Lung Cancer Segmentation Using Improved Golden Eagle Optimization with Clustering Model

Abstract

Introduction: In recent times, lung cancer has emerged as a widespread and concerning ailment, significantly impacting global health. Early detection is currently recognized as the most effective strategy to enhance the survival rate among cancer patients. The conventional methods for diagnosing lung cancer are time-consuming.

Objectives: To develop a Computer Aided Diagnosis (CAD) based lung nodule segmentation and classification.

Methods: Initially, the input lung images are pre-processed by the median filtering. Then, the lung nodules are segmented using the Enhanced Local Information Weighted Intuitionistic with Fuzzy C-means clustering) (ELWI-FCM) with Improved Golden Eagle Optimization) (IGEO) algorithm used for lesion segmentation. The IGEO is the combination of GEO and OBL (opposition based learning). Finally, the Deep Learning (DL) model DenseNet201 is utilized for classifying the lung nodules as normal and abnormal classes.

Results: The experimentation is carried out on the LIDC-IDRI dataset and achieved better accuracy and dice values of 98.7% and 98.9% respectively.

Conclusions: This work presents an automatic earlier detection of lung cancer thus reducing the risks of deaths.