Segmentation of lung nodules in CT images using weighted average based threshold and maximized variance

Abstract

Background: Lung cancer is a major health concern globally, being the primary cause of cancer-related deaths. It accounts for approximately one–sixth of all cancer fatalities. Objective: The goal of this study is to develop an effective method for the early detection of lung tumors using computed tomography (CT) images. This method aims to identify lung tumors of various sizes and shapes, which is a significant challenge due to the variability in tumor characteristics. Methods: The research utilizes CT images of the lungs in sagittal view from the LID-IDRI database. To tackle the issue of tumor variability in size, shape, and number, the study proposes a novel image processing technique. This technique involves detecting tumor clusters using a weighted average-based automatic thresholding method. This method focuses on maximizing inter-class variance and is supplemented by further classification and segmentation processes. Results: The proposed image processing technique was tested on a dataset of 315 lung CT images. It demonstrated a high level of accuracy, achieving a 98.96% success rate in identifying lung tumors. Conclusion: The study introduces a highly effective method for the detection of lung tumors in CT images, irrespective of their size and shape. The technique’s high accuracy rate suggests it could be a valuable tool in the early diagnosis of lung cancer, potentially leading to improved patient outcomes.