Application of U-Net

Abstract

Lung cancer (lung carcinoma) is a malignant tumor defined by unrestrained cell growth in lung tissues. Long-term tobacco smoking is the major cause of lung cancer. Radiographs and Computed Tomography (CT) are used to see the lung cancer. The diagnosis is performed by the process called bronchoscopy and can be confirmed by biopsy. CT is a lung screening device which is a non-nosy, painless and use low dose X-rays to display the CT images of the lung cancer. By rotating the X-ray beam, it allows the radiologist to observe exquisite stages or slices of the lungs. It finds out smaller nodules of most cancer with help of multi-slice spiral computed tomography scanner. The LIDC-IDRI dataset consists of medical images from various patients affected by lung cancer. The images are initially segmented using U-net architecture for discovering the cancer affected region. The lung nodules are further examined using Convolutional Neural Network (CNN) to identify whether it is Benign or Malignant tumor. Initial stages nodules are smaller in size and its seriousness can be monitored based on the curvature. The resultant will show the proper way to approach the treatment.