PDA

View Full Version : Re: Limb Position Measurement



Watson W
02-03-1995, 08:34 AM
Warren,

The following ms. is in press. It pretty much

>am still working on them. I would greatly appreciate more information
>with respect to the basic design, cost, weight for the patient,
>power requirements and electronic and software interfacing of Polhemus
>and Ascension electromagnetic devives.
>Where would I find publications on this subject? Thanks for your
>

Let me know if you have more specific questions after giving this
a go through.

Best,
David
+---------------------------------------------------------------------+
| David Dean, Ph.D., Assistant Professor *********************** |
| Dept.s Anatomy, Orthodontics, & Biomed Eng. * FAX: (216) 368-3204 * |
| Bolton-Brush Study, 3080 Bolton Dental Bldg * alternate: 368-8669 * |
| Case Western Reserve University *********************** |
| 10900 Euclid Avenue E-mail: dean@lucifer.cwru.edu |
| Cleveland, OH 44106-4905 USA Voice: (216) 368-1975 |
+---------------------------------------------------------------------+

-----------------------------------------------------------------------
In Press (LA Marcus, ed.): Advances in Morphometrics. NYC:Plenum

3D Data Capture and Visualization
David Dean
Departments of Anatomy, Orthodontics, and Biomedical Engineering
Case Western Reserve University
10900 Euclid Avenue
Cleveland, OH 44106-4905 U.S.A.
E-mail: dean@lucifer.cwru.edu

"The next great surge in morphometrics will involve the superb
new technology of `high-speed workstations."
Bookstein (1991)

Overview
The hallmark of the New Morphometrics is the use of landmark
coordinate data. Current computer-based digitization,
visualization, and analysis applications in morphometrics are
heavily geared toward 2D (two dimensional) data sets. While some
organisms are flatter than others, all are three dimensional. We
are accustomed to conflating these 3D (three dimensional)
organisms to 2D despite the compromises involved. If coordinates
are collected from pictorial representations (e.g., photographs,
x-rays, frame-grabbers, etc.) it is important to attempt to
register each organism in the same position before image/data
capture. This is especially true for oblique views. The lateral
view of bilaterally symmetric organisms is a very popular means
to represent 3D organisms in 2D. This approach ignores minor, but
interesting, bilateral asymmetries - a popular topic of
investigation now. Problems of data capture object registration
or asymmetry, generally, do not affect the collection of data
sets in 3D.
Organisms live in the fourth dimension. It is the task of
morphometrics to adequately describe 3D organismal variation.
This variation is due to various 4D processes: ontogeny,
phylogeny, pathology, injury, or sexual differentiation. Most, if
not all, 2D morphometric algorithms are easily extended to 3D.
Until recently three dimensional data capture devices and data
visualization software have not been readily available. This
paper discusses the recent profusion of such devices and
software.
Is it always desirable to undertake three dimensional
morphometric analyses? There is no single answer to this
question. Each morphometrician must perform their own
cost/benefit analysis. As a lure Bookstein (1991) notes that in
contrast to two-dimensional biological form, three-dimensional
form may suggest a richness of local and regional geometric
descriptors that often matches our intuitive sense of how
biological forms are usefully compared. Thus, while many
measurements can be made on two dimensional views of or cuts
through an organism, only the three dimensional surfaces give
evidence of topology and texture. When possible these important
biological shape descriptors ought not be ignored.
The focus of this chapter are tools that assist in the
analysis of macroscopic biological surfaces and volumes, (e.g.,
bones, skin, organ systems) that are readily visualized. When not
visible from the surface often radiological techniques can be
applied to visualize structures in vivo. It should be mentioned
that there are a host of problems unique to remote three
dimensional visualization, especially those related to
thresholding and isosurface construction, that are not discussed
here (Fruhauf, 1991; Magnusson et al., 1991).
Data collection hardware and 3D image reconstruction
software (i.e., from slice data) for morphometrics needs to have
a high degree of precision, at least within the acceptable limits
of direct data capture means such as calipers. There are lots of
homemade devices (e.g., Corti et al., this volume) that we will
not discuss here that obtain this level of accuracy. All of these
devices are used to collect data directly from the surface of the
study object. When the data are collected indirectly, i.e. by
scanning, one must have a 3D visualization environment that
allows quick and precise collection of landmark, surface, and/or
volumetric data; of course, creating your own visualization
environment requires a great deal of expertise.
There are primarily two industrial applications driving the
development of both direct and remote 3D digitizing devices and
visualization environments: rapid prototyping and virtual reality
modeling. Tools developed for rapid prototyping and virtual
reality visualization have already been put to use in preparing
fossil reconstructions (Kalvin et al., 1992). Medical imaging
applications have also been the motivation for many new
developments in 3D visualization (Coatrieux et al., 1990).
The following is an annotated listing of 3D data capture and
visualization hardware and software. I have first hand experience
with many of the devices and applications discussed below.
Independent review of each device or package was requested, but
if not available, summaries were based on product literature:
caveat emptor. Sections I and II discuss hand operated data
capture devices and remote laser-light and stereomicroscopic
surface imaging, respectively. Section III covers a broad range
of visualization software packages that have morphometric
applicability. Section IV finishes up with a review of currently
available rapid prototyping devices also referred to as CAMs
(Computer-Aided-Manufacturing) or 3D printers.

Data Capture Devices I: Hand digitization

These devices are most useful for capturing landmark-point
and line-tracing coordinate data. However, as compared to the
scanning devices discussed in Section II, they are not as useful
for precise capture of surface patches. Radiological slice data
(e.g., CT and MR) or histological means are the best way to
gather information on the complete volume. Several of the hand-
held devices involve a source of electromagnetic radiation which
creates a field through which a hand-held stylus (sensor) moves.
The tip of the stylus is tracked as it is applied to an object of
interest. The location of significant surface features are
recorded to disk at the operator's instruction.
These devices are in high demand in human kinematic studies,
virtual reality applications, or industrial robotic applications,
activities where three dimensional positional tracking is a
requirement. Several of these devices are designed to be
lightweight and portable. Most offer devoted software interfaces
or can be easily adapted to control/data-recording software run
from a laptop computer. Several of these devices offer interfaces
to common CAD programs so that the data can be visualized as it
is collected. Many of these interfaces allow the user to change
the point of view, and zoom in and out of the data set. Most
often the stylus tip controls the point of view, but in some
interfaces the data lie within a tumbling cube, i.e. you chose
one of the six faces as a view.
Multiple viewing options are critical when collecting large
numbers of points and or line tracings. Visualization becomes
more and more difficult as the data come to overlie each other.
When collecting large amounts of data the visual check on
positional accuracy allows mistakes to be corrected immediately
rather than discovered at an inopportune time.
In most cases it is necessary to rigidly secure the object
being digitized. If more than one view of an object is to be
digitized (e.g. the part of the object that was anchored in the
initial digitizing session), it is very helpful to take multiple
landmarks redundantly between the two session. A Procrustes fit
can then be done between the landmarks that show no error (i.e.
throw away landmarks with noise) to match up all the landmark
data from each view of the object. Several of the various hand-
held digitizers that produce 3D coordinate data are reported
below. Unlike laser light or other scanning digitizers, hand-held
stylus devices have the option for dual use for 2D digitization
of images.

Electromagnetic Devices

1.1 DigiTracker ($3395)
Visual Circuits
3309 83rd Avenue North
Brooklyn Park, MN 55443
tel (612) 560-6205
fax (612) 659-6629

As of this writing the device has not yet been released. An OEM
version of the Ascension Flock of Birds, it is reported to be a
work-alike of the Polhemus 3DRAW. Like the Flock of Birds it's
accuracy rating is a magnitude lower, 0.1" RMS. It may have
greater utility for biometric applications than an unaltered
Flock of Birds device, which Visual Circuits retails for $2575,
because of its well-designed stylus. It is supplied with a
powerful digitizing software environment, BitBlitzer. A number of
mainstream Draw, Video, and CAD software export filters are
supplied with BitBlitzer.

1.2 Flock of Birds ($2695)
Ascension Technology Corp.
P.O. Box 527
Burlington, VT 05402
tel (802) 860-6440
fax (802) 860-6439

An early split-off from the Polhemus group. The same technology
has been adapted to tracking multiple electromagnetic 6D (xyz
coordinates and normals) sensors ("a flock") within the magnetic
field emitted by a single stationary source. Popular device for
biomechanical studies of joint movement. Translational accuracy
rated to 0.1" RMS. Above price is for single sensor device.


Figure 1. Polhemus 3DRAW: an magnetic source underneath the
digitizing platform is detected by coils in the stylus. The three
dimensional position of the stylus tip is reported through a
serial port to, most probably, a portable computer. An entire
system, including portable computer, can be assembled that weighs
less than 20Kg.

1.3 3DRAW ($6500)
Polhemus
One Hercules Drive, P.O. Box 560
Colchester, VT 05446
tel (800) 357-4777/(802) 655-3159
fax (801) 655-1439

This fourth generation device has been used by several
investigators to study non-human primate and human crania (see
Dean, this volume; Lele and Richtsmeier, 1992; Corner and
Richtsmeier, 1992; Richtsmeier and Walker, 1993; Vannier and
Conroy, 1989). Originally developed to be mounted on a fighter
pilot helmet to track head movements (Raab et al., 1979), this
application was soon obsoleted by laser-based sighting devices.
It consists of a platform under which a fixed 3 axis magnetic
source is emitted. A specimen can be fixed to the surface of this
platform. This object is digitized with the aid of a hand-held
stylus.
The stylus is equipped with a 3 axis magnetic sensor that
allows a small computer to determine the location of the tip in 3
space, including xyz normal information. Because it assumes the
magnetic field is constant, the work area should be kept
reasonably clear of ferrous objects. None should be within a
meter of the digitizing envelope, and other large iron objects
such as steam heated radiators and filing cabinets should be 2-3
meters away. Given a changing ferrous environment, the device
should be recalibrated (it comes with calibration software)
before each digitizing session. Unless the stylus is used very
carefully, the normal information is not very important as part
of the morphometric data collected. However, the viewpoint of the
data collection interface can be tied to the stylus. This allows
easy perusal of the data from various points of view, a crucial
error-checking step whenever potentially overlapping coordinates
are collected.
Their is firmware control of the stylus data report; it can
be set in point or continuous mode on the fly. Continuous mode is
very useful for tracing edge features (see Dean, this volume).
The accuracy of this device has been tested internally by
Polhemus using widely accepted methods (Krieg et al., 1992) and
has shown accuracy of 0.01" RMS (Root-Mean-Square), or 0.254 mm.
While this is more than a two-fold improvement over the original
(first generation) 3SPACE device, the generation of the device
that has commonly been used with large non-human primate and
human skull material, it is still below the acceptable range for
most dental studies.
The latest version of this device, the 3DRAW, is highly
portable. The 3DRAW table, stylus, and data handling device all
together weigh less than 12 pounds. There is a padded, hard-
shell, airline check-in carrying case available.
An interface was recently released that allows the data
collected by this device to be sent to the CADKEYS generic CAD
program. This flexible PC program can be readily tailored to most
biometric applications. It is included, gratis, with the device.
Another longstanding interface for Polhemus, Metrecom, Scientific
Accessories (see the latter two below) devices is marketed by
Mira Imaging (2257 South East, Suite 1A, Salt Lake City, UT
84106, tel. 801/466-4641, fax. 801/466-4699). This interface,
HyperspaceTM, is primarily aimed at producing a surface mesh of
solid objects. The NATO morphometrics workshop attendees tested
this software and found it difficult, but not impossible, to
adapt to morphometric applications such as single point or space
curve digitization.

Laser Light Stylus-based Digitizers


Figure 2. Laser light digitizer: The stylus emits a light that is
detected by three non-collinear sensors. Using their positional
offset the location of the stylus is determined by triangulation.
These expensive devices have high accuracy and low artifact noise
(figure supplied by Pixsys, Inc.).

1.4 Flashpoint Model 3000i ($14,900)
Pixsys, Inc.
5680-B Central Avenue
Boulder, CO 80301
tel (303) 447-0248
fax (303) 447-3905

A device like this or the much more expensive Optotrak (section
1.4) is likely to be optimal for many morphometric studies. These
devices do not have the echo or humidity artefacts of acoustic
digitizers or the artefacts due to changing magnetic fields
common with electromagnetic digitizers. These devices are,
technologically, the newest, thus their high cost.

Both devices consist of one or more LEDs (Light Emitting Diodes)
or fiber optic light emitters. These light sources are detected
by an array of 3 light sensors mounted in a bracket so that they
are fixed in a non-collinear position. This fixed offset is used
to triangulate the position of the light source. The light source
can be attached to a stylus. Data can be collected in point or
continuous mode. There are device drivers for all the common CAD
programs.

The rated accuracy of the current model is 0.5 mm. Over the next
few years I expect to see rapid increases in the accuracy of this
device coinciding with price reductions. The entire device
weights 22.4 pounds and should be quite portable.

1.5 Optotrak ($57,500)
Northern Digital, Inc.
403 Albert Street
Waterloo, ON N2L 3V2, Canada
tel (800) 265-2741/(519) 884-5142
fax (519) 884-5142

This device has basically similar components and works in the
same fashion as the Pixsys. Its accuracy is rated at 0.1 mm in
the x-y plane, 0.15 mm in z, when the stylus is within 2.25m of
the sensors. Weighing approximately 90 pounds, not including a
computer to record and visualize data, it is not easily
transported.

Acoustic Stylus-based Digitizer
1.6 GP-12 ES ($3200)
Science Accessories Corporation
2 Research Drive, P.O. Box 825
Shelton, CT 06484
tel (203) 925-1661
fax (203) 929-9636

The stylus in this device produces a clicking sound. This sound
is differentially detected by three wall-mounted acoustic
sensors. With the three separate readings of the click, the
location of the source can be triangulated. The stylus should be
at least as flexible as that of the 3DRAW or DigiTracker (I have
not tried to use it), making complex concavities more reachable
by this probe than, for example, those used in fixed-base, servo-
mechanism digitizing arms (see next section). However this
maneuverability is not without problems. Convexities may,
however, prevent sound from reaching the acoustic sensors without
distortion. High humidity is also known to effect the quality of
the data. The manufacturer's customer's claim an accuracy of
0.01", 0.254 mm., in a cubic meter digitizing envelope. There is
a real-time driver that works with a large number of CAD programs
including: AutoCAD, Cadkey, Mastercam. Scientific Accessories is
also willing to develop a customized interface.

Rigid and Servo-mechanism Arms
1.7 Cyclone ($52,000)
Renishaw, Inc.
623 Cooper Court
Schaumburg, IL 60173
tel (708) 843-3666
fax (708) 843-1744

This device comes right out of the rapid prototyping industry. It
includes a computer terminal and digitizing base. This unjointed
arm digitizer sits in place, retreating in the Z direction as
protrusions on the object are passed under its probe on a
linearly moving platform. Successive cross sections are collected
in this manner to build up an isosurface of the object. It moves
slowly and gently enough during its slow traverse that there
should be little chance for damage of specimens (stiff casts
would be at less risk than friable originals). However, complex
convexities are ignored. A similar device, the Retroscan
($20,000) uses a drill bit to mill a model of a digitized object
from solid plastic blocks.

1.8 Diagraph 202 ($1472)
Seritex Inc.
450 Barell Avenue
Carlstadt, NJ 07072
tel (201) 939-4606
fax (201) 939-3468

The Diagraph, designed by Swiss manufacturer GPM, is used to
directly digitize objects (e.g., a bone) held rigidly in place.
One end of its rigid arm points at surface features while the
opposite end of the arm puts a pen on paper. The paper sits on a
horizontal plane of tracing (or graph) paper set at a particular
z-height. The Diagraph is useful for the collection of serial
sections for 3D reconstruction.

1.9 Faro Arm ($14,400)
Faro
125 Technology Park
Lake Mary, FL 32746
tel (800) 736-6063/(407) 333-9911
fax (407) 333-4181

Morphometricians have attempted to use digitizing arms for at
least the past 30 years. It has not been a successful in most
craniofacial applications because the objects of study have been
too topologically complex. Both of these devices have four
joints, giving them a great deal of flexibility. Even so it might
be difficult capturing some points in the orbits, temporal fossa,
posterior choanae, the palette and parts of the basicranium.
Tracing lines on bony tori or elsewhere on the skull is likely to
range between difficult and impossible. The accuracy of the
Bronze Series 6 is 0.012 inches, 0.3 mm, within a 6 foot
spherical diameter.

Faro manufactures a more expensive line of digitizing arms, the
Surgicom series. These are used in human surgical planning,
execution, and follow-up. A complete neurosurgical package,
including visualization software, has recently been approved for
use by the FDA is marketed by ISG Technologies (6509 Airport
Road, Mississauga, ON L4V 1S7, Canada, tel. 905/672-2100, fax.
905/672-2307). The digitizing arm is primarily used as a virtual-
reality tool in designing a surgical plan, digitizing landmarks
to set up a reference frame on the patient that matches that in
the surgical plan, and as a position tracker for surgical
instruments. Advances in the use of this and other technologies
for frameless stereotactic surgery (Zinreich et al., 1993;
Sandeman et al., 1992) will likely be furthered by the
incorporation of newer morphometric techniques.


Figure 3. Romer Model 2000: This highly portable device provides
high accuracy well within the range required for dental studies.
Because of their limited mobility, digitizing arms may not be
useful for complex surfaces. The hard-shell airline check-in case
on the right holds a portable computer, printer, and the
digitizing arm.

1.10 Romer Model 2000 ($105,000)
Romer Supratech, Inc.
5145 Avenida Encinas
Carlsbad, CA 92008
tel (619) 438-7802
fax (619) 431-7940

This model is the smallest of a line of 4 arms with digitizing
envelopes of between 6.5' and 9.8' in diameter. These arms have
rotary joints making them more maneuverable than the Faro arm.
These devices are also designed to be portable. The Model 2000
weighs 17 pounds. The small base includes a table clamp and an
i/o processing unit with a serial port. It can thus be easily
adapted to use with a portable computer. The accuracy rating is
0.00125" RMS or 0.032 mm as tested by the NIST (National
Institute of Standards and Technology).

Data Capture Devices II: Remote Digitization
In this section we will not discuss multisurface, slice-based
approaches such as CT or MR-scanning. The two types of digitizers
discussed here, laser light and optical, are primarily used to
digitize one surface, the external surface. Optical systems
calibrate the X and Y values from the power of magnifcation and
the Z can be determined by the focus level.

Laser light digitization has a reasonable possibility of
providing data that can be used to construct highly accurate
surface representations of the original object, i.e., an
isosurface. Point resolution is a function of the width of the
light source beam. Scan accuracy is determined not only by beam
width, but also camera sensitivity, surface shininess, and the
system's ability to screen out artefactual reflections. In most
cases users will benefit from importing the data into a
visualization package and digitizing the required 3D landmarks
and/or lines there.

Multiple Point Source Laser Light Scanning

2.1 Digibot II ($49,000)
Digibotics
2800 Longhorn Boulevard, Suite 102
Austin, TX 78758
tel (512) 832-6544
fax (512) 832-1163

The Digibot II uses a four-axis laser light scanner which checks
sequential points on the surface. This allows some correction for
artifacts. The four axis array also obviates focusing problems
that other systems may have.

2.2 Hirez Model 3030 ($46,400)
Cyberware Laboratory, Inc.
2110 Del Monte Avenue
Monteray, CA 93940
tel (408) 675-1440
fax (408) 675-1494

This device consists of a moving source of laser light that is
detected by a video camera at a known angle of offset. Like all
of this first group of laser light digitizers, the Cyberware
scanners use multiple laser light sources. This allows the device
to cover more surface area in less time. An entire human face can
be scanned in less than 10 seconds. This is regularly done for a
number of different morphometric applications (Cutting et al.,
1988; Wohlers, 1992). However, because of the diffuse light
source artifacts, especially of complex surfaces, tend to corrupt
the data. The recommendation of scanning a surface from different
orientations and co-registering the scans can lead to compounded
artifacts. Assuming no artifacts, landmark accuracy is
approximately 0.02 inches, 0.5 mm.

One version of this device also captures color information from
the surface which can be used to render a more realistic
isosurface. One version of this device weighs less than 50
pounds.

2.3 Hyscan ($n.a.)
Hymarc Ltd. (Ottawa, Canada)
G.A. Davis Associates, Inc.
P.O. Box 36240
19959 Vernier Avenue, Suite 2B
Harper Woods, MI 48225
tel (313) 886-4101
fax (313) 886-1107

A multiple point laser device with very high accuracy, +/-
0.001", and relatively slow data acquisition on large objects.
The Hyscan accumulates data at the rate of 10,000 points per
second.

2.4 Optica ($50,000)
3D Technology, Inc.
12 Cambridge Drive
Trumbull, CT 06611-4764
tel (203) 371-8500
fax (203) 371-6300

Using a technology developed in England, but reminiscent of the
Cyberware scanner, this vendor claims 0.0008", 0.02mm accuracy
(Vinarub and Kapoor, 1992).

2.5 Rapid Profile Sensor ($40,000)
Laser Design, Inc.
9401 James Avenue South, Suite 162
Minneapolis, MN 55431
tel (612) 884-9648
fax (612) 884-9653

This device is scheduled to be available shortly. The developers
of this device claim digitizing speeds comparable with the
Cyberware scanner with greatly improved accuracy. A whole body
version of this scanner is scheduled to ship at $145,000.
Single Point Source Laser Light Scanning

2.6 CyberScan ($30,000)
CyberOptics Corp.
2505 Kennedy Street NE
Minneapolis, MN 55413
tel (612) 331-5702
fax (612) 331-3826

Sharnoa Corp.
45901 5 Mile Road
Plymouth, MI 48170
tel (313) 925-1661
fax (313) 454-7198

This single source laser sits at a pre-set "stand-off" distance
from the surface to be digitized. It has a high accuracy, 0.001"
(0.0254 mm.), but can fail on surfaces with convexities that
block the path of the scan head. These devices would take several
hours to scan an entire human face. Sharnoa uses the same scan
head and provides integration with CNC machine milling tools for
rapid prototyping.

2.7 Laser Triangulation Probe LTP 60 ($28,700 scan head, $200,000 complete devic
e)
Carl Zeiss, Inc.
One Zeiss Drive
Thomwood, NY 10594
tel (914) 747-1800/681-7849
fax (914) 681-7454

Perhaps the most accurate digitizing head on the market, the
accuracy of this device is rated at 2.0 microns, 0.0008". The
probe collects 100 coordinates per second when positioned 115 mm
away from the object of interest. An observation beam tracks the
surface to maintain this standoff distance. The device is
supplied with a visualization environment, Holos-UX, which allows
the user to define and locate coordinates of landmarks and
surface patches on the digitized object.

2.8 Lasar and TriCam ($50,000 and $11,000, respectively)
Perceptron, Inc.
23855 Research Drive
Farmington Hills, MI 48335
tel (810) 478-7710
fax (810) 478-7059

Single laser point devices that use radar focusing. The
digitizing envelope of the Lasar is commonly a cube of 2-3 meters
with an accuracy of 0.1 mm. The TriCam is more accurate in its
smaller digitizing envelope. The TriCam weighs a portable 20
pounds.

2.9 MessTechnik OMS 400/250 ($40,000 scan head, $193,000 complete device)
Wegu Inc.
Building II
Orchard Ridge Corporate Park
Fields Lane
Brewster, NY 10509
tel (914) 277-5753
fax (914) 277-5830

A close second to the Zeiss device, the accuracy of this device
is rated to 2.2 microns, 0.000098". The rate of data capture is
5000 points per second. The full device uses a video source and
tactile probe to insure it is following the surface carefully.
The standoff distance from the object is a function of lens
magnification, from 15 to 75 mm.

2.10 OP5 ($30,000)
Renishaw, Inc.
623 Cooper Court
Schaumburg, IL 60173
tel (708) 843-3666
fax (708) 843-1744

This laser scanning probe has a rated accuracy of 0.0001". The
probe rotates about a single axis up to 120 degrees, but must be
within 14mm of the scanned surface. It collects 200 points per
second.

2.11 Surveyor ($70,000)
Laser Design, Inc.
9401 James Avenue South, Suite 162
Minneapolis, MN 55431
tel (612) 884-9648
fax (612) 884-9653

The developers of this device claim digitizing speeds comparable
with the Cyberware scanner and accuracy improvements of at least
a magnitude, i.e., 0.05 mm. This device would be optimal for
studies requiring "dental" precision levels and rapid data
acquistion.

Moire
2.12 CadEyes ($80,000)
Medar Company
38700 Grand River
Farmington Hills, MI 48335
tel (810) 477-3900
fax (810) 477-8897

As with all moire photography, light is passed through a screen
grating. The resulting topographic pattern can be used for 3D
digitization. Medar claims 0.0001", 0.00254 mm, accuracy in the Z
dimension with this device.

2.13 Mini-Moire ($39,000)
Electro-Optical Information Systems
528 Euclid Street
Santa Monica, CA 90402
tel (310) 451-8566
fax (310) 393-2453

Electro-Optical Information Systems device works on the same
principle as the CadEyes scanner. It claims about the same
accuracy level, however it is much more portable.

Convergent Photogrammetry

2.14 Photomodeler ($895)
(Windows)
Eos Systems, Inc.
2040 West 12th Avenue
Vancouver, BC V6J 2G2 CANADA
tel (604) 732-6658
fax (604) 732-4716

This software uses standard photographs or 35mm slides to produce
a precise 3D model (AutoCAD 3D *.DXF file) of any object. The
documentation shows you how to calibrate your camera s focal
length, image aspect ratio, image position, and lens distortion
for this application. There are a few more planning few steps to
take so that all views will be complementary.

Macroscopic Stereomicroscopes
2.15 Reflex Microscope ($35,000)
Reflex Measurements Limited
Hadley House
Water Lane
Butleigh, Somerset BA6 8SP, England
tel/fax (44-458) 50332

This device has been used with great success on small biological
specimens (MacLarnon, 1989). An accuracy of 0.002 mm is claimed
for the 110 x 110 x 50 mm digitizing envelope.

2.16 Tool-Makers and Digital Contracers Microscopes ($35-45,000)
Mitotoyu
M-R-O Industrial Supplies, Inc.
440 South Main Street
Manville, NJ 08835
tel (201) 368-0525/(800) 345-4135
fax (908) 707-1266

The Tool-Makers device has a 100 x 100 x 50 mm digitizing
envelope with a claimed accuracy of .02 mm. The Digital
Contracers Microscope has a 100 x 100 x 25 mm digitizing envelope
with a claimed accuracy of .02 mm. These are standard optical
digitizing devices designed for the tool and die industry but
also readily adaptable to sub-macroscopic morphometrics,
especially on micro-organisms and teeth.

2.17 Video Comparator DVC110 ($8900)
Deltronic
3900 West Segerstrom Avenue
Santa Ana, CA 92704
tel (714) 545-0401
fax (714) 641-0946

More like the Reflex in having a potentially larger digitizing
envelope: 2 x 2 x 4 . The manufacturer rates its accuracy at
0.0002 , 0.05 mm; a level of accuracy acceptable for mammalian
dental studies.

III Visualization Software

NATO conference attendees tested the shareware visualization
environment created by Dennis Slice (this volume) DSDigit that
runs under MS-DOS. It provides a view of a data "cube" from
various sides and can be used to calculate distances between
landmarks. Jeffrey Walker (SUNY-Stony Brook) demonstrated his
Morphosys package for the Mac. It provides sophisticated
morphometric image analysis tools.

The visualization environments discussed below have been
developed for different primary audiences but all have some
application useful to morphometrics. Many of these packages offer
a 3D "pick" tool to collect landmark and line tracing coordinates
from isosurface models. Many of these packages provide image
enhancement tools such as isosurface shading and lighting options
to help the user better identify the morphology to be digitized.

Shareware

3.1 Khoros 2.0 (shareware)
(Multiple UNIX platforms, NeXT, PC-XWindows, Macintosh-XWindows)
E-mail: khoros-request@chama.eece.unm.edu.
Usenet: comp.soft-sys.khoros.
FTP: Khoros may be downloaded from the following FTP Locations
(read $KHOROS_FTP/release/install.ftp first):
USA ftp.eece.unm.edu 129.24.24.119 /pub/khoros
USA ftp.uu.net 192.48.96.9 /pub/window-sys/khoros
CANADA popeye.genie.uottawa.ca 137.122.20.3 /pub/khoros
GERMANY ftp.rrz.Uni-Koeln.DE 134.95.80.5 /graph/khoros
GERMANY ftp.lrz-muenchen.de 129.187.10.35 /local/khoros
ITALY ipifidpt.difi.unipi.it 131.114.8.130 /pub/khoros
JAPAN ftp.waseda.ac.jp 133.9.1.32 /pub/khoros
UK ftp.mcc.ac.uk 130.88.203.12 /pub/cgu/khoros
UK unix.hensa.ac.uk 129.12.21.7 /pub/uunet/window-sys/khoros

A powerful visualization environment written at the University of
New Mexico. There are several tools for adjusting isosurface
thresholds using image gradient information.

3.2 NIH Image 1.54 (shareware)
(Macintosh)
ftp site:
Source Site: zippy.nimh.mih.gov
Source Path: \pub\image
Source Name: nih-image-154-fpu.hqx
nih-image-154-nonfpu.hqx
nih-image-154-docs.hqx

A plug-in (Plug-in Digitizer), written by Cyrus Daboo of
Cambridge University, now allows NIH Image to grab images using
the frame grabber built into the AV Macs or from any other
Quicktime compatible digitizer with a 'vdig' component. NIH image
has tools for measuring areas, distances and angles, and for
counting things. The digitizer patch is available as a binhexed,
self-extracting archive by anonymous FTP from zippy.nimh.nih.gov,
in the /pub/nih-image/plug-ins directory. The archive also
includes a report on the AV digitizer that points out some its
flaws, such as automatic gain control that can't be disabled. It
also includes Apple's AV Digitizer Options extension that allows
the user to switch the AV digitizer into grayscale mode, which
greatly improves the quality of images captured using grayscale
cameras.

3.3 QSH (shareware)
(multiple Unix workstations)
Dr. Marilyn Noz
Department of Radiology
New York University School of Medicine
550 First Avenue
New York, NY 10016-6402

Radiological image handling software with a wide range of
previsualization tools, including: pixel editing, thresholding,
and in-slice measurements. After the object of interest has been
isolated, a simple image stack (*.qsh file) and header file
(*.qhd file) can be built.

3.4 SGI Explorer (gratis software with standard workstation purchase)
(IRIX [SGI UNIX])
Silicon Graphics Incorporated
2011 North Shoreline Boulevard
Mountain View, CA 94039
tel (415) 390-1960/(800) 800-4744
fax (415) 390-6153

Free visualization tool that comes with an SGI workstation.
Piggybacking on the power of SGI's GL (Graphics Language), it
provides several tools to display and explore 3D data.

3.5 Viewit (freeware)
(Multiple UNIX platforms, Macintosh)
National Center for Supercomputing Applications
Computing Applications Building
605 E. Springfield Avenue
Champaign, IL 61820
ftp://ftp.ncsa.uiuc.edu.

Viewit has a minimal user interface. It can be used to produce
volumetric images from slice data. It is known to require a lot
of RAM. There are also visualization tools. The Viewit electronic
newsletter is available from viewit@ncsa.uiuc.edu. Clint Potter
is the primary author and contact at: cpotter@ncsa.uiuc.edu.

Commercial Packages

3.6 Alice ($1495, $1046.50 academic price)
(Windows, Windows NT, Macintosh, PowerPC)
Hayden Image Processing Group
1919 14th Street, Suite 405
Boulder, CO 80302
tel (303) 449-3433
fax (303) 449-3772

Originally released under the nomen: DIPStation. A
medical imaging toolkit with help for segmentation,
volumetric image rendering, measurements from slices
and volumes, and data analysis. The segmentation
tools in Alice are stronger than many more expensive
packages.

3.7 Analyze 6.2 ($16000)
(Multiple UNIX platforms, Macintosh)
Mayo Medical Ventures
200 First St. SW
Rochester, MN 55905
tel (507) 284-8452
fax (507) 284-5410

Analyze offers a large set of slice data processing tools and 3D
image-building algorithms (Robb and Hanson, 1991). Analyze
provides import/export facility for a large number of
radiological, medical imaging, and histological data formats.
Linear measurements, landmarks, and line tracings may be captured
from the resulting isosurface models.

3.8 AVS ($6500)
(Multiple Unix platforms, Macintosh, Windows NT, Windows)
Advanced Visual Systems, Inc.
300 Fifth Avenue
Waltham, MA 02154
tel (617) 890-4300
fax (617) 890-8287

This visualization environment comes with a large number of
image-handling primitives that assist in 2D and 3D image
analysis. These primitives support several hardware-based
graphics languages such as SGI GL, Xlib, GDI, XGL, and PEX,
speeding redraw of the user s private interface.

3.9 Data Explorer ($5900)
(Multiple UNIX platforms)
IBM Data Explorer Group
Product Marketing, Department 590
8 Skyline Drive
Hawthorne, NY 10532
tel (914) 784-5089
fax (914) 784-5082

Image manipulation software with applications similar to
3DViewnix. Isosurface building algorithm Alligator (Kalvin, 1991)
which derives its flexibility from the "Winged Edge" data
structure (Kalvin et al., 1991). Recently powerful data reduction
software, "Superfaces", was added to this package. The package
also includes powerful data capture pick tool and image handling
primitives (Rogowitz and Treinish, 1993). These image handling
primitives greatly speed the creation of tools for image
analysis.

3.10 IDL ($3750, $1500 to educational institutions)
(Windows, Macintosh, Multiple UNIX platforms)
Research Systems Inc.
2995 Wilderness Place, Suite 203
Boulder, CO 80301
tel 303-786-9900
fax 303-786-9909

Interactive Data Language provides tools for image data
reduction, visualization, segmentation, measurement collection,
and analysis. Instead of canned image processing tools, there is
a high level, user-friendly, image processing language, IDL.

3.11 Image Volumes ($4500)
(IRIX)
Minnesota Datametrics Corp.
1000 Ingerson Road
St. Paul, MN 55126-8146
tel (612) 482-7938
fax (612) 490-9717

A package to assist collection of measurements from volumetric
images which includes large image processing toolkit. Provides
facile collection of linear distances, surface areas, and
volumes. Also includes native tools for superimposition by
landmarks, bounding boxes, or other transformations.

3.12 MacStereology ($750)
(Macintosh)
Ranfurly Microsystems
16 Derwent Avenue
Haddington, Oxford OX3 0AP England
FTP demo copy: ftp://zippy.nimh.nih.gov/pub/nih-image/programs

This program includes tools to make measurements directly from
images or to produce 3-D reconstructions from stacks of images.
MacStereology can also accept data from digitizing tablets or
*.PICT files. The user can set magnification. Measurement tools
include area, perimeter and center of gravity. Requires only 1
megabyte of RAM.

3.13 Mocha ($1495)
(Windows)
Jandel Scientific
2591 Kerner Boulevard
San Rafael, CA 94901
tel (800) 874-1888/(415) 453-6700
fax (415) 453-7769

Primarily aimed at a histological and microbiological audience,
this package provides a large number of general image processing
and data capture tools. Analysis proceeds primarily on x-y slices
of 3D subjects, but 3D reports can be generated.

3.14 Surfacer 3.1 ($4000)
(Multiple UNIX platforms, Windows)
Imageware, Inc.
313 North First Street
Ann Arbor, MI 48103
tel (313) 994-7300
fax (313) 994-7303

Surfacer imports data from a wide variety of standard formats. In
addition to data rendering and digitization it offers a large
complement of CAD reverse engineering tools.

3.15 3DViewnix ($1000)
(Multiple UNIX platforms)
Medical Imaging Processing Group
Radiology Associates
University of Pennsylvania
Brockley Hall, 4th Floor
428 Service Drive
Philadelphia, PA 19104
tel (215) 662-6780
fax (215) 898-9145

A large number of image manipulation and data capture tools
(Udupa et al., 1993), primarily for radiological slice data
(e.g., MR and CT) but adaptable to other 3D data sets.
Isosurfaces are built from topological information encoded by a
highly adaptable data structure known as the Shell (Udupa and
Odhner, 1993).

3.16 VoxBlast ($2500)
(Multiple UNIX platforms, Macintosh, PowerMac, Windows)
VayTek, Inc.
P.O. Box 732
305 West Lowe
Fairfield, IA 52556
tel (515) 472-2227
fax (515) 472-8131

VoxBlast produces isosurfaces from a wide variety of slice data.
There are tools to take measurements from these isosurfaces and
to calculate volumes.

3.17 VoxelView ($10,000, academic price $5000, $1295 Macintosh version )
(Irix and Macintosh)
Vital Images
505 N. Fourth St., Fairfield, Iowa 52556
tel (515) 472-7726
fax (515) 472-1661

Produces 3D isosurface models from a variety of data sources. One
of the few visualization environments (other than SGI Explorer)
that takes advantage of SGI"s IRIX GL (Graphics Language) and
thus the hardwired graphics engines found in SGI workstations.
This native hardware support speeds up image rendering. Reduced
redraw times mean faster digitization from multiple views as well
as if you are changing visualization parameters (e.g. threshold,
contrast, number of light sources, etc.).

IV Rapid Prototyping Devices
Generating an average landmark configuration, especially if the
landmark density is high enough to warrant an isosurface model,
is a major part of the New Morphometrics. Similarly the form of
an average form or morphotype is best viewed as a real model.
These three dimensional printers are still very expensive but
have dropped enough in price to make requests for single copy
output affordable. This copy can then be used as a positive in
mold production if copies are needed.

4.1 FDM 1000 ($50,000)
Stratasys, Inc.
14950 Martin Drive
Eden Prairie, MN 55344
tel (612) 937-3000
fax (612) 937-0070

You select a Z axis within your 3D data set. The data are then
divided into Z slices of 0.050" thickness. An extruder begins at
the center of each Z slice and spins concentric circles emitting
a 0.050 thick filament of plastic at just over the melting
temperature. The plastic left wherever there was data in the
original data set instantly cools. The extruder finishes that Z
slide and bumps up 0.050" to the next Z slice. If an overhang is
expected in the data a bridge is built that can be removed later.
The models produces are almost seamless. It can interface with
most of the aforementioned data visualization environments.

4.2 Helisys ($95,000)
Helisys, Inc.
Building M10
2750 Oregon Court
Torrance, CA 90503
tel (310) 782-1949
fax (310) 782-8280

The volumetric data is converted to Z slices along the preferred
Z axis. Each slice is cut into a sheet of paper. The sheets of
paper are then laminated together. The models have hard-wood
resilience and are virtually seamless. Unlike laser sintering
devices there is no warping in the plastic model already produced
due to heat in the part that is being formed. It can interface
with most of the aforementioned data visualization environments.

4.3 QuickCast ($195,000)
3D Systems
26081 Avenue Hall
Valencia, CA 91355
tel (805) 295-5600
fax (805) 257-1200

A standard sterolithographic device (Jacobs, 1993) which uses a
laser to etch the desired surfaces in an epoxy resin block. This
technique is referred to as laser sintering. Built in a series of
Z slices, slice thickness can be as low as 0.005". It can
interface with most of the aforementioned data visualization
environments.

4.4 SLS Sinterstation ($350,000)
DTM Corp.
1611 Headway Circle, Building 2
Austin, TX 78754
tel (512) 339-2922
fax (512) 339-0634

This is a standard laser sintering device. DTM claims to have
made more careful study of the materials used by other laser
sintering vendors, especially for biomedical applications. It has
been used to produce skeletal casts. It can interface with most
of the aforementioned data visualization environments.

4.5 Solider ($490,000)
Cubital America Inc.
1307F Allen Drive
Troy, MI 48083
tel (810) 585-7880
fax (810) 585-7884

A standard stereolithography device developed in Israel. It can
interface with most of the aforementioned data visualization
environments.

4.6 Z-mode ($800 per model + model costs)
3D Medical Imaging Services
1241 Garden Street
San Luis Obispo, CA 93401
tel (805) 542-9281
fax (805) 542-9168

Using proprietary volumetric rendering software and the Cubital
3D printing technology from ISG (see Metrecom Arm description
above). This company accepts a wide range of 3D CT and MR data
for the production of 3D models. These models are typically used
in surgical planning or prosthesis development. The Z plane
resolution of these models is claimed to be 0.002".

Summary

Most morphometric activities discussed at the NATO conference
derived their data from the external surfaces of organisms, the
leading 3D data capture technologies are laser light based.
Laser-based digitizers are currently the most expensive devices,
therefore advances in less expensive electromagnetic devices such
as the Polhemus 3DRAW may be the near term answer for landmark
and space curve digitization. The Laser Designs device looks to
be a breakthrough in combining speed and accuracy in a single
point source laser light digitizer. Three dimensional
visualization software is now readily available from a variety
sources. High quality three dimensional printers should start
making their way into the mainstream shortly.

Acknowledgment: Thanks to Leslie Marcus and Marco Corti for their
kind invitation to present this paper at the NATO ASI Workshop.
Thanks to Jim Rohlf, Alan Kalvin (IBM TJ Watson), Chip Maguire
(Columbia Univ.), and John Russ (Univ. North Carolina) for
careful comments on the manuscript. Much of the visualization
information was obtained from "med.volviz.faq.94-08.2/3" and
"med-volviz-faq-94-10" produced by Matti Haveri (mhaveri@cc.oulu.fi).

Literature Cited

Bookstein, FL (1991) Morphometric Tools for Landmark Data:
Geometry and Biology. Cambridge: Cambridge University Press.
Coatrieux JL, Barillot C. A survey of 3D display techniques to
render medical data. In: Hohne KH, Fuchs H, Pizer SM, eds.
3D imaging in medicine. Berlin: Springer-Verlag, 1990; 175-
195.)Corner BD, Lele S, and Richtsmeier JT (1992)
Measurement error of three-dimensional landmark data. Quant.
Anthropol. 3:347-359.
Corner B and Richtsmeier JT (1992) Cranial growth in the Squirrel
monkey (Saimuri sciureus): A quantitative analysis using
three dimensional coordinate data. Am. J. Phys. Anthropol.
87:67-81.
Corti et al. this volume??
Cutting CB, McCarthy JG, and Karron DB (1988) Three-dimensional
input of body surface data using a laser light scanner.
Annals of Plastic Surgery 21:38-45.
Fruhauf M (1991) Volume visualization on workstations: image
quality and efficiency of different techniques. Comput. &
Graphics 15:101-107.
Jacobs, P (1993) Stereolithography: From art to part. Cutting
Tool Engineering 45:1-3.
Kalvin AD (1991) Segmentation and surface-based modeling of
objects in three-dimensional biomedical images. Ph.D.
thesis:New York University.
Kalvin AD, Cutting CB, Haddad B, and Noz M (1991) Constructing
topologically connected surfaces for the comprehensive
analysis of 3D medical structures. SPIE 1445:247-258.
Kalvin AD, Dean D, Hublin J-J, and Braun M (1992) Visualization
in Anthropology: Reconstruction of Human Fossils from
Multiple Pieces. In (AE Kaufman and GM Nielson, eds.):
Proceedings of IEEE Visualization '92, pp. 404-410.
Krieg JC, Jones HR, Rodgers AG, and Schneider MR (1992) A 4
millisecond, low latency, 120 Hz, electromagnetic tracker
for virtual reality applications. Internal Application Note,
Polhemus.
Lele S and Richtsmeier JT (1992) On comparing biological shapes:
Detection of influential landmarks. Am. J. Phys. Anthropol.
87:49-65.
MacLarnon AM (1989) Applications of the Reflex instruments in
quantitative morphology. Folia Primatol. 53:33-49.
Magnusson M, Lenz R, and Danielsson PE (1991) Evaluation of
methods for shaded surface display of CT volumes. Comput.
Med. Imaging Graph. 15:247-256.
Raab FH, Blod EB, Steiner TO, and Jones HR (1979) Magnetic
position and orientation tracking system. IEEE Trans.
Aerospace and Elec. Sys. 15:709-718.
Richtsmeier JT and Walker A (1993) A morphometric study of facial
growth. In (A Walker and R Leakey, eds.): The Nariokotome
Homo erectus skeleton. Cambridge: Harvard University Press,
pp. 391-410.
Robb AR and Hanson DP (1991) A software system for interactive
and quantitative visualization of multidimensional
biomedical images. Austral. Phys. and Eng. Sci. in Med.
14:9-30.
Rogowitz BE and Treinish LA (1993) An architecture for rule-based
visualization. IEEE Visualization 93:236-234.
Sandeman DR, Patel N, Chandler C, Nelson RJ, Coakham HB, and
Griffith HB (1992) Advances in image directed neurosurgery:
Preliminary experience with the ISG Viewing Wand compared
with the Leksell G frame. Clin. MRI 2:91-92.
Udupa JK and Odhner D (1993) Shell rendering. IEEE Comp. Graphics
and Applic. 27:58-67.
Udupa JK, Goncalves KI, Narendula D, Odhner D, Samarasekera S,
and Sharma S (1993) 3DVIEWNIX: An open, transportable
software system for the visualization and analysis of
multidimensional, multimodality, multiparametric images.
SPIE 1897:47-58.
Vannier MW and Conroy GC (1989) Imaging workstations for
computer-aided primatology: Promises and pitfalls. Folia
Primatol. 53:7-21.
Vinarub EI and Kapoor N (1992) 3D digitizing hidden surface using
borescopic sensor technology. Proc. App. Machine Vision
Conf. 92: MS92 160:1-11.
Wohlers TT (1992) 3D Digitizers. Computer Graphics World 2:73-77.
Zinreich JS, Tebo SA, Long DM, Brem H, Mattox DE, Loury ME,
Vander Kolk CA, Koch WM, Kennedy DW, and Bryan RN (1993)
Frameless stereotaxic integration of CT imaging data:
Accuracy and initial applications. Radiology 188:735-742.
-----------------------------------------------------------------------