Machine learning approach to predict lung cancer using CT scan images

Abstract

Machine learning methods have become a popular tool for medical researchers to find accurate prediction. Machine learning techniques can identify the patterns in complex data set and it can effectively predict cancer. In the recent medical era, medical imaging plays an essential role to perform effective diagnosis and analysis of the complete human body. Medical experts completely depend on the medical imaging results which are obtained from the image sensors. Cancer is a disease of human cells. As the World Health Organization (WHO) reported, 9 million people died cancer all around the world. Today medical diagnosis process needs computational methods to find the results with improved accuracy. Machine learning techniques are widely used to support medical practitioners to find the diseases in the earlier stage. The CT scan images are used along with computational methods gives better results than normal tests. Early cancer detection plays a vital role in human health in the healthcare industry. In the human body respiration system, lungs are the fundamental element. In all over the world, Lung cancer is a severe disease of people died. The accurate prediction of lung cancer is a very challenging task for medical practitioners. In this research work, the patient’s real CT scan images are gathered from the Lung Image Database Consortium (LIDC) archive. There are five main phases involved in this proposed research work such as image preprocessing, lung segmentation, nodules detection, feature extraction and image classification. The image segmentation is performed by using the OTSU method. OTSU is an image thresholding based segmentation method. This segmentation algorithm is used to find accurate segmentation of the lung region from the CT scan images. Machine learning-based Classification method is used to detect whether the nodule is either benign or malignant.