RT Journal Article SR Electronic T1 Screening for Chagas disease from the electrocardiogram using a deep neural network JF medRxiv FD Cold Spring Harbor Laboratory Press SP 2023.01.24.23284930 DO 10.1101/2023.01.24.23284930 A1 Carl Jidling A1 Daniel Gedon A1 Thomas B. Schön A1 Claudia Di Lorenzo Oliveira A1 Clareci Silva Cardoso A1 Ariela Mota Ferreira A1 Luana Giatti A1 Sandhi Maria Barreto A1 Ester C. Sabino A1 Antonio L. P. Ribeiro A1 Antônio H. Ribeiro YR 2023 UL http://medrxiv.org/content/early/2023/01/26/2023.01.24.23284930.abstract AB Background Worldwide it is estimated that more than 6 million people are infected with Chagas disease (ChD). It is considered one of the most important neglected diseases and, when it reaches its chronic phase, the infected person often develops serious heart conditions. While early treatment can avoid complications, the condition is often not detected during its early stages. We investigate whether a deep neural network can detect ChD from electrocardiogram (ECG) tracings. The ECG is inexpensive and it is often performed during routine visits. Being able to evaluate ChD from this exam can help detect potentially hidden cases in an early stage.Methods We use a convolutional neural network model, which takes the 12-lead ECG as input and outputs a scalar number associated with the probability of a Chagas diagnosis. To develop the model, we use two data sets, which jointly consist of over two million entries from Brazilian patients, compiled by the Telehealth Network of Minas Gerais within the SaMi-Trop (São Paulo-Minas Gerais Tropical Medicine Research Center) study focused on ChD patients and enriched with the CODE (Clinical Outcomes in Digital Electrocardiology) study focused on a general population. The performance is evaluated on two external data sets of 631 and 13,739 patients, collected in the scope of the REDS-II (Retrovirus Epidemiology Donor Study-II) study and of the ELSA-Brasil (Brazilian Longitudinal Study of Adult Health) study. The first study focuses on ChD patients and the second data set originates from civil servants from five universities and one research institute.Findings Evaluating our model, we obtain an AUC-ROC value of 0.80 (CI 95% 0.79-0.82) for the validation data set (with samples from CODE and SaMi-Trop), and in external validation datasets: 0.68 (CI 95% 0.63-0.71) for REDS-II and 0.59 (CI 95% 0.56-0.63) for ELSA-Brasil. In these external validation datasets, we report a sensitivity of 0.52 (CI 95% 0.47-0.57) and 0.36 (CI 95% 0.30-0.42) and a specificity of 0.77 (CI 95% 0.72-0.81) and 0.76 (CI 95% 0.75-0.77), respectively, in REDS-II and ELSA-Brasil. We also evaluated the model for considering only patients with Chagas cardiomyopathy as positive. In this case, the model attains an AUC-ROC of 0.82 (CI 95% 0.77-0.86) for REDS-II and 0.77 (CI 95% 0.68-0.85) for ELSA-Brasil.Interpretation The results indicate that the neural network can detect patients who developed chronic Chagas cardiomyopathy (CCC) from the ECG and – with weaker performance – detect patients before the CCC stage. Future work should focus on curating large and better datasets for developing such models. The CODE is the largest dataset available to us, and their labels are self-reported and less reliable than our other data sets, i.e. REDS-II and ELSA-Brasil. This, we believe, limits our model performance in the case of non-CCC patients. We are positive that our findings constitute the first step towards building tools for more efficient detection and treatment of ChD, especially in high-prevalent regions.Funding This research is financially supported by the Swedish Foundation for Strategic Research (SSF) via the project ASSEMBLE (Contract number: RIT 15-0012), by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by Knut and Alice Wallenberg Foundation, by Kjell och Märta Beijer Foundation, by the Brazilian Agencies CNPq, CAPES, and FAPEMIG, and by projects IATS, and CIIA-Saúde. The ELSA-Brasil study was supported by the Brazilian Ministries of Health and of Science and Technology (grants 01060010.00RS, 01060212.00BA, 01060300.00ES, 01060278.00MG, 01060115.00SP, and 01060071.00RJ). The SaMi-Trop and REDS-II cohort studies are supported by the National Institutes of Health (P50 AI098461-02, U19AI098461-06, 1U01AI168383-01). LG, SMB, ECS and ALPR receive unrestricted research scholarships from CNPq; ALPR received a Google Latin America Research Award scholarship. The funders had no role in the study design; collection, analysis, and interpretation of data; writing of the report; or decision to submit the paper for publication.Evidence before this study Chagas disease (ChD) is a neglected tropical disease, and the diagnosis relies on blood testing of patients from endemic areas. However, there is no clear recommendation on selecting patients for serological diagnosis in those living in endemic regions. Since most of the patients with Chronic ChD are asymptomatic or oligosymptomatic, the diagnostic rates are low, preventing patients from receiving adequate treatment. The Electro-cardiogram (ECG) is a widely available, low-cost exam, often available in primary care settings in endemic countries. Artificial intelligence (AI) algorithms on ECG tracings have allowed the detection of hidden conditions, such as cardiomyopathies and left ventricular systolic dysfunction.Added value of this study To the best of our knowledge, this is the first study that presents an AI model for the automatic detection of ChD from the ECG. As part of the model development, we utilise established large cohorts of patients from the relevant population of all-comers in affected regions in the state of Minas Gerais, Brazil. We evaluate the model on data sets with high-quality ground truth labels obtained from the patients’ serological status. Our model has moderate diagnostic performance in recognition of ChD and better accuracy in detecting Chagas cardiomyopathy.Implications of all the available evidence Our findings demonstrate a promising AI-ECG-based model capacity for discriminating patients with chronic Chagas cardiomyopathy (CCC). However, detecting ChD patients without CCC is still insufficient, and further developments that lead to higher performance are needed. We believe this can be achieved with the addition of epidemiological questions, and that our model can be a useful tool in helping pre-selecting patients for further testing in order to determine the infection with ChD. The use of AI-ECG-based strategies for recognizing CCC patients deserves to be tested in the clinical setting.Competing Interest StatementThe authors have declared no competing interest.Funding StatementThis research is financially supported by the Swedish Foundation for Strategic Research (SSF) via the project ASSEMBLE (Contract number: RIT 15-0012), by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by Knut and Alice Wallenberg Foundation, by Kjell och Märta Beijer Foundation, by the Brazilian Agencies CNPq, CAPES, and FAPEMIG, and by projects IATS, and CIIA-Saúde. The ELSA-Brasil study was supported by the Brazilian Ministries of Health and of Science and Technology (grants 01060010.00RS, 01060212.00BA, 01060300.00ES, 01060278.00MG, 01060115.00SP, and 01060071.00RJ). The SaMi-Trop and REDS-II cohort studies are supported by the National Institutes of Health (P50 AI098461-02, U19AI098461-06, 1U01AI168383-01). LG, SMB, ECS and ALPR receive unrestricted research scholarships from CNPq ALPR received a Google Latin America Research Award scholarship. The funders had no role in the study design collection, analysis, and interpretation of data writing of the report or decision to submit the paper for publication.Author DeclarationsI confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.YesThe details of the IRB/oversight body that provided approval or exemption for the research described are given below:This study complies with all relevant ethical regulations. CODE Study was approved by the Research Ethics Committee of the Universidade Federal de Minas Gerais, protocol 49368496317.7.0000.5149. Since this is a secondary analysis of anonymized data stored in the TNMG, informed consent was not required by the Research Ethics Committee for the present study. ELSA-Brasil was approved by the Research Ethics Committees of the participating institutions and by the National Committee for Research Ethics (CONEP 976/2006) of the Ministry of Health. Sami-Trop study was approved by the Brazilian National Institutional Review Board (CONEP), No. 179.685/2012. In both investigations, all human subjects were adults who gave written informed consent. All researchers who deal with datasets signed terms of confidentiality and data utilization.I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.YesI understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).YesI have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable.YesSaMi-Trop cohort was made openly available (https://doi.org/10.5281/zenodo.4905618}). The CODE-15\% cohort was also made openly available (https://doi.org/10.5281/zenodo.4916206). The data sets contain information about Chagas condition mortality, age, sex, the ECG tracings, and the flag indicating whether the ECG tracing is normal. The DNN model parameters that give the results presented in this paper are also available (https://doi.org/10.5281/zenodo.7371623). This should allow the reader to partially reproduce the results presented in the paper. Restrictions apply to additional clinical information on the CODE-15\% and SaMi-Trop cohorts to the full CODE cohort to the REDS-II dataset and, to the ELSA-Brasil cohort. Researchers affiliated to educational or research institutions might make requests to access the data sets. Requests should be made to the corresponding author of this paper. They will be forwarded and considered on an individual basis by the Telehealth Network of Minas Gerais, SaMi-Trop and ELSA-Brasil Steering Committee. An estimate for the time needed for data access requests to be evaluated is three months. If approved, any data use will be restricted to non-commercial research purposes. The data will only be made available on the execution of appropriate data use agreements. https://doi.org/10.5281/zenodo.4905618 https://doi.org/10.5281/zenodo.4916206