INTRODUCTION:

The Human tooth, is the first stage in the process of digestion. The main function of the teeth is to mechanically break down food by cutting and crushing it for swallowing, the rest process of digestion to continue. Basically, humans have four types of teeth: Incisors, canines, Premolars and molars each with a specific function. It is every individual’s duty to take very good care of these teeth. If these teeth are taken good care, the person will chew better which will eventually lead to good health.

Every living organism might be attacked by one or the other type of disease, so does the human teeth. The human teeth can be attacked by a number of diseases or disorders like: dental caries (cavities), gingivitis, periodontal disease etc. To overcome the above tooth disorders, regular care and cleansing should be done.

Once the tooth is decayed, proper treatment must be given. The dentist will be able to treat the tooth in the best way he/she can. Sometimes a patient wants to get a second opinion. Many times doctors do not have sound knowledge or misread the analysis. It tends to give the wrong diagnosis about the tooth to the patient. Nowadays doctors are using software to assist in the diagnosis. But at that time a patient might want to get a third party opinion about his/her tooth, whether it is in good condition or not. Hence for a better, efficient, more accurate classification and to know what exactly is wrong with the tooth, this system will come to the best of use.

The anticipated system takes the X-Ray image as an input and extracts the features to classify the input as sound tooth or a tooth that needs to be extracted. If the X-Ray image has more than one tooth, cropping of the required tooth is done. As the main concentration is on a single tooth, only that tooth must be considered.

Once the pre-processing is done, the relevant features are extracted using spatial and frequency domain. In the classification stage, the extracted features play an important role in deciding, if the tooth in the image is Normal or Abnormal. Normal meaning that there is nothing wrong with the tooth that has been processed and Abnormal means tooth has some problem. The Abnormality can be dental caries (cavities), gingivitis, periodontal disease etc. After the classification stage, the condition of the processed tooth is detected as Normal or Abnormal.

LITERATURE REVIEW:

A very few authors have worked on classification of radiolucency from the X-ray images. ¹states how important it is to classify dental carries from the diagnosis of the radiolucency. It specifies how image processing techniques used for threshold, fill holes, aluOp and count Pix operators to check the tooth X-ray image and specify the extent to which the caries are present and its type.

²Makes use of MATLAB to perform the detection of caries in three phases; the first phase is the preprocessing phase, where the rotation of the image is performed (if necessary) to understand the intensity of caries, a Histogram study of the image is carried out. The second phase is the image segmentation where the tooth is separated individually from the digital radiography; the third phase is the identification phase, where the caries is detected. To detect the caries the concepts of erosion and dilation are used. Erosion is used to erode away the region and dilate is used to enlarge the boundary region.

³Is about dental caries and the cysts of jaws that are the frequently occurring pathologies encountered in a dental practice. The work presents two distinct image processing algorithms for the detection of dental anomalies. The first part of the work is a novel approach for the detection of the dental caries using the hybridized negative transformation. The second part presents, statistical texture analysis for the dental images that consists of cysts along with dental caries. The texture analysis is done when the objects to be segmented based on texture content rather than intensities. The texture content of the panoramic image is characterized by Gray Level Co-occurrence Matrix (GLCM). The authors have extracted texture features using GLCM: energy, entropy, homogeneity, contrast and correlation. Segmenting the region around the cysts is done by texture features which find the texture boundaries.

⁴States that the dental condition was made by a two step image analysis, consisting of qualitative analysis and quantitative analysis of the dentin and the pulp. Active contour was used to segment dentin and pulp that simplified the image analysis process. With the help of the dentists visual inspection, the qualitative analysis was made. The quantitative analysis was made by computing various statistical parameters. The end result shows the variance and the intensity ratio between dentin and pulp is good enough to be the statistical parameter to differentiate the condition of the tooth.

⁵Focuses on the avoidance of the formation of the decay, by detecting plaque at its early stage. The concepts of Artificial Neural Networks (ANN) are used to detect the identification of plaque. The detection of plaque deals with four stages: Edge Detection, Segmentation, Feature Extraction and Classification. Radiographic dental images are taken as inputs. Median filter is used to remove the noise from these images and Edge Detection is done using the Canny Edge Detection. Clustering (Enhanced K-Mean) method for segmentation has been implemented. The method exhibits the lowest failure rate, when compared to the other methods.

⁶Discusses on how cavities can damage the tooth and if not taken good care of. The paper proposes a model which detects caries in X-Ray images using various image processing techniques that involve RGB to Gray conversion, Generation of Binary Images, finding the region of interest, removing the background, identifying regions, dividing image into multiple blocks and finally identifying the cavities present in the tooth via X-Ray images.

⁷Proposes a method that focuses on root canal edge detection of an X-Ray image. Pre-processing is done by using top hat bottom hat transformation along with the sharpening filter for edge enhancement. This combination gives both qualitative and quantitative assessment to dentists on the presence of cavities. Based on some morphological tools to grade severity of some metric values the caries are extracted.

⁸Presents a method to diagnose internal decay from radiographical images. Image segmentation is done using Kernel Fuzzy C-Means (KFCM) algorithm. The processed images are labeled as decay and then introduced into the cascade object detector for Diagnosis.

⁹Focuses on clinical tooth decay detection. Feature Extraction is done using Autocorrelation Coefficient and Gray-level Co-occurence Matrix (GLCM). The Support Vector Machine (SVM) and the Artificial Neural Networks of back propagation is is used on the Extracted Features separately. Comparison between SVM and the ANN is done.

¹⁰proposes two edge detection techniques to extract features from radiographs. To smoothen the images so as to highlight the defect, the Gaussian filter is used. Laplacian edge detection technique is used to to sharpen the image. This gives a slight discontinuity on the dental radiograph when compared to the original tooth.

DATA COLLECTION:

The proposed work has considered the data in the form of X-Ray images. The sets contain both normal as well as abnormal X-ray with the latter containing manifestations of radiolucency are taken. The classification uses the X-Ray images that are collected from the Department of Oral Medicine and Radiology KLE VK Institute of Dental Sciences, Belgaum, Karnataka, India.

The dataset contains 120 X-rays, where a number of normal and abnormal images (radiolucency). The set covers a wide range of abnormalities that include enamel, decay, filling.

Fig. 1: An X-Ray Image Data

The collected data of Images are cropped to a single tooth, as the system can classify only one tooth at a time. Cropping of the images is done manually using the Windows 10 Photos app. The cropped images are then sent back to the Radiology centre to understand the Defect/Radiolucency (Extraction) and the treatment that needs to be given for every tooth. The response was received as an excel sheet that contained the tooth number, the defect the tooth has (extraction / NAD).

METHODOLOGY:

Fig. 2: Flow of the System

PRE-PROCESSING

Pre-processing is done to remove the abnormalities in the X-Ray image and make it suitable for extracting the features. After the extraction of features, the numerical data has to be normalized as the data is spreads out in a large range. NaN (Not a Number) is the data that could not be normalized and hence the manual pre-processing (Normalization) is carried out. The NaN is normalized to zero (0) for all the extracted features.

Fig. 3: Cropped X-Ray Image of tooth Fig. 1

FEATURE EXTRACTION:

MATLAB is used to extract the features from the image. A total number of 23 features are extracted. These features are extracted from each tooth image. Spatial features extracted are: Intensity Mean Vector, Intensity Standard Deviation Vector, Number of Pixels, Sum of 1’s, Sum of 0’s, Mean of Binary, Column Standard Deviation, Row Standard Deviation, Diagonal Standard Deviation. Frequency based features are extracted using Fast Fourier Transformation and Discrete Cosine Transform. All these extracted features which are stored in an excel sheet. The classes considered are: Extraction and Sound Images.

These extracted features have normalized for classification. Normalization is done manually using the Weka Tool.

Algorithm:

Input: Single tooth X-Ray Image.

Output: Feature vector.

Method:

Step 1: Convert the RGB image to gray scale

Step 2: Resize the image to a standard size.

Step 3: Calculated total number of pixels in 5 different intensity ranges [0-50, 51-100, 101-150, 151-200, 201-255]

Step 4: Calculate mean and standard deviation of all the five intensity ranges.

Step 5: Convert the gray scale image into binary.

Step 6: Calculate the sum of black and white pixels.

Step 7: Calculate the standard deviation of the row and column of the binary image.

Step 8: Apply Fast Fourier Transform and calculate the Standard deviation.

Step 10: Apply DCT to the tooth image and calculate the Standard deviation.

CLASSIFICATION:

The classification is done using the Weka tool. The proposed system is evaluated on a dataset of pre-processed X-Ray images. Each X-Ray contains a single tooth. 60% of X-Ray images are considered as training and 40% as testing. The identification of the radiolucency is done using the following classifiers: Random Forest, Bagging and Random Committee, Sequential Minimal Optimization (SMO) and Multi-layered Perceptron. To test the robustness of the proposed method, k-fold cross validation was carried out with k=10.

Fig. 5: Table of results.

The above figure 4.3 shows the results of various classification methods for classifying tooth as Extraction and Sound. Multi-layered Perceptron, Random Forest, Bagging and Sequential Minimal Optimization (SMO) are used for experimenting our algorithm. Random Forest gives the highest accuracy rate of 82.3%. Bagging, Sequential Minimal Optimization (SMO) and Multi-layered Perceptron (MLP) gives an accuracy rate of 79.4118%, 76.4706% and 75.00% respectively.

ACKNOWLEDGEMENT:

The Department of Oral Medicine and Radiology KLE VK Institute of Dental Sciences, Belgaum, Karnataka, India, gratefully acknowledge us and gave their support. Dr. Vasanti Jirge, MDS, PGDHPE, Reader of the Department, Oral Medicine and Radiology, KLE VK Institute of Dental Sciences, Belgaum, Karnataka, India has extended her support to complete the anticipated work.

CONCLUSION:

The paper classifies X-Ray tooth image into extraction and sound teeth. This system is built to be an aid for Dentists, and also help the layman in getting his/her own analysis of their tooth, based on the tooth’s Radiolucency. Features are extracted using spatial domain of the image and using Random Forest method the encouraging results are obtained. The proposed system can be improved by increasing the size of the dataset and using hybrid classifiers. In future the classification of the tooth which requires Root Canal and Filling can be considered.

REFERENCES:

1. Ştefan Oprea, Costin Marinescu, Ioan Liţă, Mariana Jurianu, Daniel Alexandru Vişan, Ion Bogdan Cioc, “Image Processing Techniques used for Dental X-Ray Image Analysis.” Date of Conference: 7-11 May 2008, Published in: 2008 31st International Spring Seminar on Electronics Technology.

2. Mrs. Shubhangi Vinayk Tikhe, Mrs. Anjali Milind Naik, Dr. Sadashiv D. Bhide, Dr. T. Ssaravanan, Dr.K. P. Kaliyamurthie, “Algorithm to Identify Enamel Caries and Interproximal Caries Using Dental Digital Radiographs.” Date of Conference: 27-28 Feb. 2016, Published in: 2016 IEEE 6th International Conference on Advanced Computing (IACC).

3. Veena Divya.K., Dr. Anand Jatti, Dr. Revan Joshi, Dr. Deepu Krishna.S, “Characterization of Dental Pathologies using Digital Panoramic X-Ray Images based on Texture Analysis.” 2017 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC).

4. Anita Ayuningtiyas, Narendra Kurnia Putra, Suprijanto, Endang Juliastuti, Lusi Epsilawati,“Quantitative Image Analysis of Periapical Dental Radiography for Dental Condition Diagnosis.” Date of Conference: 7-8 Nov. 2013, Published in: 2013 3rd International Conference on Instrumentation, Communications, Information Technology and Biomedical Engineering (ICICI-BME).

5. P, Sudha, K. Thangavani, K. Visudha, S. Gnana Saravanan, “Dental Plaque Identification and Classification Using Artificial Neural Networks.”

6. Priyanca P. Gonsalves, “Diagnosis of Dental Cavities Using Image Processing.”, Date of Conference: 9-11 April 2009, Published in: 2009 ICME International Conference on Complex Medical Engineering.

7. Anupama Bhan, Garima Vyas, Sourav Mishra, Pulkit Pandey, “Detection and Grading Severity of Caries in Dental X-ray Images.”, Date of Conference: 22-23 Sept. 2016,Published in: 2016 International Conference on Micro-Electronics and Telecommunication Engineering (ICMETE).

8. Navid Khalili Dizaji, Tufan Kumbasar, “Diagnosing interdental decays in mouth radiography images using Kernel Fuzzy C means segmentation and cascade object detector.” Date of Conference: 30 Nov. - 2 Dec. 2017, Published in: 2017 10^th International Conference on Electrical and Electronics Engineering (ELECO).

9. Wei Li, Wei Kuang, Yun Li, Yu-Jing Li, Wei-Ping Ye, “Clinical X-Ray Image Based Tooth Decay Diagnosis using SVM.” Date of Conference: 19-22 Aug. 2007, Published in: 2007 International Conference on Machine Learning and Cybernetics.

10. Kanika Lakhani, Bhawna Minocha, Neeraj Gugnani, “Analysis of Edge Detection Technique for Feature Extraction in Dental Radiographs.”

Received on 10.12.2018 Modified on 18.01.2019

Research J. Pharm. and Tech. 2019; 12(5):2171-2174.

DOI: 10.5958/0974-360X.2019.00361.5