European heart journal. Digital health最新文献

{"title":"Sudden cardiac arrest prediction via deep learning electrocardiogram analysis.","authors":"Matt T Oberdier, Luca Neri, Alessandro Orro, Richard T Carrick, Marco S Nobile, Sujai Jaipalli, Mariam Khan, Stefano Diciotti, Claudio Borghi, Henry R Halperin","doi":"10.1093/ehjdh/ztae088","DOIUrl":"10.1093/ehjdh/ztae088","url":null,"abstract":"<p><strong>Aims: </strong>Sudden cardiac arrest (SCA) is a commonly fatal event that often occurs without prior indications. To improve outcomes and enable preventative strategies, the electrocardiogram (ECG) in conjunction with deep learning was explored as a potential screening tool.</p><p><strong>Methods and results: </strong>A publicly available data set containing 10 s of 12-lead ECGs from individuals who did and did not have an SCA, information about time from ECG to arrest, and age and sex was utilized for analysis to individually predict SCA or not using deep convolution neural network models. The base model that included age and sex, ECGs within 1 day prior to arrest, and data sampled from windows of 720 ms around the R-waves from 221 individuals with SCA and 1046 controls had an area under the receiver operating characteristic curve of 0.77. With sensitivity set at 95%, base model specificity was 31%, which is not clinically applicable. Gradient-weighted class activation mapping showed that the model mostly relied on the QRS complex to make predictions. However, models with ECGs recorded between 1 day to 1 month and 1 month to 1 year prior to arrest demonstrated predictive capabilities.</p><p><strong>Conclusion: </strong>Deep learning models processing ECG data are a promising means of screening for SCA, and this method explains differences in SCAs due to age and sex. Model performance improved when ECGs were nearer in time to SCAs, although ECG data up to a year prior had predictive value. Sudden cardiac arrest prediction was more dependent upon QRS complex data compared to other ECG segments.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 2","pages":"170-179"},"PeriodicalIF":3.9,"publicationDate":"2025-02-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11914729/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143665621","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An investigation into the causes of race bias in artificial intelligence-based cine cardiac magnetic resonance segmentation.","authors":"Tiarna Lee, Esther Puyol-Antón, Bram Ruijsink, Sebastien Roujol, Theodore Barfoot, Shaheim Ogbomo-Harmitt, Miaojing Shi, Andrew King","doi":"10.1093/ehjdh/ztaf008","DOIUrl":"10.1093/ehjdh/ztaf008","url":null,"abstract":"<p><strong>Aims: </strong>Artificial intelligence (AI) methods are being used increasingly for the automated segmentation of cine cardiac magnetic resonance (CMR) imaging. However, these methods have been shown to be subject to race bias; i.e. they exhibit different levels of performance for different races depending on the (im)balance of the data used to train the AI model. In this paper, we investigate the source of this bias, seeking to understand its root cause(s).</p><p><strong>Methods and results: </strong>We trained AI models to perform race classification on cine CMR images and/or segmentations from White and Black subjects from the UK Biobank and found that the classification accuracy for images was higher than for segmentations. Interpretability methods showed that the models were primarily looking at non-heart regions. Cropping images tightly around the heart caused classification accuracy to drop to almost chance level. Visualizing the latent space of AI segmentation models showed that race information was encoded in the models. Training segmentation models using cropped images reduced but did not eliminate the bias. A number of possible confounders for the bias in segmentation model performance were identified for Black subjects but none for White subjects.</p><p><strong>Conclusion: </strong>Distributional differences between annotated CMR data of White and Black races, which can lead to bias in trained AI segmentation models, are predominantly image-based, not segmentation-based. Most of the differences occur in areas outside the heart, such as subcutaneous fat. These findings will be important for researchers investigating performance of AI models on different races.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 3","pages":"350-358"},"PeriodicalIF":3.9,"publicationDate":"2025-02-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12088717/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144112595","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine-learning phenotyping of patients with functional mitral regurgitation undergoing transcatheter edge-to-edge repair: the MITRA-AI study.","authors":"Fabrizio D'Ascenzo, Filippo Angelini, Corrado Pancotti, Pier Paolo Bocchino, Cristina Giannini, Filippo Finizio, Marianna Adamo, Victoria Camman, Nuccia Morici, Leor Perl, Saverio Muscoli, Gabriele Crimi, Paolo Boretto, Ovidio de Filippo, Luca Baldetti, Giuseppe Biondi-Zoccai, Federico Conrotto, Sonia Petronio, Arturo Giordano, Rodrigo Estévez-Loureiro, Davide Stolfo, Christian Templin, Mauro Chiarito, Elena Cavallone, Veronica Dusi, Gianluca Alunni, Jacopo Oreglia, Mario Iannaccone, Marco Pocar, Matteo Pagnesi, Stefano Pidello, Ran Kornowski, Piero Fariselli, Simone Frea, Michele La Torre, Claudia Raineri, Giuseppe Patti, Italo Porto, Antonio Montefusco, Sergio Raposeiras Roubin, Gaetano Maria De Ferrari","doi":"10.1093/ehjdh/ztaf006","DOIUrl":"10.1093/ehjdh/ztaf006","url":null,"abstract":"<p><strong>Aims: </strong>Severe functional mitral regurgitation (FMR) may benefit from mitral transcatheter edge-to-edge repair (TEER), but selection of patients remains to be optimized.</p><p><strong>Objectives: </strong>The aim of this study was to use machine-learning (ML) approaches to uncover concealed connections between clinical, echocardiographic, and haemodynamic data associated with patients' outcomes.</p><p><strong>Methods and results: </strong>Consecutive patients undergoing TEER from 2009 to 2020 were included in the MITRA-AI registry. The primary endpoint was a composite of cardiovascular death or heart failure (HF) hospitalization at 1 year. External validation was performed on the Mitrascore cohort. 822 patients were included. The composite primary endpoint occurred in 250 (30%) patients. Four clusters with decreasing risk of the primary endpoint were identified (42, 37, 25, and 20% from Cluster 1 to Cluster 4, respectively). Clusters were combined into a high-risk (Clusters 1 and 2) and a low-risk phenotype (Clusters 3 and 4). High-risk phenotype patients had larger left ventriculars (LVs) (>107 mL/m²), lower left ventricular ejection fraction (<35%), and more prevalent ischaemic aetiology compared with low-risk phenotype patients. Within low-risk groups, permanent atrial fibrillation amplified that of HF hospitalizations. In the Mitrascore cohort, the incidence of the primary endpoint was 48, 52, 35, and 42% across clusters.</p><p><strong>Conclusion: </strong>A ML analysis identified meaningful clinical phenotypic presentations in FMR undergoing TEER, with significant differences in terms of cardiovascular death and HF hospitalizations, confirmed in an external validation cohort.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 3","pages":"340-349"},"PeriodicalIF":3.9,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12088727/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144112895","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comparison of machine learning and conventional criteria in detecting left ventricular hypertrophy and prognosis with electrocardiography.","authors":"Jui-Tzu Huang, Chih-Hsueh Tseng, Wei-Ming Huang, Wen-Chung Yu, Hao-Min Cheng, Hsi-Lu Chao, Chern-En Chiang, Chen-Huan Chen, Albert C Yang, Shih-Hsien Sung","doi":"10.1093/ehjdh/ztaf003","DOIUrl":"10.1093/ehjdh/ztaf003","url":null,"abstract":"<p><strong>Aims: </strong>Left ventricular hypertrophy (LVH) is clinically important; current electrocardiography (ECG) diagnostic criteria are inadequate for early detection. This study aimed to develop an artificial intelligence (AI)-based algorithm to improve the accuracy and prognostic value of ECG criteria for LVH detection.</p><p><strong>Methods and results: </strong>A total of 42 016 patients (64.3 ± 16.5 years, 55.3% male) were enrolled. LV mass index was calculated from echocardiographic measurements. Left ventricular hypertrophy screening utilized ECG criteria, including Sokolow-Lyon, Cornell product, Cornell/strain index, Framingham criterion, and Peguero-Lo Presti. An AI algorithm using CatBoost was developed and validated (training dataset 80% and testing dataset 20%). F1 scores, reflecting the harmonic mean of precision and recall, were calculated. Mortality data were obtained through linkage with the National Death Registry. The CatBoost-based AI algorithm outperformed conventional ECG criteria in detecting LVH, achieving superior sensitivity, specificity, positive predictive value, F1 score, and area under curve. Significant features to predict LVH involved QRS and P-wave morphology. During a median follow-up duration of 10.1 years, 1655 deaths occurred in the testing dataset. Cox regression analyses showed that LVH identified by AI algorithm (hazard ratio and 95% confidence interval: 1.587, 1.309-1.924), Sokolow-Lyon (1.19, 1.038-1.365), Cornell product (1.301, 1.124-1.505), Cornell/strain index (1.306, 1.185-1.439), Framingham criterion (1.174, 1.062-1.298), and echocardiography-confirmed LVH (1.124, 1.019-1.239) were all significantly associated with mortality. Notably, AI-diagnosed LVH was more predictive of mortality than echocardiography-confirmed LVH.</p><p><strong>Conclusion: </strong>Artificial intelligence-based LVH diagnosis outperformed conventional ECG criteria and was a superior predictor of mortality compared to echocardiography-confirmed LVH.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 2","pages":"252-260"},"PeriodicalIF":3.9,"publicationDate":"2025-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11914727/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143665569","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Smartphone for heart sound measurement in hospital: feasibility and influencing factors.","authors":"Yun Li, Zhanjiang Zhao, Aikeliyaer Ainiwaer, Daoju Mei, Peirong Zhang, Frits W Prinzen, Hongxing Luo","doi":"10.1093/ehjdh/ztaf007","DOIUrl":"10.1093/ehjdh/ztaf007","url":null,"abstract":"<p><strong>Aims: </strong>Smartphones have recently been utilized to measure heart sounds in the general population, but not yet in real-world hospital settings. This study aims to assess the feasibility of smartphones for heart sound measurement across various hospital departments and to identify the factors causing suboptimal heart sound measurements.</p><p><strong>Methods and results: </strong>The FonoCheck app was used to measure heart sounds from the chest of 296 hospitalized patients. Two assessors independently evaluated the quality of heart sound recordings based on the presence of the first and second heart sounds. Both environmental and patient-related factors were examined for their effects on heart sound quality. Visual assessments identified 254 (86%) good-quality heart sound recordings, with lower frequencies observed in the emergency room (67%), respiratory intensive care unit (78%), and general intensive care unit (82%). The heart sound recordings were affected by various types of noise, including respiration, conversation, motion, and interference from medical devices. However, patient demographics such as sex and body mass index were not associated with poor heart sound quality (<i>P</i> > 0.05), except for age which had a negative impact (<i>P</i> = 0.003). None of the patients' comorbidities, including atrial fibrillation, coronary artery disease, heart failure, and chronic obstructive pulmonary disease, significantly affected the heart sound measurements (<i>P</i> > 0.05).</p><p><strong>Conclusion: </strong>It is feasible to use smartphones to measure high-quality heart sounds in hospitals. However, environmental factors and patient's age may lead to suboptimal measurements. This study supports the future medical applications of FonoCheck app in hospital settings.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 3","pages":"486-495"},"PeriodicalIF":3.9,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12088729/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144112816","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Personalized app-based coaching for improving physical activity in heart failure with preserved ejection fraction patients compared with standard care: rationale and design of the MyoMobile Study.","authors":"Silav Zeid, Jürgen H Prochaska, Alexander Schuch, Sven Oliver Tröbs, Andreas Schulz, Thomas Münzel, Tanja Pies, Wilfried Dinh, Matthias Michal, Perikles Simon, Philipp Sebastian Wild","doi":"10.1093/ehjdh/ztae096","DOIUrl":"10.1093/ehjdh/ztae096","url":null,"abstract":"<p><strong>Aims: </strong>Patients suffering from heart failure with preserved ejection fraction (HFpEF) often exhibit a sedentary lifestyle, contributing to the worsening of their condition. Although there is an inverse relationship between physical activity (PA) and adverse cardiovascular outcomes, the implementation of Class Ia PA guidelines is hindered by low participation in supervised and structured programmes, which are not suitable for a diverse population of HFpEF patients. The MyoMobile study has been designed to assess the effect of a 12-week, app-based coaching programme on promoting PA in patients with HFpEF.</p><p><strong>Methods and results: </strong>The MyoMobile study was a single-centre, randomized, controlled three-armed parallel group clinical trial with prospective data collection to investigate the effect of a personalized mobile app health intervention compared with usual care on PA levels in patients with HFpEF. Major inclusion criteria were age ≥ 45 years, a diagnosis of HFpEF, LVEF > 40%, and current HF symptoms (NYHA Class I-III). Major exclusion criteria included acute decompensated HF, non-ambulatory status, recent acute coronary syndrome or cardiac surgery, alternative diagnoses for HF symptoms, active cancer treatment, and physical or medical conditions affecting mobility. Participants were recruited from hospitals, general practices, and practices specialized in internal medicine and cardiology in the Rhine-Main area, Germany. Participants underwent an objective 7-day PA measurement with a 3D accelerometer (Dynaport, McRoberts) at screening and after the 12-week intervention period. Following the screening, eligible participants were randomized into one of three groups: standard care (PA consulting), the intervention arm with app-based PA tracking and coaching, or the intervention arm with tracking but without coaching. The primary efficacy endpoint was the change in average daily step count between the average step count at baseline and at the end of the intervention, comparing standard care to a 12-week app-based PA coaching intervention.</p><p><strong>Conclusion: </strong>Exercise intolerance is a primary symptom in HFpEF patients, leading to poor quality of life and HF-related adverse outcomes due to physical inactivity. The MyoMobile study was designed to investigate the use of app-based coaching to improve PA in patients with HFpEF with a personalized, home-based intervention, focusing on simple step counts for flexibility and ease of integration into daily routines.</p><p><strong>Clinical trial registration: </strong>URL: https://clinicaltrials.gov/ct2/show/NCT04940312.</p><p><strong>Unique identifier: </strong>NCT04940312.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 2","pages":"298-309"},"PeriodicalIF":3.9,"publicationDate":"2025-01-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11914726/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143665620","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial intelligence for the analysis of intracoronary optical coherence tomography images: a systematic review.","authors":"Ruben G A van der Waerden, Rick H J A Volleberg, Thijs J Luttikholt, Pierandrea Cancian, Joske L van der Zande, Gregg W Stone, Niels R Holm, Elvin Kedhi, Javier Escaned, Dario Pellegrini, Giulio Guagliumi, Shamir R Mehta, Natalia Pinilla-Echeverri, Raúl Moreno, Lorenz Räber, Tomasz Roleder, Bram van Ginneken, Clara I Sánchez, Ivana Išgum, Niels van Royen, Jos Thannhauser","doi":"10.1093/ehjdh/ztaf005","DOIUrl":"10.1093/ehjdh/ztaf005","url":null,"abstract":"<p><p>Intracoronary optical coherence tomography (OCT) is a valuable tool for, among others, periprocedural guidance of percutaneous coronary revascularization and the assessment of stent failure. However, manual OCT image interpretation is challenging and time-consuming, which limits widespread clinical adoption. Automated analysis of OCT frames using artificial intelligence (AI) offers a potential solution. For example, AI can be employed for automated OCT image interpretation, plaque quantification, and clinical event prediction. Many AI models for these purposes have been proposed in recent years. However, these models have not been systematically evaluated in terms of model characteristics, performances, and bias. We performed a systematic review of AI models developed for OCT analysis to evaluate the trends and performances, including a systematic evaluation of potential sources of bias in model development and evaluation.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 2","pages":"270-284"},"PeriodicalIF":3.9,"publicationDate":"2025-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11914731/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143665564","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mobile/wearable digital devices for care of active cancer patients: a survey from the ESC Council of Cardio-Oncology.","authors":"Giuseppe Boriani, Jacopo F Imberti, Riccardo Asteggiano, Pietro Ameri, Davide A Mei, Michał Farkowski, Julian Chun, Josè Luis Merino, Teresa Lopez-Fernandez, Alexander R Lyon","doi":"10.1093/ehjdh/ztae082","DOIUrl":"10.1093/ehjdh/ztae082","url":null,"abstract":"<p><strong>Aims: </strong>The Council of Cardio-Oncology of the European Society of Cardiology developed an on-line anonymous survey to provide an overall picture of the current practice on the use of mobile and wearable digital devices in cardio-oncology and the potential barriers to their large-scale applicability.</p><p><strong>Methods and results: </strong>Between June 2023 and January 2024, an online anonymous questionnaire was completed by 220 healthcare professionals from 55 countries. The greatest number of respondents reported that mobile/wearable digital devices have a role in all active cancer patients for measuring heart rate (33.9%), blood pressure (34.4%), body temperature (32.0%), physical activity (42.4%), and sleep (31.2%). In the setting of atrial fibrillation detection, respondents were evenly split between applying these technologies in all patients (33.0%) or only in selected patients (33.0%). Regarding QTc interval monitoring, 30.6% reported that mobile/wearable digital devices play a role only in selected patients. The decision to use the device was taken by the patient in 56.6% of cases and the physician in 43.4%. The most important barrier reported to mobile/wearable device implementation in the setting of cardiac rhythm monitoring and QTc measurement was their cost (weighted average: 3.38 and 3.39, respectively).</p><p><strong>Conclusion: </strong>Mobile/wearable digital devices are considered to play an important role in different settings of cardio-oncology, including monitoring of patients' parameters and arrhythmia detection. Their role in monitoring physical activity and QTc interval appears more nuanced. The most important perceived barrier to mobile/wearable digital device implementation is considered their high cost.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 2","pages":"162-169"},"PeriodicalIF":3.9,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11914721/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143665548","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From marginal gains to clinical utility: machine learning-based percutaneous coronary intervention risk prediction models.","authors":"Muhammad Ibtsaam Qadir, Ravi S Hira, Fiona R Kolbinger","doi":"10.1093/ehjdh/ztaf001","DOIUrl":"10.1093/ehjdh/ztaf001","url":null,"abstract":"","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 2","pages":"159-161"},"PeriodicalIF":3.9,"publicationDate":"2025-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11914716/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143665597","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An ensemble learning model for detection of pulmonary hypertension using electrocardiogram, chest X-ray, and brain natriuretic peptide.","authors":"Risa Kishikawa, Satoshi Kodera, Naoto Setoguchi, Kengo Tanabe, Shunichi Kushida, Mamoru Nanasato, Hisataka Maki, Hideo Fujita, Nahoko Kato, Hiroyuki Watanabe, Masao Takahashi, Naoko Sawada, Jiro Ando, Masataka Sato, Shinnosuke Sawano, Hiroki Shinohara, Koki Nakanishi, Shun Minatsuki, Junichi Ishida, Katsuhito Fujiu, Hiroshi Akazawa, Hiroyuki Morita, Norihiko Takeda","doi":"10.1093/ehjdh/ztae097","DOIUrl":"10.1093/ehjdh/ztae097","url":null,"abstract":"<p><strong>Aims: </strong>Delayed diagnosis of pulmonary hypertension (PH) is a known cause of poor patient prognosis. We aimed to develop an artificial intelligence (AI) model, using ensemble learning method to detect PH using electrocardiography (ECG), chest X-ray (CXR), and brain natriuretic peptide (BNP), facilitating accurate detection and prompting further examinations.</p><p><strong>Methods and results: </strong>We developed a convolutional neural network model using ECG data to predict PH, labelled by ECG from seven institutions. Logistic regression was used for the BNP prediction model. We referenced a CXR deep learning model using ResNet18. Outputs from each of the three models were integrated into a three-layer fully connected multimodal model. Ten cardiologists participated in an interpretation test, detecting PH from patients' ECG, CXR, and BNP data both with and without the ensemble learning model. The area under the receiver operating characteristic curves of the ECG, CXR, BNP, and ensemble learning model were 0.818 [95% confidence interval (CI), 0.808-0.828], 0.823 (95% CI, 0.780-0.866), 0.724 (95% CI, 0.668-0.780), and 0.872 (95% CI, 0.829-0.915). Cardiologists' average accuracy rates were 65.0 ± 4.7% for test without AI model and 74.0 ± 2.7% for test with AI model, a statistically significant improvement (<i>P</i> < 0.01).</p><p><strong>Conclusion: </strong>Our ensemble learning model improved doctors' accuracy in detecting PH from ECG, CXR, and BNP examinations. This suggests that earlier and more accurate PH diagnosis is possible, potentially improving patient prognosis.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"6 2","pages":"209-217"},"PeriodicalIF":3.9,"publicationDate":"2025-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11914732/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143665494","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}