COMPARATIVE ANALYSIS OF SELECTED DEEP LEARNING ALGORITHMS FOR DETECTION AND CLASSIFICATION OF LUNG CANCER

Lung cancer detection through advanced image-based diagnostic tools is critical for early intervention and improved patient outcomes. Traditional approaches often face challenges with accuracy and sensitivity due to the intricate visual patterns associated with lung tumors. This study introduces a deep learning-based approach for lung cancer detection and classification using convolutional neural networks (CNNs) within a multimodal framework. The study utilizes pretrained VGG19, CNN, ResNet50 and InceptionV3 architectures to differentiate between benign, malignant and normal lung images integrating robust feature extraction to enhance detection accuracy. In this study, lung images was obtained from publicly available datasets, consisting of a 1752 images with 70% used for training and 30% for testing. The acquired datasets underwent pre-processing steps with resizing each image to 224x224 pixels, followed by noise reduction using a Gaussian blur filter, which smoothed the images and preserved the lung’s essential structure thereby enhancing image quality for segmentation and feature extraction. Performance evaluations were based on metrics such as False Positive Rate (FPR), Sensitivity, Accuracy, Precision, F1-Score, and Computation Time (CT). VGG19 achieved an accuracy of 94%, sensitivity of 94%, precision of 84%, F1-score of 93%, and CT of 58 seconds. The CNN model achieved accuracy of 96%, sensitivity of 96%, precision of 98%, and a CT of 60 seconds. InceptionV3 outperformed the others with an accuracy of 97%, sensitivity of 97%, precision of 99%, F1-score of 97%, and CT of 58 seconds. ResNet50 delivered 96% accuracy, 96% sensitivity, 92% precision, and a CT of 57 seconds. This study underscores the benefits of deep learning for lung cancer detection. The integration of diverse imaging data and advanced CNN architectures demonstrated high accuracy and efficiency indicating potential for clinical adoption to support early accurate lung cancer diagnosis.