Research on the ROI registration algorithm of the cardiac CT image time series. (February 2018)

• Record Type: Journal Article
• Authors: Meng, Qingliang; Lu, Xiaoqi; Zhang, Baohua; Gu, Yu; Ren, Guoyin; Huang, Xianwu
• Is Part Of: Biomedical signal processing and control. Volume 40(2018)
• Page Start: 71
• Page End: 82
• Publication Date: 2018-02
• Subjects: Medical image registration -- Region of interest -- Adopted K-Nearest Neighbor (KNN) Best-Bin-First (BBF) -- Random sampling consensus (RANSAC); Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2017.09.011
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The algorithm combing BBF and RANSAC algorithms for ROI registration is proposed. Large data amount, time-consuming and low registration accuracy issues are solved. New extraction principle and method of the ROI are proposed. BBF algorithm improves the repeated backtracking shortcoming of KNN algorithm. RANSAC algorithm eliminates mismatching point pairs fitting image edges precisely. Abstract: Objective: Based on the Scale-invariant feature transform (SIFT) features, a novel registration algorithm is proposed to solve the problems including the large amount of data emerged from the cardiac image registration process, time-consuming issue and the lower registration accuracy. Method: First of all, the region of interest (ROI) of the image to be registered is extracted; then, the feature points of the image are extracted by using the SIFT algorithm; finally, a novel registration algorithm which combines the adopted K-d tree Nearest Neighbor (KNN) Best-Bin-First (BBF) algorithm with the random sampling consensus (RANSAC) algorithm is employed to achieve the registration algorithm and to enhance the registration accuracy, so as to solve the high dimensionality of feature vector and easier mismatching issues. Result: The experimental results are as follows: first of all, the amount of data processed during the registration is reduced by 60%–80% after extracting the ROI without destroying the original image data. Secondly, the registration time is reduced by 50%–70%, compared with the traditional registration algorithm. Thirdly, the whole registration precision increases by 10%–20% by using the BBF algorithm to match the feature points and using the RANSAC algorithm to filter the mismatching. Conclusion: The proposed algorithm equipped with the robustness and stability can greatly reduce the time required for registration, improve the registration accuracy.