Machine Learning Based Approach for Detection of Lung Cancer in DICOM CT Image

Abstract

Lung cancer is one of the leading causes of cancer among all other types of cancer. Thus, an early and effective identification of lung cancer can increase the survival rate among patients. This method presents a computer-aided classification method in computerized tomography images of lungs. In the proposed system, MATLAB has been used for implementing all the procedures. The various stages involved include image acquisition, image preprocessing, segmentation, feature extraction and support vector machine (SVM) classification. First, the DICOM format lung CT image is passed as input which undergoes preprocessing. Then, a threshold value is calculated and image is segmented into left lung and right lung. After that 33 features of each segmented lung are taken and passed as input to the SVM. Finally, the image is classified as cancerous or non-cancerous based on the training data. This method aims to give more satisfactory results when compared to other existing systems.