Tool Localization in 3D Ultrasound Images (Demonstration Application)

Screenshot of ToolDemo Result - a 2D slice containing the tool:
Resulting slice

Author: Marián Uherčík, Center for Machine Perception, CTU in Prague & CREATIS-LRMN, INSA Lyon, December 2009
Description of the application:

The tool localization method is based on model fitting using RANSAC (RNS) and two models of tool in 3D US have been proposed (RNS+AxShp, RNS+IntDstr) in [1]. A method RNS+LineFilter does line filtering to enhance contrast between needle and background has been proposed in [2]. The models IntDstr and LineFilter were learned on training data, while the model AxShp does not use any learning. Note that for a question of speed the method RNS+LineFilter is reading line-filtered data from the file computed in pre-processing.

Download links:
  • Compiled executable for Win32 with one testing dataset - ZIP file (5 MB) + MATLAB Runtime Component for Windows (Download, 171 MB)
  • Testing 3D ultrasound datasets - ZIP file (150 MB) - Copy them to subdirectory 'USdata/'
  • Documentation & Tutorial in PDF

You need to install MATLAB Runtine Component even if you already have MATLAB on your computer.

References:
  1. Marián Uherčík, Jan Kybic, Hervé Liebgott, Christian Cachard: IEEE Transactions on Biomedical Engineering (BME), pp. 1907-1916, Aug. 2010. [pdf]
  2. Marián Uherčík, Jan Kybic, Christian Cachard, Hervé Liebgott: IEEE International Ultrasonics Symposium (IUS), pp. 594-597, September 2009 in Roma. [pdf]