Computer Assisted Image Analysis Methods for Evaluation of Periodontal Wound Healing
Daniela Bazan PALIOTO
Soh SATO
Gillian RITMAN
Luiz Fernando MOTA
Raul G. CAFFESSE
Division of Periodontics, Houston Dental Branch, University of Texas, Houston, TX, USA
Correspondence: Dra. Daniela Bazan Palioto, R. José Maria Barroca, 372, 13170-330 Sumaré, SP, Brasil. Tel: +55-19-873-1506. e-mail: dpalioto@yahoo.com
Braz Dent J (2001) 12(3): 167-172 ISSN 0103-6440
INTRODUCTION | MATERIAL AND METHODS | RESULTS | DISCUSSION | ACKNOWLEDGMENT | RESUMO | REFERENCES
The aims of this study were to determine the accuracy of the computer assisted
image analysis method and to evaluate its application for the assessment
of periodontal wound healing in dogs. Histological material was analyzed
with an optic microscope connected to a CCD color camera which transmitted
the image to a frame grabber converting the light signals into pixels from
which the measurements could be obtained. Twenty sections were read to assess
the intra- and inter-examiner reproducibility for the parameters of area
filled by new tissue, area of epithelium, area of bone and linear measurements
of the cementum. The data were statistically analyzed using the t-test to
test the hypothesis that there was no difference between and within examiners.
No statistically significant differences were noted (with a confidence interval
of 95%) for any parameter when intra-examiner reproducibility was assessed.
Similar results were achieved for surface areas when the inter-examiner readings
were computed. However, values of linear measurements for cementum showed
statistically significant differences between recorders (p<0.05). Results
were consistently uniform and the method demonstrated high accuracy when
intra-examiner readings were evaluated.
Key Words: digital image, histometry, periodontal wound, accuracy.
INTRODUCTION
Histological measurements of soft and hard tissues are considered the gold
standard for the evaluation of periodontal wound healing. Morphometry of
tissue sections has been traditionally studied to gain a better understanding
of structure and function (1). Image analysis system evaluation allows the
investigator to quickly and reproducibly identify and measure areas of interest.
Furthermore, the image can be zoomed to allow closer examination of specific
details.
With the advent of inexpensive microprocessors, high-quality cameras and
large memory storage, image processing and statistical imaging analysis are
both practical and cost-effective (2,3). Advances in software development
make these techniques accessible and comprehensible to operators with varied
experience.
The use of guided tissue regeneration procedures has intensified the
need to determine the types and quantity of tissues formed in healing. Regeneration
is differentiated from new attachment in one basic, but very important manner:
alveolar bone formation with new inserted periodontal fibers in new cementum
is a prerequisite for the regeneration of periodontium (4-6). Areas of new
cementum deposition without adjacent bone, and vice-versa, and areas of ankylosis,
for example, do not satisfy the criteria for regeneration (7). Furthermore,
recent advances in the understanding of functions and mechanisms of action
of growth factors to regulate the healing process have provided evidence
that these proteins may serve as therapeutic agents to enhance the healing
of periodontal wounds (8-11). Sigurdsson et al. (12) found a significant
enhancement in periodontal regeneration using the rhBMP-2 in dog models.
They described limited root resorption and ankylosis generally limited immediately
apical to the CEJ. Lynch et al. (13,14) demonstrated a synergism when platelet-derived
growth factor and insulin-like growth factor were combined to enhance periodontal
wound healing. An image analyzing system will allow measurements to determine
how much of the original defect has been repaired and what type of new tissue
has formed.
Previous methods available to analyze periodontal wound healing were not
very accurate and were extremely time consuming. Measurements of epithelium
from a photograph could be accomplished by positioning a piece of string
along the contours, cutting the string, straightening and measuring it. These
methods were slow and tedious but could supply useful data. Similarly, grids
could be overlaid on a microscope field and the number of cells counted within
grid areas. With a simple image analysis system using a stylus and digitizing
table, interfaced with a computer, length and volumes can be obtained from
the x and y coordinates recorded as the feature is traced. The advantage
of the modern systems is that photographic rendition is unnecessary and image
analysis software can allow zooming, filtering, image enhancement, panning
and other techniques to be applied to the image.
Image processing is the manipulation of the digitized image to enhance and
evaluate information contained in the original image. In this way, it is
possible to enhance certain features in the digitized image not readily seen
in the original form and multiple fields may be processed. The field of the
digital image analysis is based on the assumption that images are two-dimensional
representations of objects that can be interpreted by analyzing an array
of discrete numerical units, or picture elements: pixels (1). These computer
manipulations enhance the visual qualities of the image as it appears in
the new form by varying the color range to enhance values of the individual
pixels but not the relative color values of adjacent pixels (3).
When analyzing the amount of staining in tissue sections, the image analysis
system must be able to mimic the many compensatory mechanisms of a trained
professional by simultaneously making allowances for process variables such
as section thickness variability, staining irregularities, irrelevant background
staining and poor representation of the lesion (3).
An additional advantage of the computer system is the easy statistical processing
of primary data (15). Data analysis and interpretation are the final stages
in the process that begins with a microscopic image and ends with an analytic
result data analysis that may have widely diverse objectives. Quantitative
measurements allow detection and documentation of the significance of very
small differences or very subtle change.
The aims of this study were to determine the accuracy of the computer assisted
image analysis method and to evaluate its application for the assessment
of periodontal wound healing.
MATERIAL AND METHODS
Fifteen purebred female beagle dogs, 2-3 years old, with no periodontal disease
were used in this study and maintained on a hard diet, except for the two
weeks after surgery, when a soft diet was administered. Standardized periodontal
defects were created in each of the left and right mandibular quadrants of
the 15 dogs. These defects were created around the 2nd, 3rd and 4th premolars
and 1st molar teeth under general anesthesia.
Sulcular incisions and elevation of buccal and lingual mucoperiosteal flaps
were made and alveolar bone was removed around these teeth with chisels and
water-cooled rotating burs. The defect involved the full circumference of
the teeth including the furcation area. The defect height from the cemento-enamel
junction to the bone margin was approximately 5 mm. The flaps were positioned
and sutured in an apical position allowing exposure of the surgically denuded
root surfaces to periodontitis-stimulating conditions for the subsequent
four months. After this period, the defects were subjected to reconstructive
surgery. Prior to the surgery, however, teeth in these animals were scaled
and polished. Plaque control was maintained by topical application of 0.12%
chlorhexidine gluconate (Peridex, Procter & Gamble, Cincinnati, OH, USA),
3 times weekly. Surgical treatment of the animal started 2 weeks after scaling
and root planing.
For the surgery, each dog was sedated with iv ketamine HCL (25 mg/ml) followed
by isofluorane gas anesthesia. The surgical area was locally infiltrated
with a 2% xylocaine solution containing epinephrine (1:50,000) to reduce
hemorrhage. Reference notches were placed at the bone level on the roots
and extended interproximally and into the furcation area as deep as the furcation
permitted.
The roots of the teeth in one of these quadrants received a combination of
1 µg each of recombinant PDGF-B and IGF-1 in methylcellulose gel (Sigma-Aldrich,
Chicago, IL, USA). A second quadrant received a placebo consisting of methylcellulose
gel only. Interproximal sutures were then placed through the flaps assuring
they covered 1/3 of the clinical crown of each tooth.
During the two weeks following surgery, all dogs were fed a soft diet and,
during the first week, tooth brushing was suspended in order to prevent unnecessary
disruption of the flap healing.
Seven days after surgery, the dogs were anesthetized with isofluorane gas
for rubber cup prophylaxis. In the following weeks, the surgical sites were
maintained by brushing with 0.12% chlorhexidine solution, every other day.
Three months after surgery, the animals were again anesthetized and sacrificed
by bleeding. The heads of the animals were perfused with 10% buffered formalin
solution and then refrigerated for 1 to 2 days. The jaws were dissected free
and placed in formalin for further fixation.
In order to enhance the speed of demineralization, the jaws were further
sectioned into tooth blocks. Demineralization was accomplished with 10% tri-fluoracetic
acid (TFA). Following demineralization, the tissue specimens were washed,
dehydrated, infiltrated and embedded in paraffin and then sectioned in 6-micron
intervals. Twelve to 20 nonserial histologic sections were made of each treated
tooth, cut in a mesio-distal direction, 30 microns apart. Sections were stained
with hematoxylin and eosin, Mallory’s trichrome or by silver impregnation.
Statistical evaluation was performed to assess the intra-examiner and inter-examiner
accuracy of the image analysis system. Twenty sections from different specimens
were read by one expert examiner without calibration before the measurements.
Two hours later the same examiner read all 20 sections again to evaluate
the intra-examiner reproducibility. The same 20 sections were then read by
the second examiner to assess the inter-examiner reproducibility. The data
were analyzed using the t-test to test the hypothesis of no difference between
and within examiners.
Technique
The input for the image analysis system was an Olympus BH-S optical microscope.
This system was set up for Kohler illumination and images were obtained with
1X and 4X objective lens.
A CCD color camera (Sony, Tokyo, Japan) was connected to the microscope and
the transmitted image was processed by a Color Frame Grabber Vision Plus
AT (Sony). The Frame Grabber is an imaging board that accepts a video signal,
performs analog to digital conversion and stores the image. This frame image
was transmitted to a microcomputer (Intel Pentium P90 computer system) that
was then visualized on a high resolution screen. The image was then processed
using Image Pro-Plus software (Media Cybernetics, Silver Spring, MD, USA).
Once the image was acquired, manipulation of contrast and colors was performed
by selecting the appropriate illumination and settings. The image was then
processed by thresholding and filtering. Thresholding is one of the simplest
algorithms and helps to select the area of interest from the background.
A pixel property was selected and a threshold set, all pixels below the threshold
were not counted, pixels above the threshold were marked. In the image analysis
system, often an interactive task, the operator observes the effect of the
threshold as it is applied and the threshold can be increased or decreased
to exclude or include more pixels.
Another type of thresholding operation is the erosion or dilation of the
pixels to separate or enlarge selected areas. This procedure can, for example,
separate two cells that touch, a process which can be interactive in that
single pixels may be added or subtracted until the desired separation or
blending is produced.
Various filters are available which help to increase the contrast between
the area of interest and the background. Background reduction and elimination
of uneven illumination can greatly enhance the image. The area of interest
may also be electronically defined by the user, thus eliminating extraneous
information.
Calibration of the system was performed with a stage slide micrometer that
assigns an actual size per pixel of 0.0203 mm. The measurements were traced
electronically using the notch as the reference point. In the absence of
the notch, the beginning of newly formed tissues detected by the examiners
was taken as a reference point. Measurements were taken in relation to the
total area filled by new tissues, area of epithelium, area of new bone and
linear measurements of the cementum were obtained by subtracting the area
without new cementum from the total extension of the furcation (Figure
1).
Data were automatically transferred from the image analysis system to a spread
sheet (Excel). This exchange was accomplished by Dynamic Data Exchange (DDE).
The Student t-test was used to test the hypothesis of no differences between
mean values for the following parameters: area filled by new tissue, area
of epithelium, area of new bone and linear measurements of the cementum,
in order to determine agreement or disagreement (non-variation or variation)
between and within examiners. A confidence interval of 95% was adopted.
RESULTS
The mean values for all parameters are reported in Tables 1 and
2. There
were no statistically significant differences between the intra- or inter-examiner
readings for the area filled by new tissue, area of epithelium or area of
new bone (Tables 1 and 2). However, there was a statistically significant
difference for inter-examiner reading for the linear measurement of the cementum,
but no difference for intra-examiner readings.
DISCUSSION
Healing of periodontal wounds is a very complex process, being the result
of the interaction of calcified and soft tissues: epithelium, connective
tissue, cementum and bone. Recent research findings have shown that biological
substances such as growth factors may enhance periodontal wound healing.
These findings lead us to ask two critical questions that should be answered
to assess the predictability of these therapies: the type of new tissues
formed and the amount of resolution of the original defect.
Ideal evaluation of histological sections should be based on quantitative
and qualitative observations. Computer assisted analysis methods permit accurate
evaluation of the tissue sections and allow the investigator to quickly and
reproducibly identify and measure areas of interest.
For the purpose of this study, areas of epithelium and connective tissue
attachment, as well as the amount of new bone and new cementum, were quantified
using areas and linear measurements. An intra-examiner and an inter-examiner
reproducibility test was performed to ensure the reliability of the measurements.
It was possible to identify and quantify each tissue type since this method
allows examination of specific areas by zooming. For example, if a root notch
was not clearly identified in the section, increased magnification of this
specific landmark allowed closer observation of the newly formed tissues.
Statistical analysis demonstrated no significant difference when the same
investigator read all the sections at different times for area filled by
new tissues, area of epithelium, area of bone and linear measurements of
the cementum (p>0.05). These results were consistently uniform and the
methodology demonstrated high accuracy.
Similar results were found when inter-examiner reproducibility was tested
for area filled by new tissues, area of epithelium, and area of new bone.
There was a statistically significant difference, however, between the readings
of the two examiners for the linear measurements of the cementum. It may
be that the two different examiners were not well calibrated to identify
the exact extension of the cementum particularly where areas of root resorption
or ankylosis could be seen intertrimed with cementum in some sections. Another
possibility is that the linear measurements could be more difficult to read
than the areas since only an inter-examiner statistically significant difference
for the linear measurements of the cementum was found. This hypothesis should
be evaluated further.
Other methods, such as the use of photographs scanned into a digital image
and highlighting of each tissue using different colors, are valuable devices
to assess the volume of new tissue formed (16). However, methodologies utilizing
photographs are very time consuming and do not allow the examiner to answer
the critical question as to the quality of the tissue formed.
The advent of sophisticated and affordable microprocessors, cameras and image
analysis of microscopic images in the medical field provides a means to quantify,
in a small way, the complex, natural image processing capabilities of the
human brain. Image processing is considered cost-effective, accurate, labor-saving,
and reproducible (3). In the past, as reported by Jarvis (17), a single color
image of 512 x 512 pixels at a brightness resolution of 8 bits or 256 levels
for each of the red, green and blue components, consumed 768 Kb which exceeded
the entire memory capacity of many microcomputer systems. Such disparity
in memory requirements, and the early design of video digitizers (requiring
triple hardware installation), inevitably limited the system configuration
for natural color display and the results that could be achieved with the
software.
Multicenter studies, largely based on the data generated by tracing measurements,
aim to evaluate the reproducibility and prognostic value of such assessments
(17). With digital image methods, double-blind or observer-blind multicenter
trials are feasible (17,18). By representing an image as a series of numbers,
the image can be stored permanently, simply by recording the numbers that
represent the image. Numbers do not fade, change colors, or become scratched
or damaged. Digital images can be transmitted to a distant site by sending
the series of numbers that represents the image via modem or computer network
(19).
Another advantage of this system is that this software can be used by operators
with only a limited knowledge of the background theories involved. However,
to appreciate and usefully implement the many applications of a image analyzer,
it helps to understand the distinct vocabulary, basic algorithmic tools and
limitations of image processing (3).
The development of a software that could clinically measure periodontal wound
healing would be of enormous value in the regeneration field.
ACKNOWLEDGMENT
We thank Dr. Walter Bretz for help with statistical analysis.
RESUMO
Palioto DB, Sato S, Ritman G, Mota LF, Caffessee RG. Imagem digitalizada
para avaliação de cicatrização periodontal. Braz
Dent J 2001;12(3):167-172.
O objetivo desse estudo foi determinar a precisão de um método
de imagem digitalizada e avaliar sua aplicação na cicatrização
periodontal. Material histológico de um estudo realizado em cães
foi analisado em microscópio óptico conectado a uma câmera
que adquire as imagens e transmite os sinais luminosos para um computador
na forma de pixels e dessa forma, permite a realização de medidas.
Vinte campos foram medidos para acessar a reprodutibilidade intra e inter
examinador para os parâmetros: área preenchida por novo tecido,
área de epitélio, área de osso e medidas lineares de
cemento. Os números foram analisados pelo test-t para testar a hipótese
de não diferença entre as medidas executadas pelo mesmo examinador
e entre examinadores diferentes. Nenhuma diferença estatística
significativa num intervalo de confiança de 95% foi encontrada quando
as medidas intra-examinadores foram analisadas (p>0,05). Resultados similares
foram encontrados para as medidas interexaminadores quando analisadas medidas
de área. Entretanto, houve diferença estatística significativa
quando as medidas lineares de cemento foram analisadas (p<0,05). Os resultados
foram consistentes e uniformes e o método demonstrou ser bastante
preciso quando medidas intra-examinadores foram avaliadas.
Unitermos: imagem digital, histometria, cicatrização periodontal,
precisão.
REFERENCES
1. Marchevsky A, Bartel P. Image analysis as a primer
for pathologists. New York: Raven Press; 1994.
2. Donat P. Image analysis in biology. Boca Raton:
CRC Press; 1992.
3. Wells WA, Rainer RO, Memoli VA. Basic principles
of image processing. Am J Clin Pathol 1992;98:493-501.
4. Bowers GM, Chadroff B, Carnevale R, Mellonig J,
Corio R, Emerson J, Stevens M, Romberg E. Histologic evaluation of
new attachment apparatus formation in humans. Part I. J Periodontol 1989;60:664-674.
5. Bowers GM, Chadroff B, Carnevale R, Mellonig J,
Corio R, Emerson J, Stevens M, Romberg E. Histologic evaluation of
new attachment apparatus formation in humans. Part II. J Periodontol 1989;60:675-682.
6. Bowers GM, Chadroff B, Carnevale R, Mellonig J,
Corio R, Emerson J, Stevens M, Romberg E. Histologic evaluation of new attachment
apparatus formation in humans. Part III. J Periodontol 1989;60:683-693.
7. Lynch SE. Methods for evaluation of regenerative
procedures. J Periodontol 1992;63:1085-1092.
8. Wirthlin MR. Growth substances: potential use
in periodontics. J West Soc Periodontol Periodontal Abstr 1989;37:101-125.
9. Dennison DK, Vallone DR, Pinero GJ, Rittman B,
Caffesse RG. Differential effect of TGF-beta 1 and PDGF on proliferation
of periodontal ligament cells and gingival fibroblasts. J Periodontol 1994;65:641-648.
10. Position Paper. The potential role of the growth and
differentiation factors in periodontal regeneration. J Periodontol 1996;67:545-553.
11. Howell TH, Martuscelli G, Oringer RJ. Polypeptide growth
factors for periodontal regeneration. Curr Opin Periodontol 1996;3:149-156.
12. Sigurdsson TJ, Lee MB, Kubota K, Turek TJ, Wozney JM,
Wikesjo UM. Periodontal repair in dogs: recombinant human bone morphogenic
protein-2 significantly enhances periodontal regeneration. J Periodontol
1995;66:131-138.
13. Lynch SE, de Castilla GR, Williams RC, Kiritsy CP,
Howell TH, Reddy MS, Antoniades HN. The effects of short term application
of a combination of platelet-derived and insulin-like growth factors on periodontal
wound healing. J Periodontol 1991;62:458-467.
14. Lynch SE, Williams RC, Polson AM, Howell TH, Reddy
MS, Zappa UE, Antoniades HN. A combination of platelet-derived and insulin-like
growth factors enhances periodontal regeneration. J Clin Periodontol 1989;16:545-548.
15. Fredriksson M, Zimmerman M, Martinsson T. Precision
of computerized measurement of marginal bone height from bite-wing radiographs.
Swed Dent J 1989;13:163-167.
16. Rajnay ZW, Butler JR, Lindsay RR, Vermino AR. The measurement
of molar furcation defect fill using digital computer technology report on
a new technique. Int J Periodontics Restorative Dent 1996;16:30-39.
17. Jarvis LR. The microcomputer and image analysis in
diagnostic pathology. Microsc Res Tech 1992;21:292-299.
18. Mekkes JR, Westerhof W. Image processing in the study
of wound healing. Clin Dermatol 1995;13:401-407.
19. Kenet RD. Digital imaging in dermatology. Clin Dermatol
1995;4:381-392.
Accepted June 18, 2001
