RADIOGRAPHIC QUANTITATION METHOD FOR ASSESSING PERIODONTAL BONE LOSS
FIELD OF THE INVENTION The present invention relates generally to methods and devices for measuring bone density and bone loss, and particularly to an improved system and method for making objective measurements whereby detected changes may be measured accurately.
BACKGROUND OF THE INVENTION Currently, radiographic techniques are implemented for measuring bone deformities such as loss of bone density. For example, with respect to teeth, bone changes are the hallmark clinical sign of periodontitis, i.e., tooth decay. As known, however, a radiograph (i.e., an x-ray) is a two dimensional image of a three dimensional structure. Thus, radiograph images lack the dimension of depth and, consequently, cannot reveal morphology of bone deformity. Thus, traditional techniques involve obtaining a sequence of images over time, and, analyzing the images which typically requires a subjective analysis. To perform a completely objective analysis of bone deformity using radiographs, two technical hurdles must be overcome: 1) ensuring that the same area of bone is being compared in sequential radiographs; and, 2) ensuring that the sequential radiographs are qualitatively similar (i.e., that comparable techniques are implemented when obtaining the sequential radiographs). With respect to animal teeth, e.g., canines, this is a highly difficult task as this manipulation currently requires extensive equipment (e.g., "jigs"), made individually for each animal in a study, to assure that the x-ray source, subject animal's head, and radiograph sensor are aligned identically for each image. This time-consuming step is procedurally very difficult, and is logistically impossible, for example, when necessary to conduct a large trial. It would be highly desirable to provide a system, method and computer program product for obtaining and processing radiographic images that facilitates the measure of bone deformities objectively. It would further be desirable to be able to provide a system, method and computer program product for processing a series of radiographic images that will facilitate the identification of bone changes with an improved degree of confidence.
SUMMARY OF THE INVENTION The present invention is directed to a system, method and computer program product for accurately determining and objectively quantifying bone deformities such as bone density loss. According to the present invention, bone deformities in a three-dimensional area of bone (e.g., periodontal bone) can be accurately determined and objectively quantified based on novel techniques applied to two-dimensional measurements (e.g., radiographs). That is, according to the invention, novel applications of existing hardware and computer software programs are implemented to manipulate and analyze digital two- dimensional radiograph images for determining bone deformities in a three-dimensional area of bone. First, digital radiographs of the bone (e.g., an animal's tooth) to be analyzed are taken. Then, at subsequent time intervals, further digital radiograph images were taken of the 3-d bone area under analysis. Then, the subsequent radiographs are morphed to perfectly match the initial image's framing, magnification and angle. Further, according to the invention, once identical images are created, an assurance is made that the image technique (amount of x-ray energy the sensor is exposed to) is comparable for each image. One solution is to generate a histogram of the pixel intensities observed in each radiograph to standardize each exposure to the initial (first) image. This histogram is a visual representation of the distribution of each of the different shades of gray recognized by the computer, for example, in an 8-bit radiograph image. Each pixel gets assigned a pixel intensity value between 0 and 255, with 0 and 255 corresponding to absolute black and absolute white, respectively, and each shade of gray being a number in between. The histogram distribution of these values may vary in shape, but the bandwidths at each extreme (e.g., representing the white of tooth enamel, and the black of air) are representative of the exposure of the radiograph sensor to x-ray energy, and therefore can be used to standardize or calibrate the images to one another. Once this step is applied to each set of two morphed radiograph images, a region of interest is outlined and a mean pixel intensity value determined. That is, the two images can be compared with confidence that only bone changes are being identified. After identifying a region of interest, an intensity value for each pixel included in the delineated area of an image is determined. Preferably, the area of interest is determined and delineated according to anatomical references, e.g., portions of roots of the teeth treated in the model, as well as the majority of bone surrounding them. This procedure yields a number, which is the mean of all pixel intensities within the area of radiograph representing bone surrounding treated teeth. This number is representative of the bone density as measured on the radiograph. A final manipulation step is required to make this number applicable to the visual representation of bone. That is, as a change in pixel intensity can not be directly,
linearly, converted to a visual change in terms of magnitude, this is corrected for by applying a transformation function, e.g., a third-degree polynomial transformation, to the pixel intensity values, which creates a linear relationship and allows changes over time to be compared in a meaningful way. The result of this methodology is that changes in the bone density can be accurately and objectively measured in a way which is directly applicable to a visual change in radiographs. This is a marked improvement over the completely subjective manner in which radiographs are traditionally measured. BRIEF DESCRIPTION OF THE DRAWINGS The objects, features and advantages of the present invention will become apparent to one skilled in the art, in view of the following detailed description taken in combination with the attached drawings in which: Figure 1 illustrates the methodology for obtaining and processing radiographic images that facilitates the measure of bone deformities objectively according to the present invention; Figure 2 illustrates the outlined region of interest for which a bone reactivity score is determined according to the present invention; and, Figure 3 illustrates the results of the bone reactivity analysis resulting from the inventive radiograph analysis techniques of the invention performed for an example study.
DETAILED DESCRIPTION OF THE INVENTION The system and quantitative methodology used for objectively analyzing digital radiographs according to the invention is now described. In the description, by way of example, bone density loss in the bone surrounding canine teeth, e.g., caused by periodontal disease, was measured as part of a model procedure for testing a vaccine for treating bone periodontitis, e.g., in canines, such as described in United States Patent Application Number 10/851965. It is understood that the technique of the present invention can be used to measure bone density loss in any type of bone matter (e.g., humans, mammalian, etc.). Figure 1 illustrates the methodology 10 for obtaining and processing radiographic images that facilitates the measure of bone deformities objectively according to the present invention. A first step 15 includes the step of obtaining sequential radiographic images of the bone to be analyzed. Preferably, at least two radiographic images spaced apart in time is required. For example, in a study utilizing the radiographic quantitation method of the invention, a subject canine had four (4) sequential digital radiograph images obtained (e.g., at weeks 0, 3, 6, and 9) using a Schick CDR® (Computerized Digital Radiograph) capture system such as provided by Schick Technologies, Inc. (Long Island City,
N.Y.) with standard radiographic techniques. Schick CDR® is a computerized imaging system that utilizes an electronic sensor instead of X-ray film for producing radiographic images. It is understood that any number of radiographic images (i.e., greater than two) can be obtained at subsequent periodic or aperiodic time intervals. Skilled artisans will appreciate that the Schick CDR may run on computer platforms such as PC-compatible computers running Pentium-based CPUs or processors of like processing speeds running Windows 98, Me, or 2000, including an adequate amount of RAM and hard disk space for storing captured X-ray images in most supported image formats including TIFF (uncompressed), GIF, JPEG, BMP and ASCII, or as raw data, and further comprising a display device adapted for displaying the captured X-ray images at a resolution such as, e.g., 800x600x256 colors. In an example embodiment described in greater detail herein, the digital radiographs were taken of canines in test groups as part of a procedure for testing the effectiveness of a periodontitis vaccination applied to a test group of canines. Of the test groups, one group (T01 ) was vaccinated intramuscularly (IM) with 1 ml of the vaccine, while the second (T02) and third (T03) groups were sham vaccinated with 1 ml of sterile saline. Further, in a procedure described in greater detail below, a challenge material (a pathogenic organism) was introduced into the root canal of molars of canines in the T01 and T02 groups following application of the vaccinations. Digital radiographs were then taken of the canine's teeth at approximately three, six and nine weeks following the procedure. In a second step 20, the obtained sequential images (e.g., obtained at weeks 0, 3, 6 and 9 in the example embodiment described) are first registered against each other using Image J® (NIH shareware, v1.28), which is a public domain Java image processing and analysis program for platforms including Linux, Macintosh OS 9, Mac OS X and Windows (available at http://rsb.info.nih.gov/ii/index.htmlK More particularly, used in conjunction with ImageJ® program is a plug-in software TurboReg® (available at http://biqwww.epfl.ch/thevenaz/turboreq/) which enables registration of images, i.e., alignment of two images, one of them being called the "source" image and the other the "target" image. According to the invention, by way of the example embodiment described, the target image is the original week 0 image and determines the reference to which the source image will be aligned/matched/registered. Source images will include the subsequent images (e.g., at weeks 3, 6, and 9) and each is registered separately to fit with the original week 0 image to enable a perfect match of the initial image's framing, magnification and angle and form a morphed registered image set. In the example embodiment described, the TurboReg® plug- in software for ImageJ® is implemented utilizing a scaled rotation technique. Thus, as a result of this step, in the example embodiment described, four morphed image sets would be obtained. It is understood that, using the TurboReg® software, other techniques can be used to register the images to form an image set including translation, rigid body, affine and bilinear
techniques. The use of software applied in the manner according to the invention advantageously eliminates the need for jigs. In a third step 25, each of the registered image sets is then calibrated using histogram standardization. In this step, the population of gray shades in the registered radiograph image sets were compared and corrected for. This step may not be necessary as conceivably all radiographs may be captured using the identical equipment (x-ray source, sensor, and computer) and technique (e.g., 7.0 kVp/0.5 mAs). However, it was felt important to eliminate any potential minor variation in technique or exposure due to changes in minute- minute line power or equipment wear. Ideally, no maintenance or adjustments are made to the equipment during the experimental period. Additionally, pixel intensity does not have a linear relationship with what the eye considers a white-to-black scale. Thus, as will be described in greater detail herein, this non-linear relationship can be corrected for by converting the pixel shades to visual grays through a 3rd -degree polynomial function. Thus, to ensure that identical images are created, assurance is made that the image technique (e.g., the amount of x-ray energy the sensor is exposed to) is comparable for each image of the sequence. As mentioned, while this need may be minimized by using the same radiographic equipment and settings at each successive image, it may be further necessary to ensure that the images could be validly compared. One possible solution contemplated by the present invention is to introduce into each radiograph taken a calibration step-wedge (usually a piece of metal with different, known thicknesses). Because of the very small size of dental radiograph sensors (approximately 2.5 x 4.0 cm for the case of a canine tooth in the example embodiment described herein), the size of the step-wedge may reduce the amount of bone which could be analyzed. It is understood that use of the calibration step- wedge may be suitable for calibrating radiographic images for analysis of larger bone structures. A preferred second solution is to generate a histogram of the pixel intensities observed in each radiograph to standardize each exposure to the initial (i.e., week 0) image. This histogram is a visual representation of the distribution of each of the 256 different shades of gray, for example, recognized by the computer in an example 8-bit radiograph image. Since an image's "bit" rate is determined by how many binomial pieces of data each pixel is described by, the shades of gray in a black and white image can be calculated by 2 raised to the power of the "bit" rate. Therefore, an 8-bit image is represented in the computer by a series of pixels, each being described by an eight-number string of 0's or 1 's, of which there are 256 combinations. Each pixel gets thereby assigned a pixel intensity value between 0 and 255, with 0 and 255 corresponding to absolute black and absolute white, respectively, and each shade of gray being a number in between. The histogram distribution of these values may vary in shape, but the bandwidths at each extreme (representing the white of tooth enamel, and the black of air) are representative of the exposure of the radiograph sensor to x-ray energy, and therefore can be used to standardize or calibrate the images to
one another. Once this step is applied to each set of two radiographs, a region of interest can be outlined and mean pixel intensity can be determined. It is thus now ensured that the two images can be compared with confidence that only bone changes are being identified. Thus, following step 25, a next step 30 includes outlining the area of bone surrounding treated teeth, excluding teeth and air. Then, at step 35, a mean density calculation of that area of specific interest is performed and recorded for each image set. Then, a value is generated representing the average "whiteness" of the bone surrounding the tooth roots, for example in the case of a canine tooth periodontitis in the example embodiment described. This number is termed the "bone reactivity score" and is an accurate, objectively- derived, representation of the mean bone density in the area of interest. This region of interest can be identified by another plug-in within the ImageJ® image processing and analysis software tool that lists the intensity value for each pixel included in any delineated area of an image. As shown in Figure 2, when applied to the example images of canine bone teeth 50, the area of interest 75 was determined and delineated according to anatomical references. As shown in Figure 2, an overwhelming amount of black (representing air dorsal to the crown of the tooth) and white (representing the unchanging enamel and dentin of the tooth itself) in the image dilutes any change being affected in the bone surrounding the tooth. Therefore, as shown in Figure 2, only a part of each radiograph representing the bone 75 from the level of the apex of the caudal root of the fourth premolar to the rostral root of the second premolar was quantified. As shown, this area 75 includes portions of all six roots of the teeth treated in this model, as well as the majority of bone surrounding them. Application of this procedure yields a number, e.g., between 0-255 for the 8-bit pixel gray scale example, which is the mean of all pixel intensities within the area of radiograph representing bone surrounding treated teeth. This number is representative of the bone density as measured on the radiograph. Returning to Figure 1 , in a further manipulation step 40, it is required to render the bone density number applicable to the visual representation of bone. This final manipulation step is required because the bone density number describing the pixel intensity does not correspond to a visual gray scale in a linear fashion. This is because the human visual gray scale (e.g., available commercially from photography suppliers) response varies as a sigmoid function. That is, given pixel intensities as the x-axis, and a human visual gray scale as the y-axis, a sigmoid-shaped curve is produced. Therefore, a change in pixel intensity would not be directly, linearly, converted to a visual change in terms of magnitude. According to the invention, this is corrected for by applying a third-degree polynomial transformation to the pixel intensity values, which creates a linear relationship and allows changes over time to be compared in a meaningful way. The grey-scale standard curve for this conversion can be described by the 3rd degree polynomial function: y= a + bx + ex2 + dx3 where, in an example embodiment, a = 2.063653,
b = -0.021063, c= 0.000091, and d= -0.0000001. The result of these modeling steps is that changes in the bone density surrounding treated teeth can be accurately and objectively measured in a way which is directly applicable to a visual change in radiographs. This is a marked improvement over the completely subjective manner in which radiographs are traditionally measured. As described herein, the technique of the present invention was implemented as part of a procedure to measure the effectiveness of a vaccine for treating periodontitis dental disease in canines, particularly by objectively measuring bone density loss of affected teeth over time. In the model procedure, a pathogenic organism was administered locally to the canine tooth roots and observations made over time of the effectiveness of a prior administered vaccine as described in commonly-owned, co-pending United States Patent Application Serial Number 10/851965 the whole contents and disclosure of which is incorporated by reference as if fully set forth herein. A model endodontic procedure was performed by a veterinarian who had received advanced training in endodontic techniques. An access to the canine root canal was created using a pear-shaped burr on a high-speed, pneumatic, water-cooled drill. One tooth was challenged at a time, with each tooth being thoroughly scrubbed with chlorhexidene solution, then covered with a fenestrated, sterile surgical drape. The surgeon used a new burr and a fresh pair of surgical gloves for each tooth. Every effort was made to maintain the sterility of the procedure. Once the root canal was accessed, root material was extirpated using a combination of barbed broaches and endodontic files. The root canal was then sterilized, regardless of treatment group, with a 10% bleach solution injected into the root canal. In order for the restoration (sometimes referred to as a "filling") to form an attachment with the enamel of the tooth, the surface enamel was etched with 40% sulfuric acid gel placed at the access opening for 20 seconds. The acid etching gel, as well as any residual bleach, was thoroughly rinsed from the root canal and tooth by copious volumes of sterile saline. The root canal was then dried completely by multiple sterile paper points being passed to the canal apex. When the root canal was completely prepared, either challenge material or sterile media (depending on the treatment group to which an animal was assigned) was instilled. Challenge material, grown in batches so that a fresh culture was available each day animals were challenged, was injected into the canal using a 27 gauge endodontic needle. The surface was dried, then the access was immediately closed using glass ionomer and light-cured epoxy with standard techniques. While this procedure and these materials may have had bacteriocidal activity, they assure the canal remaining closed for the entire observation period. Additionally, the high concentration of challenge organisms in the challenge material assures that some number of organisms would survive the filling process. Part of the impetus behind developing this technically difficult and time- consuming model is the subjective nature of other models of canine dental disease. Bone
changes are the hallmark clinical sign of periodontitis and only with completely objective measurements could such changes be measured accurately. Figure 3 illustrates the results of the bone reactivity analysis resulting from the inventive radiograph analysis technique for the example study of three groups T01 , T02 and T03, for each of the three time intervals, post challenge. As described herein, the first group T01 included canines subject to vaccination and challenged, group T01 included canines not subject to vaccination and challenged, and group T03 included canines not subject to vaccination and not challenged. In the result depicted in Figure 3, the animals reacted to the challenge procedure with increased mean bone reactivity scores. This means that the radiographs became darker, corresponding to a decrease in bone density. While the radiographs (not shown) indicated significant lesions in both the T01 and T02 groups, the bone in the non-vaccinated animals (group T02) was much less dense overall than the bone in the vaccinates. At nine weeks, mean bone reactivity scores of the non-vaccinated dogs in group T02 were significantly different than the vaccinated group T01 (p = 0.05). The present invention provides a way to measure radiographs completely and objectively. However, it should be understood that, while the invention provides an accurate objective measure of bone density loss over time, because it is a mean value of an entire region of bone, mixed changes are not completely reflected. That is, areas of bone which have lytic changes (i.e. become darker on radiograph) may be minimized or even nullified through this system by other areas which have sclerotic changes (i.e. become lighter on radiograph). Since changes in response to infection are usually mixed lytic/sclerotic reactions, changes must be overwhelmingly lytic to register as an increase in bone reactivity score. This aspect of the model should be viewed as an increase in the stringency applied to changes in this model. Because of the large number of animals used in the example study, any change in the mean values of a group must therefore represent an overwhelming biological change. The fact that the scores representing the animals in the non- vaccinated/challenged (e.g., T02) group became significantly more than the scores in the vaccinated/challenged (e.g., T01 ) group indicates that the bone in dogs in group T02 was much more lytic, as would be expected during nine weeks of an acute, ongoing infection. The minimal change from baseline in the vaccinated animals represents animals which were, on average, able to ameliorate bone change. The endodontic procedures applied in this model are acknowledged to have an impact on the bone surrounding the tooth, regardless of whether a pathogenic organism is infused into the root canal as part of the treatment. This is the reason behind the inclusion of groupT03 in the study design as a procedural control. This group had neither exposure to vaccine, nor exposure to the challenge material, so any change observed in the animals in this group can be concluded to be solely from the procedure. The vaccinated group closely mimics the changes seen in the procedural control group, which is further evidence that the
administered vaccine effectively mitigates the impact of the pathogenic strain of P. gulae introduced into the root canal. While the present invention has been particularly shown and described with respect to preferred embodiments thereof, it will be understood by those skilled in the art that the foregoing and other changes in forms and details may be made without departing from the spirit and scope of the present invention. It is therefore intended that the present invention not be limited to the exact forms and details described and illustrated, but fall within the spirit and scope of the appended claims.