 |
Description  |
|
|
FIELD OF THE INVENTION
The present invention relates to the field of radiographic analysis of the
human body and, in particular, to a method of measuring and displaying
bone mineral density adjacent to prosthetic bone implants,
BACKGROUND OF THE INVENTION
Bone prostheses are commonly provided to patients having bone disease or
injury. Such prosthetic devices are manufactured of durable materials such
as metals, ceramics, or dense plastics, and are attached to the remaining
bone to replace the function of defective or missing bone and joint. In a
hip replacement, for example, the ball-shaped head of a femur may be
replaced with a prosthetic ball attached to the proximal end of the femur
by a shaft fitted within the femur shaft.
The loosening of such implants over time and with use is a significant
concern. Many different techniques have been employed to try prevent such
loosening. Examples include cementing the implants to the bone, the use of
porous coatings on the implant to stimulate bone in-growth, and custom fit
implants.
A good fit between the implant and the bone will impart a pattern of stress
onto the bone which may cause it to regenerate. If the bone grows adjacent
to the implant there will be less chance of loosening.
Alternatively, loss of bone around the implant may indicate poor fit
between the bone and implant in certain areas and provide an early
indication of future loosening or failure of the implant. For this reason,
the implanted bone is often monitored after the implant is surgically
implanted. Such monitoring may be performed by conventional radiographic
studies, however large changes in bone density of up to 30% may be
necessary before such changes become apparent on the radiograph.
Preferably, digital radiographic techniques are used to provide a
quantified measurement of bone density. Such techniques include dual
energy x-ray absorptiometry ("DEXA") in which a measurement of bone
mineral density is derived from the varying absorption of the bone of
x-rays at different energies, and dual photon absorptiometry ("DPA") where
a similar measurement is made using radioisotopes. Such densitometers
provide quantitative in-vivo measurement of bone mineral density ("BMD").
Other digital radiographic techniques such as computed tomography ("CT")
may also provide measurements of bone density, however, the metal of the
prosthesis may create image artifacts in a CT image rendering the
measurement of bone density in the neighborhood of the prosthesis
problematic.
Conventional DEXA or DPA equipment, when used to monitor changes in bone
density, may obscure subtle changes of the bone near the implant. The
region of interest ("ROI") that is isolated and evaluated by such
equipment may include irrelevant bone, tissue and other artifacts.
Further, the ability to determine bone loss over time is limited, with
such equipment, because of the difficulty in matching the data between two
different measurement periods.
Recently there has been increased interest in implants constructed of
composite materials, such as carbon fibers and various matrix materials,
having less stiffness than the ceramic and metal materials presently used.
An implant whose flexibility more closely matches that of the bone in
which it is implanted is thought to eliminate "stress shielding" in which
the bone around the implant is shielded from normal stresses, and thus
benefit from the effects of such stress in bone remodeling and bone
strengthening. A flexible implant, in contrast to stiffer implants, may
pass stress through to the surrounding bone.
Many composite materials considered for implants are essentially
transparent to x-rays making it difficult to accurately locate the
interface between the bone and the implant for post-operative evaluation
of the fit of the implant and the health of the surrounding bone.
SUMMARY OF THE INVENTION
The present invention provides a method of reproducibly evaluating bone
density measurements in a region of interest conforming to the interface
between a radiolucent implant and neighboring bone.
Specifically, a matrix of bone density data values is analyzed to identify
the location of radio-opaque reference markers embedded in the implant. A
stored template, having stored reference marker data, is fit to the
location of the reference markers and provides a template edge used to
deduce an implant boundary within the matrix of data values. A measurement
boundary translated from the implant boundary by a predetermined distance
along a translation axis is then determined and bone mineral density
within a plurality of segments following the path of the measurement
boundary are calculated. A plot of segment values versus distance along
the implant boundary is then displayed.
One object of the invention, therefore, is to provide an accurate
indication of the state of the bone immediately adjacent to the implant.
The use of a conforming ROI prevents bone from outside of the implant area
from influencing or obscuring the measurements of bone near the implant.
The result is improved sensitivity and easier measurement, The use of a
template allows accurate determination of the implant edge, preventing the
bone measurement from being influenced by inclusion of the low density
implant.
Another object of the invention is to provide a readily reproducible
measurement that may be compared to other later studies. Fiducial points
may be established with respect to the bone and the implant to form a
reference for the plot of bone density. The conforming ROI is referenced
from the edge of the implant and the fiducial points and therefore may be
accurately and repeatably located.
Plots of bone density showing the medial and lateral sides of the implant
aligned in the proximal/distal direction also may be displayed
simultaneously.
Yet a further object of the invention, then, is to provide a display method
that highlights possible implant related bone resorption. With the medial
and lateral sides of the bone measured and displayed simultaneously,
symmetrical or non-symmetrical bone loss is readily apparent, such as may
be caused by abnormal stress patterns from the implant.
Other objects and advantages besides those discussed above shall be
apparent to those experienced in the art from the description of a
preferred embodiment of the invention which follows. In the description,
reference is made to the accompanying drawings, which form a part hereof,
and which illustrate one example of the invention. Such example, however,
is not exhaustive of the various alternative forms of the invention, and
therefore reference is made to the claims which follow the description for
determining the scope of the invention.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a schematic view of a dual x-ray bone densitometer suitable for
collecting bone density data over a scan for use with the present
invention;
FIG. 2 is a flow chart showing the steps of separating tissue, bone, and
implant and identifying fiducial points in the collected data;
FIG. 3 is a flow chart showing the steps of the "identify implant" step
from the flowchart of FIG. 2;
FIG. 4 is a flow chart showing the steps of the "identify bone" step from
the flowchart of FIG. 2;
FIG. 5 is an illustration of the location of the landmarks in the femur and
implant used in the process of FIG. 2;
FIG. 6 is a flow chart showing the steps of creating a histogram of bone
density;
FIG. 7 is a flow chart showing the steps of creating a trans-prosthetic
profile of bone density;
FIG. 8 is a pictorial representation of a screen display of a bone density
image together with a periprosthetic histogram;
FIG. 9 is a pictorial representation of a screen display of a bone density
image together with a trans-prosthetic profile;
FIG. 10 is an anterior/posterior view of a composite implant for a femur
showing a template edge derived from the implant, the location of
radiopaque reference markers forming stored reference marks of a template
including the template edge, and measured reference mark positions to
which the template may be fit;
FIG. 11 is a flow chart showing the steps of employing the template of FIG.
10 in creating an implant boundary; and
FIG. 12 is a simplified cross-sectional view of the implant of FIG. 10
taken along lines 12--12 showing the effect of rotation of the implant on
the relative spacing of the reference markers.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
The present invention provides the ability to analyze in vivo bone mineral
density within one or more regions of interest, or ROI, that conforms to
the outline of the prosthesis implanted into a patient. This flexibly
defined ROI, which will thus not normally be rectangular, permits critical
areas of bone density to be examined free from influence by other areas
and structures. The ROI may also be referenced to a fiducial point, to
permit the same ROI to be evaluated by repetitive scans taken over long
periods of time to generate reliably comparative values.
The values obtained from bone density measurement in the ROI may be
displayed in a histogram of average or cumulative bone density values
taken within a series of segments along the implant. The display of such a
histogram on a CRT screen can provide to the clinician an instantaneous
picture of bone density over the length of the implant. Alternatively, a
profile of bone density along an individual segment may be examined to
make a quantitative measurement of bone adsorption or hypertrophy in the
neighborhood of the bone. Both medial and lateral histograms and profiles
may be displayed simultaneously to permit the rapid identification of
symmetrical effects that may indicate their origin in stress or lack of
stress from the implant.
Instrument
Referring to FIG. 1, a bone densitometer 10, of the preferred type for use
with the present invention, projects a dual-energy x-ray beam 12 from a
x-ray source 14 along a ray 16 through a patient 18 to an opposed detector
20. The x-ray source 14 and detector 20 are mounted on a carriage 22 to
move in unison in a raster scan pattern 24 by means of stepper or servo
motor (not shown). The raster scan pattern 24 sweeps the ray 16 over a
rectangular area of the patient 18 by alternately directing it along one
of two perpendicular axes x, and y of a Cartesian coordinate system with
the x-ray beam 12 parallel to a third orthogonal z-axis.
Preferably, in the case of the evaluation of a hip implant, the carriage 22
is positioned to sweep an area including the neck and shaft of the femur
26, including the trochanter, and prosthetic implant 27. The orientation
of the raster scan pattern 24 is preferably such that the shaft of the
femur 26 lies substantially parallel to the y-axis of the raster scan and
the data is acquired in successive scan lines along the x-axis. The
scanning, data analysis, and data display procedures described here may
also be performed on prostheses in other parts of the body in which case
the orientation, location and length of the raster scan may be suitably
adjusted.
The movement of the carriage 22 is controlled by a motor controller 28
receiving signals from a computer 30. The x-ray source 14 includes an
x-ray tube along with a K-edge filter to create two narrow energy bands of
x-ray emissions. The signal from the detector 20 is sampled and digitized
by data acquisition system ("DAS") 34 which may transmit the digitized
samples to the computer 30 which stores the data in computer memory (not
shown) or on mass storage device 36. An operator may provide inputs to the
computer 30 by means of keyboard 38 and trackball 40 (or mouse) which
allow positioning of a cursor on display screen 42 as is understood in the
art. The display screen 42 also provides a means of displaying information
obtained from the raster scan.
At a continuous series of discrete data points over the raster scan pattern
24, the signal from the detector 20 is sampled by the DAS 34 at each of
two x-ray energy levels produced by the x-ray source 14, as filtered by
the K-edge filter. Thus at each location, two samples, having values
corresponding to the absorption by the patient 18 of x-rays 12 at both of
the x-ray energy levels, may be collected. Each pair of samples may be
identified to the x and y coordinate of the location at which the samples
were acquired. Together, the sample pairs over the entire raster scan
pattern form elements of a data matrix, whose matrix coordinates
correspond to the ray 16 coordinates. On presently available DEXA
machines, as shown in FIG. 1, a spatial resolution of 0.6 mm between
samples may be obtained over a raster scan area of about 12 by 14 cm.
Referring to FIG. 2, the acquisition of this data matrix is represented by
process block 50. As is understood in the art, the sample pair taken at
each of the two energy levels together provides an indication of bone
density or mineral content of the bone along ray 16. The data matrix
therefore represents the density of the tissue and bone over the scanned
area of the patient's femur 26. It will be understood to one of ordinary
skill in the art that the data matrix of bone density samples also may be
obtained by other densitometers, such as those based on DPA and CT as
described above. The data matrix is stored in a file on computer 30.
Image Analysis
Upon completion of the acquisition of data matrix, the data elements of the
matrix are analyzed to differentiate data elements associated with bone
from data elements associated with tissue and the implant. This process is
illustrated by the flow chart of FIG. 2, illustrating the process of image
analysis, which begins with the step of acquiring the matrix of data
values from the scan of the patient. Included within this step, indicated
at 50 in FIG. 2, is the combining of the two values of each sample to
create a non-calibrated value corresponding to the total density along the
associated ray 16 of each sample. The data points thus created are
referred to as PBM, for pseudo bone mineral content. The numbers are
pseudo values because they are non-calibrated and therefore dimensionless.
At this point in the analysis, therefore, only the relative differences
between the data points are significant, not their absolute values. While
the calibration for each value could be done at this point, it is
consumptive of computer resources, and thus is deferred at this point, and
the PBM values are used. The calibration of the final values to correspond
to standard physical values is performed as a last step. This technique is
well known in the art.
The remaining steps in the flowchart of FIG. 2 will be described first in
overview and then in detail. The data values of PBM are processed at step
52 or a template is used to identify the implant. Then at step 54, the
values are analyzed again to identify the bone. At step 56, the bone
values are again analyzed to identify the bone landmarks. Based on that
analysis, the regions of interest in the bone are identified at step 58.
From the analysis of those regions, baselines are derived at step 60.
Finally the results are calculated and calibrated and the results are
displayed at step 62.
Identify Implant
This process is intended to identify regions of the data matrix of PBM
values which correspond to the physical implant. Since the values are at
this point dimensionless, the analysis of the values to determine the
values which correspond to the implant must be done on the basis of
relative comparison of values rather than absolute values. To begin the
process, which is illustrated in FIG. 3, the matrix of data values is
filtered with a low band pass filter to remove high frequency noise
components from the data at step 70. Next the data values must be analyzed
to generally identify the area of the implant.
Metal and Ceramic Implants
Expressed in absolute values, the normal biological range of bone mineral
density in human femurs ranges up to values less than 3.0 grams per square
centimeter. An examination of the density values of common hip implants
reveals that most metal and ceramic implants have a density value much
greater than 3.0. Thus, working with the PBM data matrix, a comparative
analysis is performed to identify a region where the values are abnormally
high when compared to other parts of the data matrix. This step is
illustrated at 72 in FIG. 3, which illustrates the procedure for the
implant identification step 62 of FIG. 2. This analysis results in the
generation of a threshold value, above which all data points are defined
to represent implant, and those data values are tagged by the computer as
representing implant.
The next step is to fill any voids or defects in the representation of the
implant. For some implants, the matrix of data values has been found to
have occasional values inside the area of the implant with inappropriately
low PBM values. At step 74, all data points that are surrounded on all
four sides within a defined distance (5 mm) by implant are also defined to
be implant. This step has the effect of filling in a solid area in the
matrix as implant area. The edges of the implant area are then adjusted at
step 76. The absolute edges of the implant are identified by locating the
points in the data matrix at which the greatest differences exist between
adjacent values. These points of greatest change are defined to be the
edge of the implant. This calculated true edge is slightly extended
outward in this step to compensate for shadowing or partial edge effects
caused by the sharp edge. This is done by moving the defined edge outward
until the change between adjacent data point values becomes almost zero.
At this step in the process a rough outline of the implant exists in the
matrix of values.
It is now possible to smooth out the rough outline of the implant created
to this point. This step, indicated at 78 in FIG. 3, involves breaking up
the identified edges of the implant into a series of small sections which
are defined between "nodes." The nodes are defined to be the locations of
changes in direction of the implant perimeter, such as corners. Since
presently used implants are known to have certain shapes, certain nodes
can be "forced" onto the data, since the system knows that a discontinuity
exists in the real implant at a certain point. Then between each set of
nodes thus defined a best fit curve routine is used to find a high order
(fourth degree) curve which will have a best fit with the measured data to
define the edge of the implant. The curves that best represent the data
are then incorporated into the final outline of the implant at step 80.
Once the outline of the implant is derived, all data point values in the
matrix of data points inside the curve thus defined are defined and tagged
to be implant, at step 82. This, in essence, isolates all the implant
values from the values derived from the areas of soft tissue and bone. The
values thus defined as implant are then highlighted for the ultimate
display at step 84 and are excluded from further data analysis to save
processor time. A manual override is provided so that correction for
misapplication of the implant defining procedure can be implemented by the
operator if necessary.
Composite Implants
When the implant 27 is a composite material rather than a metal or ceramic,
the comparative analysis described above in which the implant 27 is
distinguished from surrounding bone by establishing a threshold of PBM
values is unsuccessful. The PBM values of the composite implant 27 are
generally lower than or equal to the bone itself.
Nevertheless, the composite implant may not simply be ignored in the
calculation of bone quality near the implant. It is important to isolate
the implant from the measurements of the bone for a number of reasons.
First, the implant edge serves as a reference to ensure the same region of
interest is being measured when a series of measurements are made spanning
a year or more. Second, determining the implant edge allows the region of
interest to be on the critical bone implant interface area. An overly
inclusive region of interest may obscure small changes in this important
interface area. Third, exact identification of the edge of the implant
prevents calculations of average bone density from being diluted by
over-inclusion of low PBM implant data.
Thus, an alternative method is adopted for determining the implant boundary
if the implant is a composite material, as identified by the operator
through the keyboard 38 of the bone densitometer 10.
Referring to FIG. 10, radiopaque tantalum marker beads 31 or other
reference markers may be placed in the shaft 29 of the implant 27 to
provide data as to the location of the implant, such as may be detected in
radiographic examination of the implant 27. The marker beads 31 are
embedded in the surface of the shaft 29 so as to be removed from regions
of contact with bone and tissue.
Two marker beads 31 and 31' may be placed at the proximal end of the shaft
29 on the medial and lateral sides respectively. A third bead 31" may be
placed at the distal end of the shaft 29. Thus, for an anterior/posterior
radiographic image of the implant 27, three reference markers 31, 31' and
31" should be separately visible.
Ideally, the distal reference marker 31" will be generally along the axis
35 of symmetry of the shaft 29 with reference markers 31 and 31' being
displaced equally on either side of the axis 35 to provide an indication
of any rotation about axis 35.
The dimensions of the implant 27 and the coordinates of the medial and
distal edges, 37 and 39 respectively, of the shaft 29, are stored as a
numerical template to be accessible by computer 30. The template is formed
of a set of coordinate points, representing the edges 37 and 39,
referenced to coordinate points of the centers of the reference markers
31, 31' and 31". Accordingly, once the position of the reference markers
31, 31' and 31" are identified within the matrix of acquired data, the
particular matrix elements along the medial and distal edges, 37 and 39,
may be rapidly identified.
Referring also to FIGS. 2 and 11, step 52 of identifying the implant 27
data, in the case of composite implants, begins with the identification of
the reference markers 31 shown as process block 85. The identification of
the reference markers 31 may be done simply by manually locating a cursor
within an image of the implant 27 where the reference markers 31 will be
visible as small circles of high contrast.
Alternatively, the high contrast reference markers 31 may be readily
identified by the same thresholding procedure used to identify metallic
implants. That is, a comparative analysis is performed of the PBM data
matrix to identify small clusters of data values where values are
abnormally high when compared to other parts of the data matrix. In the
event that more than three areas are identified with such high PBM values,
areas having an exact number of data values equal to the known size of the
reference markers 31 are preferentially selected.
The center of mass of the each of the qualifying data values is selected as
the coordinates of the corresponding reference markers 31.
Next, as indicated by process block 86, stored reference marks 33, 33',
33", corresponding to reference markers 31, 31', and 31", forming part of
the template accessible by the computer 30, are fit to the measured
reference marks 31. This fitting involves incrementing or decrementing the
coordinates of each of the stored reference mark 33, 33", and 33", until
the cumulative separation between the measured reference markers 31, 31',
and 31" and the stored template reference markers 33, 33', and 33" is
minimized as represented by the sum of the square of the absolute
differences. This procedure has the effect of shifting the template over
the measured values until a best fit (without rotation) is realized.
This translative best fit procedure is followed by a rotation of the axis
35 of the template on the measured data, again with an eye toward reducing
the sum of the magnitude of the distances between the measured reference
marks 33, 33', and 33" and the actual reference marks 31, 31', and 31"
within the data matrix.
This process of translating the template with respect to the data matrix
and rotating the template with respect to the data matrix is repeated for
a predetermined number of iterations after which the template and actual
data should be accurately aligned in a best fit. If the difference between
the measured markers 31, 31', and 31" and the markers 33, 33', and 33" of
the template is greater than a predetermined error value, a signal is
provided to the operator indicating a possible problem with matching the
data such as may be caused by misidentification of the proper template to
be used, or artifacts within the image that were mis-identified as
reference markers.
If the difference between the actual measured reference markers and the
reference markers of the template is less than the predetermined error
value, the program proceeds to process block 87 and the data of the data
matrix is marked to indicate which data values are of implant 27 as
opposed to the bone. This step provides a tagging of all the data point
values in the matrix of data points inside the implant essentially
identical to that of step 84 previously described. Again, a manual
override is provided so that correction for misapplication of the implant
to filing procedure can be made by the operator.
Referring to FIG. 12, in a further embodiment multiple templates may be
stored or generated by computer 30 for each implant 27, each template
providing a projection of the implant 27 at slightly different angles of
rotation about the implant axis 35. New medial and lateral boundaries 37'
and 39' may be generated based on the projection of the rotated implant 27
on the image plane.
After the best fit process of process block 87, remaining deviation between
the measured reference markers 31, 31' and 33' and the template reference
markers 33, 33' and 33" is interpreted as rotations of the implant 27
about the implant axis 35. The direction of rotation is immaterial
provided the shaft 29 is essentially symmetrical about axis 35. The
foreshortening caused by rotation is used to deduce the rotation and to
effectively shrink or expand the boundaries 39 and 37 according to the
other stored templates to provide an improved matching of the template.
The deduced angle of the implant 27 may be used to generate a notification
to the operator that the implant is being imaged at an angle perhaps
different from that originally measured in a previous benchmark.
Identify bone
Regardless of the type of implant used, at this point the matrix of PBM
values has a set of values which correspond only to bone and tissue, and
these two must be discriminated. To determine the edge of the bone, a raw
threshold determination, such as that used for finding the edge of the
implant, was found not to be accurate due to wide variations in the bone
densities of actual patients, particularly at the bone margins and, in
some cases, in the middle. Therefore a more sophisticated technique was
adopted which begins with an edge detection procedure. The whole process
is illustrated in FIG. 4, which corresponds to step 54 of FIG. 2.
In step 90, a density distribution curve of the values over the entire
curve is generated. This curve is a plot of all PBM data values,
regardless of location. The curve is examined at step 92 to find the first
valley in the curve, which is the starting point for the threshold value
to separate bone and soft tissue. This threshold is then used at step 94
to locate the approximate edges of the bone. The edges are then refined at
step 96 through a process of l | | |
