Leveraging Radon Descriptors and Machine Learning to Anticipate Left Ventricular Endocardial Scar Tissue Patterns

EasyChair Preprint 11440

Abstract

Introduction Scar tissue within the left ventricular (LV) endocardium is a significant factor contributing to the development of ma-lignant ventricular arrhythmias in patients with myocardial infarction, leading to potentially fatal cardiac outcomes. In this study, our objective was to evaluate the pattern of scar tissue in the LV endocardium using a Radon descriptor-based machine learning approach. Methods To achieve our objective, we employed an automated LV segmentation technique to identify the LV endocardial wall. Morphological operations were then performed to delineate the regions of scar tissue on the endocardial wall. We fo-cused on utilizing a Radon descriptor-based machine learning approach. Specifically, we selected CT images of the heart from 17 patients and extracted patches from these images. These patches were subsequently categorized into two groups: "endocardial scar tissue" and "normal tissue." Ten feature vectors were extracted from each patch using Radon descriptors, which were then fed into a traditional machine learning model. Results Our results demonstrated that a decision tree model exhibited the highest performance, achieving an accuracy of 98.07%. This study represents the first attempt to employ a Radon transform-based machine learning method for distinguishing patterns between "endocardial scar tissue" and "normal tissue" groups. Conclusion In conclusion, our study leveraged a Radon descriptor-based machine learning approach to predict and analyze left ventricular endocardial scar tissue patterns. The results obtained using the decision tree model demonstrated high accuracy in distinguishing between scar tissue and normal tissue. This research methodology holds promise for further advancements in intervention strategies for patients with myocardial infarction and can contribute to improving patient outcomes.

Keyphrases: Heart, Left ventricular, Radon, machine learning

@booklet{EasyChair:11440,
  author    = {Yashbir Singh and Deepa Deepa},
  title     = {Leveraging Radon Descriptors and Machine Learning to Anticipate Left Ventricular Endocardial Scar Tissue Patterns},
  howpublished = {EasyChair Preprint 11440},
  year      = {EasyChair, 2023}}