Lung Tumor Segmentation Accelerated by CUDA

In the last few years the research done in automatising of medical diagnosis systems have increased significantly despite the fact that it is a slow process to gain trust in such a domain where every detail can make the difference between life and death. Lots of methods based on classic programmable algorithms and machine learning methods gave very good results in lung segmentation and tumor segmentation tasks which can lead to a big increase in early prevention of such dangerous diseases. The goal of this paper is to prove that a classic programmable algorithm can detect lung tumors on CT scans with very good precision and can run as fast as a neural network. To achieve it we implemented a segmentation algorithm using the CUDA API and managed to obtained important results in both precision and speed of detection.