Dicom Image Handling for Medical Analysis and the ViVa Project

Marco Donizelli and Andrea Giachetti CRS4, Cagliari, Italy

Introduction

In this paper we describe the facilities for image retrieval, display and segmentation used both for direct medical analysis of the images and the 3d reconstructions requested by the CRS4 ViVa (VIrtual VAscular) project. The target of this project is the realization of a simulated vascular environment obtained by reconstructing the 3D vessel structures from processed diagnostic images and then performing blood flow simulations or guided navigation in the ``virtual'' vessels.

Methods

We built a system able to retrieve the images directly from a DICOM storage station (based on CTN routines) able to get and receive images from DICOM modalities connected on the net. On this station an image server program (DIRSI) is installed and can accept requests for images by name or by patient information, sending back the slices in the format requested by the Java or C/C++ viewing station. The scheme of the interaction is shown in Fig. 1.

[fig 1. not shown]

C++/Java libraries to make requests to the server have been realized and segmentation tools for local workstations realized at CRS4 use this approach to get the images.
The main display/segmentation facility developed at CRS4 is the OODISS, a Java Object Oriented display and segmentation tool. This tool allows the user to load multiple images, perform almost any kind of basic image processing and several segmentation procedure based on edge following, a balloon model and the Generalised snakes routine (GSNAKE) through a client-server CORBA connection with the original C program. The object oriented approach and the multi-display system will allow the insertion of other processing tools (both with Java computational kernel or C kernel on particular platforms through the client-server CORBA connection if the computational weight is huge), the use of Java makes the viewer platform-independent.

Results

The system built is ready to be applied in an hospital environment with DICOM modalities and can allow an easy image retrieval and examination from any pc or workstation connected on the local area network.
For the ViVa project, the system is able to extrect and save 2D contours from the slices in the format required by the geometry reconstruction tool developed at CRS4 and the first simulation works have been realised and presented in conferences (SIMAI 98, MEDICON 98). In the future the segmentation tool will be directly interfaced using CORBA with the reconstruction programs. Support for direct 3D segmentation and reconstruction will be also added.

Corresponding Author:

Andrea Giachetti

CRS4, Via N. Sauro 10

09123 Cagliari

e-mail giach(at)crs4.it

Oral presentation at EuroPACS'98, Barcelona, Spain