Abstract

Info

Title:: Chest radiograph registration technique based on segmentation mask obtained by deep learning and its application

Author(s):: LURE Fleming Yuanming1; WANG Xiaoyi2; GUO Lin1; XIA Li1; QUAN Shenwen1; QIAN LingJun1; LI Hongjun3; 1. Shenzhen Zhiying Medical Imaging Co., Ltd. Shenzhen 518000, China 2. The Fourth Peoples Hospital of Qinghai Province, Xining 810007, China 3. Beijing Youan Hospital, Capital Medical University, Beijing 100069, China

Keywords:: Keywords: chest radiograph deep learning segmentation mask image registration subtraction analysis

Abstract:: Abstract: Medical image registration technique is of vital importance for clinical diagnosis and treatment. The deep learning based registration method improves the accurate and speed of registration when compared with conventional registration methods. In order to apply deep learning algorithm to do chest radiograph registration and subsequent subtraction analysis, the original chest radiographs is preprocessed with the use of the mask obtained by deep learning, and the chest radiograph registration is achieved with mask images as input, ResUnet as registration structure. The evaluation of the registration results showed that the model developed by mask and registration technique based on deep learning has high image registration accuracy in chest radiograph registration. The proposed registration model can be well applied to the subtraction analysis of chest radiographs.