Abstract 4342556: Rhythm Identification of Wide Complex Tachycardia on 12-Lead ECG Using a Convolutional Neural Network
Bibliographic record
Abstract
BACKGROUND: Wide complex tachycardia (WCT) remains one of the most challenging dilemmas in electrocardiogram (ECG) interpretation. Supraventricular tachycardia (SVT) and ventricular tachycardia (VT) have different prognostic implications and have drastically different management. There are numerous algorithms that summarize criteria for EKG or QRS pattern recognition such as the Brugada and Vereckei criteria. However, validation studies show that depending on the training of the interpreter, the ability to recall and apply such criteria varies and affects the outcome. Artificial intelligence, especially deep learning excels at pattern recognition, making it an attractive candidate to interpret EKGs. METHODS: Using raw ECG data of WCT (QRS > 120ms, heart rate > 120 bpm), a convolutional neural network (CNN), specifically ResNet, was developed for classification. The model characterizes WCT into SVT, VT, or paced rhythm. An initial base model was trained on all available training ECGs. The base model was fine tuned on an Electrophysiologist (EP)-adjudicated training set. The final model was validated against a holdout dataset of 400 EKGs on the basis of definitive findings to distinguish VT from SVT with aberrancy. The models were interrogated using explainability techniques including LIME and GradCAM. RESULTS: We extracted a total of 13792 WCT ECGs, of which 608 were VT, 12106 SVT and 1078 were pacing. The two-stage ResNet achieved an average area under the receiver operating characteristic curve (AUC) of 0.946 (95% CI 0.920-0.967), 0.950 (95% CI 0.929-0.968) and 0.975 (95% CI 0.957-0.991) for SVT, VT, and paced groups. Performance metrics for the test set were as follows: SVT – sensitivity 86.9% (±6.4%), specificity 88.5% (±3.9%); VT – sensitivity 74.6% (±7.9%), specificity 94.6% (±2.7%); pacing – 94.6% (±4.2%), sensitivity 95.0% (±2.7%). Explainability analysis revealed that the model looks at various parts of the QRS including depolarization and repolarization. CONCLUSION: The developed ResNet model is excellent at pattern recognition, which can identify any given WCT as VT, SVT, or paced rhythm with similar performance to Cardiologist read. AI is a valuable tool in assisting clinicians with interpretations of challenging ECG rhythms such as WCT and optimizing the clinical outcomes.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".