3D Female Pelvic Organ Models:
Comparison of the Visible Human Female with a Reproductive Age Pelvis

Wm. LeRoy Heinrichs, M.D., Ph.D., Ann Pothen, B.S., Rory Mather, B.A., Philip Constantinou, B.S., Matthew Lewis, B.S., Robert A. Chase, M.D., and Parvati Dev, Ph.D.

Department of Gynecology and Obstetrics (WLH), Division of Anatomy (RAC), and Stanford University Medical Media Information Technologies (SUMMIT), Stanford University,
Stanford, CA
MATERIALS

The pelvic organs, concentrating on the internal reproductive tract of the Visible Human Project-Female (VHP-F) and a unique, cryo-preserved human female pelvis cadaver of a 30 year old woman, the Stanford Visible Human-Female (SVH-F) are being extracted for developing 3D models. Among the pelvic organs of primary interest to gynecologic surgeons, the uterus, fallopian tubes and ovaries are compared in this initial study.

METHOD

A rectangular volume of data, enclosing the reproductive organs of interest, was selected from each of the two datasets. From the VHP-F, this volume comprised 155 slices, 450 pixels wide x 500 pixels high (pixel size ~0.343mm, slice separation 0.33mm). From the SVH-F, this volume comprised 30 slices, 475 pixels wide x 525 pixels high (pixel size ~0.25 mm, slice separation 1.5mm).
Each pixel in the volume was converted from a 24-bit RGB color to an 8-bit grey value using the formula:

I = 0.2989 * R + 0.5870 * G + 0.1140 * B

where, R, G and B are the pixel intensity in the red, green, and blue channels respectively, and I is the intensity of the resulting grey scale pixel.

The grey scale volume was processed through volume rendering software, X3D, provided by CEMAX-ICON, Inc., running on a Sun Ultra 1. Cut planes were defined so as to eliminate overlying structures such as bowel, and to make visible the ovaries and the uterus. The rendering software allowed us to define a range of grey scale intensities that highlighted the desired organs, separating them from the surrounding fat and other tissues.

Prior to conversion of slice data into three-dimensional surface models, it was necessary for a content expert (WLH) to preprocess every slice and mask out all except the relevant structures. This manual step was necessary because efforts at automated segmentation would always cause unwanted nearby structures of similar intensity to be included along with the desired organ.

Three dimensional surface models were constructed using the Visualization Toolkit (vtk) software, available from General Electric Comapny, and the Tcl interpretive language, operating on a SGI Indigo 2 workstation. For generating the models, marching cubes and various filtering algorithms were applied to transform the slice data into outlines of objects, and then into a triangulated surface. Further refinement was accomplished by converting the models (written in STL format) into Inventor format and applying VisModel CAD modeling software from Engineering Animation, Inc.

RESULTS

Images, presented as a movie for each dataset (see links in the Materials section), represent the comparable slices. Differences are clearly visible in the shape, size, and position of the reproductive organs. We measured the antero-posterior dimension of the uterus and ovary. In the SVH-F, a woman of reproductive age, the uterus measures 4.6cm and the ovaries are 1.8cm. In the VHP-F, a post-menopausal woman, the uterus is 3.7cm and the right ovary is 1.0cm. Another movie presents a rendered view of a 3D surface model of the ovaries and uterus (SVH-F-3D) derived using the methods above.

CONCLUSION

These procedures permit development of authentic, integrated, 3D surface and solid organ models that can form the fundamental virtual anatomic regions of surgical interest (AROSI) essential for simulations of videoendoscopic surgery.
© SUMMIT, Stanford University