How to have a structured database for lung cancer segmentation using deep learning technologies

Abstract

CAD systems have achieved a good results in detection and diagnosis of different diseases such as lung cancer, breast cancer, liver cancer, brain cancer.. This success is due to the use of new technologies of artificial intelligence, which are machine learning and deep learning, these new technologies depend strongly with the use of large and structured databases. In lung cancer field, most of databases have a manual annotation but they are not structured in the form of images and masks, while this structure is necessary in the training of deep learning models especially in the task of pulmonary nodule segmentation. We present in this paper the necessary steps to have a structured database, then we apply classical segmentation techniques in order to compare the resulting masks with those of our database.