| Keywords
Visible Human, anatomy, physiology, simulation,
anatomical database
Introduction
Anatomy is the basis for understanding classical
medicine and in the past has provided, through cadaver dissection,
a view of the human body similar to that of "open"
surgical procedures. Surgical tools including scalpels, forceps
and hand-manipulated instruments were used for both training
with cadavers and treatment of the living. Diagnostic imaging
of the internal anatomy of the living became available more
than a century ago and required a three-dimensional understanding
of anatomy in order to analyze the "shadows" of
superimposed anatomical structures in a radiograph. Now, the
delivery of anatomical knowledge requires a new format to
match the wide variety of medical diagnostic imaging and therapeutic
procedures that have evolved to include image guided, robobtic,
remote, and minimally invasive procedures and surgeries. Anatomy
needs to be presented in a format similar to that used in
these procedures. Now, anatomy must often be understood from
the viewpoint of the observer traveling through the lumen
of a vessel or into a volume classically defined as a potential
space. It must be not only understood from the familiar viewpoint
of reflected visible light but also as the structure defining
an image synthesized from reflected ultrasound, emitted radiofrequency
electromagnetic radiation or transmitted x-rays or gamma rays.
The Visible Human provides computerized, photorealistic human
anatomy that can be manipulated to mimic the presentation
and the format of today's high tech medicine. The Visible
Human can now be observed in the classical presentation of
open surgical anatomy, viewed in cross-section through ANY
plane and presented from the internal perspective of the observer
(camera) inside the body.
The Visible Human can be manipulated to
provide a computer database that appears, feels, and sounds
like a human body. As knowledge, traditionally presented in
the physiology laboratory, is integrated with the structure
of the Visible Human, the database will become active and
reactive like a living person. As we extend the concept, the
database will be rendered dynamically in time as well as space
so that the student can witness and intervene in development
and aging and the progression (or regression) of disease.
The Visible Human Project originated in
1986 when the National library of Medicine (NLM) in its long
range planning saw the need to establish libraries of digital
images that would compliment their biographic and data services
(NLM Long Range Plan, 1986). It surmised that there would
be an increasing role for electronically represented images
in clinical medicine and biomedical research. A planning panel
convened by the NLM Board of Regents to study the matter in
1989 made the following recommendation: NLM should undertake
a first project building a digital image library of volumetric
data representing a complete adult male and female. This Visible
Human Project (VHP) will include digitized photographic images
from cryosectioning, digital images from computerized tomography
(CT) and digital magnetic resonance images (MRI) of cadavers.
The project was to be a cornerstone for future collections
of related image datasets and digital-image libraries (NLM
Board, 1990).
In August of 1991, the NLM contracted with
the University of Colorado Center for Human Simulation to
carry out Phase One of the VHP. This first step included the
acquisition of optimum cadaveric specimens. MR images (axial
of the head and neck and coronal images through the rest of
the body) were obtained at approximately 4 mm intervals throughout
the entire body of candidate male and female specimens. Radiographs
(14) and CT images (at 1 mm intervals) surveyed the whole
body specimens. A panel, convened by the NLM, reviewed data
from six specimens and selected the Visible Human Male and
Female. The specimens were then reduced transaxially, interval
by interval (1 mm intervals for the male and 0.33 mm intervals
for the female), and the remaining specimen was photographed.
Transverse photographic images were aligned one to another
and then registered, slice by slice, with the CT image dataset.
The digital dataset for the male was completed
and delivered to the NLM in the spring of 1994 and made available
over the Internet in November of that year. The dataset for
the VH female was finished in December of 1995 and is also
obtainable over the Internet from the NLM (Fig. 1).
Methods
The Visible Human Project was based on the
anatomy of intact adult human cadavers, free of infectious
or invasive disease or trauma. The dataset includes CT images
at one millimeter intervals, throughout the specimen, congruent
with the photographic cross sections. The dataset also includes
MRI and radiographs of the same cadaver.
The search for suitable cadavers proved
to be the most time consuming part of Phase One and took nearly
2.5 years to identify the male and female. Human bodies for
purposes of medical research and teaching are obtained in
most states through donations by their citizens to the State
Anatomical Board (SAB). These boards have been established
to consummate arrangements for donations and to oversee the
proper handling and appropriate distribution of such material.
The search for cadavers as "normal" as possible
was pursued by a consortium of SABs in Colorado, Texas and
Maryland through whose combined resources approximately 3000
gifts a year were received. The final choices for the VHP
were made from sample sets of three cadavers of each sex after
reviewing their medical records and analyzing survey radiographs,
CT scans and MR images.
Processing of potential specimens commenced
immediately upon their arrival at the University of Colorado
Health Sciences Center (UCHSC) since it had been determined
that optimum CT and MR images could best be obtained from
unfrozen material. MR images were obtained as soon as possible
since these images demonstrate the greatest change, post mortem,
of all the modalities. In order to obtain a fixed position
for the unfrozen cadaver that could be maintained from CT
through cryosectioning, each candidate specimen was positioned
in a specially constructed box that was fastened to the CT
gantry. When the CT scout views demonstrated that the cadaver
was correctly situated, the specimen was cocooned in foam
that rapidly hardened to stabilize it. After CT images were
obtained at 1 mm intervals through the length of the entire
body, the box and cadaver was removed from the scanner, frozen
and stored.
In order to cut the cadaver at intervals
of one millimeter or less, a special cryomacrotome was constructed
to section a block as large as 14 inches AP x 20.3 inches
high x 22 inches in width. To use this cryomacrotome, the
frozen embedded cadaver was segmented into blocks no larger
in height than 20.3 inches. For this purpose a backsaw was
constructed utilizing a very thin blade at high tension. The
specimen was frozen to the temperature of dry ice and segmented
with the saw into four blocks of approximately equal size
(Fig. 2). The precise location for the desired saw cut was
established by CT and its location was marked on the body
surface. After a block was isolated, the hardened foam surrounding
it was partially removed and the unit was prepared for embedding
in a 3% gelatin solution. An aluminum walled mold of sufficient
size to accommodate individual blocks was used as an embedding
receptacle. Adjustable pedestals situated in the floor of
the mold allowed the block to be mounted within the cavity
so the block's surface was parallel to the cryomacrotome's
cutting plane. After this alignment was achieved, the mold
was filled with the gelatin solution and frozen to -70 C.
Slicing the frozen gelatin-embedded specimen
commenced by fixing the mold's base plate to the table of
the cryomacrotome. The top surface of the block was removed
(1 mm for the VHM, 0.33 mm for the VHF) when the table on
which it was mounted translated beneath the spinning blade.
As the cryomacrotome table continued its traverse beyond the
blade, it transported the block, with its newly exposed surface,
into a photographic chamber. After image capture, the table
and block were moved back to their initial position and raised
by the desired interval in the z-axis preparatory to the next
cut. This entire process was initiated and remotely controlled
by a cryomacrotome operator.
In the photographic chamber, each freshly
cut surface of the block was flushed with compressed air and
examined for defects. Uncut fragments of tendons, fascia,
etc., were removed and exposed cavities filled with latex.
When the surface was judged to be suitable for photography,
it was sprayed with absolute alcohol and surrounded with a
black mask that also displayed the slice number and a calibrated
gray scale (Fig. 3).
Digital and photographic images of the prepared
surfaces were captured under computer control. The sequence
first recorded a digital image of the surface with a Leaf
digital camera back on a Hasselblad camera body using Leaf
software. This image was examined on the image acquisition
computer's monitor to confirm the surface was suitable for
film capture. If judged satisfactory it was compressed to
a 24 bit image (8 each for red, green and blue) and cropped
to anterio-posterior dimensions of the anatomy (1216 pixels).
These digital images later were aligned with the aid of vertically
situated fiduciary rods that were embedded with the cadaver
specimens. Original film images were also captured with 35
mm and 70 mm Rollieflex cameras successively positioned by
the computer over the masked block surface. Strobe lights
were used to illuminate the block's surface during the digital
and film imaging process. After the three camera sequence
was completed, the block surface was refrozen, prior to the
next cycle, by contact with an aluminum tray containing dry
ice.
Acquisition of the anatomical cross sectional
images at 1 mm intervals for the Visible Human Male required
one month per block (over a nine month period). The data set
delivered to the NLM included CT, MR and radiographic images
in addition to the photographic images. Axial MR images of
the head and neck and coronal sections through the rest of
the body were obtained at 256 pixel by 256 pixel resolution.
Each pixel has 12 bits of gray tone. CT data consisted of
axial scans through the entire body at one mm intervals. CT
images are 512 pixels by 512 pixels where each 12 bit pixel
value is related to electron density or Hounsfeld number.
The CT axial images have been aligned with the anatomical
cross sections and their image files numbered similarly.
The Visible Human Female data set has the
same characteristics as the male with one exception. The axial
anatomical images were obtained at 0.33 mm intervals. It took
twelve months (three months per block) to obtain the 5,189
cross-sectional photographs that comprise the female dataset.
RESULTS
The Visible Human Dataset of over 13,000 images (including
both male and female) is available on the Internet from the
National Library of Medicine. Access to the data requires
a license from the NLM (the license is also available via
the Net). To date, over 1,500 licenses have been granted to
individuals, companies and schools in over 30 countries for
use or development of the images. As a result, the Visible
Human Dataset is now the most widely circulated compilation
of digital human anatomy anywhere. The image database for
the two cadavers occupies 64 gigabytes of pixel-based information
making it the largest collection anatomical images ever assembled.
Many applications that incorporate the data
are now available and have been summarized at the biennial
Visible Human Project Conferences (1996, 1998 and 2000). The
first generation of products presented the images with navigation
interfaces (some examples include Female Visible Human CD,
1997; Complete Visible Human Male Laserdisc and CDs, 1995;
Digital Humans, 1996, Virtual Human CD, 1995; Visible Human
Explorer CD, 1997; Cross-sectional Anatomy Tutor, 1996; Coronal
Man Poster, 1995). The next generation of educational resources
incorporated labels and the latest generation includes fully
segmented and classified data (Segmented Visible Human, 1997;
Virtual Human 1997). This Journal, VHJOE, includes presentation
of the photographic data and the identification masks that
classify each pixel. In other applications, such as knee arthroscopy
simulation, the image volume database is being modified to
reflect the ligament and cartilage changes accompanying flexion
and extension of the knee.
Development of the database at the University
of Colorado Center for Human Simulation (http://www.visiblehuman.org)
includes the segmentation and classification of the entire
Visible Human Male Dataset (completed in 1997). Each of the
1,877 images in the VHM Dataset was filtered to enhance the
contrast of selected tissue interfaces. Edge detection processing
isolated the high contrast edges which were edited for each
tissue type. The result of this process was a mask image,
16 bits deep, for each of the 1,877 slices. 11 bits of the
16 bit mask image were utilized to assign one of the over
1,700 anatomical objects a unique number. A master database
of these unique mask numbers and associated anatomical objects
also associates each structure with a system and region of
the body and other properties of anatomical importance such
as function and innervations. For the classification scheme,
anatomical names generally conform to standard texts such
as Gray? Anatomy. These masks can now be utilized as identifiers
for each voxel in the anatomical images and one use for these
identifiers is for selection of the voxels to render as a
three-dimensional representation of a desired body part, region
or system (Fig 4).
We have also utilized Lorensen's arching
cubes algorithms (Lorensen and Cline, 1987) to reduce the
outermost voxels that define each anatomical entity to polygon
representations of the surfaces of those structures. These
polygon representations can be three-dimensionally rendered
on specialized graphical engines in real-time. We have pushed
the envelope of utility for these polygons by texture mapping
the color intensities from the voxels most closely associated
to these polygons. Utilizing this technique preserves the
visual reality of the tissues in the original cut slices,
including the appearance of any structures that are deep to
the surface. All this reality would be lost with the typical
mapping of a fabricated clay-like texture to the anatomical
surfaces or even the mapping of a realistic texture from a
different subject.
For some regions of the body, e.g., the
posterior abdomen and the knee we have begun the process of
assigning mechanical properties to the identified tissues.
We can visualize as three dimensional objects, each voxel
(or polygon) having their mechanical properties defined and
also interact with these objects through a haptic interface.
We have utilized a haptic feedback to feel the tissues (voxels
or polygons) of the Visible Human (Reinig, 1996; Reinig et.
al 1996). In the case of a scalpel - this feeling of cutting
through the tissues of the Visible Human is provided by reactive
forces applied to the scalpel as it encounters different tissues
(voxels or polygons) in its path (Coin et. al., 1996; Drain
et. al., 1993). The user feels the drag of the scalpel just
as though they were cutting through real tissue. Of course,
the tissue of the Visible Human must now react as though it
were cut by the scalpel so there are new surfaces where the
scalpel cut was made. These techniques are the beginnings
of the generalized surgical simulator to be available in the
ideal anatomy teaching laboratory (Muller et. al., 1997).
At the Center for Human Simulation we have
built prototype simulators for medical procedures that utilize
this type of serial section data for dynamic, user controllable,
3D presentation and also allow the user to feel the anatomy
through the instruments utilized in clinical practice. Each
procedure was chosen because of its dependence on a thorough
understanding of complex 3D anatomy or because feeling is
a major source of information defining the progress of the
procedure. Our prototype simulators are based on real time
visualization and haptic feedback. One of our first simulators
was a partial task simulator that provides a feel for surgical
incisions. The student, holding a scalpel handle can feel
each tissue interface as a cut is made through the Visible
Human male. A second simulator uses the haptic interface to
produce the force one encounters with a needle procedure.
The student, performing a celiac plexus block and holding
the injection needle, can feel each tissue the needle encounters
in its path from the skin surface of the T12-L1 area all the
way to the neighborhood of the abdominal aorta at the celiac
plexus. As a simulator, the goal of the procedure is well
defined but it does not prevent the user from directing the
needle into a position far from the desired target. In the
event of a misdirect, the student can be brought back to the
proper path or allowed to error along a path that e.g., might
be more appropriate for a renal biopsy. This simulator also
incorporates clinically available cues such as fluoroscopy
and aortic pulsations to confirm the position of the needle
(Fig. 5). These simulators of living human anatomy will be
coupled with other simulators of both anatomy and physiology
to provide a complete operating room environment for team
teaching all the players involved in a given OR procedure.
The future of these theaters will mimic the reality of the
cockpit used for flight instruction in modern flight training
schools.
Summary
Use of the Visible Human Dataset to visualize
or simulate the human body requires that each distinct tissue
be identified and assigned the properties that characterize
it. This massive task has been started with the identification
of every voxel (volume element) in the male image database
and in some regions, mechanical properties have also been
defined (e.g., the speed of sound or the mechanical resistance
to a needle or scalpel). Clinical identification of the internal
and external surfaces of the body provides the traditional
views of classical "surface anatomy? and now video windows
to structures deep inside the body. Patient imaging has changed
dramatically with recent technological innovations such as
miniature video cameras, fiber optic image transmission and
even untethered "pill" cameras. Topographically,
external surfaces of the body that can now be visualized include
surfaces of the GI, GU and respiratory systems. With bronchoscopes,
laryngoscopes, colonoscopes, sigmoidoscopes, etc., we are
now able to remotely view external body surfaces such as the
stomach lining, the carina or the cardiac sphincter - in the
living patient. With minimal invasion, endoscopes, arthroscopes
and vascular instrumentation open up another expanse of internal
surface anatomy of the living human. The Visible Human affords
us the opportunity to investigate and interpret views of these
surfaces inside and outside the body with the additional information
(labeled voxels) and views (cross-sections, 3D renderings,
simulated images etc.) not available from a patient.
Cadaver dissection contributes significantly
to the understanding of anatomical structure and function
but has changed very little during the last century. Cadaver
anatomy (embalmed), however, bears little resemblance in either
color or feel to the living body and access to cadaver dissection
has become more restricted due to competing curriculum demands
and increased environmental protection. The use of a human
simulator with haptic feedback may overcome these deficiencies
and actually enhance the experience obtained from dissection.
Computer simulation, made possible by the introduction of
virtual reality computer technology coupled with a computerized
anatomical database such as the Visible Human, affords opportunities
for developing a human simulator to transform anatomical education
as dramatically as flight simulation has transformed pilot
training. A virtual cadaver can exhibit the look and feel
of a living patient, a difficult charge for the real cadaver.
The understanding of human anatomy that
radiological applications provide is dramatized by the fact
that in February of 1896, just three months after Konrad Roentgen
announced the discovery of x-rays, the first American clinical
x-ray image was produced in a physics laboratory at Hanover,
New Hampshire (Spiegel, 1995). This rapid clinical application
of his discovery for the visualization of internal human anatomy
by a noninvasive method foreshadowed the massive amount of
information on internal structures that was to come. With
the appearance of the CT in the early 1970s and MRI in the
eighties, anatomical structures in every part of the body
were being revealed in astonishing detail. The latter instrument
also could display cross sectional anatomy in any plane through
the body. The massive accumulation of information on human
morphology that has resulted from the use of these powerful
clinical imaging modalities has led to the possibility of
visualizing the entire human living body in three dimensions.
The training required to interpret these images and practice
acquiring them can be provided by the Visible Human without
the hazards of ionizing radiation.
Although cross sections have been utilized
for centuries to teach anatomy, Computed Tomography (CT),
from the 1960's, provided the first major clinical need to
teach cross-sectional anatomy. Anatomists were able to meet
the challenge with classical cadaver techniques, sectioning
bodies in the same plane (transaxial) as the diagnostic imaging
modality. Direct correlation of the physical cross-sections
and the CT images provided an understanding of the unknown
by comparison to the familiar. Reconstruction of sagittal
or coronal cross-sections from multiple CT slices became available
but was not prolific. Three-dimensional renderings from multiple
CT serial sections also have been utilized in limited applications.
Recent technological advances in CT such as fast scanning
(helical scanning and multiple simultaneous slices) and auto-segmentation
provide more volumetric data and associated reconstructions
for applications such as virtual colonoscopy. The Visible
Human Dataset includes CT images at 1 mm intervals and is
registered to the photographic cross-sections and their associated
identification masks. This data has been utilized for simulating
radiographs and CT scans of the Visible Human. As the configuration
of the Visible Human becomes more user-controllable these
simulated images will provide training for radiographic technique
and positioning.
Magnetic Resonance Imaging (MRI) presented
a problem for the anatomist wanting to produce relevant teaching
materials. MRI is able to produce cross-sections at any angle
to the long axis of the body and in fact may be utilized to
directly collect volumetric image data. Early anatomical atlases
included standard transverse, coronal, sagittal, and anecdotal
oblique planes but they were from different bodies and consequently
correlation of the anatomy was difficult. With segmentation
and graphical rendering these MRI data can take on familiar
3D surface representations of anatomical entities but, again,
correlation to photographic anatomy can not be done on the
same specimen. The Visible Human provides a photographic,
anatomical database that can be manipulated the same way as
the MRI data. It was sectioned transaxially but can now be
sliced in coronal or sagittal planes or at user defined oblique
angles. This anatomy (and future data at higher resolution)
provides the basis for understanding the MRI that presents
structures invisible in previous generations of the imaging
technology.
Ultrasound generates cross sectional images
based on the acoustic impedance of the imaged tissue. Like
MRI, the ultrasound images are generated as cross-sections
at ANY user-defined angle. Ultrasound, however, provides the
added complication that the images are produced at 30 frames
per second. Images are generated from nonionizing radiation
and, therefore, can be used in vulnerable populations such
as pregnant women, fetuses and children. The modality is enjoying
increased usage in remote applications much the same as the
video-oscopies mentioned earlier. In these applications, the
ultrasound transducer is threaded into lumens, cavities, and
vessels in order to place it closer to an area of interest.
We have recently built a web-based application that allows
a user to interactively define a plane through the photographic
data of the Visible Human similar to a plane of interest imaged
with ultrasound. An oblique slice through the labeled, photographic
data of the Visible Human is extracted and presented to the
user. This 3D labeled atlas of human anatomy can be used for
any cross-sectional correlation and will drive a need for
extending the atlas to include higher resolution and structural
variation.
All of these anatomical teaching or imaging
methods will some day be incorporated into a human simulator.
We will be able to view the surface anatomy of a cyber-human,
image the selected virtual subject with any radiological modality
and dissect any part of the same body. Furthermore, all of
the techniques we used or images we saw or tissues we felt
could be recorded and analyzed for constructive criticism
or examination. In other words, the instructor can see or
experience exactly what the student saw or experienced in
their self-directed laboratory. Perturbations caused by the
study are reversible, a luxury of the virtual world that is
unavailable in the real world.
Computer simulation when based on photographic
images such as those of the Visible Human, can duplicate not
only surface anatomy but also internal anatomy revealed only
by radiology or dissection in the real world. Most importantly,
it does not inconvenience or compromise patient care. Many
advances required for these simulators are anticipated to
come from other fields. e.g., computer technology utilized
in Hollywood for special effects (deformation, 3D rendering
and animation) and movie colorization and rotoscoping all
contribute to the many tools required for simulating the scientifically
correct human body. The mathematics and computational speed
required for morphing a man to a car may someday provide the
ability to atrophy a muscle or inflame a pancreas. Rendering
tools provide the ability for the student to visualize objects
in three dimensions, from their own perspective. Animation
tools may provide physiological and kinematic extensions of
this database and colorization and rotoscoping technology
has already contributed to the process of feature extraction
from photographic anatomical images. Theme park entertainment
and video arcades will most likely drive the haptic interface
industry to make devices that give realistic feeling at a
substantially reduced cost. Of course, the aircraft industry
can contribute the maturity of a well-developed, studied,
and accepted utilization of computer simulations for the training
of highly skilled professionals.
Some major demonstrations of the concepts
already available for computer simulation of medical procedures
have convinced us that the human body can be duplicated well
enough to capture the student imagination in situations that
simulate human anatomy and limited physiology. The fabrication
is close enough to the clinical situation that the student
can practice and learn effectively with such a simulator.
These procedure simulators have provided responses judged
realistic by experts in their fields of use and formal evaluation
of these simulators is in progress. The extension of the concept
to other bodies, including a young female, fetal an pediatric
specimens is also ongoing at the CU Center for Human Simulation.
With the maturation of computer technologies,
including 3D visualization, and haptic and audio interfaces,
combined with volumetrically defined anatomy such as the Visible
Humans we are now able to learn in immersive environments
simulating the reality of the living, human body. Computer
simulations will support dynamic systems and are rapidly growing
in sophistication and abilities. The photographic cadaver
images such as those comprising the Visible Human anatomical
volume will become more sophisticated and more lifelike as
the image data is personified with characteristics and properties
that bring it ?to life? in a virtual reality laboratory and
afford the user an opportunity to study and interact with
what seems to be the living tissues of a human being - all
this, without the hindrance of inconveniencing or jeopardizing
living patients or volunteers.
CONCLUSIONS
The Visible Human Dataset is a demonstration that a whole
body can be prepared in such a way that it can be volumetrically
reconstructed in cyberspace for visualization and simulation.
The data are being utilized all over the world as a resource
for human anatomy visualization, modeling, simulation, training,
morphometrics and entertainment. At the CU Center for Human
Simulation segmentation and classification of the VH male
has been completed and applications providing user defined
oblique slices and a dissectible body have been developed.
Simulations of ultrasound images, radiographs and CT images
from the Visible Human database are in progress. Procedural
simulators relating to arthroscopy, arthrocentesis, and other
needle-based procedures such as anesthetic blocks are also
being developed. Changes in the volumetric image data reflecting
variations in body position are occurring. The extension of
this database to include anatomical variation and pathology
will enhance the utility of this cyberdata for education and
training of health care providers. The impact of computer
simulation of the human body on the understanding of structure
and function through visualization and procedure simulation
are being formally evaluated. In summary, the combination
of Visible Human data and the technologies contributing to
immersive computer simulations will transform medical education
and evaluation in the near future and the gastrointestinal
community is positioned to lead this educational metamorphosis.
ACKNOWLEDGMENTS
The author wishes to thank Dr. Donald A.B. Lindberg, Director
of the NLM and Drs. Michael J. Ackerman and Donald P. Jenkins,
Program Project Officers, for their support. I also wish to
thank the team of young investigators who make this project
successful and especially acknowledge those citizens who donate
their remains to medical research and teaching.
References
1. Center for Human Simulation, U. of
Colorado http://www.uchsc.edu/sm/chs
2. Cotin, S., H. Delingette, M. Bro-Nielsen, N. Ayache, J.M.
Clément, V. Tassetti, and J. Marescaux 1996 S.J. Weghorst,
H.B. Sieburg, and K.S. Morgan, eds., IOS Press and Ohmsha,
Washington DC., pp 139-151 Geometric and Physical Representations
for a Simulator of Haptic Surgery. In: Health Care in the
Informatics Age. IOS Press and Ohmsha, Washington
DC., pp 139-151.
3. National Library of Medicine (U.S.) Board of Regents.
Electronic imaging: Report of the Board of Regents (1990),
U.S. Department of Health and Human Services, Public Health
Service, National Institutes of Health. NIH Publication
90-2197. National Institutes of Health. NIH Publication
90-2197.
4. Reinig, K.D., C.G. Rush, H.L. Pelster, V.M. Spitzer, and
J.A. Heath 1996 Real-Time Visually and Haptically Accurate
Surgical Simulation. In: Health Care in the Informatics
Age. S.J. Weghorst, H.B. Sieburg, and K.S. Morgan, eds.,
IOS Press and Ohmsha, Washington DC., pp 542-545.
|
