Artificial Intelligence to Interpret Wide-Complex Tachycardia—Trust the Machine?
Bibliographic record
Abstract
In this issue of the Canadian Journal of Cardiology, Chow et al. present their experience using an artificial intelligence (AI) algorithm to interpret wide-complex tachycardia (WCT). 1 Differentiating ventricular tachycardia (VT) from supraventricular tachycardia (SVT) is crucial, because VT is a lifethreatening condition that requires immediate treatment, whereas SVT is generally less dangerous.Currently, various diagnostic algorithms are used to distinguish between VT and SVT, but their accuracies range from 68.8% to 77.5%. 2 The authors tested their AI algorithm on a group of 3330 electrocardiograms (ECGs) with WCT, and showed that their model was very good at differentiating VT from SVT, with performance that was better than nonelectrophysiologist (non-EP) cardiologists, and similar to EP cardiologists. 1 The authors are congratulated in this impressive work; their findings represent the study using the largest data set to date differentiating VT and SVT and builds on a rapidly growing body of literature in using AI for ECG interpretation.In their study, Chow et al. incorporated 3330 WCT ECGs including 2906 SVT ECGs and 424 VT ECGs, with all ECGs adjudicated by at least 2 cardiologists (including 1 EP cardiologist), and a second EP cardiologist when VT was identified by one of the adjudicators.Furthermore, the authors defined "ground truth' for AI training by means of a hierarchic classification beginning with all readers agreeing and down to a consensus across readers.Following adjudication, ECGs were converted to scalable vector graphic files, and put into a convolutional neural network (ZeroLess-DARTS [differentiable architecture search method]).In their internal test set, the authors reported sensitivity, specificity, and accuracy of 93.0%, 91.8%, and 91.9%, respectively.In an external test set of 354 selected WCT ECGs, the authors reported sensitivity, specificity, and accuracy of 80.5%, 81.6%, and 81.1%, respectively.
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 | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.003 |
| 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".