Lung Cancer Detection from Computed Tomography (CT) Scans using Convolutional Neural Network

With increasing patients of Lung Cancer every year, it is important to detect Lung Cancer so as to give proper medical treatments. Low dose CT Scan images are used for the detection of Lung Cancer. The first step is to detect the pulmonary nodules in lungs. LUNA16 data has 888 CT scans with annotated nodules in the CT scans. The annotation has coordinates of the lung nodules. A 32 × 32 × 32 cube is made around the nodules with nodule as the centre. A 3D Convolutional Neural Network (CNN) is used to detect nodules using these cubes. For Lung Cancer detection, Data Science Bowl 2017 Kaggle competition data is used. It has 1595 CT scans. Lung nodules are predicted on this data using the nodule detector by running on the CT scans as grids. An ROI mask for lungs is applied to the CT scan using Image Processing. The predicted nodules coordinates are used to make cubes around nodules as the same size as before and a second 3D CNN is used to predict cancer using it. The model achieves precision and recall of 89.24% and 82.17% respectively.