Medical Ultrasound Processing applied to Ground Penetrating Radar

摘要

Ground penetrating radar (GPR) is a technique for non-destructive detection and illustration of buried objects and cavities as well as soil layers and water table. GPR is based on actively emitting electromagnetic waves in a medium, usually in form of impulses. Along its path, the radar signal is scattered, reflected and diffracted at discontinuities, as well as attenuated through absorption. The running time of the signal is a measure for the spatial depth of objects. The reflection data is registered and edited to form a so called radargram. The major problem facing the GPR technique is the difficulty associated with the visualization and interpretation of the radargrams. Currently, this is still done manually and, therefore, subjective and time consuming. Available software packages support only basic manipulation of the image data, which includes background removal, noise filters, and edge enhancement. Therefore, improving the processing of GPR data is still a major field of research. Contrarily, medical three-dimensional (3D) ultrasound is likewise based on reflection signals, but automated analysis, visualization, segmentation and interpretation of such data has already been well established. Within this paper, a converter software is introduced that aims to interface GPR data to medical ultrasound and apply medical image visualization and processing techniques to GPR data. From the GPR radargrams stored in DZT format, corresponding ultrasound files conformant with the Digital Imaging and Communications in Medicine (DICOM) standard are generated and rendered into volumes with ImageJ and the Medical Imaging and Interaction Toolkit (MITK). As a result, hybrid visualization is obtained where direct volume rendering is combined with model-based surface reconstruction of cylindrical pipes, pipe systems, cavities, and other basic structures. Furthermore, our interface allows the application of medical image processing and analysis to GRP data.