Gender-Based Analysis of Electrocardiographic Parameters: Utilizing Machine Learning Feature Selection Methods in Classifying Cases of Heart Failure with Preserved Ejection Fraction


Cellat Z. F., Öz E.

International Sustainability In Life Congress, Aydın, Türkiye, 19 - 22 Mayıs 2024, ss.76

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: Aydın
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.76
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

Heart failure is a significant health issue worldwide, causing high morbidity and mortality. Cases of heart failure with preserved ejection fraction (HFpEF), particularly among women, are on the rise (1, 2). The dataset used in this study includes information on 26 different parameters for 52 male and 66 female patients (n=118) over the age of 18, who presented to the Cardiology Department of Samsun Training and Research Hospital between November 2022 and August 2023. Gender-specific electrocardiographic (ECG) parameters were used to classify HFpEF patients using machine learning methods (Gradient Boosting, K-Nearest Neighbors, Logistic Regression, Random Forest ve Support Vector Machines (3). Various model dependent and model agnostic feature selection techniques were applied to the ECG parameters to create different feature sets (4). The highest classification accuracy (79%) was achieved using the Gradient Boosting method. The most influential features in this classification were smoking, P-wave duration, age, P-wave amplitude and left ventricular end-systolic diameter. Machine learning has the potential to enhance accuracy in diagnosing by highlighting the importance of gender-specific ECG parameters in HFpEF patients, offering personalized approaches in disease management. The use of this technology will enable the development of more precise diagnostic and treatment strategies in cardiology practice.