Nowadays, 3D ultrasound systems are available and more widely used. Consequently, with an appropriate 3D image-based needle detection technique, needle guidance and interventions may significantly be improved and simplified. Accurate information about the position of the needle with respect to important vessel or nerve structures in the body can minimize risks to the patient and improve health outcomes. Various needle tracking systems are proposed such as robot-assisted navigation, optical needle localization, electromagnetic tracking, and photoacoustic techniques. However, they are not widely used in practice as they require additional equipment in the operating room, request specific skills to operate the additional systems and they add costs to the US system and the needle or catheter. Therefore, image-based localization techniques are more appealing for intervention support, as they can potentially overcome these limitations while the manual skills are significantly simplified after the system returns the best scan plane with respect to the needle.
In this research, we investigate on image-processing algorithms to efficiently and robustly detect and visualize the needle in 3D US volumes for various US-guided procedures. The introduced methods need to maintain a balance between detection accuracy and computational complexity to permit real-time implementations for live intervention support. One proposed image-based needle localization technique is based on extraction of instrument voxels using 3D structure directions, i.e. 3D Gabor transformation and robust approximation of the tool axis. Our methods are able to successfully detect the needle with accuracy in the sub-millimetre domain. Moreover, the scan plane presented to the medical specialist always contains the instrument and its tip.
Partners involve in the project are: Eindhoven University of Technology, Philips Research Eindhoven, and Catharina Hospital Eindhoven. Researchers involved are: Dr. Sveta Zinger, Ir. Arash Pourtaherian and Peter H.N de With.