A Computerized System to
Manufacture Prostheses for Amputees in Developing Countries
Nicolas E. Walsh, M.D.
Jack L. Lancaster, Ph.D.
Virgil W. Faulkner, C.P.O.
William E. Rogers, M.S.
Introduction
The purpose of this project is to design and develop a computer aided design/computer aided manufacture (CAD/CAM) system for a lightweight, all plastic prosthesis
capable of being fabricated remotely from the site of evaluation of the amputee. Essential to the design of this prosthesis are characteristics which will allow final fitting
and modification to be accomplished by technicians in underdeveloped countries who may not have all of the equipment normally located in a modern North American
prosthetic facility.
This ongoing, multistage endeavor involves: (1) the development of a non-contact imaging device capable of capturing a three-dimensional topographical record
of the amputees' residual limb, (2) the development of a computer software system that would allow the prosthetist to design a prosthetic socket that is biomechanically
correct for production by a numerically computer controlled milling machine, (3) the development of production molding tools for the manufacture of lightweight
prosthetic plastic components, and (4) field testing the new lightweight all plastic system.
Statement of the Problem
In the United States, approximately 60,000 lower extremity amputations are
performed every year.1 In addition, there are approximately 400,000 living amputees.2 Most of these people are candidates for ambulation and will be fitted with
prostheses.
While most people in the Western world have access to modern prosthetic rehabilitation, many who become amputees in other countries do not. As reported by Jacobs
and Murdoch,3 traditional approaches toward rehabilitation from the industrial world "have failed." At the World Congress of the International Society of Prosthetics
and Orthotics, Dr. Emer Helander of the World Health Organization stated, "There is a need to train 70,000 fitters [prosthetists] if the rest of the world is to have access
to the same prosthetic services that exist in the Western world."3 In order to meet this need, multiple investigators have encouraged research into computer-aided design
and computer-aided manufacture (CAD/CAM) as a new technology in the design and fabrication of prostheses for amputees worldwide.3,4-8 To enhance the impact
of CAD/CAM on the prosthetic industry, it is necessary to develop new plastic technology that can be used by developing as well as industrialized nations.
The rationale for providing a prosthesis based on thermoplastic components was presented in the 1982 Annual Report of the Department of Mechanical Engineering,
University College London (UCL).9 The use of thermoplastic components in modular
prostheses enables the inherent characteristics of plastics (lightweight, low cost, corrosion resistance, water resistance, and low density) to be used to the advantage of
the amputee and the prosthetist. Thermoplastic materials can meet the requirements of rapid, easy assembly and ease of adjustment.
The Rehabilitation Engineering Laboratory (REL), located at the University of Texas Health Science Center at San Antonio, is committed to becoming a center
dedicated to the development of a low cost, all plastic prosthesis that can be designed, manufactured, and fitted by field workers anywhere in the world.
Regardless of the setting in which a traumatic or surgical amputation occurs, a prosthesis is required to restore the amputee to a normal life. The prosthesis allows
the amputee to stand, run, walk, and climb by providing a mechanical extension to the residual limb. The prosthesis consists of an artificial foot, shaft, hand crafted socket
into which the residual limb is inserted, and a suspension mechanism.
Because each amputee's residual limb is different, the current practice is to custom make each prosthesis. This is a time consuming and expensive artisan skill. The
most important aspect of the lower extremity prosthesis is socket design;10 the socket is the interface between the human leg and the mechanical support system.
Ultimately, the design and fit of the socket is what determines the patient's acceptance, comfort, suspension, and energy expenditure.3,11 All of these factors in unison
determine the utility and acceptance of the prosthesis.
The patellar-tendon-bearing (PTB) prosthesis is especially suited for ambulation of below-knee amputees.12 Fitting procedures for this socket are non-repeatable and
extremely labor intensive, requiring a highly skilled practitioner. The basic methodology for making PTB prostheses has not changed for 25 years: a better method must
be found if prosthetic rehabilitation is to be extended to amputees worldwide. The use of new methods of imaging combined with CAD/ CAM technology may open the
way to a new and better solution to this problem. An automated prosthetic fabrication system utilizing multiple remotely located shape sensing devices and a central
fabrication facility may
provide for rapid, simplified prosthetic design and fabrication. Geographically dispersed, less highly trained field workers could make final adjustments and perform
maintenance on the prosthesis made at the central fabrication center.
Work to Date
Imaging
The greatest barrier to the implementation of CAD/CAM in the field of prosthetics is the inability to capture the exact topographical shape of the amputee's residual
limb. For a CAD/CAM system to work effectively, the residual limb topography must be accurate, otherwise the computer axiom of "garbage in/garbage out" will apply.
The most important aspect of a lower extremity prosthesis is the socket design: the socket is the interface between the human leg and the mechanical support system.
The socket is often the component of the prosthesis necessitating replacement when a patient requires a new limb. The average amputee will require three to five
prostheses during the first five years following amputation due to the changes in the residual limb.13 Thus, being able to quickly and inexpensively make a new socket
that fits well and is comfortable is especially important.
Lack of topographical description and quantification of the changes in an amputee's residual limb is a major impediment to improvement in socket design. None of
the measurement techniques reported to date have provided a satisfactory solution for immediate or long-term evaluation of the surface topography or internal bone
configuration of the below-knee residual limb.13 Without reproducible sequential measurements, it has not been possible to scientifically predict typical changes that occur
in the residual limb, thus deterring the development of an improved socket design which may have better and longer optimal fit.3,8
Investigators have utilized multiple techniques to obtain an accurate mathematical description of selected body regions. 14-19 Methods such as computer assisted
tomography and magnetic resonance imaging provide adequate data but these techniques are neither cost effective nor readily available throughout the world. Other
methods are
often inaccurate or less adaptable for use in the region of the knee.
A laser scanning device in our laboratory was specifically designed to accommodate the variabilities of contours and geometrical complexity of the below-knee residual limb.15 This particular laser device was designed to require
little maintenance and be easily transportable. Minimal skills are required to operate the scanning device. It requires only a minor calibration device that can be adapted to operate in underdeveloped countries.
Capabilities required for laser scanning include: (1) the ability to scan from the tip of the residual limb to above the knee, with profiles adequate to describe the critical topographical features essential for prosthetic design, (2) the
ability to scan the residual limb in less than ten seconds, (3) the ability to provide accurate and reproducible measurements that include utilization of anatomical markers for repeat evaluations and orientation references, and (4) ease
of operation and accommodation to the patient.
The operation of the REL laser shape sensing device is based on oblique observations of a laser line projected onto the surface of the residual limb. The laser beam is directed as a vertical line which establishes a reference plane
along a radial axis. The intersection of the projected laser line with an opaque object (such as a residual limb) forms a profile of the surface of the object. The video scan coordinates of a profile point are used to determine the spatial
coordinates of a corresponding point on the surface profile because of the positional encoding afforded by oblique orientation of the video camera.
Surface geometry of the residual limb dictates that camera viewing angles between 40° and 60° be used for most body surfaces to minimize obstruction by adjacent surface structures while maintaining a reasonable degree of
positional encoding.
For three-dimensional data acquisition, multiple profiles are acquired around the residual limb with the laser and camera mounted on a turntable and the laser beam project-
ed through the axis-of-rotation. The angular coordinates of profile are determined by the turntable position. Acquiring data while the turntable is rotated about a centrally mounted amputee produces a set of profiles describing the surface
of the residual limb. A group of profiles acquired with sufficiently small angular increments provide an accurate 3-D representation of the surface of an object (Figure 1)
.
The video laser line detection (VLDD) hardware is the key component of the laser scanning system and was designed and constructed to perform five major tasks: (1) track the turntable position, (2) generate and control horizontal
video cursor, (3) detect, organize, and store laser profile data,
(4) buffer profile data, and (5) generate hardware level interrupt when profile data acquisition is complete.
The VLDD provides a horizontal cursor which is used together with the projected laser line to acquire and store surface anatomical reference marks. Anatomical landmarks are visually selected and denoted by the intersection of
the laser line and the horizontal cursor superimposed on the surface point of interest. By recording three distinct anatomical marks, data from different scans of the same object can be easily oriented to each other for comparison.
Anatomical
marking is useful when performing repeat scans on the same individual for long term comparative studies.
A scan is performed by selecting the starting angle, ending angle, and angular increment. The operator is informed of the number of profiles which will be acquired and prompted for a file name for data storage. When a file name
is entered, the computer positions the turntable 10° prior to the starting angle and waits for the operator to initiate data acquisition or abort and restart. Scans of 360° are currently being evaluated since the entire surface cannot be
scanned with less than 360°. When acquisition is complete, the raw data profiles are automatically stored on disk for later use and another scan can be initiated immediately, if desired. A raw data file of 44K bytes is used to store
the data acquired with 85 uniformly spaced profiles separated by approximately 4.2°
During the scan set-up, the residual limb is positioned such that its lowest point is near the axis of rotation of the turntable (Figure 2)
. This is done by viewing the laser profile at turntable positions of 0° and 90°, and by positioning
the residual limb so that the laser strikes the lowest point from both views. This positioning provides more symmetric sampling of the bottom tip region. During this time, up to 32 anatomical marks can be recorded and stored in the data file header. The current version of the laser scanner
uses a single camera with two lasers to cover a radial extent of -0.5" to +3.5" from the axis of rotation and a height of 4" to 12" above the turntable. The laser can scan
around an object for a full 360° in 1° increments if needed. This corresponds to a cylindrical scanning region 8" high by 7" diameter. The mean radial sample spacing
ranges from 33 samples per inch at the top to 50 at the bottom of the scan region. The mean height sample spacing is about 21 samples per inch across the scan region.
The angular spacing can be as small as one degree, but acquisition time has to be lengthened to greater than 10 seconds.
An alignment/calibration jig was used to assess absolute accuracy of the scanner in locating points in the radial and height directions. Radial positions of -0.25",
1.25", and 2.75" and height positions of 4" to 12" in 1" increments were used. This provided 27 reference marks spread throughout the allowable scan region which
were not part of the original calibration grid. The worst case expected error of the system is + /- sample space which is approximately 0.03" for radial samples and +/-0.05" for height samples. The observed maximum deviation from the expected value at each of the three radial positions was very close to these expected values. The
mean value of the radius determined from the nine height positions at each radius was within 0.02" of the expected mean value with a sample standard deviation of
0.02".
To test the positional reproducibility of anatomical marking in a realistic setting, surface locations were recorded for three widely spaced surface reference marks
on a below-knee model at three different orientations. The model was removed and repositioned between each of the three orientations. The surface marks were chosen
to yield distances between each of about 3.5" to 4.0". Three-dimensional locations (x,y,z) of each surface mark was determined using the horizontal video cursor, the
laser line, and the turntable orientation to input raw data positional information. Distances (D1,2,D1,3 and D2,3 between the surface reference marks were calculated for
each position of the model. Since distance between surface marks does not
change with orientation, the calculated distances at different orientations can be used to assess reproducibility of anatomical marking when an identical object is scanned
at a different position within the allowable scan region. This test is relevant to assess the ability to reorient surface data when successive scans of the same patient are
made. This evaluation also indicates the overall measurement accuracy of the laser scanner at different object orientations. The mean value of D1,2 was 3.65" with a
standard deviation of 0.03" and a range of observations of 0.06". The mean value of D2,3 was 4.05" with a standard deviation of less than 0.01" and a range of only 0.01".
The mean value of D1,3 was 3.99" with a standard deviation of 0.03" and a range of 0.05". These results indicate that positional reproducibility of surface markers for
the laser scanner system is very good and is close to the positional errors expected from sample spacing.
To assess the capability to scan and build a positive model of a residual limb, a scan was taken of a positive model of a below-knee limb and the data forwarded to
the Prosthetic Research Study Group at Seattle, Washington20 for fabrication of a positive model. The surface data was converted to concentric rings (44 samples each)
with a spacing of 1/8" based on requirements of the fabrication software.Figure 3 shows (a)
the original model, (b)
the profiles produced by the scanner, (c)
using a solid
and wireframe model, and (d)
the positive model which was reconstructed using a numerically computer controlled milling machine.
Computer Aided Design Software
A computer assisted design program has been developed which allows modification of a computer model of the residual limb in much the same manner the prosthetist
modifies a plaster model before creating a socket. The program has been created with the following criteria: (1) the program needs to be able to accept three-dimensional
data from a scanning device and create a computer model of the residual limb, (2) the shape of the model then will be modified to the biomechanical shape of a good socket
in an interactive manner, (3) the modified shape must be output in a form compatible with popular
CAD programs so that a model for making the socket can be machined or the socket can be directly made, and (4) this program must be able to run on a relatively
inexpensive microcomputer workstation.
Initially, efforts were directed at using a commercial CAD program on an IBM/AT compatible personal computer to perform the shape modifications. The CAD
program proved to be unwieldy and far too slow to be an interactive tool. At this point, it was decided to develop a custom program to perform the shape modifications.
The hardware development environment for the program was dictated by a combination of having a microcomputer as a target machine and by the computers that
were available for program development. Three-dimensional computer graphics are both memory and central processing unit intensive. An IBM/AT running MS-DOS
cannot address enough memory to deal with an object consisting of more than a few hundred 3D data points. A few thousand data points are necessary to create an
effective 3D model of a residual limb. Program development was switched to a SUN/3160 workstation to
increase flexibility of data manipulation. Care was taken so that the final program developed on the 3D workstation could be transferred to a lower cost microcomputer
and it is expected that the final software will run on a PC.
Software development tools were selected so that the finished program could be run on a number of different machines including an IBM/AT with the proper plug-in
board. The development tools consisted of a C-compiler and a graphics sub-routine package. The program was written in the C-programming language because it is
portable across many different computer types. In order not to have to write a device-dependent graphics routine, a graphic sub-routine package called HOOPS was
purchased. HOOPS from Ithaca Software (Ithaca, NY) is available for SUN workstations, the Apple Macintosh II, and the IBM/AT. Most important for the IBM/AT,
there is a HOOPS version which runs on the Nth Graphics Engine, which is a plug-in board for the AT. The Nth board has its own graphics memory and graphics
processor which allows very large models to be created by using an AT.
This program uses software tools in a manner analogous to the way a prosthetist uses tools to modify a plaster model. Instead of tools, a mouse is used to select areas
of the model to modify and to pick the appropriate software tool to perform the modification. The top of the screen is divided into a work area on the left and an area on
the right showing the original and modified shapes superimposed (Figure 4)
. Each of these two areas has a menu of buttons for changing the viewing orientation of the
models. On the bottom of the screen, there are two rows of buttons. The top row selects the tool to use for modification and the bottom row contains global operations.
Above the buttons is a status line which shows the current function and what action to perform next.
The model shapes can be created from a set of vertical profiles or from a set of horizontal contours. Both profiles and contours are displayed creating a wireframe
image of the model. Colored contours are used to aid in orientation and to mark the front, back, left, and right side of the model. The initial orientation displays the front
of the model to the right.
All viewing functions are controlled by the mouse. The views displayed are determined by the position of a "camera" (Figure 5A)
. Positioning the camera relative to the model determines the point of view. The spin button will
orbit the camera about the model. When the spin button is pushed determines the direction and speed of the camera movement. As long as the button is held down, the camera will spin (Figure 5B)
. The view button selects an absolute
camera position relative to the initial orientation. This allows the user to select a known orientation quickly (Figure 5C)
. The "reset camera" button returns the camera to the initial orientation (Figure 5D)
. The "solid toggle" button
toggles between the wire frame model and a solid shaded model. The solid model shows surface features better than the wire frame but it takes considerably longer to draw. On the right hand menu, the buttons marked Original and
Modified toggle the original and modified models on and off (Figure 6)
.
Each of the modification functions is used
to emulate operations performed by the prosthetist on his plaster model. The object of these modifications is to increase socket contact pressure in weight-bearing areas and to relieve pressure on sensitive areas.
In the patellar-tendon area, the "plaster model" is cut away leaving a channel that will create a protuberance in the socket. The software tool "Rattail" is used to make this modification. A profile is selected in the center of the patellar-tendon area. A window showing only this profile is displayed. The profile shape is modified to the shape of the desired channel (Figure 7)
. The channel is then cut through the model in this shape (Figure 8)
.
The "patch" function can be used both to
increase socket pressure and to relieve pressure around sensitive areas. Areas to be modified are first outlined with the mouse. The mouse is then used to select a control point for the point of greatest modification. The distance to move
the control point is entered at the keyboard (Figure 9)
. For pressure relief areas such as the head of the fibula, a bump is created with a smooth transition from the control point to the edge of the patched area (Figure 10)
. The patch
function can also be used to increase socket pressure by entering a negative distance to move the control point. This function can be used in the popliteal area (Figure
11)
.
In areas like the medial tibial flare, plaster is shaved off in order to increase socket pressure. In using the "shave" function, the area of interest is first outlined using
the mouse, then modified (Figure 12)
.
The "smooth" function can be used like sandpaper to remove rough edges from the model. It is generally used when modifications to the model are complete.
Utility programs have been written to convert data from scanners to a format that the modification program can use. In addition, there are utility programs that convert
files to a format that can be read by CAD programs so that models for making sockets can be produced or so that sockets can be produced directly. The output files can be sent electronically by phone or computer network to a central facility for machining.
Thermoplastic Lightweight Prosthetic Components
The production of thermoplastic prosthetic sockets by shrink-forming is receiving widespread and growing interest. This technique makes use of the "memory proper-
ties" of certain thermoplastics which describes the tendency of a hot stretched and cooled sample to retain its stretched form until reheated, whereupon, shrinkage will occur. Heat shrinkable, conical preforms for socket production
are generally produced by reheating, expanding, and cooling a smaller intermediate molding to build-in high shrinkage stresses which are subsequently relieved during shrinkforming.21,22 In one method of socket manufacture, a
rectified
plaster model of the patient's residual limb is placed inside a heat shrinkable preform (Figure 13)
which is subsequently heated to soften and shrink it so it conforms closely to the model's contours. Final shaping is performed by applying
a vacuum so that the hot plastic is drawn into contact with the model and is maintained while the plastic cools and hardens.
The manufacture of thermoplastic sockets for artificial limbs from heat shrinkable preforms has several advantages over existing techniques. It is potentially simpler and more cost effective than drape-forming from a flat sheet
and reduces the manual skill requirement of the process. As a result, the socket wall thickness is more uniform. A further advantage is that preshrink wall thickness can be easily adjusted so that a minimum value of post-shrink wall
thickness can be assured in the socket. Post-forming shrinkage should also be reduced. Com
pared with the "Rapidform" system for producing polypropylene sockets, 23 shrink-forming eliminates the requirement for an expensive, "purpose-built" vacuum forming machine, a definite advantage in some developing countries.
Two new methods for producing heat shrinkable thermoplastic preforms have recently been described.24 One method is based on established heat shrink technology and special crosslinked thermoplastics. The second method
is based on blow-molding which simplifies preform manufacture by reducing it to a single stage operation with concomitant time and cost savings. Instead of reheating and expanding an intermediate mold, the extruded tube of molten
thermoplastic (parison) produced by the blowmoldmg machine is expanded to the final preform dimensions defined by the mold cavity and cooled by contact with the mold surface.
(Figure 14)
It is considered that blowmolded, heat shrinkable preforms present the greatest potential for economic production and immediate application. As a result, blowmold tools have been designed and manufactured for the REL for
production of heat shrinkable preforms suitable for shrink forming below-knee and above-knee sockets. Work in progress will extend the material range of heat shrinkable preforms available to the prosthetist, to improve the match
between tissue and socket compliance and expose this exciting and promising socket forming technique to a wide population of end users.
Field Testing
Field testing of the system will begin in 1989. This portion of the project will evaluate the hypothesis that the lightweight all plastic prosthesis will decrease energy expenditure, increase duration of daily pros-
thetic usage, and improve overall amputee function and satisfaction.
We will invite below-knee adult amputees who normally wear their prostheses all day to take part in the project. We will conduct a baseline evaluation with the original prostheses to determine suitable candidates. After selection
of the appropriate candidates, an image of the residual limb will be obtained utilizing the laser/video system. The computer data of the residual limb will be modified using the computer aided design software. Positive models will
be obtained using a computer numerically controlled milling machine. A socket will be fabricated using heat shrink plastics. Each subject will be fitted with a lightweight all plastic prosthesis.
At one-month, two-month, and threemonth followup visits, the investigators will collect data to determine the subject's energy expenditure, usage of prosthesis, and
functional status. The subject with the new prosthesis will provide information regarding duration of daily prosthetic usage, overall functional status, and degree of satisfaction with the prosthesis. Thus, the lightweight all plastic
prostheses will be worn for three months. The data collected will be analyzed to determine whether the new prostheses improved function and daily use.
Conclusion
The capability to fabricate current state-of-the-art prostheses in underdeveloped countries has profound limitations. Our approach to solving this problem is the utilization of a remote residual limb scanning device to provide
topographical data which can be sent via telephone modem to a centralized automated prosthetic fabrication facility. The REL laser scanner has the accuracy and other attributes necessary to function as this remote device. The lightweight
all plastic
prosthesis resulting from this process will help address the need for prosthetic limbs in underdeveloped countries.
|
