A novel artificial intelligence (AI) model correctly identified patients at near-term risk of sustained ventricular tachycardia (VT) who could potentially benefit from preemptive interventions to prevent sudden cardiac death (SCD).
The AI-model utilises a single-lead electrocardiogram (ECG) screening tool that could offer physicians a new approach to SCD risk management, researchers have said, presenting the findings during a late-breaking clinical science session during the 2023 Heart Rhythm Society annual meeting (19–21 May, New Orleans, USA).
As traditional mechanisms for predicting and preventing mid- and long-term SCD are limited, the study sought to understand if AI could be leveraged to better identify near-term occurrences of VT using data from Holter ECG recordings.
The authors of this study developed a deep learning-based model using the first 24 hours of extended Holter monitor recordings, a type of portable electrocardiogram, to predict the risk of sustained (≥30 sec) ventricular tachycardia (VT) over two weeks.
The model used 78,294 unselected Holter recordings collected across the USA, UK, France, Czech Republic, South Africa and India. Among 59,302 recordings used for validation, the mean age of patients were 61.3±17.3 years and 40% were male. A total of 222 recordings presented sustained VT with a mean rate of 157±38 bpm, and median duration 62 seconds [IQR 42, 173]), with the vast majority (98%) being monomorphic.
“Current methods for predicting SCD are extremely limited. By leaning on artificial intelligence, we hope to revolutionise the way physicians monitor, prevent, and predict SCD, improving the lives of patients while generating cost savings for our healthcare system,” said Laurent Fiorina (Institut Cardiovasculaire Paris Sud, Ramsay, France). “For high-risk patients who suffer from multiple conditions including hypertension, obesity, older age and diabetes, this technology could be lifesaving to help more accurately predict sustained VT and offering physicians important insights to offer early SCD prevention interventions.”
On the internal validation dataset, the model achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 0.939 with a sensitivity of 83.3% and a specificity of 88.7%. On the external validation dataset, the AUC was 0.911 with a sensitivity and specificity of 78.9% and 81.4%, respectively. The AI-model correctly predicted VT occurrence in 88% of patients with rapid VT (≥180 bpm). Lastly, the reference model revealed an internal validation AUC of 0.833.
The authors are currently looking to validate the model in future prospective clinical studies. They would also like to extend near-term prevention through ECG monitoring to hospital monitoring or wearable sensors with potential applicability to larger populations.
