Sectioned Images of the Cadaver Head Including the ... - Visible Korean

2 downloads 64741 Views 2MB Size Report
enabled us to prepare quality 7.0-T MRIs, sectioned images, and segmented images of the head. ... 2009 and June 7, 2009. Current version published November 18, 2009. .... the entire cadaver was adjusted for an axial plane. Although the axial .... checked for quality using the Adobe Photoshop CS3 ex- tended version 10 ...
CONTRIBUTED P A P E R

Sectioned Images of the Cadaver Head Including the Brain and Correspondences With Ultrahigh Field 7.0 T MRIs Very detailed images of the brain, able to reveal small brain structures that are important for learning and teaching anatomy, are being generated by computer modeling. By Jin Seo Park, Min Suk Chung, Dong Sun Shin, Dong-Hwan Har, Zang-Hee Cho, Young-Bo Kim, Jae-Yong Han, and Je-Geun Chi

| Unlike computed tomographic images and mag-

images. In this research, advanced techniques and equipment

netic resonance images (MRIs), sectioned images of the human

enabled us to prepare quality 7.0-T MRIs, sectioned images,

body with real color and high resolution have certain advantages in learning and teaching anatomy. Comparisons between

and segmented images of the head. These images are expected to contribute to our understanding of the topographic neuro-

sectioned images of the brain and MRIs are useful in many

anatomy of the head and to aid interpretations of MRIs and CTs

ways. Therefore, we prepared 312 MRIs at ultrahigh field 7.0 T

of the human brain.

ABSTRACT

(axial direction 0:4  0:4  0:4 mm3 voxel size) of a cadaver brain, 2343 sectioned images (axial direction, 0.1 mm intervals,

KEYWORDS | Cadaver brain; sectioned images; segmented

0:1  0:1 mm2 pixel size, and 48 bits color) by serial-sectioning

images;

the cadaver head, 234 segmented images in which brain

7.0-T MRIs

regions were separately delineated (1 mm intervals and 0:1  0:1 mm2 pixel size) by outlining 64 head structures in sectioned images. Three-dimensional images of 64 head structures were made by volume reconstruction from sectioned

Manuscript received February 6, 2009; revised April 14, 2009 and June 7, 2009. Current version published November 18, 2009. This work was supported by the Dongguk University Research Fund of 2008. J. S. Park is with the Department of Anatomy, Dongguk University College of Medicine, Gyeongju, South Korea (e-mail: [email protected]). M. S. Chung (corresponding author) is with the Department of Anatomy, Ajou University School of Medicine, Suwon, South Korea (e-mail: [email protected]). D. S. Shin is with the Department of Electrical Engineering, Seoul National University College of Engineering, Seoul, South Korea (e-mail: [email protected]). D.-H. Har is with the Graduate School of Advanced Imaging Science, Multimedia and Film, Chungang University, Seoul, South Korea (e-mail: [email protected]). Z.-H. Cho and J.-Y. Han are with the Neuroscience Research Institute, Gachon University of Medicine and Science, Incheon, South Korea (e-mail: [email protected]; [email protected]). Y.-B. Kim is with Department of Neurosurgery, Gachon University Medicine and Science. He is also with Neuroscience Research Institute, Gachon University of Medicine and Science, Incheon, South Korea (e-mail: [email protected]). J.-G. Chi is with the Department of Pathology, Seoul National University College of Medicine, Seoul, South Korea (e-mail: [email protected]). Digital Object Identifier: 10.1109/JPROC.2009.2025524

1988

three-dimensional

images;

0.1-mm-sized

voxel;

I . INTRODUCTION Serial-sectioning of an organ or a whole body has certain advantages in learning and teaching human anatomy. Unlike magnetic resonance images (MRIs) and computed tomographic images (CTs), sectioned images (SIs) offer visualization in real color at high resolution and can be used to make realistic three-dimensional (3-D) images [1]–[8]. High-resolution SIs are essential in understanding human body structures. Therefore, the SIs of complete cadavers were created in the U.S. Visible Human Project (0:33  0:33  0:33 mm3 -sized voxels, 24-bit color) [1], the Visible Korean (0:2  0:2  0:2 mm3 -sized voxels, 24-bit color) [3], and the Chinese Visible Human (0:2  0:2  0:2 mm3 -sized voxels, 24-bit color) [6]. Three-dimensional virtual dissection software based on the SIs were made and referred to as Voxel-Man of the Visible Human Project [2] and as virtual lumbar puncture of the Visible Korean [8]. However, the SIs of the Visible

Proceedings of the IEEE | Vol. 97, No. 12, December 2009

0018-9219/$26.00 Ó 2009 IEEE

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

Park et al.: Sectioned Images of the Cadaver Head

series did not satisfy neuroscientists because the voxel size of the SIs was insufficient to observe small structures within the brain. Moreover, although the Visible series included 1.5-T MRIs (intervals and pixel size 1.0 mm) [1], [3], [6], the quality of the MRIs was not satisfactory either. Therefore, ultra-high-resolution SIs and MRIs of the brain are needed for use by neuroscientists. At present, ultrahigh field 7.0-T MRIs are used for brain research because brain structures can be observed in detail [9]. Therefore, many clinicians use 7.0-T MRIs to study brain areas of interest. We believe that SIs corresponding to 7.0-T MRIs of an important region or organ, such as the brain, will be appreciated by clinicians. Moreover, given the advantages of neuroimaging technologies, there is an increased need for SIs of the brain that can be compared with high-resolution 7.0-T MRIs. The objective of this research was to present 7.0-T MRIs and upgraded SIs of a cadaver head, which would be helpful in creating a brain atlas, virtual dissections, and virtual operations using 3-D images. To achieve this objective, a normal cadaver head, including the whole brain, was scanned with a 7.0-T MR unit to produce MRIs with 0:4  0:4  0:4 mm3 sized voxels, and also serially sectioned in the axial direction to produce SIs with 0:1  0:1  0:1 mm3 sized voxels in 48-bit color, which were precisely aligned and had constant brightness. Significant head structures in the SIs were outlined to create segmented images and 3-D images.

Fig. 1. (a) Cadaver head in the 7.0-T magnetic resonance scanner, (b) showing the axial and sagittal lines on the cadaver’s face.

II . MATERIALS AND METHODS The donated cadaver of a 67-year-old Korean man, 1620 mm in height and weighing 45 kg, was selected. He had died of cardiorespiratory arrest and had myasthenia gravis. No fixatives or dyes that would alter the tissue colors of the brain, muscle, or other tissues [1], [6] were injected. Four hours after death, the brain was scanned using a 7.0-T MR scanner (Magnetom, Siemens AG, Berlin, Germany) consisting of a 90-cm bore superconducting magnet (Magnex Magnet Technology, Oxford, U.K.) and equipped with a Siemens Syngo console [Fig. 1(a)] [9]. A home-built 8-channel phased array coil was utilized for imaging. A 7.0-T MRI study involving living humans is being performed, but a 7.0-T MRI study has not been performed involving cadavers. Therefore, we should determine the parameter of 7.0-T MR scanning for cadavers through trial and error. As a result, the 7.0-T MR scanning parameters for the cadaver head were as follows: T2weighted, TR ¼ 39 ms, TE ¼ 17 ms, BW ¼ 60, flip angle (FA) ¼ 30 , and matrix size ¼ 480  576  104. In addition, three imaging sessions were performed for the whole brain coverage, and the total matrix size was 480  576  312. As a result, 7.0-T MRIs of the cadaver brain were acquired with 0:4  0:4  0:4 mm3 -sized voxels [Fig. 2(a) and Table 1].

Fig. 2. (a) 7.0-T MRIs, (b) sectioned images, and (c) segmented images of the cadaver. The third row (a) 7-T MRIs and (b) sectioned images show both the (1) anterior and (2) posterior commissures.

Vol. 97, No. 12, December 2009 | Proceedings of the IEEE Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

1989

Park et al.: Sectioned Images of the Cadaver Head

Table 1 Features of the 7.0-T MRIs, Sectioned Images (SIs), and Segmented Images

During MR scanning, the axial plane of brain was identified. In previous studies [1], [3], [6], the direction of the entire cadaver was adjusted for an axial plane. Although the axial plane of the entire body was sought, the axial plane of the brain was not identified. Therefore, in the study described in this paper, the direction of the head was adjusted for an axial plane. The axial plane was determined referring to the anterior and posterior commissures, which were shown in an MRI [Fig. 2(a)]. After scanning, horizontal and sagittal lines were drawn on the cadaver face along the laser indicator of the MR machine [Fig. 1(b)]. After acquiring MRIs, the cadaver was placed into a freezer. A week later, the head, together with the upper part of the thorax (head block), was separated from the main body with a saw. A small block of the thorax (thorax block) with similar width and height diameters to the axial planes of the head block was separated [Fig. 3(a)], and the two blocks were refrozen.

Fig. 3. (a) Head and thorax blocks in the embedding box for the main and preliminary experiments. (b) Embedding agent is being poured into the box.

1990

The head and thorax blocks were placed in a heavy embedding box with a steel base and wooden side boards [3]. The head block was placed for grinding from the vertex of the head to the chin. Subsequently, the direction of the head block was adjusted in the embedding box until the axial and sagittal lines drawn on the face were parallel to the short and long side boards of the embedding box. The thorax block was placed on the top of the head block for a preliminary trial, and the position of the thorax block was adjusted just above the head block [Fig. 3(a)]. The embedding agent (1000 ml of water, 30 g of gelatin, and 0.5 g of methylene blue) was poured around the head and thorax blocks to fix them rigidly in the embedding box. Subsequently, a small amount of embedding agent was poured into the box and frozen. This procedure was repeated to prevent rapid volume increases of the embedding agent until the embedding box was filled with the embedding agent [Fig. 3(b)] [3]. As a preliminary trial, the thorax block in the embedding box was milled with a cryomacrotome; the milling machine was remodeled in Hanwon Corporation, South Korea. In the cryomacrotome with a moving error of less than 0.001 mm, the cadaver could be serially sectioned at constant thicknesses. The rotating speed of the cutting disc and the moving speed of the embedding box were adjusted to acquire good quality sectioned surfaces [Fig. 4(a)]. For the same reason, during the preliminary trial, we optimized the blade changing cycles in the cutting disc. Accordingly, we were able to achieve section intervals of 0.1 mm, as compared with the 0.2 mm sections achieved previously [3]. In the preliminary experiment using the thorax block, the optimal photographic conditions were determined. We used a Canon EOS 5D digital camera with a Canon 50 mm micro lens. Resolution and color depth were 4368  2912 pixels and 48-bit color, respectively. The distance from the digital camera to the sectioned surface was adjusted to the photograph area (436:8  291:2 mm2 of the sectioned surfaces) [Fig. 4(b)]. As a result, SIs with

Proceedings of the IEEE | Vol. 97, No. 12, December 2009 Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

Park et al.: Sectioned Images of the Cadaver Head

Fig. 4. (a) The embedding box and the cryomacrotome with cutting disc during sectioning at 0.1 mm. (b) Sectioned surface, which is being photographed using a digital camera.

Fig. 5. A serially sectioned image taken (a) without and (b) with polarizing filters showing the elimination of scattered reflected lights.

pixel sizes of 0:1  0:1 mm2 were acquired. The positions and directions of two Elinchrom Digital S strobes were adjusted to maintain a constant brightness across the sectioned surfaces. The digital camera lens and the strobes were covered with polarizing filters to reduce scattered reflected lights (Fig. 5). The sectioned surfaces were photographed using strobe lights to optimize the filter settings and shutter speed [3]. The main experiment was conducted as follows. The head block on the cryomacrotome was serially ground from the vertex to the chin at 0.1 mm intervals to obtain 2343 sectioned surfaces. Subsequently, the same block was ground from the chin to the superior border of the sternum at 0.5 mm intervals to produce 157 additional sectioned surfaces of the neck [Fig. 4(a) and Table 1].

After wiping away frost and particles from the sectioned surface, it was photographed such that the photograph included the surrounding embedding agent, gray scale, and color patch [Fig. 4(b)]. The photograph was digitalized: Image data were saved as both a joint photographic coding experts group (JPEG) file in 24-bit color and a Canon raw 2 (CR2) file format in 48-bit color to create the SIs. On the computer monitor, the JPEG SI was checked for quality using the Adobe Photoshop CS3 extended version 10 (Photoshop). After checking, serialsectioning was continued. Finally, SIs were converted from the CR2 format to the tag image file format (TIFF) with 48-bit color using Canon EOS utility software. On Photoshop, SIs were converted right-to-left to conform with radiologic convention. It took 50 days in the winter to acquire the 2500 SIs of the cadaver head and neck [Fig. 2(b) and Table 1]. Using self-developed software, a column of every axial SIs of the head was stacked to produce a sagittal image. Using a similar method, the rows of all axial SIs were stacked to produce coronal images [Fig. 6(a) and (c)]. Photographic errors of the SIs were repaired using Photoshop. Incorrectly aligned SIs were identified by referring to postmortem MRIs [Fig. 2(a)], sagittal SIs, and coronal SIs [Fig. 6(a) and (c)]. Incorrectly aligned SIs were automatically aligned to neighboring SIs using the Bload files into stack. . .[ tool in Photoshop (Fig. 7). SIs with inconstant brightness were identified by referring to gray scale and color patch in SIs [Fig. 4(b)] and to sagittal and coronal SIs [Fig. 6(a) and (c)]. Incorrectly bright and dark SIs were manually corrected using Photoshop. We outlined head structures in 234 SIs at 1 mm intervals (Table 1). Sixty-four important head structures were chosen for outlining. In the case of the vascular system, the luminal outlines of the arteries and veins, not their mural outlines, were decided to outline (Table 2). In a previous study, we developed the basic outlining procedures that each structure in the SIs was outlined using a magnetic lasso tool or a lasso tool in Photoshop to make segmented images [4], [10]. In the study reported in this paper, the basic outlining procedures of SIs were similar but were refined as compared with those used in a

Fig. 6. Sagittal (a) sectioned and (b) segmented images. Coronal (c) sectioned and (d) segmented images.

Vol. 97, No. 12, December 2009 | Proceedings of the IEEE Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

1991

Park et al.: Sectioned Images of the Cadaver Head

Fig. 7. (a) Incorrectly aligned continuous sectioned images and (b) images automatically realigned on Photoshop.

previous study. On SIs, selected anatomic structures, identified by referring to an atlas [11], were outlined automatically using the Bmagic wand[ tool in Photoshop. When automatic outlining was not possible, semiautomatic outlining was performed using the Bquick selection[ tool; after roughly outlining a small area of a structure, the outline was dragged until the expanded outline automatically fitted the area of the structure (Fig. 8). The Wacom tablet pen was more helpful than a computer mouse for outlining. The segmented images in which structure outlines were filled with different colors were saved in 8-bit color TIFF [Fig. 2(c) and Table 1]. Sagittal and coronal segmented images were also prepared using self-developed software to verify outlining [Fig. 6(b) and (d)]. As a preliminary step before volume reconstruction, the intervals of SIs were changed from 0.1 to 1 mm, the pixel

size was changed from 0:1  0:1 mm2 to 1  1 mm2 , and the bit depth was changed from 48-bit color to 8-bit gray. Using the segmented images, the outside of the head skin in SIs was automatically erased using Photoshop. On MRIcro software version 1.4 (MRIcro), the SIs were stacked and reconstructed by volume modeling to acquire a 3-D image of the head with 1  1  1 mm3 voxel size in 8-bit gray. The 3-D image was arbitrarily sectioned to display sectional planes. Likewise, after the outside of the internal head structures was erased, 3-D images of the structures were made and then rotated (Fig. 9).

I II . RESULTS Three hundred twelve 7.0-T MRIs of the cadaver brain were acquired. Although the cadaver head was scanned

Table 2 Sixty-Four Outlined Head Structures, Which Are Categorized According to the Systems

1992

Proceedings of the IEEE | Vol. 97, No. 12, December 2009 Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

Park et al.: Sectioned Images of the Cadaver Head

Fig. 8. (a) Outline of a small region of the globus pallidus, (b) which was expanded to fit the whole of the globus pallidus using mouse drag by ‘‘quick selection’’ tool on Photoshop.

Fig. 9. (a) Three-dimensional images of head, sectioned to display sectional planes; (b) the brain, eyeballs, and optic nerves; and (c) lentiform nuclei, caudate nuclei, thalami, and brainstem.

using a head coil [Fig. 1(a)] the MR scan sequences focused only on the brain to enable brain structures to be well observed at a voxel size of 0.4  0.4  0.4 mm3 [Fig. 2(a) and Table 1]. The MRIs showed fine structures, such as gray matter, white matter, optic radiation, and corpus striatum in the cerebrum. The MRIs showed also fine cerebral arteries without MRI contrast media [Fig. 10(c)]. High-quality MRIs were obtained using a 7.0-T MR machine [Fig. 2(a)]. A total of 2500 SIs of the cadaver head were obtained without any technical difficulty. The cadaver head was sectioned by grinding instead of being sliced to obtain SIs at intervals of 0.1 mm. The main results of the serialsectioning were 2343 SIs of the head from the vertex to the chin at 0.1 mm intervals, and additional ones were 157 SIs of the neck at 0.5 mm intervals [Fig. 2(b) and Table 1]. The SIs showed some postmortem changes. The main trunks of the cerebral arteries remained patent, although small arteries and all the veins collapsed, except the dural venous sinuses. Cerebrospinal fluid had largely disappeared from the ventricles, whereas fluid in the subarachnoid space had pooled in the cisterns. The SIs showed atherosclerosis of the cerebral arteries and mild subcutaneous edema and congestion in the soft tissues of the skull base, but there were no other specific pathologic findings. In SIs, anastomosis was noted between the occipital sinus and the left internal jugular vein, but we did not find any additional remarkable anatomic variations [Fig. 2(b)]. Accordingly, we regarded the cadaver as an appropriate subject for reference purposes despite the patient’s age and the presence of myasthenia gravis. In the axial and coronal SIs, the bilateral cerebral and cerebellar hemispheres were nearly symmetric, and the anterior and posterior commissures on the SIs were shown simultaneously [Figs. 2(b) and 6(c)], as was observed in 7.0-T MRIs [Fig. 2(a)]. It was possible because axial and sagittal lines were drawn on the cadaver face using the MRIs and direction of the head block was adjusted using the axial and sagittal lines on the cadaver face [Fig. 1(b)] before sectioning the head. Axial SIs corresponding to axial 7.0-T MRIs could be referred to for the clinical interpretation of MRIs and CTs of the head. The final SIs maintained correct alignment and constant brightness that could be verified by the corresponding MRIs [Fig. 2(a)] and sagittal and coronal SIs [Fig. 6(a) and (c)]. Incorrect alignment was easily revised by the new technology on Photoshop (Fig. 7). In SIs of a previous study [3], outlines of each structure in the cerebellum [Fig. 11(a)], brainstem [Fig. 11(b)], and ear structures [Fig. 11(c)] were not shown clearly. The SIs of this study showed fine structures, such as the dentate nucleus and the emboliform nucleus in the cerebellum [Fig. 11(d)], the substantia nigra and red nucleus in the brainstem [Fig. 11(e)], and fine structures of the ear [Fig. 11(f)]. Good quality SIs were obtained after preliminary optimizations and by using an advanced digital Vol. 97, No. 12, December 2009 | Proceedings of the IEEE

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

1993

Park et al.: Sectioned Images of the Cadaver Head

Fig. 10. 1.5 T and T2-weighted magnetic resonance images of (a) Visible Human Project and (b) Chinese Visible Human; (c) 7.0-T and T2-weighted magnetic resonance images of this study. The MRIs of this study showed more fine cerebral arteries (arrows of C), lentiform nuclei, gray matter, and white matter (c) than previous studies (a) and (b).

camera that allowed a 0.1  0.1 mm2 pixel size in 48 bits of color. Stereoscopically, SIs showed structures as small as 0.1 mm for the 0:1  0:1  0:1 mm3 -sized voxels. Furthermore, small-sized intervals and pixels enabled us to prepare sagittal and coronal SIs of high quality, which showed no difference in detail with the original axial SIs [Fig. 6(a) and (c)]. Two hundred thirty-four segmented images were prepared at 1 mm intervals [Fig. 2(c) and Table 1]. Sixty-four head structures were outlined (Table 2) more semiautomatically using the advanced techniques in Photoshop

(Fig. 8) than in a previous study [4]. Atherosclerosis was observed in the cerebral arterial walls, and the luminal outlines of the cerebral arteries were outlined more easily. The segmented images were satisfactory in quality, which was verified using sagittal, coronal segmented [Fig. 6(b) and (d)], and 3-D images of structures. Three-dimensional images of the head structures were made by volume reconstruction, and then the 3-D images were sectioned and rotated (Fig. 9). This was performed conveniently without the help of the computer programmers using MRIcro (a free software package with a user-friendly interface).

Fig. 11. Zoomed-in sectioned images of the (a) cerebellum, (b) midbrain, and (c) left ear of our previous study [3]; Zoomed-in sectioned images of this study of the (d) cerebellum, (e) midbrain, and (f) left ear showing detailed structures.

1994

Proceedings of the IEEE | Vol. 97, No. 12, December 2009 Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

Park et al.: Sectioned Images of the Cadaver Head

IV. DISCUSSION In the study described in this paper, a 7.0-T MR scanner contributed to an improvement of MRIs. In MR scanning for high-quality MRIs, the most important condition was the strength of the magnetic field of the MR scanner, which uses the Tesla unit [9]. In the Visible Human Project [1], the Visible Korean [3], and the Chinese Visible Human [6], a 1.5 T MR scanner was used; in this paper, a 7.0-T MR scanner was used [Fig. 1(a)]. The results showed that the images were much improved (Fig. 10). In addition, the cadaver was MR-scanned and then serially sectioned, and thus the 7.0-T MRIs could be directly compared with the SIs (Fig. 2). Because of these added improvements, we anticipate that the SIs and the corresponding 7.0-T MRIs will be more useful for clinical neuroanatomy. In our previous research, the pixel sizes of the SIs were 0:2  0:2 mm2 [3]; however, this was reduced to 0:1  0:1 mm2 in this paper. Additionally, here 48-bit color was used, whereas 24-bit color was used in our previous study (Table 1). These two improvements resulted in the production of higher quality SIs (Fig. 11). Therefore, the important structures were exactly identified and easily outlined [Fig. 2(c) and Table 2]. In this paper, 0.1 mm intervals, together with the smaller pixels of the SIs, contributed to the improvement of 3-D images as well. After stacking the SIs, volume reconstruction was performed to produce 3-D images composed of voxels (Fig. 9). In our previous studies, a voxel size of 0.2  0.2  0.2 mm3 was used [3], [4], whereas in this paper, a voxel size of 0.1  0.1  0.1 mm3 is able to be used (Table 1). Accordingly, all structures with a dimension > 0.1 mm can be identified in the 3-D image. The voxel size used in this paper would be the last trial in digitalized images of gross anatomy. Voxel sizes G 0:1  0:1  0:1 mm3 enter the microscopic level and have limited importance with respect to gross anatomy. The SIs of the complete cadaver indicate that the 3-D images of any region can be produced. In particular, continuous 3-D images from the brain to the peripheral nerve in the foot can be acquired. However, whole-body SIs present many technical problems. First, it would take approximately one year to obtain whole-body SIs at 0.1 mm intervals; indeed, 50 days were required to prepare the head alone. Furthermore, because our laboratory is not equipped with a large cold room, serial-sectioning can be performed only in the winter. So we demand several years. Secondly, an ad-

REFERENCES [1] V. M. Spitzer, M. J. Ackerman, A. L. Scherzinger, and D. G. Whitlock, BThe Visible Human male: A technical report,[ J. Amer. Med. Inf. Assoc., vol. 3, pp. 118–130, 1996. [2] T. Schiemann, J. Freudenberg, B. Pflesser, A. Pommert, K. Priesmeyer, M. Riemer, R. Schubert, U. Tiede, and K. H. Ho¨hne, BExploring the Visible Human using the

vanced digital camera is needed to acquire whole-body SIs with a pixel size of 0.1  0.1 mm2 . In this research, we used a digital camera with a 4368  2912 pixel resolution because the cadaver head in the sectioned surface was smaller than 436:8  291:2 mm2 (Fig. 4). Accordingly, because shoulder sections would be larger, a digital camera with a higher resolution would be necessary. Thirdly, wholebody SIs with a 0:1  0:1  0:1 mm3 voxel size would involve an enormous amount of data, which in itself causes distribution and processing difficulties. In our opinion, the segmentation of whole body SIs would require almost ten years to complete [4], [10]. We chose the head because it is relatively small, but important. We believe that the SIs of the head will contribute not only to understanding the topographic neuroanatomy of the head but also to the interpretation of MRIs and CTs of the brain. Accordingly, we have been labeling detailed brain structures on the SIs to allow them to be used as a reference source by neuroradiologists. The newest version of Photoshop contains tools that can be used to enhance automatic processing. To automatically correct mal-aligned SIs, the Bload files into stack. . .[ tool was used (Fig. 7); and to automate segmentation, the Bmagic wand tool[ and Bquick selection tool[ were utilized (Fig. 8). Using the Photoshop technique in this paper, more detailed segmented images of other selected structures, such as the specific cerebral gyrus or specific thalamic nucleus, could be produced effectively. It has proven to be difficult in the past to perform volume reconstructions without computer programming, but MRIcro (a free software package) enables investigators to perform volume reconstruction to make 3-D images. The 3-D images helped us verify the alignment of the SIs and the accuracy of segmentation (Fig. 9). Naturally, free software packages of this type are limited, and thus there is a need to devise software that can better produce real color 3-D images from SIs and segmented images. The 7.0-T MRIs, SIs, segmented images, and 3-D images of the cadaver head produced during this study are available free of charge for noncommercial applications. After obtaining permission from our group, the data will be provided to users worldwide either on- or offline. These high-resolution images will hopefully stimulate researchers to develop suitable medical simulation tools for the brain and contribute to improvement of neuroimaging studies in clinical practice. h

VOXEL-MAN framework,[ Comput. Med. Imag. Graph., vol. 24, pp. 127–132, 2000. [3] J. S. Park, M. S. Chung, S. B. Hwang, Y. S. Lee, D. H. Har, and H. S. Park, BVisible Korean Human: Improved serially sectioned images of the whole body,[ IEEE Trans. Med. Imag., vol. 24, pp. 352–360, 2005. [4] J. S. Park, M. S. Chung, S. B. Hwang, Y. S. Lee, and D. H. Har, BTechnical report on semiautomatic segmentation using the

Adobe Photoshop,[ J. Digit. Imaging, vol. 18, pp. 333–343, 2005. [5] V. M. Spitzer and A. L. Scherzinger, BVirtual anatomy: An anatomist’s playground,[ Clin. Anat., vol. 19, pp. 192–203, 2006. [6] S. X. Zhang, P. A. Heng, and Z. J. Liu, BChinese Visible Human project,[ Clin. Anat., vol. 19, pp. 204–215, 2006. [7] J. S. Park, M. S. Chung, S. B. Hwang, B. S. Shin, and H. S. Park, BVisible Korean

Vol. 97, No. 12, December 2009 | Proceedings of the IEEE Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.

1995

Park et al.: Sectioned Images of the Cadaver Head

Human: Its techniques and applications,[ Clin. Anat., vol. 19, pp. 216–224, 2006. [8] J. S. Park, Y. W. Jung, J. W. Lee, D. S. Shin, M. S. Chung, M. Riemer, and H. Handels, BGenerating useful images for medical applications from the Visible Korean Human,[ Comput. Methods Progr. Biomed., vol. 92, pp. 257–266, 2008. [9] Z. H. Cho, Y. D. Son, H. K. Kim, K. N. Kim, S. H. Oh, J. Y. Han, I. K. Hong, and

Y. B. Kim, BA fusion PET-MRI system with a high-resolution research tomograph-PET and ultra-high field 7.0 T-MRI for the molecular-genetic imaging of the brain,[ Proteomics, vol. 8, pp. 1302–1323, 2008. [10] J. S. Park, D. S. Shin, M. S. Chung, S. B. Hwang, and J. Chung, BTechnique of semiautomatic surface reconstruction of the Visible Korean Human data using

commercial software,[ Clin. Anat., vol. 20, pp. 871–879, 2007. [11] V. M. Spitzer and D. G. Whitlock, Atlas of the Visible Human Male: Reverse Engineering of the Human Body. Sudbury, MA: Jones and Bartlett, 1998.

ABOUT THE AUTHORS Jin Seo Park received the B.S. and M.S. degrees from Chungang University, South Korea, in 1998 and 2001, respectively, and the Ph.D. degree from Ajou University School of Medicine, South Korea, in 2003. He currently is an Assistant Professor in the Department of Anatomy, Dongguk University College of Medicine, South Korea. Since 2001, he has taught gross anatomy to medical and nursing students and studied gross anatomy. He is working on the Visible Korean project (MRIs, CTs, sectioned images, and segmented images of the entire and a region of body) and its applications.

Min Suk Chung, M.D., corresponding author, is Professor in Department of Anatomy, Ajou University School of Medicine, Suwon, South Korea. He received his B.S., M.S., and Ph.D. degrees from Yonsei University, Seoul, South Korea. For his master’s and doctoral theses, he studied clinical anatomy by dissecting cadavers. After Ph.D. acquisition (1996), he became interested in virtual dissection and performed preliminary experiments for Visible Korean. By the financial support of Korea Institute of Science and Technology Information, he launched Visible Korean (2000) to improve quantity and quality of the initial Visible Human Project.

Dong Sun Shin received his B.S. degree in computer science from Korea National Open University and then an M.S. degree in cognitive science from Sungkyunkwan University. Dr. Shin received his Ph.D. degree in gross anatomy from Ajou University School of Medicine. He is currently a postdoctoral fellow at the Graphics and Media Lab of Seoul National University. His interests include: 3-dimensional modeling of human structures from the Visible Korean and human animation for virtual fashion shows.

Dong-Hwan Har is Professor in Graduate School of Advanced Imaging Science, Multimedia and Film, Chungang University, Seoul, South Korea.

1996

Zang-Hee Cho received the B.S. and M.S. degrees from Seoul National University (South Korea) in 1960 and 1962, respectively and Ph.D. from Uppsala University (Sweden) in 1966. He has been faculty at the University of Stockholm and University of California-Los Angeles. In 1979, he moved to Columbia University as a Professor of Radiology (Physics) and served as a co-Director of Columbia University Imaging Center until 1984 before he joined UCI. Since 1985, he was the professor of Radiological Science and head of the MRI Center at University of California at Irvine until the current position, the University professor and Director of the Neuroscience Research Institute, Gachon University of Medicine and Science, Incheon, Korea. Professor Cho has been a pioneer in Positron Emission Tomography (PET) and Magnetic Resonance Imaging since the inception of the computerized tomography (CT) in 1972. He was the first one who pioneered mathematical algorithms related to CT scanners and subsequently developed world’s first BRing PET[ in 1975 and its nuclear detector BBismuth Germanate Oxide (BGO)[ in 1976, both of which revolutionized modern brain-imaging, and recently 7.0T MRI at NRI Gachon University where he performed current 7.0T MRI imaging. Young-Bo Kim received the B.S., M.S. and Ph.D. degrees from Hanyang University College of Medicine, Seoul, South Korea, in 1986, 1993 and 1996. He currently is a Professor in the Department of Neurosurgery and Neuroscience Research Institute, Gachon University Medicine and Science, Incheon, South Korea.

Jae-Yong Han received both the B.S. and M.S. degrees from Kyung-Hee University, South Korea in 1999 and 2001 respectively. Also, he received his Ph.D. degree in Biomedical engineering from the same university in 2005. He is currently a postdoctoral fellow at Neuroscience Research Institute, Gachon University of Medicine and Science. He is working in the field of Ultra High field MRI. His main interests are image acquisition and data processing. Je-Geun Chi, M.D. is a graduate of Seoul National University College of Medicine in 1962, and got his Ph.D. at Seoul National University Graduate School. He studied pediatric pathology and neuropathology at Boston Children’s Hospital and Beth Israel Hospital, Boston, USA, during 1970 to 1976. He came back to his alma mater as assistant professor of pathology in 1976, and worked there until his retirement in 2003. Currently he is a distinguished visiting scientist at Neuroscience Research Institute of Gachon University of Medicine and Science, Korea.

Proceedings of the IEEE | Vol. 97, No. 12, December 2009 Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on November 22, 2009 at 01:30 from IEEE Xplore. Restrictions apply.