline ECG, leveraging the AI-EF model's insights, and developed the AI-CTRCD model using transfer learning on the AI-EF model. In 1011 anthracycline-treated patients, 8.7% experienced CTRCD. High AI-CTRCD scores indicated elevated CTRCD risk (hazard ratio (HR), 2.66; 95% CI 1.73-4.10; log-rank p < 0.001). This remained consistent after adjusting for risk factors (adjusted HR, 2.57; 95% CI 1.62-4.10; p < 0.001). AI-CTRCD score enhanced prediction beyond known factors (time-dependent AUC for 2 years: 0.78 with AI-CTRCD score vs. 0.74 without; p = 0.005). In conclusion, the AI model robustly stratified CTRCD risk from baseline ECG.

Deep learning-based model detects atrial septal defects from electrocardiography: a cross-sectional multicenter hospital-based study.

Background: Atrial septal defect (ASD) increases the risk of adverse cardiovascular outcomes. Despite the potential for risk mitigation through minimally invasive percutaneous closure, ASD remains underdiagnosed due to subtle symptoms and examination findings. To bridge this diagnostic gap, we propose a novel screening strategy aimed at early detection and enhanced diagnosis through the implementation of a convolutional neural network (CNN) to identify ASD from 12-lead electrocardiography (ECG). Methods: ECGs were collected from patients with at least one recorded echocardiogram at 3 hospitals from 2 continents (Keio University Hospital from July 2011 to December 2020, Brigham and Women's Hospital from January 2015 to December 2020, and Dokkyo Medical University Saitama Medical Center from January 2010 and December 2021). ECGs from patients with a diagnosis of ASD were labeled as positive cases while the remainder were labeled as negative. ECGs after the closure of ASD were excluded. After randomly splitting the ECGs into 3 datasets (50% derivation, 20% validation, and 30% test) with no patient overlap, a CNN-based model was trained using the derivation datasets from 2 hospitals and was tested on held-out datasets along with an external validation on the 3rd hospital. All eligible ECGs were used for derivation and validation whereas the earliest ECG for each patient was used for the test and external validation. The discrimination of ASD was assessed by the area under the receiver operating characteristic curve (AUROC). Multiple subgroups were examined to identify any heterogeneity. Findings: A total of 671,201 ECGs from 80,947 patients were collected from the 3 institutions. The AUROC for detecting ASD was 0.85-0.90 across the 3 hospitals. The subgroup analysis showed excellent performance across various characteristics Screening simulation using the model greatly increased sensitivity from 80.6% to 93.7% at specificity 33.6% when compared to using overt ECG abnormalities. Interpretation: A CNN-based model using 12-lead ECG successfully identified the presence of ASD with excellent generalizability across institutions from 2 separate continents. Funding: This work was supported by research grants from JST (JPMJPF2101), JSR corporation, Taiju Life Social Welfare Foundation, Kondou Kinen Medical Foundation, Research fund of Mitsukoshi health and welfare foundation, Tokai University School of Medicine Project Research and Internal Medicine Project Research, Secom Science and Technology Foundation, and Grants from AMED (JP23hma922012 and JP23ym0126813). This work was partially supported by One Brave Idea, co-funded by the American Heart Association and Verily with significant support from AstraZeneca and pillar support from Quest Diagnostics.

Pediatric chronic graft-versus-host disease-related dry eye disease and the diagnostic association of potential clinical findings.

Pediatric graft-versus-host-disease (GVHD)-related dry eye disease (DED) is often overlooked due to a lack of subjective symptoms and reliable testing, leading to irreversible corneal damage. To study the clinical findings contributing to the accurate detection of pediatric GVHD-related DED, a retrospective study of pediatric patients treated with hematopoietic stem cell transplantation (HSCT) at Keio University Hospital between 2004 and 2017 was conducted. Association and diagnostic values of ophthalmological findings for DED were analyzed. Twenty-six patients who had no ocular complications before HSCT were included in the study. Eleven (42.3%) patients developed new-onset DED. The cotton thread test showed excellent diagnostic accuracy in detecting DED (area under the receiver operating curve, 0.96; sensitivity, 0.95; specificity, 0.85) with a cut-off of 17 mm, which was higher than the conventional threshold of 10 mm. Additionally, the presence of filamentary keratitis (FK) and pseudomembranous conjunctivitis (PC) were significantly associated with the diagnosis of DED (p value, 0.003 and 0.001 for FK and PC, respectively) and displayed good diagnostic performance (sensitivity, 0.46 and 0.54; specificity, 0.97 and 0.97 for FK and PC, respectively). In conclusion, the cotton thread test with a new threshold, the presence of PC and FK, could be helpful for promptly detecting pediatric GVHD-related DED.

Multicenter prospective validation study for international chronic ocular graft-versus-host disease consensus diagnostic criteria.

PURPOSE: To validate the international chronic ocular graft-versus-host disease (GVHD) diagnostic criteria (ICCGVHD) compared to the National Institute of Health diagnostic criteria 2014 (NIH2014) for chronic ocular GVHD. METHODS: Between 2013 and 2019, the study enrolled 233 patients with or without chronic ocular GVHD combined with the presence or absence of systemic chronic GVHD in an internationally prospective multicenter and observational cohort from 9 institutions. All patients were evaluated for four clinical parameters of ICCGVHD. RESULTS: The relation between the ICCGVHD score (0-11) and NIH2014 eye score (0-4) was relatively high (r = 0.708, 95% CI: 0.637-0.767, p < 0.001). The sensitivity and specificity of ICCGVHD for NIH 2014 for 233 patients were 94.3% (95% CI: 89.6%-98.1%) and 71.7% (95% CI: 63.0-79.5%), respectively (cutoff value of the ICCGVHD score = 6). The positive predictive value was 77.1% (95% CI: 71.1%-82.1%), and the negative predictive value was 87.0% (95% CI:81.6-92.5%). For the patients with systemic GVHD (n = 171), the sensitivity and specificity were 94.2% and 67.2%, respectively (ICCGVHD-score cutoff value = 6). By receiver operating characteristic (ROC) curve analysis, the area under the curve (AUC) was 0.903 (95% CI: 0.859-0.948). For patients without systemic GVHD (n = 62), the sensitivity and specificity were 100% and 76.7%, respectively (ICCGVHD-score cutoff value = 6). The AUC was 0.891 (95% CI 0.673-1.000). CONCLUSIONS: Good sensitivity, specificity, predictive value and correlation were found between ICCGVHD and NIH2014. ICCGVHD scores ≥6 can be useful to diagnose ocular GVHD with or without systemic GVHD for clinical research.

Multinational Federated Learning Approach to Train ECG and Echocardiogram Models for Hypertrophic Cardiomyopathy Detection.

BACKGROUND: Novel targeted treatments increase the need for prompt hypertrophic cardiomyopathy (HCM) detection. However, its low prevalence (0.5%) and resemblance to common diseases present challenges that may benefit from automated machine learning-based approaches. We aimed to develop machine learning models to detect HCM and to differentiate it from other cardiac conditions using ECGs and echocardiograms, with robust generalizability across multiple cohorts. METHODS: Single-institution HCM ECG models were trained and validated on external data. Multi-institution models for ECG and echocardiogram were trained on data from 3 academic medical centers in the United States and Japan using a federated learning approach, which enables training on distributed data without data sharing. Models were validated on held-out test sets for each institution and from a fourth academic medical center and were further evaluated for discrimination of HCM from aortic stenosis, hypertension, and cardiac amyloidosis. Last, automated detection was compared with manual interpretation by 3 cardiologists on a data set with a realistic HCM prevalence. RESULTS: We identified 74 376 ECGs for 56 129 patients and 8392 echocardiograms for 6825 patients at the 4 academic medical centers. Although ECG models trained on data from each institution displayed excellent discrimination of HCM on internal test data (C statistics, 0.88-0.93), the generalizability was limited, most notably for a model trained in Japan and tested in the United States (C statistic, 0.79-0.82). When trained in a federated manner, discrimination of HCM was excellent across all institutions (C statistics, 0.90-0.96 and 0.90-0.96 for ECG and echocardiogram model, respectively), including for phenotypic subgroups. The models further discriminated HCM from hypertension, aortic stenosis, and cardiac amyloidosis (C statistics, 0.84, 0.83, and 0.88, respectively, for ECG and 0.93, 0.94, 0.85, respectively, for echocardiogram). Analysis of electrocardiography-echocardiography paired data from 11 823 patients from an external institution indicated a higher sensitivity of automated HCM detection at a given positive predictive value compared with cardiologists (0.98 versus 0.81 at a positive predictive value of 0.01 for ECG and 0.78 versus 0.59 at a positive predictive value of 0.24 for echocardiogram). CONCLUSIONS: Federated learning improved the generalizability of models that use ECGs and echocardiograms to detect and differentiate HCM from other causes of hypertrophy compared with training within a single institution.

The effect of diabetes on surgical versus percutaneous left main revascularization outcomes: a systematic review and meta-analysis.

BACKGROUND: The optimal method of coronary revascularization for diabetes mellitus (DM) patients with left main coronary artery disease (LMCAD) is controversial in the drug-eluting stent (DES) era. METHODS: We performed a systematic review and meta-analysis comparing DES-based percutaneous coronary intervention (PCI) to coronary artery bypass grafting (CABG) for LMCAD in DM patients and tested for effect measure modification (EMM) by diabetes for adverse events. We included all randomized controlled trials (RCTs) and observational studies comparing CABG to DES-based PCI including DM patients with LMCAD published up to March 1, 2021. We completed separate random-effects meta-analyses for four RCTs (4356 patients, mean follow-up of 4.9 years) and six observational studies (9360 patients, mean follow-up of 5.2 years). RESULTS: In RCTs among DM patients, DES-based PCI, compared to CABG, was associated with a 30% increased relative risk (RR) (RR 1.30, 95% CI 1.09-1.56, I2 = 0%), while among non-DM patients, there was a 25% increased relative risk (RR 1.25, 95% CI 1.07-1.44, I2 = 0%) for the composite endpoint of all-cause mortality, myocardial infarction, stroke, and unplanned revascularization (MACCE). There was no evidence of EMM (p-value for interaction = 0.70). The mean weighted SYNTAX score was 25.7. In observational studies, there was no difference between DES-based PCI and CABG for all-cause mortality in patients with DM (RR 1.13, 95% CI 0.91-1.40, I2 = 0%). CONCLUSIONS: CABG was superior to PCI for LMCAD in RCTs in DM patients for MACCE. Heart teams may consider DM as one of the many components in the clinical decision-making process, but may not want to consider DM as a primary deciding factor between DES-based PCI and CABG for LMCAD with low to intermediate anatomical complexity in the other coronary arteries. STUDY REGISTRATION: CRD42021246931 (PROSPERO).

Importance of external validation and subgroup analysis of artificial intelligence in the detection of low ejection fraction from electrocardiograms.

Aim: Left ventricular systolic dysfunction (LVSD) carries an increased risk for overt heart failure and mortality, yet treatable to mitigate disease progression. An artificial intelligence (AI)-enabled 12-lead electrocardiogram (ECG) model demonstrated promise in LVSD screening, but the performance dropped unexpectedly in external validation. We thus sought to train de novo models for LVSD detection and investigated their performance across multiple institutions and across a broader set of patient strata. Methods and results: ECGs taken within 14 days of an echocardiogram were obtained from four academic hospitals (three in the United States and one in Japan). Four AI models were trained to detect patients with ejection fraction (EF) <40% using ECGs from each of the four institutions. All the models were then evaluated on the held-out test data set from the same institution and data from the three external institutions. Subgroup analyses stratified by patient characteristics and common ECG abnormalities were performed. A total of 221 846 ECGs were identified from the 4 institutions. While the Brigham and Women's Hospital (BWH)-trained and Keio-trained models yielded similar accuracy on their internal test data [area under the receiver operating curve (AUROC) 0.913 and 0.914, respectively], external validity was worse for the Keio-trained model (AUROC: 0.905-0.915 for BWH trained and 0.849-0.877 for Keio-trained model). Although ECG abnormalities including atrial fibrillation, left bundle branch block, and paced rhythm-reduced detection, the models performed robustly across patient characteristics and other ECG features. Conclusion: While using the same model architecture, different data sets produced models with different performances for detecting low-EF highlighting the importance of external validation and extensive stratification analysis.

Phenomapping in patients experiencing worsening renal function during hospitalization for acute heart failure.

AIMS: The impact of worsening renal function (WRF) on the prognosis of patients with acute heart failure (AHF) remains controversial. We aimed to identify phenotypically distinct subgroups among individuals with both AHF and WRF using cluster analysis. METHODS AND RESULTS: Overall, the data of 483 patients with both AHF and WRF enrolled in the West Tokyo Heart Failure Registry were analysed. Using cluster analysis, we identified three phenotypically distinct subgroups (phenogroups 1, 2, and 3). We assessed the impact of WRF on the prognosis of each phenogroup by comparing the incidence of composite endpoints, including all-cause death and re-hospitalization due to heart failure, with those of a propensity score-matched, non-WRF control group. Participants in phenogroup 1 (N = 122) were the youngest (69.3 ± 13.7 years), had relatively preserved estimated glomerular filtration rate (eGFR, 70.0 ± 27.7 mL/min/1.73 m2 ), and reduced left ventricular ejection fraction (LVEF) (41.8 ± 13.7%). Conversely, participants in phenogroup 3 (N = 122) were the oldest (81.7 ± 8.5 years), had the worst eGFR (33.0 ± 20.9 mL/min/1.73 m2 ), and had preserved LVEF (51.7 ± 14.8%). The characteristics of the participants in phenogroup 2 (N = 239) were between those of phenogroups 1 and 3. The propensity score matching analysis showed that WRF was associated with a higher incidence of composite endpoints in phenogroup 1, whereas this association was not observed in phenogroups 2 and 3. CONCLUSIONS: Using cluster analysis, we revealed three phenotypically distinct subgroups of patients with both AHF and WRF. WRF was associated with worse clinical outcomes in the subgroup of younger patients with reduced LVEF and preserved renal function.

Two Medullary Hemorrhage Cases Complicated by Respiratory Distress in the Early Phase.

Medullary hemorrhage is quite rare among brain stem hemorrhage cases, thus the clinical course remains unclear. In the medulla oblongata, respiratory centers are located and previous reports indicate that medullary lesions have possible relationship with acute respiratory distress syndrome. This kind of respiratory failure is commonly caused by neurogenic pulmonary edema (NPE), which is defined as noncardiac noninfectious acute respiratory distress syndrome with changes in intracranial condition including cerebrovascular events. However, to date, very few reports have described cases with medullary hemorrhage accompanied by NPE. We experienced 2 patients with medullary hemorrhages. A 65-year-old man presented with sudden onset of headache, whose head computed tomography showed right medullary hemorrhage. Another 76-year-old woman was transferred because of sudden limb weakness and diagnosed with left medullary hemorrhage. Digital subtraction angiography showed the presence of arteriovenous fistula in the medulla oblongata and drainer veins in the second case. Both cases were complicated by acute pulmonary edema in the early phase, suggesting the possible association of the medullary hemorrhage with NPE.