Design and implementation of a medical image knowledge base for pulmonary nodules diagnosis

Abstract

In order to achieve pulmonary nodule computer-aided diagnosis (CAD) more effectively in the context of big data and deep learning, we designed and implemented a medical image knowledge base (KB) to store the case data of thoracic computed tomography (CT) scanning. To guarantee this medical image KB more flexible and easy to expand, its two MySQL relational databases (DICOM medical image database and expert diagnosis database) were designed to be independent logically, but be stored in the same database. We used Apache Web Server to implement the medical image KB. Then we utilized PHP scripting language to manage and maintain the KB. We employed Lung Image Database Consortium (LIDC) dataset and designed some test cases to test our medical image KB. Summarily, the medical image KB presented in this paper is capable of storing thoracic CT image and its diagnostic information effectively and structurally for pulmonary nodule diagnosis; and it is high potential for realizing the CAD of pulmonary nodule in the background of big data and deep learning.