Received: 27 November 2020 Revised: 25 March 2021 Accepted: 27 March 2021

DOI: 10.1002/cpe.6323

RESEARCH ARTICLE

Detection of periodontal bone loss in mandibular area from

dental panoramic radiograph using image
processing techniques

M. S. Antony Vigil1 V. Subbiah Bharathi2

1 Department of Computer Science and

Engineering, SRM Institute of Science and Abstract

Technology, Chennai, India Medical imaging plays the vital role in diagnosis of abnormalities. Periodontitis is a
2 Department of Electronics and
common chronic inflammatory disease damaging the soft tissue that untreated can
Communication Engineering, SRM Institute of
Science and Technology, Chennai, India lead to loss of bone, which supports the teeth. The severity of periodontitis is cor-
related to pocket depth in alveolar area. Even if analyses of the pocket depth can be
Correspondence
M. S. Antony Vigil, Department of Computer manually performed, an automatic assistive tool can drastically help radiologists con-
Science and Engineering, SRM Institute of
duct more accurate analyses. In the proposed research, image processing algorithms
Science and Technology, Ramapuram Campus,
Chennai, India. were developed to compute pocket depth and diagnose the periodontitis. Radiologists
Email: [email protected]
manually segmented 350 panoramic radiographic images, dividing them into normal
and periodontitis. The same dataset is used in our work to validate the Classification
algorithm. The images are preprocessed with median filter and histogram equalization
to improve the contrast and then segmented using two-dimensional-Otsu threshold-
ing method into teeth and bony area in the mandibular region. Normal pocket depth of
3 mm as reported by American Academy of Periodontology is equivalently converted
to pixel height in the radiographic images. Based on this pocket depth rule based clas-
sification method classifies the images into normal and periodontitis. The proposed
work achieved 91.34% accuracy, 92.8% sensitivity, and 95.47% F-score in classifying
the dental panoramic radiography.

KEYWORDS
2D-OTSU threshold, dental panoramic radiographs, histogram equalization, median filter,
morphological operation, periodontitis, rule based classification

1 INTRODUCTION

Radiological imaging in the medical field has been continuously evolving in diagnosing the bone diseases. Dental radiographs are used by the dentist
for diagnosing the oral disease like caries, tooth decay, impacted teeth, age estimation, etc. Dental radiographs are taken with lower radiations to
capture image of teeth and gums which helps the dentist to diagnose the cavities, tooth decay and other teeth abnormalities.1
There are several types of intraoral and extraoral dental radiographs which help to view the hard tissues of mouth.2 Extra oral radiographs
like panoramic and cephalometric focuses on the view of jaw and skull. Intraoral radiographs like bitewing, periapical and occlusal are used to
view the different aspects of the teeth. Bitewing radiograph depicts the crowns of lower and upper teeth. Periapical radiograph depicts the full
image of the teeth and helps in detecting the damages in root and bone areas. Occlusal radiograph shows the developmental stage of the teeth.
Panoramic radiographs are used for scanning the lower jaw and upper jaw whereas cephalometric captures the full structure of the skull and
the jaws.

Concurrency Computat Pract Exper. 2021;e6323. wileyonlinelibrary.com/journal/cpe © 2021 John Wiley & Sons, Ltd. 1 of 10
https://doi.org/10.1002/cpe.6323
2 of 10 VIGIL AND BHARATHI

Initially dental radiographs are subjected to noise, uneven exposure and low contrast. However for accurate diagnosis of a disease it has to be
processed by image processing techniques thereby improving its quality that can help with the segmentation procedure. The processing of dental
radiographs helps the dentist in detecting and diagnosing the oral diseases effectively. Techniques like preprocessing, enhancing, segmentation,
extraction of region of interest, feature detection and classification3 are used to implement the proposed work.
Periodontitis is one of the common dental inflammatory diseases of tooth supporting tissue which is most probably affecting the adults.4 It is
the gum disease caused by the infection of tissues holding the teeth. Alveolar bone loss is a periodontal disease which affects the soft tissue and
then to loss of bone supporting the teeth which further leads to loss of teeth. According to the task force report on the update of 1999 periodontal
disease classification by American Academy of Periodontology (AAP) the probing pocket depth of more than 3 mm is considered as periodontitis.
Bone density loss of radiographic images can be measured using various technologies. Digital radiographs overcome the limitations of conventional
radiographs5 and are effective in evaluating the bone loss of human body.
Panoramic Radiograph helps to evaluate the bone and teeth structure of the upper and lower Jaw structures. Digital imaging generates pixel
which helps for gathering detailed information about alveolar bone loss. The role of radiographs in diagnosing periodontal disease is loss of lamina
dura along the alveolar crest, widening of periodontal ligament of interdental crestal bone. Characteristics of bone loss are of vertical, horizontal,
generalized and localized.6
The panoramic radiograph plays a vital role in the identification of alveolar bone loss7 but they are lacking in the detection of bone disease due
to less quality of image. Diagnosing the normal and periodontal bone loss disease from the panoramic radiograph is not an easy task. The anatomical
structure of every tooth and bone area of human being is different.8 The structure of the bone and the density of the bone marrow differs from age
to age and gender to gender.
Diagnosing of alveolar bone loss in mandibular area can be done by enhancing the dental panoramic radiographs. The enhancement
is achieved by applying proper pre-processing and filtering algorithms.9 Segmentation and classification algorithms play the major role
in diagnosing the periodontal disease from the radiograph. Segmentation algorithm helps in dividing the image into separate sections
based on features or pixels. In dental radiograph, segmentation deals with the separation of areas based on the similar attributes
or features.10
There are many image processing methodologies or algorithms used for radiographic segmentation. Each segmentation process specializes in
a particular process. Morphological operations and filtering algorithms are used for preprocessing and enhancement of the input image.11 Otsu
threshold method removes the unwanted regions around the teeth and other intermediate regions which helps to classify the periodontal bone loss
in mandibular area images from normal images.

2 RELATED SURVEY

Digital image analysis in radiography helps the radiologists to better diagnosis of bony diseases. Advancement in image analysis and classification
of dental caries and bony diseases from dental radiography shows the new innovation in technologies. Research has been done to find the optimal
results of dental diseases using various image processing technologies. Advanced image processing techniques like image filtering, image enhance-
ment, morphological processing can improve the detection of periodontal diseases.12 For the classification of medical images as abnormal and
normal cases morphological segmentation plays the major role.13
Periodontitis is the disease caused due to improper personal habits and oral hygiene. Yellow colored soft thin layers are formed in surrounding
gums. It is necessary to clean the gum area if not it destroys the alveolar bone.14 Possibilities of periodontal disease can be predicted using machine
learning based on the habitual and clinical findings. Radiographs fails to disclose the actual structure of the dental anatomy since it provides the
two-dimensional view of three-dimensional structure. It shows less destruction than the actual loss. Early stage of periodontal disease known as
the condition of gingiva cannot be predicted from the radiograph. It do not provide the relationship of soft and hard tissue and thus depth of bone
loss detection is not accurate.15
Periodontal diseases can be caused due to the action of chemical mediators by inflammatory process in the surrounding structure of the teeth.
Bitewing and interproximal radiographs helps the dentist to identify the bony defects. But evaluation of alveolar bone height and alveolar bone
architecture is not exactly analyzed using these radiographs. Radiographs are helpful in the treatment planning decision of periodontal rather than
diagnosing the periodontal disease.16 Digitized radiographs can be analyzed using the digital subtraction method for diagnosing and measuring the
bone density.
Cone-Beam Computed Tomography (CBCT) can provide better solution for the bony area analysis. Three-dimensional anatomical struc-
ture can be viewed and accessed using the CBCT image which overcomes the issues of radiography.17 Three-dimensional radiographs of CBCT
give the exact measurement of normal bone height and abnormal bone level. Since the cost of CBCT is more and the availability of this tech-
nique is less in most of the clinics, radiographs are preferred to image analysis. Researchers use periapical radiographs to find the depth of
bone loss in alveolar region.18 Periodontal diseases are not estimated and the results are not accurate in periapical radiographs when compared
to CBCT.
VIGIL AND BHARATHI 3 of 10

Automatic Subtraction algorithm is applied to the Dental Radiography to detect the periodontal disease. Region of interest is calculated using
the summation method which helps to detect the minute changes in the radiography.19 Edge-based matching was done to analysis the structure of
teeth which gives the accurate result using automatic subtraction algorithm.
An image enhancement framework MedGA based on Genetic Algorithms was presented to improve the threshold selection for MR images.20
Particle Swarm optimized texture-based histogram equalization was proposed to enhance the MRI Image of brain.21 Computer-Aided Densitometric
Image Analysis and Digital Subtraction Radiography tools were used for periodontal diagnosis and management of surgically induced bone loss from
bitewing radiograph.22
Cementoenamel Junction location method is used to measure the bone loss of periodontitis which used cemetoenamel junction, alveolar
crest and apex of tooth root as the landmarks. This method estimates the horizontal bone loss in radiographs with the variation of 25% from the
ground truth.23
Faster regional Convolution Neural Network (faster R-CNN) model was used to train the images based on the ground truth annotated by
experts. This model achieved a sensitivity of 0.84, a specificity of 0.88 and an F-measure of 0.81 in detecting the periodontally comprised teeth
which reduces the diagnostic effort by automatic assistance.24
Deep Neural Transfer Network (DeNTNet) was applied to the panoramic dental radiographs to detect the periodontal bone loss which applies
Convolution Neural Network using transfer learning. A fully automated method for periodontal bone loss detection with numbering of tooth reduces
the workload of dental clinicians which achieves the average performance of 0.69 and F1 score of 0.75.25
In a similar work reported by Joachim Krois26 Deep Convolution Neural Network (CNN) have been employed to detect periodontal bone loss
with mean (SD) accuracy of 81%.
The diagnosis of Periodontitis stages are based on the severity of pocket depth with reference to root of the tooth.27 Even though image seg-
mentation and analysis algorithms are improved, exposure of radiation is the considerable factor.28 Artificial Intelligence-based dental diagnosis
helps the dentist in diagnosing the diseases which needs a lot of image samples to train the model.29

3 PROPOSED METHODOLOGY

The aim of the proposed research is to classify the dental panoramic radiographic image as normal image or periodontal image. Figure 1 shows the
block diagram of the proposed work.
Dental panoramic radiography images are used in this work from various subjects with different intensity and dimensions. The collected images
are categorized as normal and abnormal with the help of dentists. The images are preprocessed using histogram equalization and median filter to
remove noise and sharpened for further processing. Erosion morphological operation is performed to obtain the boundaries of the teeth and bony
area. Two-dimensional Otsu (2D Otsu) threshold segmentation is used to achieve the expected result.

3.1 Data acquisition

Dental panoramic radiograph shows the full view of the mouth than other radiographs. The study is about the detection of periodontal disease in
which it is difficult to locate the bone loss in mandibular region of alveolar bone from DPR because of different structure and intensity. The study
was carried out for 350 dental panoramic radiographs collected from subjects aged between 20 and 50.
The dataset used in our research are collected from SRM Dental College, Chennai, Tamil Nadu, India. More than 500 radiographs are collected
and 350 are used for our research that were segregated by dentists. The images are grouped as manually with the help of dentists as normal images
and periodontal images which helps the researchers to use the image for further processing.

3.2 Data preprocessing

Preprocessing is the initial step in image processing which aims to enhance the image by removing the noise and unwanted entities that makes the
image ready for further processing. Initially the collected images are converted to gray scale image which enables to do various processing. The
conversion of RGB to grayscale image can be done using the mathematical formula 0.2989 * R + 0.5870 * G + 0.1140 * B, where R, G, B represents
the intensities of Red, Green, and Blue colors.
Median filter noise removal algorithm is used to remove the noise30 of the dental panoramic radiography. Since the structure of each tooth
is different non-linear filtering method is proposed in this work.31 Median filter works by scanning the image through pixel and replace every
pixel value with the median of neighboring pixels. This filter uses a window of size 9×9 in which the target pixel is replaced with the median of the
window pixels.
4 of 10 VIGIL AND BHARATHI

FIGURE 1 Proposed block diagram for detection of periodontitis

The alveolar region of mandibular is considered as the region of interest and it is extracted from the image by fixing the specified dimensions
like height and width. Since the dimension of each image is different the region of interest is fixed by cropping the image region using the size of the
crop and position of the pixels. The initial threshold value is set to the filtered image to sharpen the edges and boundaries.
To make the edges and boundaries of teeth and bone area clear the image is enhanced by mapping the intensity values of grayscale images to
high contrast, sharpening and then undergone histogram equalization for better quality and contrast of tooth image.

3.3 Morphological operation

Morphological operation is the image processing technique that can be processed based on shapes and structures (32 ). It can be applied to grey scale
images in which the intensity and the pixel values are unknown. Radiograph of dental images has different teeth and bone shapes with unknown
intensity. Morphological technique uses a small shape known as structuring element of size 7 × 7 in which it is compared with the corresponding
neighboring pixels to test whether the structure is fits or hits the neighbor.
VIGIL AND BHARATHI 5 of 10

Dilation and Erosion are the most morphological operations used in digital images in which erosion deletes the pixels on the boundaries or edges
of the object while dilation adds pixels in the boundaries.33 In the proposed research of medical application erosion morphological operation is used
to locate the sharp boundary of the Dental anatomical structure.

3.4 2D-Otsu threshold segmentation

The process of partitioning the image into small partitions based on region of interest or features is known as Segmentation. The use of Image
Segmentation is to locate or identify the object and its boundaries. The processed grayscale images are converted to binary images for segmentation.
Better result of segmentation of radiographs can be achieved by threshold method. The selection of threshold value is the major and difficult task34
in the case of panoramic radiography. Otsu method is used to calculate threshold value automatically based on the intensity value of the radiographs
thus enables the researchers to locate objects of interest from the image.
The threshold value is used to locate the area of interest from its background which can be calculated when the variance of two classes is max-
imum. The obtained binary images are segmented images where object of interest is represented as white and background as black. The threshold
value generated using the one-dimensional Otsu method was used to segment the teeth area35 but segmenting the alveolar area of dental radio-
graphic images is not as expected. Extension of the threshold values calculation is needed to locate the exact area of interest. 2D Otsu’s thresholding
method is used in the proposed research for better segmentation of dental radiographs. In this segmentation method each pixels intensity value is
compared with the average intensity of neighborhood pixels.
2D Otsu threshold method is implemented in the proposed work based on the values of two-dimensional histogram values.36 The total number
of occurrences of two-dimensional bins can be denoted as f ij for a pair (i,j) and N be the total pixels in image N, then the probability mass function in
2D histogram:

Pij = fij ∕N.

Let p0 and p1 represent the probabilities and m0 , m1 , mT represents the intensity mean values and total mean vector of two classes.
The interclass discrete matrix is given by

∑
1
Sb = pk [(mk − mT )(mk − mT )T ].
k=0

Discrete matrix trace can be defined as T r (Sb )

Tr (Sb ) = (mTi p0 − mi )2 + (mTj p0 − mj )2 ∕p0 (1 − p0 ),

where

∑ ∑
s−1 t=1
mi = iPij .
i=0 j=0

∑ ∑
s−1 t=1
mj = jPij .
i=0 j=0

The value of threshold (s,t) is obtained by maximizing T r (Sb ).

Let the optimal threshold is (s,t) which can be obtained by maximizing T r (Sb ). The implementation steps for proposed algorithm is

Step1: Compute the probabilities and histogram of every intensity level of the input image.
Step 2:Set the value of s,t, and max as zero.
Step 3: Evaluate the value of T r (Sb ).
Step 4: Set max = T r (Sb ) if the threshold value reaches the maximum.
Step 5: Iterate the values of s and t till T r (Sb ) reaches the maximum.
Step 6: Update the maximum threshold value as T r (Sb ).

The threshold value obtained by the proposed 2D Otsu Segmentation algorithm helps the researchers for better segmentation result of dental
panoramic radiographs to detect the periodontal disease.
6 of 10 VIGIL AND BHARATHI

3.5 Rule-based classification

The segmented image contains the teeth and bony area as black and remaining area as white pixels. The segmented areas are assigned into labels
and undergo rule-based classification (RBC) algorithm as given below which classifies the abnormality from normal based on the pocket depth of
about 3 mm. Pocket depth is calculated based on the landmark of occlusion and height of each labels from the occlusion. Since the proposed work is
for Mandibular region, the areas below Occlusion is considered for Classification. Based on the abnormality level labeled by Dentist in many images
48 pixels from bottom point of occlusion is assigned as Normal pocket Depth which is equalized to the standard value 3 mm certified by AAP.
The proposed RBC Algorithm works as follows.
Step 1: Assign the segmented areas below occlusion as Labels.
Step 2: Set the value of Normal Pocket Depth (X).
Step 3: Calculate the pocket depth of each labels (Y i , i = 1,2, … n) using the properties like start_pixel, end pixel and height of each labels.
Step 3.1: Assign bottom_point of occlusion = Start_pixel of label, bottom_point of label = End_pixel of label.
Step 3.2: Calculate Height_label (Y i ) = End_pixel of label − Start_pixel of label.
Step 4: Compare the Value of Normal Pocket Depth(X) with the pocket depth of each labels(Y i ).
Step 5: If Y i > X, goto step 6 else go to step 4 till all labels are processed.
Step 6: Classify the image as Abnormal (Periodontitis).

The proposed RBC helps the researchers for better detection of bone loss in mandibular periodontal area.

4 EXPERIMENTAL RESULTS AND DISCUSSION

The proposed model was tested for 350 dental panoramic radiographs where 250 normal and 100 abnormal datasets which were collected from
different subjects. The collected images are manually segregated by dentists and tested with our proposed automatic model. The following figures
show the output of the input image in each steps of processing. As certified by American Academy of Periodontitis (AAP) the probing pocket depth of
more than 3 mm is diagnosed as periodontitis. The experiment was carried out for the mandibular area to detect the bone loss in periodontal region.
The first image is the sample of normal image and the second image is abnormal image which has bone loss in alveolar area of mandibular region.
Figure 2 shows the screenshots of segmentation and classification of periodontal disease. Figure 2(A)–(D) shows the input image (DPR), his-
togram plot of input image, image after histogram equalization and the plot after histogram equalization respectively. Figure 2(E),(F) shows the
region of interest in mandibular area where the researchers used for processing. Figure 2(G),(H) shows the preprocessed and enhanced images using
median filter and histogram equalization algorithms. 2D Otsu threshold segmentation is applied for this extracted image and Figure 2 (I),(J) shows
the segmented image of mandibular region. The segmented image consists of black and white pixels where the bone and teeth area of the image is
in black and the regions other than teeth and bone in oral cavity are in white.
Since the gingival pocket depth of less than 3 mm is considered as normal, the images are segmented by considering this factor. The proposed
rule based classifier combined with 2D Otsu thresholding and median filter classifies the images with 93.7% accuracy in detection of periodontitis.
Figure 2 (K),(L) shows the final output which classifies the image as normal and abnormal based on the proposed image processing approaches. The
proposed research work is implemented using the software MATLAB 2020a and we have proposed RBC which classifies the DPR by considering
the occlusion, cementoenamel junction and root apex as the landmarks for each tooth.
Table 1 shows the result of classification and performance of the proposed work which shows 93.7% images are classified correctly and 6.3%
are classified incorrectly where the normal images achieve 92.8% and abnormal images achieve 96% accuracy.
Figure 3 shows the Classification of images as Normal and Abnormal. 250 normal and 100 abnormal images are segregated by dentists and those
images are tested with the proposed model which classifies 232 normal and 96 abnormal correctly. Figure 4 shows the performance of the proposed
method as 92.8% normal and 96% abnormal images are classified correctly and 7.2% normal and 4% abnormal images are classified incorrectly.
As discussed in the literature survey in the similar work by Jaeyoung Kim et al., 800 radiographs were used and achieved the average perfor-
mance of 0.69 and F1 score of 0.75. In the work by Joachim Krios, 1750 radiographs were employed with the mean (SD) accuracy of 81%. In our
proposed work more than 500 dental panoramic radiographs are collected from SRM Dental College, Chennai, Tamil Nadu, Chennai and for the 350
images filtered and labeled by Dentist our method achieves the performance of 93.7%.
Table 2 shows the dice similarity coefficient and intersection over union using the manually segmented images and the images obtained
by the automatic segmentation approaches and Figure 5 shows the contours of a couple of examples of the segmented images (manually and
automatically) overlapping the images. Table 3 shows the confusion matrix which summarizes the performance of proposed model with the
ground truth.
VIGIL AND BHARATHI 7 of 10

F I G U R E 2 Segmentation
and classification of periodontal
disease. (A) Dental panoramic
radiograph; (B) Image after
histogram equalization; (C)
Histogram plot of input image; (D)
Histogram plot after histogram
equalization; (E) Normal image
ROI mandibular area; (F)
Abnormal image ROI mandibular
area; (G) Normal image after
enhancement; (H) Abnormal
image after enhancement; (I)
Normal image after
segmentation; (J) Abnormal image
after segmentation; (K)
Diagnosing as normal image; (L)
Diagnosing as abnormal image

TA B L E 1 Classification and performance of proposed model

Dataset Total number of images Classified correctly Classified incorrectly % Classified correctly % Classified incorrectly

Normal 250 232 18 92.8 7.2

Abnormal 100 96 4 96 4

Total 350 328 22 93.7 6.3

8 of 10 VIGIL AND BHARATHI

F I G U R E 3 Classification of images as normal and abnormal

(periodontitis)

F I G U R E 4 Performance of dental panoramic images classified

correctly and incorrectly

TA B L E 2 Dice similarity coefficient and

Model Intersection over union Dice similarity coefficient
intersection over union using the manually
Deep neural transfer network 67.94 80.91 segmented images and the images obtained by
the automatic segmentation approaches
Deep-CNN 81.55 89.84

Proposed method 90.63 95.08

FIGURE 5 Examples of the contours of the segmented images (manually and automatically) overlapping the images
VIGIL AND BHARATHI 9 of 10

TA B L E 3 Confusion matrix of proposed model Ground Truth by Dentist 

Actual by Proposed Model 
Dataset Normal Abnormal

Normal 232 18

Abnormal 4 96

The proposed automatic model developed using robust Image Processing Techniques achieves the accuracy of about 91.34%, sensitivity of
92.8% and F-score of 95.47% which is better compared to the existing methods. This automatic model helps the dentist as an assistive tool for
detecting the periodontal bone loss in mandibular area by reducing their workload and easy detection.

5 CONCLUSION

The research work proposes an automated model that uses median filter and histogram equalization for enhancement of the dental panoramic
radiograph. Fixing the region of interest and segmentation were the great challenge for the researchers due to the different size and orientation of
the images. Erosion morphological operation and 2D Otsu threshold-based segmentation are used along with the proposed RBC method gives good
accuracy of about 91.34%, sensitivity of 92.8% and F-score of 95.47%. Our automatic detection model can be used by Dentist as an assistive tool to
classify the dental panoramic radiograph as normal and abnormal based on periodontal bone loss in mandibular which reduces the diagnostic effort
and workload of the dentist. This work can be extended further to identify the periodontal compromised teeth which will fall eventually based on
the severity of the periodontal bone loss and detection of periodontitis in maxillary area.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

ORCID
M. S. Antony Vigil https://orcid.org/0000-0002-3362-4763

REFERENCES
1. Wang C-W, Huang CT, Lee JH, et al. A benchmark for comparison of dental radiography analysis algorithms. Med Image Anal. 2016;31:63-76.
2. Shah N et al. Recent advances in imaging technologies in dentistry. World J Radiol. 2014a;6(10):794-807.
3. Jamil U, Sajid A, Hussain M, Aldabbas O, Alam A, Shafiq MU. Melonoma Sementation using bio-medical image analysis for smarter mobile healthcare.
J Ambient Intell Humaniz Comput. 2019;10:4099–4120. https://doi.org/10.1007/s12652-019-01218-0.
4. Hamzah N, Aziz SA, Fauzi AR, et al. Effects of gelam honey (Melaleuca cajuputi) on alveolar bone loss in experimental periodontitis. J Dental Surg.
2014b;907970:1-8.
5. Bansal GJ. Digital radiography. A comparison with modern conventional imaging. Postgrad Med J. 2006;82:425-428.
6. Vijay G, Raghavan V. Radiology in periodontics. J Indian Acad Oral Med Radiol. 2013;25(1):24-29.
7. Hou G-L, Hung C-C, Yang Y-S, Shieh T-Y, Tsai C-C. Radiographic alveolar bone loss in untreated Taiwan Chinese subjects with adult periodontitis measured
by the digital scanning radiographic image analysis method. Dentomaxillofacial Radiol. 2003;32:104-108.
8. Hovorakova M, Lesot H, Peterka M, Peterkova R. Early development of the human dentition revisited. J Anat. 2018;233:135-145.
9. Rahmi-Fajrin H, Puspita S, Riyadi S, Sofiani E. Dental radiography image enhancement for treatment evaluation through digital image processing. J Clin
Exp Dent. 2018;10(7):629-663.
10. Poonsri A, Aimjirakul N. Teeth segmentation from dental X-ray image by template matching. Paper presented at: Proceedings of the Biomedical
Engineering International Conference. Laung Prabang, Laos; 2016; IEEE.
11. kumar A, Bhadauria HS, Kumar N. Segmentation of periapical dental X-ray images by applying morphological operations. Int J Eng Res Comput Sci Eng.
2018;5(2):371-375.
12. Yoon DC, Mol A, Benn DK, Benavides E. Digital radiographic image processing and analysis. Dent Clin N Am. 2018;62:341-359.
13. Saravanan S, Karthigaivel R, Magudeeswaran V. A brain tumor image segmentation technique in image processing using ICA-LDA algorithm with ARHE
model. J Ambient Intell Humaniz Comput. 2020. https://doi.org/10.1007/s12652-020-01875-6.
14. Lakshmi TK, Dheeba J. Digitalization in dental problem diagnosis, prediction and analysis: a machine learning perspective of periodontitis. Int J Recent
Technol Eng. 2020;8(5):67-74.
15. Prafiadi H, Putra NK. Image analysis for correlation between dental panoramic and MicroCT to measure bone density. Paper presented at: Proceed-
ings of the International Conference on Instrumentation, Communications, Information Technology, and Biomedical Engineering. Bandung Institute of
Technology, Indonesia; 2013:359-362.
16. Corbet EF, Ho DKL, Lai SML. Radiographs in periodontal disease diagnosis and management. Aust Dent J. 2009;54:s27-s43.
10 of 10 VIGIL AND BHARATHI

17. Dawood SP, Brown J. Cone beam CT in dental practice. Br Dent J. 2009;207:23-28.
18. Kripal K, Dileep A. Role of radiographic evolution: an aid to diagnose periodontal disease. Periodontal Disease - Diagnostic and Adjunctive Non-surgical
Consideration. London, UK: IntechOpen; 2019:1-17.
19. Ahn Y, Chae O. Automatic subtraction radiography algorithm for detection of periodontal disease in internet environment. Paper presented at:
Proceedings of the International Conference on Computational Science and Its Applications. Atlanda, GA; 2005:732-740.
20. Rundo L, Tangherloni A, Cazzaniga P, et al. A novel framework for MR image segmentation and quantification by using MedGA. Comput Methods Prog
Biomed. 2019;176:159-172.
21. Acharya UK, Kumar S. Particle swarm optimized texture based histogram equalization (PSOTHE) for MRI brain image enhancement. Optik.
2020;224:1–12.
22. Bragger U, Pasquali L, Rylander H, Carnes D, Kornman KS. Computer-assisted densitometric image analysis in periodontal radiography. a methodological
study. J Clin Periodontol. 1988;15(1):27-37.
23. Lin PL, Huang PY, Huang PW. Automatic methods for alveolar bone loss degree measurement in periodontitis periapical radiographs. Comput Methods
Prog Biomed. 2017;148:1-11.
24. Thanathornwong B, Suebnukarn S. Automatic detection of periodontal compromised teeth in digital panoramic radiographs using faster regional
convolutional neural networks. Imaging Sci Dent. 2020;50:169-174. https://doi.org/10.5624/isd.2020.50.2.169.
25. Kim J, Lee H-S, Song I-S, Jung K-H. DeNTNet: deep neural transfer network for the detection of periodontal bone loss using panoramic dental radiographs.
Sci Rep Nat Res. 2019;9:17615. https://doi.org/10.1038/s41598-019-53758-2.
26. Krois J, Ekert T, Meinhold L, et al. Deep learning for the radiographic detection of periodontal bone loss. Sci Rep Nat Res. 2019;9:8495. https://doi.org/10.
1038/s41598-019-44839-3.
27. Tonetti MS, Greenwell H, Kornman KS. staging and grading of periodontitis: framework and proposal of a new classification and case definition. J Clin
Periodontol. 2018;45(Suppl 20):S149-S161. https://doi.org/10.1111/jcpe.12945.
28. Hermann CD, Hyzy SL, Olivares-Navarrete R, et al. Craniosynostosis and resynostosis: models, imaging, and dental implications. J Dent Res. 2016;95:1-7.
https://doi.org/10.1177/0022034516643315.
29. Tohnak S, Mehnert AJH, Mahoney M, Crozier S. Synthesizing dental radiographs for human identification. J Dent Res. 2007;86(11):1057-1062.
30. Tao P, Pei Y, Celenk M, Fu Q, Wu A. Adaptive image enhancement method using contrast limitation based on multiple layers BOHE. J Ambient Intell Humaniz
Comput. 2020;11:5031–5043. https://doi.org/10.1007/s12652-020-01810-9.
31. Zhu Y, Cheng H, Zhu Y, Cheng H. An improved median filtering algorithm for image noise reduction. Phys Proc. 2012;25:609-616.
32. Rad AE, Rahim MSM, Norouzi A. Level set and morphological operation techniques in application of dental image segmentation. Int J Comput Inf Eng.
2014;8(4):182-185.
33. Na’am J, Harlan J, Madenda S, Wibowo EP. Identification of the proximal caries of dental X-ray image with multiple morphology gradient method. Int
J Adv Sci Eng Inf Technol. 2016;6(3):343-346.
34. Upadhyay P, Chhabra JK. Multilevel thresholding based image segmentation using new multistage hybrid optimization algorithm. J Ambient Intell Humaniz
Comput. 2020;12:1081–1098. https://doi.org/10.1007/s12652-020-02143-3.
35. Setianingrum AH, Rini AS, Hakiem N. Image segmentation using the Otsu method in dental X-rays. Paper presented at: Proceedings of the International
Conference on Informatics and Computing. Papua, Indonesia; 2017.
36. Zhang J, Hu J. Image segmentation based on 2D Otsu method with histogram analysis. Paper presented at: Proceedings of the International Conference
on Computer Science and Software Engineering. Wuhan, China; 2008:105-108

How to cite this article: Vigil MSA, Bharathi VS. Detection of periodontal bone loss in mandibular area from dental panoramic radiograph
using image processing techniques. Concurrency Computat Pract Exper. 2021;e6323. https://doi.org/10.1002/cpe.6323