Comparison of Feature Extraction Filters For Lung Cancer Prediction

Abstract

Lung cancer is a deadly form of cancer that is difficult to detect. It is more important for care to inspect nodules soon and correctly since it generally causes death in both men and women. The number of people diagnosed with lung cancer is directly proportionate to the number of chain smokers. As a result, a variety of procedures have been developed to identify lung cancer in its early stages. A comparative analysis of multiple machine learning-based approaches for lung cancer diagnosis has been reported in this progress. In addition, to use image recognition to identify lung cancer nodules, many classifier methods are linked with numerous segmentation algorithms. [1] CT scan images have been discovered to be more acceptable for having reliable results in this investigation. The experiment is divided into two parts: Identify the most important feature used in lung cancer analysis by CT scan and map it to a computer-related format in the first step. In the second phase, machine learning techniques are used to pick and extract features. As a result, CT scan images are commonly utilized to diagnose cancer. The classification method, Random Forest Classifier was used to investigate lung cancer prediction. The results reveal that classification accuracy improves in the vast majority of situations, indirectly proving reliability. For feature extraction, we used three filters – Gabor, Local Binary Pattern, and Histogram of Oriented Gradients. Hence, their comparisons are shown in this paper. Keywords- Gabor, Histogram of Oriented Gradient, Local Binary Patterns, Lung Cancer Prediction, Random Forest.