A MARS-Based Approach for PVC Localization Using Clinical Data Klinik Veriler Kullanilarak PVC Lokalizasyonu i in MARS Yakla simi

Almus C., SERİNAĞAOĞLU DOĞRUSÖZ Y., Svehlikova J., Ondrusova B.

33rd IEEE Conference on Signal Processing and Communications Applications, SIU 2025, İstanbul, Turkey, 25 - 28 June 2025, (Full Text) identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/siu66497.2025.11112491
  • City: İstanbul
  • Country: Turkey
  • Keywords: clinical data, Electrocardiographic Imaging, inverse problem, MARS, PVC
  • Middle East Technical University Affiliated: Yes

Abstract

Electrocardiographic Imaging (ECGI) is a noninvasive technique for identifying premature ventricular contraction (PVC) origins, which aids radiofrequency ablation (RFA). This study evaluates a framework based on Multivariate Adaptive Regression Splines (MARS) across different geometric models, including homogeneous, inhomogeneous, and hybrid models, using clinical data. Patient-specific heart surface potentials were simulated, and corresponding body surface potentials (BSPs) were computed using the boundary element method (BEM). These BSPs, along with the heart surface potentials, were used to train the regression model. Activation times were computed, and the earliest activation point was identified as the estimated origin. The performance was evaluated using localization error (LE), and compared with Tikhonov regularization. The best-performing geometric model demonstrated better or comparable performance compared to Tikhonov regularization.