Lung Cancer Detection using CNN with Contour Guided Visualization

Lung cancer continues to be one of the most common and deadly illnesses globally, with enhanced survival rates largely relying on prompt and precise diagnosis. This research presents an automated system that combines deep learning and conventional image preprocessing methods to provide accurate and understandable diagnostic results. The approach consists of a series of preprocessing actions, such as converting to grayscale, binarizing, applying thresholding, segmenting, and extracting contours. Through contour detection, the system marks the main lung outline in green and additionally accentuates this area with a blue bounding box on the segmented binary image. Although this stage does not specifically isolate the tumor area, it clearly outlines the lung region, making abnormal growths or irregular formations more apparent. In the end, an enhanced VGG16 convolutional neural network is employed to classify the examined areas into three categories: benign, malignant, or normal, ensuring a dependable method for early lung cancer identification. An accuracy of approximately 96% was demonstrated in the experimental evaluation, demonstrating that the combination of CNN-based classification and contour-guided visualization provides a computationally effective and clinically interpretable method. In resource-constrained settings, this method supports initial lung cancer screening and provides a feasible alternative to complex segmentation models.