Effective Lung Cancer Detection Framework Employing Adaptive Histogram Equalization and Deep CNN

This paper presents a fast image segmentation technique in medical imaging to reduce the time consumed by doctors in analyzing CT scans. The manual analysis of modern medical imaging is hardly possible because of the very huge size and picture' complexity. The accuracy and convergence time are two critical parameters that affect the performance of the segmentation algorithms. Because of the existing challenges in medical image segmentation, the need to study and apply new evolutionary techniques is an imperative. Lung carcinoma is the most frequently occurring cancer in men worldwide and, therefore, calls for early detective measures to enable time-critical, life-saving therapies. Among the numerous medical imaging techniques, CT has shown very high efficiency in detecting lung cancer. This paper evaluates five optimization algorithms' performance in extracting tumors from lung CT images using a deep convolutional neural network for picture classification. Comparatively, the strategies of median, adaptive-median, and average filters have been evaluated; therefore, the adaptive median filter has been selected as the best possible enhancement methodology for medical CT images. Moreover, adaptive histogram equalization is used to enhance the contrast of images. The improved images pass through four different algorithms before being displayed. The research validates results in Python on a dataset of 20 lung images. In this study, it is noteworthy that during validation, the GCPSO algorithm hit the maximum accuracy of 98.78%.