Enhanced Lung Cancer Diagnosis via Attention-based Deep Belief Network segmentation and GWO-EHO Based Deep Learning Classification

Abstract

Lung cancer is the biggest contributor to cancer-related deaths around the world, and early identification is critical to improving patient outcomes.  Nonetheless, early cancer detection is a significant difficulty, especially in low-resource areas where access to healthcare facilities and skilled radiologists is limited. The input CT lung cancer images are collected from DICOM file format dataset. To reduce noise in the resulting dataset, pre-processing is conducted to the images. A  Adaptive optimum weighted mean filter (AOWMF) is applied for the pre-processing phase. The Attention-based Deep Belief Network (ADBN) segmentation approach is used to process the segmentation and determine the afflicted area of the lung cancer. Utilizing the DenseNet-201 Model, the feature is extracted. The proposed research uses the Grey Wolf Optimization based Elephant Herding Optimization (GWO-EHO) architecture to classify brain cancer as normal, benign and malignant”. The proposed EHO model offers improved precision and a better rate of brain tumor classifies with a high accuracy..