Deep learning unmasks the ECG signature of Brugada syndrome — TEI-tagged view

Luke MelopersName, Giuseppe CicontepersName, Ashton ChristypersName, Gabriele VicedominipersName, Luigi AnastasiapersName, Carlo PapponepersName, Edward GrantpersName (2023date). “Deep learning unmasks the ECG signature of Brugada syndrome.” PNAS Nexustitle·j.

Object of study — detecting Brugada syndrometerm·dom from baseline ECGsterm·mth with a deep neural networkterm·mth

Research question — whether a deep neural networkterm·mth can detect the signature of BrSterm·dom in a conventional ECGterm·mth when it is invisible to the eye

Contribution — an ECGterm·mth screening algorithm for BrSterm·dom that competes with the sodium-channel-blocker challenge without its side effects

Full text — tagged

Deep learning unmasks the ECGterm·mth signature of Brugada syndrometerm·dom Abstract One in 10 cases of sudden cardiac deathterm·dom strikes without warning as the result of an inherited arrhythmic cardiomyopathy, such as Brugada Syndrometerm·dom (BrSterm·dom). Normal physiological variations often obscure visible signs of this and related life-threatening channelopathies in conventional electrocardiogramsterm·mth (ECGsterm·mth). Sodium channel blockersterm·mth can reveal previously hidden diagnostic ECGterm·mth features, however, their use carries the risk of life-threatening proarrhythmic side effects. The absence of a nonintrusive test places a grossly underestimated fraction of the population at risk of SCDterm·dom. Here, we present a machine-learning algorithm that extracts, aligns, and classifies ECGterm·mth waveforms for the presence of BrSterm·dom. This protocol, which succeeds without the use of a sodium channel blockerterm·mth (88.4% accuracy, 0.934 AUCterm·key in validation), can aid clinicians in identifying the presence of this potentially life-threatening heart disease.

Keywords: artificial intelligence, Brugada Syndrometerm·dom, deep neural networkterm·mth, electrocardiogramterm·key, Sudden Cardiac Deathterm·dom Sudden cardiac deathterm·dom (SCDterm·dom) persists as a significant public health problem, claiming millions of lives every year, including many young and otherwise healthy individuals. Among patients without evident structural heart disease, SCDterm·dom often occurs as the first clinical manifestation of an inherited disorder, such as Brugada Syndrometerm·dom (BrSterm·dom). Unfortunately, such patients seldom show overt signs of disease. The present study demonstrates that a deep neural networkterm·mth can detect the signature of BrSterm·dom in an ECGterm·mth, even in the frequent case when these signs are undetectable by the human eye. This tool, when applied to routine clinical ECGsterm·mth, will alert clinicians to potentially life-threatening conditions that would otherwise go undiagnosed.

Introduction Sudden cardiac deathterm·dom (SCDterm·dom) kills more than 4.2 million people worldwide every year ( 1 ). Among young and otherwise healthy individuals, SCDterm·dom often appears without warning as the first clinical manifestation of an inherited arrhythmic disease ( 2–4 ), such as Brugada Syndrometerm·dom (BrSterm·dom) ( 5–8 ). Two-thirds of patients afflicted with BrSterm·dom show no symptoms of cardiovascular disease, and SCDterm·dom often serves as its first clinical manifestation ( 5 ).

The hidden and dynamic nature of this disease underlines the need for an improved diagnostic test ( 9 ). The incidence of confirmed BrSterm·dom varies by country ( 5 ); but, the difficulty of diagnosis coupled with the complex genetic architecture ( 10 , 11 ) and inheritability of this disease suggests that its true prevalence is unknown. Most affected patients present a normal ECGterm·mth, and nearly half of BrS-afflicted individuals who survive a cardiac arrest fail to show a diagnostic type 1 patternterm·key in a routine ECGterm·mth ( 12 ). Both factors severely hinder the recognition of this lethal disease.

Where permitted, cardiologists may administer one of several sodium channel blockersterm·mth (SCBterm·mth) to unmask the BrSterm·dom type 1 patternterm·key ( 5 , 13 , 14 ) as demonstrated in Fig. 1 . However, these drugs can cause life-threatening proarrhythmic effects, which limit their use to those individuals with a known family history of SCDterm·dom ( 15 , 16 ). And ajmalineterm·key, the most effective diagnostic SCBterm·mth, is unavailable for clinical use in many countries. including the United States of America and Japan. Thus, the development of an easily applied, nonintrusive BrSterm·dom diagnostic test remains an important and pressing objective.

Representative heartbeatsterm·key recorded for V1, V2, and V3 in ( left ) baseline electrocardiogramsterm·mth of three typical patients diagnosed as spontaneous type 1 BrSterm·dom( + ) ( top ), BrSterm·dom( + ) ( middle ), and BrSterm·dom( − ) ( bottom ). Electrocardiogramsterm·mth are recorded during ajmalineterm·key administration for patients not presenting a spontaneous type 1 patternterm·key ( right ). Note the ST-elevation feature signifying BrSterm·dom in the type 1 ECGterm·mth traces. In our study, only 17% of BrSterm·dom( + ) patients exhibit a spontaneous type 1 patternterm·key, while the remaining 83% of patients required an ajmalineterm·key challenge to reveal a hidden type 1 patternterm·key.

The interpretation of an electrocardiogramterm·key for medical diagnosis requires the application of expert knowledge in a full consideration of the many interconnected factors that influence ECGterm·mth waveform patterns. This central problem of learning and retention lends itself well to an approach that classifies on the basis of a deep neural networkterm·mth (DNNterm·mth), tied to a very large library of digital waveform standards ( 17 ). An extensive body of effort has applied artificial intelligence with the processing power of modern computers to recognize diagnostic features of gross rhythm disorders such as atrial fibrillation, and certain common cardiomyopathies for which large digital ECGterm·mth databases exist ( 18–30 ). However, thus far, machine learning has yet to establish digital signatures for more subtle, but no less-threatening conditions such as BrSterm·dom, often harbored by asymptomatic individuals who present apparently normal ECGsterm·mth. The difficulty of assembling a large-scale digitized dataset of expertly diagnosed ECGsterm·mth presents a significant challenge to progress ( 31 ).

Here, we describe the development of a robust noninvasive method to diagnose the electrophysiological signature of BrSterm·dom from the digital analysis of a routine clinical electrocardiogramterm·key. This work differs substantially from the many other machine-learning approaches to ECGterm·mth classification, particularly those that target BrSterm·dom with only a capacity to identify affected individuals who present a visible type 1 patternterm·key in their baseline ECGterm·mth ( 32–35 ).

Rather than surveying a continuous ECGterm·mth for abnormalities, much like a medical professional might perform a visual assessment, our approach reduces a standard multi-lead ECGterm·mth trace to a set of high-quality representative single heartbeats, one for each lead, and then classifies upon this basis. This preprocessing strategy better isolates the persistent baseline function of the heart’s electrical conduction system by overcoming transient uncorrelated beat-to-beat variance present in the raw traces. This dimensionality reduction helps the neural network recognize the persistent ECGterm·mth features that most consistently encode for BrSterm·dom.

Methods Study population The present study draws upon an existing database of 1,455 consecutive patients undergoing electrophysiological evaluation for the diagnosis of BrSterm·dom at the Arrhythmia Department of IRCCS Policlinico San Donato (San Donato Milanese, Milano, Italy), beginning in April 2015 (Fig. 2 and Table S1 ).

(A) Flowcharts illustrating the organization of electrocardiogramsterm·mth into one cohort for the purposes of training and validating DNNterm·mth models, and into a second cohort for independent validation. (B) Distributions of patient clinical statistics, including medical and family history in the Claris training cohort. RBBB refers to right bundle branch block. Complete details can be found in the Supplementary material .

A second set of ECGsterm·mth, collected in a separate clinic beginning in March 2020, serves as a second independent validation cohort, consisting of 405 consecutive new patients in the Brugada screening program (Table S2 ). The data collected for each patient include a physical examination, complete personal and family medical histories, and a complete set of digital ECGterm·mth traces.

These populations each contain two overarching classification groups. Patients in the Brugada Group BrSterm·dom( + ) present evidence of a spontaneous or drug-induced type 1 BrSterm·dom ECGterm·mth pattern. The Control Group BrSterm·dom( − ) is comprised of patients, all of whom presented a negative sodium channel blocker testterm·mth, defined as an ECGterm·mth that does not convert to yield a type 1 patternterm·key after a pharmacological provocation test (ajmalineterm·key injection at the maximal dose of 1 mg/kg, according to the current consensus statement ( 5 )). Although we obtained ECGsterm·mth for eligible patients during ajmalineterm·key administration to confirm BrSterm·dom diagnoses, we trained and tested our classification algorithms exclusively on baseline ECGsterm·mth, recorded without ajmalineterm·key.

The study patients presented as candidates for ajmalineterm·key infusion owing to diagnostic indicators of risk, associated with the possibility of BrSterm·dom for at least one of the following reasons: (i) Family history of SCDterm·dom or family history of BrSterm·dom, (ii) unexplained syncope, (iii) nondiagnostic Brugada pattern or suspicious ECGterm·mth, (iv) aborted cardiac arrest, (v) unexplained ventricular arrhythmias. The exclusion criteria for the ajmalineterm·key test included the presence of a spontaneous type 1 patternterm·key ( n = 106 ; 7.3 % ), an established diagnosis of an overt structural cardiomyopathy, and severe conduction disturbances (e.g. left or right bundle branch block with QRS > 140 ms ). During the study period, a small number of patients ( n = 53 ; 3.5 % ) refused to undergo the ajmalineterm·key test. This retrospective study subscribed to the plan described in a protocol approved by the Ethical Committee of the I.R.C.C.S. Policlinico San Donato. All participants provided informed consent.

Acquisition and storage of ECGterm·mth traces All ECGterm·mth records were digitally acquired and stored in the Arrhythmology Department digital repository at the IRCCS Policlinico San Donato. The principal calibration and validation effort relied upon measurements made using a set of Workmate Claris instruments (Abbott Laboratories, Abbott Park, IL, USA). The hospital outpatient clinic afforded ECGsterm·mth for the supplemental validation cohort recorded using a Mortara ELI 350 system (Mortara Instrument, Inc., Milwaukee, WI, USA).

Both systems were configured to simultaneously acquire and digitize two limb leads (L1, L2) and six precordial leads (V1 through V6), and automatically reconstruct the 12-lead ECGterm·mth. The Claris system records the ECGterm·mth with a sampling frequency of 2,000 Hz and a resolution of 5 μ V, while the Mortara system records at a sampling frequency of 1,000 Hz and a resolution of 1 μ V. All ECGterm·mth records were collected using a customized high precordial lead placement optimized for the detection of the BrSterm·dom ECGterm·mth pattern. The six precordial leads were placed at the left and right parasternal positions in the second (V1 II ICS, or V1; V2 II ICS, or V2), third (V1 III ICS, or V3; V2 III ICS, or V4), and fourth intercostal spaces (V1 IV ICS, or V5; V2 IV ICS, or V6).

All patients exhibited stable sinus rhythm during baseline ECGterm·mth acquisition without premature ectopies. All baseline ECGterm·mth records were analyzed by three trained and independent cardiologists, who each provided an initial index classification: (i) normal (i.e. no evident abnormalities resembling a BrSterm·dom pattern), (ii) spontaneous type 1 BrS patternterm·dom, and (iii) suspicious BrSterm·dom pattern (i.e. ECGterm·mth abnormalities including but not limited to complete or incomplete right bundle branch block, type 2, or type 3 morphology).

Routine diagnosis calls for the acquisition of a baseline ECGterm·mth, followed by one recorded prior to and during an ajmalineterm·key challenge for all cases, except for patients whose baseline ECGterm·mth exhibited a spontaneous type 1 BrS patternterm·dom. Approximately one minute of an ECGterm·mth recording prior to ajmalineterm·key administration was recorded, anonymized, and exported as a set of text files from the hospital database. Additional patient clinical characteristics, including age, sex, personal history of syncope and cardiac arrest, and family history of sudden death and BrSterm·dom diagnosis, were also logged in the export files as additional factors of variation. This information was archived in an SQL database. ECGterm·mth traces recorded on the Mortara system were collected for a new set of patients upon admission to the hospital and reserved for independent validation. ECGterm·mth traces recorded from both systems are subject to the same ECGterm·mth trace processing, outlined below.

ECGterm·mth trace processing We reduce the pulse train in each lead to a single de-noised and representative heartbeatterm·key. This procedure, which is detailed in the Supplementary material , combined with down-sampling to 200 Hz, yields a much sparser representation (Fig. 3 a). Averaged, single-lead heartbeats, concatenated end-to-end, forms a fused, one-dimensional representation of the entire electrocardiogramterm·key for each patient. The results presented in this study classify on the basis of a 9-lead ECGterm·mth (all leads excluding the unipolar limb leads aVF, aVR, and aVL). In the Supplementary material , we additionally explore individual and other lead combinations in detail.

(A) Flowchart illustrating the ECGterm·mth data processing workflow from trains of heartbeats as monitored by each ECGterm·mth lead, to segmentation, outlier discrimination, and averaging to form a set of single representative beats. (B) Structure of the neural network used for training the Brugada syndrometerm·dom classifier including input, hidden, and output layers. The hidden layer unit includes a fully connected layer followed by Gaussian noise, ReLU activation, and batch normalization. The hidden layer unit is repeated three times, followed by a fully-connected output layer with sigmoid activation. (C) Data partitioning strategy for training a neural network. The overall data partitioning scheme implements leave-one-out cross-validationterm·mth (LOOCVterm·mth) to assess the performance of independent neural networks constructed for every patient. For each patient, a 7-fold cross-validation separates the dataset into seven training and validation subsets, each tested on the holdout patient. This analysis yields an average score and standard deviation for each patient across the 7-fold DNNsterm·mth.

The results reported here stem from the analysis of retrospective clinical databases. The ECGsterm·mth in these databases were not recorded for the purpose of digitally training a machine-learning algorithm, but rather to enable a clinician to assess visually for the presence of diagnostic features. Consequently, we found it necessary to screen for measurement artifacts and deemed some recordings unsuitable for multivariate analysis. We performed this screening manually prior to any application of machine learning. As diagrammed in Fig. 2 a, 1,455 unique patients yielded a set of 1,154 high-quality ECGsterm·mth. A total of 370 of 405 ECGsterm·mth met acceptance standards in the Mortara validation cohort.

Neural networks The representative heartbeatsterm·key for each patient, along with the attendant factors of variation populate a design matrix X E C G . We define a disease-state vector Y by assigning 0 or 1 to every ECGterm·mth, representing Brugada-negative and-positive, respectively, determined by the presence or the absence of a type 1 ECGterm·mth pattern, confirmed where necessary by a positive response to ajmalineterm·key.

We train a deep neural networkterm·mth (DNNterm·mth), consuming the elements of the design matrix X E C G . supervised by the disease state Y , to optimize the parameters of a classification function F ( X E C G ) = Y (Fig. 3 b). Details about the training, optimization, and validation procedures for our neural networks can be found in the Supplementary material . All computer codes developed for ECGterm·mth processing, neural network training, and statistical analysis are written in-house using Python 3.9 and Tensorflow 2.

Independent validation of DNNsterm·mth Two distinct tests confirm the validity with which the DNNterm·mth models developed in this work classify newly encountered ECGsterm·mth for BrSterm·dom. A large-scale process of leave-one-out cross-validationterm·mth (LOOCVterm·mth) (Fig. 3 c) independently gauges the degree to which ensemble learning correctly classifies each and every individual patient in the training dataset.

In a procedure repeated for all N patients in the database, we withhold one patient in turn and use the N − 1 other patients to train a distinct set of seven DNNterm·mth models that are blind to the withheld patient. We employ ensemble learning for each N − 1 training subset to minimize sampling bias and overfitting. Consequently, a unique set of seven DNNterm·mth models trained on exclusive N − 1 populations subsampled by 7-fold cross-validation independently classify each patient.

By design, the corresponding set of seven independent DNNterm·mth models assigns a set of seven numbers between 0 and 1 to every withheld ECGterm·mth, which we term the DNN scoreterm·mth, or y ^ , which could be interpreted as a probability of positive diagnosis. We compute the final DNN scoreterm·mth for each validated patient as the average of the scores produced by the corresponding seven exclusive DNNsterm·mth. We further test the utility of a DNNterm·mth model trained by all the Claris patients by deploying it to classify a second independent cohort of validation data containing 370 patients (Mortara dataset).

Interpreting statistical results We gauge the validation success of the DNNterm·mth models by means of the areas under curve (AUCterm·key) of receiver-operating characteristics (ROC) plots ( 36 ). DNN scoresterm·mth close to 0 or 1 indicate high classification certainty for the absence or the presence of a BrSterm·dom pattern, respectively, while scores close to the decision threshold of 0.5 point to ECGsterm·mth that yield a low classification certainty. The decision threshold of 0.5 is determined using Youden’s J statisticterm·stat ( 37 , 38 ).

To assess our results, we compare continuous variables using a Student t -test and binomial variables with χ 2 or Fisher’s exact test, as appropriate. We compute the AUCterm·key of ROC curves of DNNterm·mth performance, determining confidence intervals by DeLong’s method ( 36 ), and recognizing statistical significance for p < 0.05 . Further details on statistical analyses and methods for minimizing overfitting are described in the Supplementary material .

Results Independent validation within the claris training dataset The DNNterm·mth, applied in LOOCVterm·mth within the Claris (training) dataset, correctly predicts the ajmaline responseterm·var for 84.0% (969 of 1,154) of patients, with an AUCterm·key of 0.912 ± 0.017 (Fig. 4 ), a sensitivity of 85.4%, and a specificity of 82.4% (Tables 1 , S1 and S5 ). Furthermore, the DNNterm·mth classifies patients showing a spontaneous type 1 patternterm·key with an accuracy of 100% (103 of 103).

Receiver–operator characteristic curve for the Claris (training, green, plotted foreground) and Mortara (independent validation, orange, plotted foreground) cohorts.

Classification metrics of the trained neural networks on independent holdout cases in the claris (training) and mortara (independent validation) cohorts.

We determine that the 9-lead fused representative ECGterm·mth trace yields the best classification performance by comparing AUCterm·key values for various lead combinations (Table S4 ). The incorporation of clinical factors of variation such as age, gender, and family history of cardiac events did not significantly change the DNNterm·mth performance (Table S6 ). In a thorough exploration of classification strategies, we observed superior performance of the training dataset DNNterm·mth to other popular supervised learning techniques such as a convolutional neural network (CNN) (Tables S8 and S9 ), decision trees, Naive Bayes, support vector machines (SVM), and k-nearest neighbors (KNN), among others (Table S12 ).

Cross-platform independent validation We performed a second independent validation of a seven-DNN ensemble trained on the basis of the 1,154 patients in the Claris (training) cohort by applying this model to classify for BrSterm·dom in a new prospectively-studied cohort. Routine examinations at the IRCCS Policlinico San Donato outpatient clinic using several available Mortara ELI 350 instruments afforded a supplemental set of 370 ECGsterm·mth, which served to define a separate, independent validation cohort. Table S3 details the composition and clinical characteristics of this cohort.

The DNNterm·mth model, entirely calibrated and validated using data from the Claris platform, correctly identifies BrSterm·dom( + ) patients in the independent Mortara cohort with an accuracy of 85.1% (327 of 370), an AUCterm·key of 0.934 ± 0.027 (See Fig. 4 ), a sensitivity of 79.6%, and a specificity of 93.6% (Table 1 ). The 9-lead classification metrics of this cross-platform independent validation compare with those of the DNNterm·mth applied to the Claris cohort training data (Table 1 ). Additionally, the DNNterm·mth correctly classified 100% of the spontaneous BrSterm·dom type 1 patients. Table S8 further details the accuracy with which the DNNterm·mth classifies the Mortara cohort.

Comparative manual assessment of the mortara cohort As a gauge of the present machine learning approach against standard clinical practice, we compiled the diagnoses of two clinicians at the IRCCS Policlinico San Donato, who visually inspected and classified the ECGterm·mth records in the Mortara cohort, without knowledge of the accompanying ajmaline-confirmed BrSterm·dom diagnoses. Table 1 summarizes the results. An expert cardiologist with more than 20 years of experience in clinical arrhythmology and electrophysiology (M.D. 1) classified the Mortara cohort ECGsterm·mth with an accuracy of 72.4%, a sensitivity of 71.5%, and a specificity of 73.0%. Table S10 provides details. A resident clinician in training (M.D. 2) performed the same classification with an accuracy of 58.9%, a sensitivity of 55.5%, and a specificity of 60.9%. See Table S11 . Most tellingly perhaps, AUCterm·key values for MD1 and MD2, 0.722 ± 0.047 and 0.582 ± 0.052 , compare with 0.934 ± 0.027 found in validation of the DNNterm·mth by the Motara cohort.

DNNterm·mth feature importance Figure 5 shows a heatmap indicating the importance of ECGterm·mth features as perceived by the DNNterm·mth. The type 1 BrS patternterm·dom is characterized by a “coved-type” ST segment, while the type 2 pattern is characterized by a “saddle-back” shape with an ST elevation of ≥ 1mm in leads V1 to V3. As such, a neural network classification model that seeks only the type 1 or 2 patterns would likely place a significant weight on the ST-segment of leads V1 to V3. However, while our DNNterm·mth model does assign some importance to these regions, it Classifies ECGsterm·mth on the basis of a complex multivariate relation among waveform amplitudes, including a nonlinear combination of the ST-segments from V1 and V2, and the QRS complexes in leads I, III, V1, V2, and V6. We did not provide our DNNterm·mth model with any input weights, nor did we label any regions of the ECGterm·mth as important. Thus, the regions of the fused representative ECGterm·mth indicated in Fig. 5 detail a positive response to the SCBterm·mth challenge ab initio . Importantly, note that the DNNterm·mth model assigns virtually no weight to the eight junctions at which we have concatenated the representative beats from each lead. This DNNterm·mth indifference affirms the robustness of our preprocessing approach.

DNNterm·mth feature importance shown on a 9-lead fused ECGterm·mth trace, representative of the DNNterm·mth training data. The colormap displays the magnitude of the weights applied by the DNNterm·mth when predicting ajmaline responseterm·var. The weights range from 0.0001 to 0.3021; higher–weighted regions have a greater influence on the DNNterm·mth model.

Discussion The present study demonstrates a neural network model that learns to recognize the signature of Brugada Syndrometerm·dom hidden in a conventional ECGterm·mth in the absence of any proarrhythmic drug challenge. These results establish that an appropriately trained DNNterm·mth model responds to subtle ECGterm·mth abnormalities that are undetectable to the human eye. The algorithm correctly classifies a majority of ECGsterm·mth by predicting the response of a patient to an ajmalineterm·key challenge without the need for other clinical information. Despite the relatively limited size of this database, the neural network assesses apparently normal ECGsterm·mth with an AUCterm·key of 0.912 ± 0.017 (accuracy of 84.0%) and 0.934 ± 0.027 (accuracy of 88.4%) in the training and validation datasets respectively.

To date, when evaluating a patient at an arrhythmia clinic, the clinician must decide whether to administer a sodium channel blockerterm·mth in an effort to unmask the BrSterm·dom phenotype ( 5 , 13 , 14 , 39 , 40 ). This drug challenge exposes the patient to the possibility of a life-threatening arrhythmia; but, a decision not to challenge risks overlooking a patient with hidden BrSterm·dom. At present, owing to the possibility of life-threatening side effects, only patients with a family history of SCDterm·dom or a suspicious ECGterm·mth undergo a drug challenge.

Thus, many people who show no visible sign of BrSterm·dom may well remain undiagnosed ( 9 ). The identification of BrSterm·dom is particularly limited in countries, including the United States of America and Japan, where ajmalineterm·key is generally unavailable, and the SCBterm·mth challenge is limited to less effective agents such as procainamide and flecainide. Moreover, several conditions expressly contraindicate this diagnostic approach. For example, a fatal ventricular arrhythmia calls for the assessment of family members for BrSterm·dom; but, clinicians may hesitate to administer ajmalineterm·key to children. This limitation may leave many young carriers of BrSterm·dom undiagnosed. These drugs are also contraindicated in cases of pregnancy, liver diseases, and advanced cardiac conduction disturbances.

Sodium channel blockersterm·mth have negative inotropic properties that preclude use for patients suffering from cardiomyopathies, particularly those associated with mutations in the SCN5A gene and other conditions overlapping with BrSterm·dom, such as hypertrophic cardiomyopathy (HCM) and arrhythmogenic right ventricular cardiomyopathy (ARVC) ( 41–44 ). Dynamic effects of these drugs vary a great deal from individual to individual, and this clouds the significance of any particular test result. The direct hepatotoxic effect of these agents further excludes their diagnostic applicability as a screening tool.

Thus, significant considerations limit the diagnostic application of SCBterm·mth drugs compared with an instrumental measure of heart function, such as an electrocardiogramterm·key, which does not rely upon the action of an agonist. While work still remains to validate machine-learning classification algorithms for patients with accompanying conditions not considered in the present study, the accuracy demonstrated thus far, approaching that normally achieved in SCBterm·mth tests, offers a potential of great benefit as a safe first step in screening a wider population.

To this purpose, several studies have reported the usefulness of some ECGterm·mth parameters in the discrimination of BrSterm·dom patterns ( 45–49 ). However, intra- and inter-observer variability in specific measurements limit the power of univariate pattern-recognition methods. More generally, the ability to identify ECGterm·mth predictors of BrSterm·dom depends on a clinician’s level of clinical experience. Very recently, a convolutional neural network approach showed success in recognizing the characteristic ST elevation of a type 1 BrS patternterm·dom ( 33 ). However, only a small fraction of BrSterm·dom( + ) individuals spontaneously present this plainly visible ECGterm·mth feature.

Indeed, in the present cohort, roughly 90% of patients showed a near-normal or merely suspicious ECGterm·mth waveform, challenging to evaluate even by experienced cardiologists. Thus, the development of this advanced DNN-based ECGterm·mth interpretation algorithm promises to aid in the diagnosis of BrSterm·dom by providing a discerning and objective tool to assist the clinician prior to the possible use of potentially harmful SCBterm·mth challenges. This tool will add considerable insight when clinicians, in keeping with current guidelines, combine the results of an ECGterm·mth with a full scope of other clinical data to establish a final diagnosis of BrSterm·dom ( 5 ). The Supplementary material offers additional details concerning classification success upon manual versus DNNterm·mth analysis with reference to patient demographics and clinical characteristics.

The sensitivity and specificity of our DNNterm·mth model compare favorably with other diagnostic tests used in cardiology, such as exercise electrocardiography to diagnose coronary artery disease ( 50 ), and exceed other medical screening tests, such as the BNP immunoassay for heart failure (AUCterm·key 0.60–0.70) ( 51 ), the Papanicolaou smear for cervical cancer (AUCterm·key 0.70) ( 52 ), and the CHA2DS2-VASc Score for stroke risk (AUCterm·key 0.57–0.72) ( 53 ). Moreover, the relatively high negative predictive valueterm·mth (NPVterm·mth) of the DNNterm·mth classification will lend reassurance in decisions to forego the invasive and risky step of an ajmalineterm·key challenge. The current findings will translate to a clinical application that informs clinicians who are considering ordering an SCBterm·mth challenge for suspected Brugada Syndrometerm·dom.

Of 1,455 patients in the present cohort, 7% presented an ECGterm·mth with a spontaneous type 1 patternterm·key. Using standard clinical protocols, a conventional BrSterm·dom screening program would have found only these patients. Among asymptomatic individuals in this study, we diagnosed more than 49% as having BrSterm·dom. Most of these patients had no idea that they carried this condition. They were referred for screening only because of the sudden and unexplained death of a relative, reinforcing the idea that the actual prevalence of BrSterm·dom is likely underestimated.

Remarkably, the DNNterm·mth model correctly classifies every patient who presented a spontaneous type 1 ECGterm·mth. This marker signals a heightened danger of cardiac arrest ( 5 , 54 , 55 ). But, nearly half of BrSterm·dom patients who survive cardiac arrest do not show a spontaneous type 1 patternterm·key, underlining the critical need to clinically identify patients that harbor a hidden disease signature and face a substantial probability of SCDterm·dom ( 12 ). In young adults, the first clinical manifestation of BrSterm·dom is often atrial fibrillation ( 56 ). This suggests that BrSterm·dom may represent a complex and diffuse phenotype, sharing genetic mutations with various supraventricular arrhythmias ( 43 , 57 , 58 ), as well as with other cardiomyopathies, such as HCM ( 44 ), ARVC ( 10 ) and early repolarization syndrome ( 59 ).

This broader relevance points to an additional reach of noninvasive BrSterm·dom screening as a means to reduce the risk of life-threatening arrhythmias and sudden death. This is particularly relevant for patients who present a nondiagnostic ECGterm·mth pattern that may mask a clinically aggressive phenotype requiring further cardiac investigation. The classification success of the DNNterm·mth suggests that machine learning will enable clinicians to identify these patients even from an apparently normal ECGterm·mth.

Furthermore, as plainly evident in Fig. 4 , a completely independent cross-platform dataset validates the DNNterm·mth with virtually the same accuracy (AUCterm·key 0.93 vs 0.9) as LOOCVterm·mth applied to the training dataset. This not only proves again that the algorithm generalizes well to classify previously unseen data, but also to successfully classify ECGterm·mth data acquired on different machines. A larger ECGterm·mth dataset and multiparametric diagnostic assessment, performed by combining ECGterm·mth machine learning and genetic information ( 60 ), could provide better accuracy, and advance substantially in diminishing the devastating impact of a family of lethal diseases by screening the general population. Additionally, next-generation electrocardiographs and new implantable or wearable devices could support operating systems that incorporate artificial intelligence built upon this DNNterm·mth model for output processing ( 61 ).

A DNNterm·mth algorithm for assisting the diagnosis of BrSterm·dom offers several advantages compared with current practices: (i) reduction of test-related adverse events, such as life-threatening arrhythmias and ajmaline-induced liver injury, (ii) reduction of the cost compared with an SCBterm·mth test, which must be performed in a hospital setting, and (iii) identification of family members at risk of BrSterm·dom in centers with limited access to drug testing or no access to ajmalineterm·key.

Interpreting DNNterm·mth feature importance Understanding where a DNNterm·mth places the greatest importance in the input data is crucial for interpreting the decisions a model makes (see Fig. 5 ). By identifying which ECGterm·mth features the DNNterm·mth weights most heavily in its classification model, we can understand which aspects of the ECGterm·mth signal play the greatest role in identifying the BrSterm·dom phenotype. This information can help guide the selection of new data for training, improve the interpretability of the model, and explain why the model might fail to correctly identify the BrSterm·dom phenotype in some cases.

We observe that the DNNterm·mth’s interpretation largely aligns with the cardiologists’ approach, as it considers not only ST elevation, but also other aspects of the ECGterm·mth, including QRS interval, QT interval, J point depression, and the presence of an incomplete right bundle branch block. This convergence is significant, as it illustrates how the DNNterm·mth learns to recognize important ECGterm·mth features in a completely unbiased data-driven approach, supervised solely on a patient’s response to the ajmalineterm·key challenge.

DNNterm·mth multivariate classification evaluates ECGterm·mth data by means of thousands of nonlinear decision-making pathways, allowing it to identify patterns and features that escape notice in conventional approaches. This enables the DNNterm·mth model to recognize subtle covariance of diagnostic significance. We believe that feature importance patterns established by our DNNterm·mth approach could help refine and expand the understanding of Brugada Syndrometerm·dom, potentially leading to the development of more robust diagnostic criteria.

The substantial data preprocessing method described here reduces a raw, noisy ECGterm·mth trace to yield a representative heartbeatterm·key as a single, averaged complete cardiac cycle (750 ms in duration) centered near the R peak. The high signal-to-noise representative heartbeatterm·key yields a persistent, reproducible waveform for classification by the DNNterm·mth. Our segmentation strategy overcomes uncorrelated beat-to-beat variance in the raw traces. This transformation aids the DNNterm·mth in recognizing persistent ECGterm·mth morphology that consistently encodes for BrSterm·dom while ignoring transient abnormalities.

Conclusion The present work demonstrates an ECGterm·mth screening algorithm for BrSterm·dom that competes with the sodium-channel-blocker challenge while presenting the clear advantage of no life-threatening side effects. This mode of computer-assisted analysis can be applied to digital ECGterm·mth traces from all current electrocardiographs. The broad application of this methodology will enable clinicians to test widely for this deadly disease. We foresee that this machine learning approach to ECGterm·mth analysis will extend beyond the diagnosis of BrSterm·dom to broader risk stratification and mitigation of SCDterm·dom.

Supplementary Material Click here for additional data file.

Acknowledgments The authors wish to express their gratitude to the I.R.C.C.S. Policlinico San Donato for providing resources and a conducive environment facilitating the success of this research. Thanks in particular go to Flavio Mastrocinque and Roberto Manuello for technical assistance in ECGterm·mth database management.

Contributor Information Luke MelopersName, Department of Chemistry, University of British Columbia, Vancouver, BC V6T 1Z1, Canada.

Giuseppe CicontepersName, Arrhythmia and Electrophysiology Center, IRCCS Policlinico San Donato, Milan 20097, Italy.

Ashton ChristypersName, Department of Chemistry, University of British Columbia, Vancouver, BC V6T 1Z1, Canada.

Gabriele VicedominipersName, Arrhythmia and Electrophysiology Center, IRCCS Policlinico San Donato, Milan 20097, Italy.

Luigi AnastasiapersName, Stem Cell Laboratory for Tissue Engineering, Università Vita-Salute San Raffaele, Milan 20132, Italy.

Carlo PapponepersName, Arrhythmia and Electrophysiology Center, IRCCS Policlinico San Donato, Milan 20097, Italy; Department of Cardiology, Università Vita-Salute San Raffaele, Milan 20132, Italy.

Edward GrantpersName, Department of Chemistry, University of British Columbia, Vancouver, BC V6T 1Z1, Canada.

Supplementary Data Supplementary material is available at PNAS Nexus online.

Funding This work was partially supported by Ricerca Corrente funding from the Italian Ministry of Health to IRCCS Policlinico San Donato. L.M., A.C and e.g. gratefully acknowledge financial support from the Natural Sciences and Engineering Research Council of Canada.

Author Contributions L.M., G.C., and A.C. contributed equally to this research. G.C., G.V., and C.P. gathered electrocardiogramsterm·mth and diagnosed patients. L.M. and A.C. developed data processing methods and wrote DNNterm·mth multivariate analysis programs, e.g. and L.A. defined strategic avenues of inquiry and interpreted logistic regression results. All authors took an active part in writing and revising this manuscript.

Data Availability Summary demographic and diagnostic data are included in the manuscript and Supplementary material . Anonymized data created for the study are available in a persistent repository: MelopersName, Luke (2023). Deep learning unmasks the ECGterm·mth signature of Brugada Syndrometerm·dom [Dataset]. Dryad. https://doi.org/10.5061/dryad.s1rn8pkd9 .

References Associated Data Supplementary Materials Click here for additional data file.

Data Availability Statement Summary demographic and diagnostic data are included in the manuscript and Supplementary material . Anonymized data created for the study are available in a persistent repository: MelopersName, Luke (2023). Deep learning unmasks the ECGterm·mth signature of Brugada Syndrometerm·dom [Dataset]. Dryad. https://doi.org/10.5061/dryad.s1rn8pkd9 .