Overview of the Visible Korean Project

Purpose and Contents of the projet

The purpose of this project is to prepare MRIs, CTs, sectioned images, segmented images, which are reconstructed to build three-dimensional images. The images would be the source of virtual dissection, virtual operation, virtual endoscopy to contribute in the medical, informatics, and industrial fields. To achieve the purpose, four cadavers were serially sectioned and anatomical disectioned image, MRI and CT scanned image data stored as follows; Whole body of male (Visible Male) (2000 year -)
Head of male (Visible Head) (2006 year -)
Pelvis of female (Visible Pelvis) (2007 year -)
Whole body of female (Visible Female) (2008 year -)
Whole body of male and female (3D visualization and development of viewer/soft ware) (2000 year -)


Information on affiliations and investigators

1. Affiliation: Department of Anatomy, Ajou University School of Medicine

2. Principal investigator: Prof. Min Suk Chung (dissect@ajou.ac.kr)

3. Co- investigators / affiliations
Jin Seo Park (park93@dongguk.ac.kr) / Department of Anatomy, Dongguk University College of Medicine
Sung Bae Hwang (hsb@kbc.ac.kr) / Department of Physical Therapy, Kyungbuk College
Dong-Hwan Har (dhhar@cau.ac.kr) / Graduate School of Advanced Imaging Science, Multimedia & Film, Chung-Ang University

4. Instruments (in Department of Anatomy, Ajou University School of Medicine)
Embedding box: to embed whole body of cadaver
Freezer: to freeze whole bodies of two cadavers (- 70 degree C)
Cryomacrotome: to serially section whole body of cadaver with 0.1 mm, 0.2 mm intervals (error: 0.001 mm) (Hanwon, Korea)
Digital camera: to make sectioned images with 0.1 mm sized pixels (resolution: 12 mega pixels) (Canon EOS 5D)


Procedure in General for obtaining MRI, CT and sectioned image data

1. Cadaver
A male cadaver, who died of leukemia on March 26th, 2001, was chosen for the final experiment; he was young (33 years old), and the body size (1,640 mm, 55 kg) was Korean average

2. MRI and CT scanning
The entire body of the cadaver was MR scanned followed by frozen, after that CT scanning was performed. The procedure kept was on purpose to achieve in comparison between the three images, such as anatomical, MRI and CT images.

2-1. MRIs
The entire body of the cadaver was MR scanned by 1.0 mm intervals on March 27th, 2001. The immobilizing box containing the cadaver was placed on the bed of the MR machine (GE Signa Horizon 1.5Tesla MRI System, Milwaukee, WI) parallel to the longitudinal axis of the bed; a laser light guided parallel positioning. Using body coil, the cadaver was horizontally MR scanned at 1 mm slice thickness and 0 mm interslice gap to produce 1,718 MR images (intervals: 1 mm). Since it was impossible to do an MR scanning of the entire body at once, it was processed in two series: from head to knee and from knee to toe. After MR scanning the entire body, two series of the MR images were combined and aligned on apersonal computer. Field of view and resolution of the MR images was 480 mm X 480 mm and 512 X 512, respectively, so the pixel size of the MR images was approximately 1.0 mm. While MR scanning, a makeshift T1 method was used to distinguish the various anatomical structures. Repetition time was fixed at 800 msec and echo time was fixed at 8 msec to increase the signal/noise ratio. Interleave method was used to remove interference between neighboring MR images.

2-2. CTs
The entire body of the cadaver was also CT scanned by 1.0 mm intervals. The immobilizing box containing the cadaver was placed on the bed of the CT machine (GE High Speed Advantage, Milwaukee, WI) parallel to the longitudinal axis of the bed; a laser light guided parallel positioning. The entire body of the cadaver was horizontally CT scanned at 1 mm slice thickness and 0 mm interslice gap to produce 1,718 spiral CT images at 1 mm intervals. The first CT image series, from head to knee,and the second CT image series, from knee to toe, were scanned separately, then combined and aligned as the same procedure of MR scanning. Field of view and resolution of the CT images was 480 mm X 480 mm and 512 X 512, respectively, so the pixel size of the CT images was approximately 1.0 mm. While CT scanning, standard algorithm was used, voltage was fixed at 120 kV, and the electric current time was fixed at 280 mAs in order to make the soft tissue distinct. As soon as CT scanning was completed, the immobilizing box was again frozen in the deep freezer.

3. Sectioned images
The cadaver was put into the embedding box. Blueembedding agent (gelatin: 30 g, methylene blue: 0.5 g, distilled water: 1000 ml) was poured into the embedding box until the agent filled about a quarter of the embedding box; the temperature was maintained at -70C in the freezer. After flattening the upper surface of the frozen embedding agent, the cadaver was taken out of the immobilizing box and laid down in the embedding box without changing the cadaver's direction. A thread was attached to the embedding box in a longitudinal direction to verify symmetry of the cadaver's head, body, and limbs.

The cadaver was embedded and frozen. A small quantity of embedding agent was poured into the embedding box and frozen to -70C in the freezer. This process was repeated until the embedding box was fully filled withthe embedding agent. These repeated processes were necessary in order to prevent the freezing embedding agent from pressing the cadaver and the alignment rods. Wood panels connected the upper parts of the two sideboards in order to prevent the freezing embedding agent from widening the two sideboards.

The embedding box was placed on the cryomacrotome and firmly fixed. Due to the heavy weight (oneton) of the embedding box, a cart was used to transfer it from the freezer to the cryomacrotome and a crane was used to place it on the cryomacrotome. When the embedding box was placed on the cryomacrotome, the footboard was directed to face the cutting blade. As a result, serial sectioning was performed feet to head so that the direction of the sectioned surfaces became the same as that of MR and CT image. The embedding box was carefully placed on the cryomacrotome parallel to the longitudinal axis of milling table and firmly fixed using several holes and screws. Footboard of the embedding box was removed before the first serial sectioning.

The process of serially sectioning the embedding box was performed for five months (November, 2001 - March, 2002) to make the sectioned surfaces. The embedding box on the milling table was moved toward the cutting blade at 0.2 mm interval. During the cutting blade rotated constantly at optimal speed, the embedding box was moved at optimal speed parallel to the cutting blade; as a result, the embedding box was serially sectioned at 0.2 mm interval. After each serial sectioning, every sectioned surface was moved to a constant position for photographing the sectioned surface. These movements of the embedding box and cutting blade were performed repeatedly by a program composed of numerical control language in the automated control box of the cryomacrotome. After a day's serial sectioning, the embedding box was returned to the freezerand the cadaver debris and embedding agent debris were collected to be sent to the crematory.

The sectioned surfaces were photographed using the digital camera to produce anatomical images, which were transferred to the personal computer. Every sectioned surface was photographed under constant conditions (F value: 10, shutter speed: 1/250 s, focusing: manual) of the digital camera while two strobe lights were flashed. The anatomical image made by photographing the sectioned surface was transferred to the personal computer and its quality (brightness, color, focus) was verified on the computer monitor. Then the anatomical image was saved as TIFF format on two personal computers before the next serial sectioning. This photographing was continuously performed after serial sectioning, which resulted in 8,590 anatomical images. Constant brightness of the anatomical images was verified by checking the red, green, blue values of the gray scale and color patch in the serial anatomical images. Alignment of the anatomical images was verified by examining four alignment rods and each anatomical structure's contours in the serial anatomical images.

4. Outlined images
Sectioned images of cadavers enable the creation of realistic three-dimensional (3D) models. In order to build a 3D model of a structure, the structure has to be outlined insectioned images. As this outlining is time consuming, users want to be provided with outlined images, and the more detailed structures are outlined, the more widely can the outlined images be utilized. In the Visible Korean, sectioned images (intervals 0.2 mm) of a male cadaver's whole body were prepared. In the available 1,702 sectioned images (intervals 1 mm), 902 structures were outlined interactively on Photoshop over a period of seven years. The outlined images were changed into black-filled images of each structure; the black-filled images were decided to distribute because of small file sizes. We investigated whether black-filled images have potential to be applied in different ways. Outlines of these images were interpolated to produce new images at 0.2 mm intervals, and outlines were filled with different colors to acquire color-filled images of whole structures. Volume and surface reconstructions of these black-filled images were used to build satisfactory volume and surface models. The black-filled images accompanied by corresponding sectioned images could provide a source of 3D models for medical simulation systems.

5-1. Volume modeling
The anatomical and segmented images were stacked and volume-reconstructed to produce 3D images. All 8,590 anatomical images (3,040 X 2,008 pixels) were stacked at 0.2 mm intervals and subsequently volume-reconstructed to produce a 3D image, which consisted of 3,040 X 2,008 X 8,590 voxels. The segmented images were made into a 3D image in a likewise manner.
The 3D images of the anatomical and segmented images were sectioned to produce coronal and sagittal images. The 3D image of anatomical images was coronally sectioned to produce 2,008 coronal anatomical images (3,040 X 8,590 pixels) and sagittally sectioned to produce 3,040 sagittal anatomical images (2,008 X 8,590 pixels). Likewise, 2,008 coronal segmented images and 3,040 sagittal segmented images were made of the 3D image of segmented images .
In the coronal and sagittal anatomical images, smoothness of the four alignment rods and each anatomical structure's contours were examined to verify alignment of the anatomical images and correctness of the segmented images.
3D images of selected anatomical organs were displayed and rotated. From the 3D image of anatomical images, the 3D image of an anatomical organ was extracted and displayed with reference to the 3D image of the corresponding segmented image. Likewise, 3D images of several anatomical organs were extracted and displayed; some anatomical organs were semitransparently displayed. The 3D images were rotated at free angles.

5-2. Surface modeling
Most currently available three-dimensional surface models of human anatomic structures have been artistically created to reflect the anatomy being portrayed. We have recently undertaken, as part of our Visible Korean studies, to build objective surface models based on cross-sectional images of actual human anatomy. Objective of the present study was to elaborate surface models of the structures that are helpful to medical simulation. The structureswere outlined and reconstructed from sectioned images of the Visible Korean (a computer database containing the digitized transverse sectional images of a 33-year old Korean man). The outlining procedure was supported by computational filtering and interpolation using commercially available software. The structures were surface reconstructed to produce surface models. The surface models produced will hopefully facilitate the development of interactive simulations for a variety of virtual surgical procedures or other educational programs. In addition, it is hoped that the improved outlining and surface reconstruction techniques described will encourage other researchers to construct similar surface models based on images obtained from different subjects.