Comprehensive evaluation of multivariate adaptive regression spline-based approaches for solving the inverse electrocardiography problem with experimental datasets

Date

2023-12-05

Author

Koçak, Muhammed Saadeddin

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

110
views

52
downloads

Cite This

The inverse problem of electrocardiography (ECG), defined as the reconstruction of electrical activity of the heart from body surface potential (BSP) measurements remains a challenging endeavor in biomedical engineering. In recent years, data-driven multivariate adaptive regression spline (MARS) based approaches have demonstrated superior accuracy compared to the methods that constrain the data. Therefore, a thorough investigation of MARS-based solution approaches is imperative. The primary objective of this study was to explore the effectiveness of MARS-based approach. To achieve this, the training and test data have been altered to reflect clinical scenarios. The results of extensive runs suggest that the MARS-based model shows comparable accuracies for depolarization and repolarization region reconstructions (p > 0.05). In addition, the diversity of noise levels and geometric error in the training data increases the model’s success at estimating heart potentials for erroneous test beats, proving its suitability for clinical applications. The second stage of this thesis explored the reaction of the MARS-based approach to flow and pacing rate variations in different interventions through a novel experimental data set. When exposed to atypical flow rates, the model's proficiency in accurately reconstructing signal patterns diminishes, especially at elevated pacing rates. Besides, for flow rates as low as 5 ml/min, median spatial and temporal estimation correlations experience major drops compared to the higher flow rates. The findings of this study contribute to a deeper understanding of MARS-based solution approaches for inverse ECG, unveiling their potential in accurately mapping cardiac electrical activity under challenging conditions. The nonparametric, adaptive, and data-driven nature of MARS is further validated by the incorporation of a contemporary experimental data set while providing insights into the impact of flow and pacing rates.

Subject Keywords

Electrocardiographic imaging, Inverse problem, Multivariate adaptive regression splines

URI

https://hdl.handle.net/11511/107727

Collections

Graduate School of Natural and Applied Sciences, Thesis