Analysis of Cardiac and Vascular Images

CVRMed-MRCAS'97

Volume 1205 of the series Lecture Notes in Computer Science pp 213-222

Date:

3D multi-scale line filter for segmentation and visualization of curvilinear structures in medical images

  • Yoshinobu SatoAffiliated withSurgical Planning Laboratory, Department of Radiology, Harvard Medical School and Brigham and Women's Hospital
  • , Shin NakajimaAffiliated withSurgical Planning Laboratory, Department of Radiology, Harvard Medical School and Brigham and Women's Hospital
  • , Hideki AtsumiAffiliated withSurgical Planning Laboratory, Department of Radiology, Harvard Medical School and Brigham and Women's Hospital
  • , Thomas KollerAffiliated withCommunication Technology Laboratory, Image Science, ETH-Zentrum
  • , Guido GerigAffiliated withCommunication Technology Laboratory, Image Science, ETH-Zentrum
  • , Shigeyuki YoshidaAffiliated withDept of Radiology, Osaka University Medical School
  • , Ron KikinisAffiliated withSurgical Planning Laboratory, Department of Radiology, Harvard Medical School and Brigham and Women's Hospital

* Final gross prices may vary according to local VAT.

Get Access

Abstract

This paper describes a method for the enhancement of curvilinear structures like vessels and bronchi in 3D medical images. We develop a line-enhancement filter based on the eigenvalues of Hessian matrix aiming at both the discrimination of line structures from other structures and the recovery of original line structures from corrupted ones. The multi-scale responses of the line filters are integrated based on the equalization of noise level at each scale. The resulted multi-scale line filtered images provide significantly improved segmentation of curvilinear structures. The line-filtered images are also useful for the direct visualization of curvilinear structures by combining with a volume rendering technique even from conventional MR images. We show the usefulness of the method through the segmentation and visualization of vessels from MRA and MR images, and bronchi from CT images.