Artificial Intelligence ECG Analysis in Patients with Short QT Syndrome to Predict Life-Threatening Arrhythmic Events

Q: What are the key findings of this study?

Shallow neural networks using QT and T-wave ECG features achieved a test AUC of 0.81, 77.4% accuracy, and up to 95% specificity in discriminating short QT syndrome patients with arrhythmic events.

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Short QT syndrome (SQTS) (n=104). Intervention: Artificial intelligence-based 12-lead ECG analysis (shallow neural networks using manually extracted ECG features) vs. Deep learning models on scanned ECG images/signals and classical machine learning algorithms. Primary outcome: Discrimination of patients with documented life-threatening major arrhythmic events (sudden cardiac death, aborted sudden death, or unexplained syncope).

Q: What does this research mean for the field?

Shallow neural networks using QT and T-wave ECG features can discriminate short QT syndrome patients at risk for life-threatening arrhythmic events with moderate accuracy (AUC 0.81, 77.4% accuracy) and high specificity. Novelty: ClaimNovelty.NOVEL_FINDING. Consensus alignment: ConsensusAlignment.ESTABLISHES_NEW_DIRECTION.

Eros Pasero; Fiorenzo Gaïta; Vincenzo Randazzo; Sergio Cannata; Philippe Maury; Carla Giustetto

doi:10.3390/s23218900

November 1, 2023SensorsOpen Access

Artificial Intelligence ECG Analysis in Patients with Short QT Syndrome to Predict Life-Threatening Arrhythmic Events

View Full Paper

Ask AI

Bookmark

Share

Key result

Shallow neural networks discriminate short QT arrhythmic risk from ECG features with ~81% AUC.

n=104

Figures

Why the study?

Arrhythmia risk stratification is presently unsatisfactory in asymptomatic SQTS patients, and AI-based ECG analysis had never been applied to refine risk stratification in this population.

Can artificial intelligence-based 12-lead ECG analysis accurately discriminate between short QT syndrome patients with and without life-threatening arrhythmic events?

Key Points

To evaluate whether machine learning and deep learning models can analyze electrocardiograms to discriminate short QT syndrome patients with and without life-threatening arrhythmic events.
Analyzed electrocardiograms from 104 short QT syndrome patients (37 with documented major arrhythmic events) partitioned into training, validation, and test sets.
Trained five shallow neural networks on subsets of 13 cardiologist-measured features, comparing them against classical machine learning algorithms.
Tested deep learning models (including Vision Transformer, Swin Transformer, and 1-D CNNs) directly on scanned electrocardiogram images and digitized signals without manual feature extraction.
Shallow neural networks utilizing manually extracted features achieved a test accuracy of 77.4% across most configurations and an area under the curve of 0.81 for combined QT and T-wave features.
Test sensitivity ranged from 36.4% to 63.6%, specificity reached 85% to 95%, maximum positive predictive value was 83.3%, and negative predictive value peaked at 81%.

Study Design

Type

Cohort (n=104)

Structured PICO

Can artificial intelligence-based 12-lead ECG analysis accurately discriminate between short QT syndrome patients with and without life-threatening arrhythmic events?

P

Population

104 patients with short QT syndrome (37 with documented major arrhythmic events and 67 without events) evaluated by artificial intelligence-based 12-lead ECG analysis.

E

Exposure

Artificial intelligence-based 12-lead ECG analysis using shallow neural networks (multi-layer perceptron) trained on 13 cardiologist-measured ECG features, as well as deep learning models (Vision Transformer, Swin Transformer, CNNs, Capsule Networks) applied directly to scanned ECG images and digitized V1-V3 signals.

C

Comparator

Comparison across different AI architectures (shallow neural networks vs deep learning vision/signal models vs classical machine learning models like SVM, KNN, decision trees, and logistic regression) and different ECG feature subsets.

O

Outcome

Classification performance (sensitivity, specificity, PPV, NPV, accuracy, F1 score, and AUC) in discriminating between SQTS patients with and without documented life-threatening major arrhythmic events (SCD, aborted sudden death, and/or unexplained syncope).hard clinical

Limitations

Limited sample size (104 patients), preventing statistically significant subgroup analyses
Manual extraction and calculation of ECG features by cardiologists, which is prone to measurement errors
Implicit bias in the selection of the 13 ECG features based on current medical knowledge
Poor quality and heterogeneity of scanned ECG images (resolution, format, color, and grid differences)
Manual cropping of ECG images for lead isolation and signal digitization
Lack of validation in an external SQTS population from different regions or medical centers
Limited number of patients (n=104), precluding subgroup analysis
Manual calculation of ECG features by cardiologists is prone to measurement errors
Implicit selection bias regarding the 13 chosen ECG features
High heterogeneity and noisy, poor quality of scanned ECG images
Manual image cropping for lead isolation and signal digitization
Lack of external validation in an independent SQTS cohort

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

Shallow neural networks analyzing manually extracted QT and T-wave ECG features show promise in identifying low-risk SQTS patients (specificity up to 95%, AUC 0.81), though suboptimal negative predictive value (<=81%) currently limits standalone clinical use for ruling out life-threatening arrhythmias.

Abstract

Short QT syndrome (SQTS) is an inherited cardiac ion-channel disease related to an increased risk of sudden cardiac death (SCD) in young and otherwise healthy individuals. SCD is often the first clinical presentation in patients with SQTS. However, arrhythmia risk stratification is presently unsatisfactory in asymptomatic patients. In this context, artificial intelligence-based electrocardiogram (ECG) analysis has never been applied to refine risk stratification in patients with SQTS. The purpose of this study was to analyze ECGs from SQTS patients with the aid of different AI algorithms to evaluate their ability to discriminate between subjects with and without documented life-threatening arrhythmic events. The study group included 104 SQTS patients, 37 of whom had a documented major arrhythmic event at presentation and/or during follow-up. Thirteen ECG features were measured independently by three expert cardiologists; then, the dataset was randomly divided into three subsets (training, validation, and testing). Five shallow neural networks were trained, validated, and tested to predict subject-specific class (non-event/event) using different subsets of ECG features. Additionally, several deep learning and machine learning algorithms, such as Vision Transformer, Swin Transformer, MobileNetV3, EfficientNetV2, ConvNextTiny, Capsule Networks, and logistic regression were trained, validated, and tested directly on the scanned ECG images, without any manual feature extraction. Furthermore, a shallow neural network, a 1-D transformer classifier, and a 1-D CNN were trained, validated, and tested on ECG signals extracted from the aforementioned scanned images. Classification metrics were evaluated by means of sensitivity, specificity, positive and negative predictive values, accuracy, and area under the curve. Results prove that artificial intelligence can help clinicians in better stratifying risk of arrhythmia in patients with SQTS. In particular, shallow neural networks' processing features showed the best performance in identifying patients that will not suffer from a potentially lethal event. This could pave the way for refined ECG-based risk stratification in this group of patients, potentially helping in saving the lives of young and otherwise healthy individuals.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Pasero et al. (2023) conducted a cohort in Short QT syndrome (SQTS) (n=104). Artificial intelligence-based 12-lead ECG analysis (shallow neural networks using manually extracted ECG features) vs. Deep learning models on scanned ECG images/signals and classical machine learning algorithms was evaluated on Discrimination of patients with documented life-threatening major arrhythmic events (sudden cardiac death, aborted sudden death, or unexplained syncope). Shallow neural networks using QT and T-wave ECG features achieved a test AUC of 0.81, 77.4% accuracy, and up to 95% specificity in discriminating short QT syndrome patients with arrhythmic events.

synapsesocial.com/papers/6a8423376b6022c7d8c52325 https://doi.org/10.3390/s23218900

View Full Paper

Ask AI

Bookmark

Share

Artificial Intelligence ECG Analysis in Patients with Short QT Syndrome to Predict Life-Threatening Arrhythmic Events

Figures

Key Points

Study Design

Structured PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Artificial Intelligence ECG Analysis in Patients with Short QT Syndrome to Predict Life-Threatening Arrhythmic Events

Figures

Key Points

Study Design

Structured PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading