alper.selver@deu.edu.tr

This study provides a necessary step for extending repeatability and mobility capabilities of 3D medical imaging by developing a DICOM compatible object, which covers visualization parameters together with compressed binary segmented data for efficient transmission in telemedicine applications.

In the current state of the DICOM standard, there are no modules that provide the possibilities of parametric representation by 2D Presentation States (2DPR) to 3D imaging. Once the final 3D rendering is obtained, current methods use video/image exporting to save the rendering result. To increase the utility of 3DPR in clinical practice, it is important to attach the available segmentation results, which are in binary form. Moreover, for effective use in telemedicine applications, this binary volumetric data should be compressed in a lossless way. In previous studies, several lossless compression methods were applied to various segmented anatomical organs (aorta, skeleton, skull, liver and kidneys etc.). Unfortunately, the results show the lack of 3D processes for lossless compression. Thus, in this study, the compression ratios of an existing effective method (i.e. Run Length Encoding - RLE) pipeline is improved by offering various scan forms (such as spiral and chevron) which are morphologically coherent with the organs in the binary image.

A large set of scanning possibilities provide low-entropy sequences for coding. Thereby, the compression performance of the RLE based system is improved. The developed system is applied to 20 segmented liver data sets having 70 to 110 images. Each image is processed by JPEG lossless (JPEG-LS) and lossy versions, RLE based compression scheme with column wise, row wise, zigzag, chevron and spiral scan orders. The result has shown that RLE based technique with spiral order give the best compression performance. The average compression efficiencies by total size is found as 22.25 MB (uncompressed), 3.04 MB (JPEG-LS), 0.52 MB (JPEG), and 0.23 MB (RLE-Spiral).

Extending RLE based methods with 3D scanning provide a significant compression performance over 95% while the performance of the 2D techniques remains under %90.