Early Detection of Pulmonary Nodules in Lung Cancer Using Image Processing

Lung cancer is considered to be a worldwide threat due to its severity and late detection. There are various imaging techniques used for screening and detecting lung cancer. However, Computerized Tomography (CT) scans have proved to be the most accurate. Many patients only seek screening after noticing symptoms of the disease, making early cancer detection more challenging. This paper proposes image preprocessing and image segmentation approach which uses computer vision algorithms to detect the presence of pulmonary nodules accurately and provide patients with immediate treatment in the earlier stages. Median filter and CLAHE are used to preprocess the image quality. Once the image is preprocessed, image segmentation methods are used such as MeanShift Clustering, graph-laplacian operator, Region growth, and watershed were compared with manually segmented lung image to check the efficiency of each model in segmenting the lung regions. After segmentation, the efficiency of each algorithm was compared through Performance metrics such as Accuracy, Precision, and Recall. The method achieved an accuracy of 78 % with the Mean Shift Algorithm, along with a precision of 80% and recall of 90 %.