Current Development of PET Image Analysis Techniques in NPCC.

Liang-Chih Wu, Jiunn-Kuen Wang, Chin-Lung Yu, and Ren-Shyan Liu.

National PET/Cyclotron Center, Taipei Veterans General Hospital.

 

As Positron Emission Tomography (PET) can provide physiological and biochemical information in a quantitative way, the image analysis techniques play an important role in the aid of diagnosis.  Our center has been working on the development of tools that are helpful to both the clinicians and researchers.  Some of the current development will be discussed.

The interpretation of whole body PET (WB-PET) studies is time-consuming because of the noisy property and the massive number of images.  We have therefore developed a PC-based multi-dimensional image display system that can improve the localization of lesion and the visual interpretation of WB-PET data.  Our system provides an integrated environment that can simultaneously display the three-dimensional (3D) volume-rendering emission images(EMIs), surface-rendering transmission images(TRI), transaxial images, coronal images, and sagittal images. In this program, both 3D and 2D EMIs was fused with TRIs to enhance the structural appearance.  During the rendering procedure of 3D images, the depth value and the coordinates of voxels were reserved in depth, x, y and z-buffers.  These buffers were used for interactive operation of 3D object merging and display. Whenever the operator clicked one pixel in the 3D maximum-value-projection (MVP) images, the corresponding axial and 3D images could be created.  With the feature, operator could easily define the geometric position of lesion and compare the structure location in 2D and 3D images.  The software renders the MVP into 15-bits image buffers instead of 8-bits depth. Such consideration can preserve the accuracy and provide the flexibility for the adjustment of window value range to decrease the influence of high activity value such as those in bladder or injection site.

Due to the technology complexity and cost consideration, not many PET centers have been setup in our country.  Accordingly, few investigators are currently working on PET imaging.  To promote the utilization of PET technology and nuclear medicine techniques, it is in need to collaborate for the development of image analysis tools.  We are therefore studying the feasibility of remote image analysis and web-based collaborative computing environment.  The system was implemented in a 3-tier client/server model.  Users could select, view and process PET images with a Java-enabled browser.  The processing tools written in Java applets were transferred from the web server that communicated with the database server using ODBC protocol.  Users can process PET images with their own browser and get the quantitative results by drawing regions of interest in the windows of browser.  For quantitative analysis, the FDG and FDOPA dynamic PET studies were implemented for testing the collaborative design.  The advantages of such a web-based collaborative computing environment are (1) ease of software maintenance, (2) reduction of network traffic, and (3) ease of collaboration and sharing of development tools/results.