First Order Gradient Derivative Features Based Classification of Lung Lesion Using Computed Tomography Images

Exhaustive exploration of biomedical images is required for better diagnosis and treatment planning at low resource settings. Biomedical images possess treasured statistics and evidence which can be coupled to predict the structure or physiology of the respective part of human body. Those treasured statistical evidence used in proposed research is radiomic features. Lung lesions are annotated and radiomic features are extracted from the labelled region to perform machine learning to classify normal and abnormal region. Also radiomic features are analyzed by visualization to identify suitable feature to classify squamous cell carcinoma, adenocarcinoma and normal cases. Support vector machine classifier is used with different kernels. SVM classifier along with Sigmoid filter turned to be promising result with better precision and f1 score compared to Radical Base Filter and Polynomial kernel.