Lung Nodule Segmentation Using Adaptive Thresholding and Watershed Transform

Lung cancer, according to the study conducted by World Health Organization in 2017 is the important reason for cancer-related death globally. It was determined that nearly one in every 6 deaths is due to cancer. The cancer nodules present in the lungs of the patient causes the disease. The detection of these nodules is very important for medical researchers. Early detection of these nodules increases the patient's survival rate significantly. Computed Tomography or CT scans of the lungs in axial view, from the publically available LIDC-IDRI database, are used for the study. The lung nodules vary in size, shape, and density. All these varying conditions make segmentation more challenging. Our approach combines several image processing techniques and efficiently detect the lung nodules of any shape and size. The proposed method uses adaptive thresholding and watershed segmentation for detection of the nodules. The techniques and algorithms were implemented using MATLAB. The method was tested for 50 cases and an accuracy of 96 percent was achieved.