Predicting takotsubo syndrome subtypes: An interpretable machine learning model for differentiating emotional versus...

Area of Science:

• Cardiology
• Artificial Intelligence
• Machine Learning

Background:

• Takotsubo syndrome (TTS) is an acute coronary syndrome involving reversible left ventricular dysfunction.
• TTS is categorized as primary (emotional triggers) or secondary (physical stress).
• Distinguishing TTS aetiologies is crucial for effective patient management.

Purpose of the Study:

• To develop and validate a machine learning model for classifying TTS aetiologies.
• To identify key clinical predictors differentiating emotional from physical TTS triggers.

Main Methods:

• A logistic regression model was trained on a dataset of 320 TTS patients.
• The model utilized admission-based clinical markers for classification.
• External validation was performed on a cohort of 121 TTS patients.

Main Results:

• The model achieved 74% accuracy and an AUC of 0.78 in the primary cohort.
• Key differentiating features included chest pain, dyspnoea, atrial fibrillation, sex, COPD, heart rate, and cancer.
• External validation yielded 74% accuracy, 77% precision, and 91% recall, with an AUC of 0.62.

Conclusions:

• An interpretable machine learning model can effectively distinguish between emotional and physical TTS causes.
• The model highlights significant clinical factors for aetiology classification.
• This tool, using admission data, can assist clinicians in practice.