/ml-iraf-analysis

Machine Learning Immediate Recurrence of Atrial Fibrillation (IRAF) Susceptibility

Primary LanguageJupyter NotebookBSD 2-Clause "Simplified" LicenseBSD-2-Clause

Machine Learning Immediate Recurrence of Atrial Fibrillation (IRAF) Susceptibility

Immediate Recurrence of Atrial Fibrillation (IRAF) occurs after synchronized cardioversion with a prevalence of 5-26%. [1] Recurrence decreases cardioversion efficacy and puts patients through the unnecessary risk of unsuccessful cardioversion. Better knowledge of patient susceptibility factors will allow clinicians to pre-treat these patients to reduce IRAF incidence. This project aims to shed more light on IRAF susceptibility factors through analysis of a cardioversion database.

Data

  • Cardioversion Database (not yet publicly avaliable)

Methods

  • Gradient Boost Machine (XGBoost)
  • Deep Learning (TensorFlow, ...)
  • Verification with scikit-learn (Random Forest, ...)
  • Planned
    • kernel Support Vector Machine (and other supervised learning to seperate susceptibility groups)
    • Clustering