Distributed quantitative evaluation of 3D patient specific arterial models A. Giachetti M. Tuveri CRS4 - VI Strada Ovest, Z.I. Macchiareddu, Uta(CA), Italy e-mail giach,mtuveri,zag @crs4.it G. Zanetti Abstract In this paper we describe a new system for the 3D recon- struction and distribution on the net of models for vessels structures. The system is specifically designed to support measurements of medical interest. We describe 2D and 3D segmentation methods implemented and the procedure used to build interactive VRML97 models. The experimental sec- tion presents a comparison between segmentation methods, and a first application to surgical planning for endovascu- lar repair of Abdominal Aortic Aneurysms. 1. Introduction In the near future a large amount of medical data will be transmitted across Internet for remote diagnosis, electronic patient records retrieval, home care, etc.. Images, annota- tions and texts can be easily transmitted with standard pro- tocols and can be viewed by medical doctors everywhere by using common web browsers. Many examples of web based telemedicine have been proposed and presented in confer- ence papers [2, 3, 4]. Moreover, there are few examples in literature relative to the web distribution of 3D models of organs, and they are not usually tailored to real clinical needs or diagnostic purposes. Typically, these models are used only for qualitative study or to implement simple sur- gical simulators [8], albeit current technology limitations force a strong simplification of the procedures being simu- lated. In this paper, we show that it is possible to use current web technology standards, i.e., VRML, to assemble patient specific models of vessels that allow surgeons to measure inherently 3D quantities that are important, for example, to surgical planning, e.g., the angles between vessels at the il- iac bifurcation. The models are reasonably compact in size, with an average size of few hundred kilobytes, and can be quickly transferred across Internet. The data sent is essen- tially the vessel surface geometry plus embedded code and pre-computed tables to support the measurements. More- over, the model is structured so that by selecting any point on a centerline it is possible to require, to a dedicated web server, a slice across the original volumetric dataset(s) or- thogonal to the local direction of the centerline, thus allow- ing the doctor to consult at the same time the original medi- cal imaging data. The paper is organized as follows: Section 2 will describe different methods for 3D segmentation, Sec- tion 3 will describe the VRML generation, Section 4 will present the application to Abdominal Aortic Aneurysm en- dovascular treatment, Section 5 will present experimental results. 2 Segmentation and 3D reconstruction 2.1 The arterial tree data structure In order to perform measurements on vascular geome- tries (and also to support other virtual reality and numeri- cal simulation applications) we designed a specialized data structure that we named “Arterial Tree”. An Arterial Tree is defined as the union of a complete surface mesh describing the vessels internal surface, and a skeleton joining series of 1D lines representing the vessels centerline. This data structure allows us to realize precise measurements of vessel length; simplify the estimate of a plane perpendicular to the vessel; provide a geometrical model that can be used to build a volume mesh inside the vessel for numerical simula- tion of blood flows [1]. We implemented three methods for the reconstruction of the AT structure and developed an user friendly interface to control the use of these methods and other image processing tools (see Fig.1). Methods are: