Lung Cancer Classification Using Image Enhancement and CNN-Based Pre-Trained Models

Lung cancer, a prevalent and life-threatening disease, necessitates accurate and timely diagnosis for effective treatment. This study evaluates the impact of a combination of image enhancement techniques such as Median Filter (MF), Contrast Limited Adaptive Histogram Equalization (CLAHE), and Unsharp Masking (USM) on feature extraction and classification of lung cancer images. Additionally, pre-trained models including VGG16, VGG19, ResNet50, and DenseNet-201 are fine-tuned. The dataset comprises three types of lung cancer CT images namely Adenocarcinoma, Large cell carcinoma, Squamous cell carcinoma, and normal CT images. Experimental results demonstrate that the CLAHE&USM combination yields superior outcomes, enhancing image quality, contrast, and sharpness, consequently improving feature extraction. The fine-tuned pre-trained models exhibit exceptional performance in classifying lung cancer. Results show a classification accuracy of 98.5% achieved by VGG16, outperforming existing methods. This research holds significant implications for the development of automated diagnostic tools, enabling early detection and treatment of lung cancer, thereby improving patient outcomes and potentially saving lives. Further exploration can encompass generalizing the model to diverse datasets and assessing its clinical feasibility.