Skip to main navigation Skip to search Skip to main content

归一化互信息与多分辨率融合的2D-3D配准方法

Translated title of the contribution: A 2D-3D registration method based on normalized mutual information and multi-resolution fusion
  • Long Chen
  • , Fengfeng Zhang*
  • , Lingtao Yu
  • , Lining Sun
  • , Minfeng Gan
  • , Wei Zhan
  • *Corresponding author for this work
  • Soochow University
  • Harbin Engineering University
  • The First Affiliated Hospital of Soochow University

Research output: Contribution to journalArticlepeer-review

Abstract

To address the problems of registration accuracy and low efficiency in the registration of minimally invasive spine surgery, a new two-dimensional-three-dimensional (2D-3D) registration method is proposed. A method based on the regional contribution of normalized mutual information combined with multi-resolution strategy is used. The registration accuracy and registration time are analyzed by changing the registration step size. The overall registration accuracy has been increased by 40% and the registration time has been decreased by 53%. When the registration step size is decreased from 0.1 to 0.05, the registration accuracy is increased by 8% and the registration time is increased by 12%. Thus, the registration step size is decreased, whereas the registration accuracy and registration time are relatively increased. The algorithm can achieve the registration of 3D and 2D images, and can effectively improve the registration accuracy and registration efficiency, which meet the needs of image registration in the surgical procedure of doctors.

Translated title of the contributionA 2D-3D registration method based on normalized mutual information and multi-resolution fusion
Original languageChinese (Traditional)
Pages (from-to)243-249
Number of pages7
JournalHarbin Gongcheng Daxue Xuebao/Journal of Harbin Engineering University
Volume41
Issue number2
DOIs
StatePublished - 5 Feb 2020
Externally publishedYes

Fingerprint

Dive into the research topics of 'A 2D-3D registration method based on normalized mutual information and multi-resolution fusion'. Together they form a unique fingerprint.

Cite this