AN EFFICIENT DEEP LEARNING BASED

ORAL LESION DETECTING USING

Presented By
Xxxxxxx (962xxxxxxxxx)
Xxxxxxx (962xxxxxxxxx)
Xxxxxx (962xxxxxxxxx)
Xxxxxx (962xxxxxxxxx)
Final Year Biomedical Engineering
Udaya School of Engineering
 ABSTRACT
• Discovering oral cavity cancer (OCC) at an early stage is an effective way to
increase patient survival rate.
• However, current initial screening process is done manually and is expensive for
the average individual, especially in developing countries worldwide.
• This problem is further compounded due to the lack of specialists in such areas.
• Automating the initial screening process using deep learning techniques to detect
pre-cancerous lesions can prove to be an effective and inexpensive technique that
would allow patients to be triaged accordingly to receive appropriate clinical
management.
 ABSTRACT(Continued)
• We aimed to develop a rapid, non-invasive, cost-effective, and easy-to-use deep
learning approach for identifying oral cavity squamous cell carcinoma (OCSCC)
patients using photographic images.
• A novel strategy of clustering algorithm based on deep neural networks were used
to build automated systems, in which complex patterns were derived for tackling
this difficult task.
• Using the initial data gathered in this project, image classification will be done
based on RF classifier and the accuracy can be achieved greater than 85% for
identification of images.
 INTRODUCTION
• Oral cancer is one of the most common cancers worldwide and is characterized by late diagnosis,
high mortality rates and morbidity. GLOBOCAN estimated 354,864 new cases and 177,384 deaths
in 2018.
• Tobacco use, in any form, and excessive alcohol use are the major risk factors for oral cancer. A
factor most prominent in South and Southeast Asia is the chewing of betel quid which generally is
comprised of areca nut, slaked lime, betel leaf and may contain tobacco.
• Late diagnosis does not need to be a defining attribute as oral cancer is often preceded by visible oral
lesions termed as oral potentially malignant disorders (OPMDs) which can be detected during
routine screening by a clinical oral examination (COE) performed by a general dentist. If a
suspicious lesion is identified the patient is referred to a specialist for confirmation of diagnosis and
further management. Previous studies in India reveal screening has resulted in early diagnosis,
down-staging of the disease and reduction in mortality amongst individuals who use tobacco and
alcohol.
• This work is devoted to the detection of oral cancer.
 LITERATURE SURVEY 1
TITLE OF THE PAPER
Automated Diagnosis and Severity Measurement of Cyst in
Dental X-ray Images using Neural Network
AUTHOR, JOURNAL
Banumathi.A, Praylin Mallika.J , Raju.S, Abhai Kumar.V,
AND YEAR OF
Int.J. of Biomedical Soft Computing and Human Sciences, 2009.
PUBLICATION
METHOD Input Image: Dental X - Rays
Techniques: Neural Networks, Image Processing
Algorithms used: Contrast stretching, Radial Basis Function
ADVANTAGE
The severity of the cysts is calculated using circularity values.
Increases the diagnostic ease of dental surgeon

DISADVANTAGE
Threshold vlaue is set by prediction.
 LITERATURE SURVEY 2
Quantitative analysis of histopathological features of
TITLE OF THE PAPER precancerous lesion and condition using Image Processing
Techniques
Jadhav. A.S, S.Banerjee, P.K.Dutta, R.R. Paul, M. Pal, P. Banerjee, K.
AUTHOR, JOURNAL
Chaudhuri, J. Chatterjee
AND YEAR OF
Proc.of the IEEE Symposium on Computer-Based Medical Systems,
PUBLICATION 2006
Input Image: Histopathological OSF images
Techniques: Image Processing
METHOD
Algorithms used: Region Growing, Hybrid Segmentation
Algorithm
ADVANTAGE Misclassification rate were calculated for both the algorithms.
Requires Hybrid Segmentation method for segmentation of cancers in
DISADVANTAGE
OSF images.
 LITERATURE SURVEY 3
TITLE OF THE PAPER Performance Analysis of Artificial Neural Networks and
Statistical Methods in Classification of Oral and Breast
cancer stages
AUTHOR, JOURNAL
HariKumar.R, Vasanthi.N.S, Balasubramani.M
AND YEAR OF
Int. J. of Soft Computing and Engineering, 2012
PUBLICATION
METHOD Input Image: Dental X - Rays
Techniques: Neural Networks, Image Processing
Algorithms used: Contrast stretching, Radial Basis Function
ADVANTAGE
Classification accuracy is compared

DISADVANTAGE
Requires more number of input variables for classification
 LITERATURE SURVEY 4
Active Contour models: Application to oral Lesion detection
TITLE OF THE PAPER in color images‖, IEEE Conference on Systems, Man, and
Cybernetics
AUTHOR, JOURNAL
Ghassan Hamarneh, Artur Chodorowski, Tomas Gustavsson
AND YEAR OF
IEEE Conference in Systems, Man and Cybernetics, 2000
PUBLICATION
Input Image: True Color Images
METHOD Techniques: Image Processing
Algorithms used: Active Contour Model (Snakes)
Segmentation of oral lesion is obtained in single band images from
ADVANTAGE
true color images
To further automatize and improve segmentation, additional or
DISADVANTAGE enhanced energy terms and more human knowledge should be
incorporated
 LITERATURE SURVEY 5
A novel wavelet neural network based pathological stage
TITLE OF THE PAPER
detection technique for an oral precancerous condition
AUTHOR, JOURNAL Ranjan Rashmi Paul, Anirban Mukherjee, Pranab K. Dutta, Swapna
AND YEAR OF Banerjee, Mousumi, Pal, Jyotirmoy Chatterjee and Keya Chaudhuri,
PUBLICATION Journal of Clinical Pathology, 2005
Input Image: Histopathological Images, TEM images
Techniques: Wavelets, Neural Networks
METHOD
Algorithms used: Multi Layered Perceptron, Feed Forward Neural
Network
ADVANTAGE False detection rate is high
Unwanted features obtained resulting from noise that can mislead both
DISADVANTAGE
the training process and the decision making process.
 LITERATURE SURVEY 6
TITLE OF THE PAPER

AUTHOR, JOURNAL
AND YEAR OF
PUBLICATION
METHOD

ADVANTAGE

DISADVANTAGE
 LITERATURE SURVEY 7
TITLE OF THE PAPER

AUTHOR, JOURNAL
AND YEAR OF
PUBLICATION
METHOD

ADVANTAGE

DISADVANTAGE
 LITERATURE SURVEY 8
TITLE OF THE PAPER

AUTHOR, JOURNAL
AND YEAR OF
PUBLICATION
METHOD

ADVANTAGE

DISADVANTAGE
 LITERATURE SURVEY 9
TITLE OF THE PAPER

AUTHOR, JOURNAL
AND YEAR OF
PUBLICATION
METHOD

ADVANTAGE

DISADVANTAGE
 LITERATURE SURVEY 10
TITLE OF THE PAPER

AUTHOR, JOURNAL
AND YEAR OF
PUBLICATION
METHOD

ADVANTAGE

DISADVANTAGE
 PROPOSED SYSTEM
• The oral images are the main input images for the proposed work.
• The image preprocessing task are the prime work because the ROI of the proposed
work may be subjected into occlusion.
• The exact cropping or some other adjustment are necessary before the
classification implementation.
• The deep features are extracted from the ROI portion and the features are given as
a input to the Random forest classifier.
• The RF classifier can produce the output such as, No lesion, No referral needed,
refer for the other reasons, refer – low risk OPMD and Refer – High risk OPMD.
 PROPOSED SYSTEM
ADVANTAGES

• High predication rate

• Reduced false alarm

• High accuracy
 PROPOSED BLOCK DIAGRAM

Preprocessing ROI Deep features Random forest

1. No lesion
2. No referral needed
3. Refer for the other reasons
4. Refer – low risk OPMD
5. Refer – High risk OPMD.
 REFERENCES
1. Banumathi.A, Praylin Mallika.J , Raju.S, Abhai Kumar.V, 2009. Automated Diagnosis and
Severity Measurement of Cyst in Dental X-ray Images using Neural Network, Int.J. of Biomedical
Soft Computing and Human Sciences, 14(2): 103 – 108.
2. Jadhav. A.S, S.Banerjee, P.K.Dutta, R.R. Paul, M. Pal, P. Banerjee, K. Chaudhuri, J. Chatterjee,
2006. Quantitative analysis of histopathological features of precancerous lesion and condition
using Image Processing Techniques, Proc.of the IEEE Symposium on Computer-Based Medical
Systems.
3. HariKumar.R, Vasanthi.N.S, Balasubramani.M, 2012. Performance Analysis of Artificial Neural
Networks and Statistical Methods in Classification of Oral and Breast cancer stages. Int. J. of Soft
Computing and Engineering. 2(3): 263 – 269.
4. Ghassan Hamarneh, Artur Chodorowski, Tomas Gustavsson, 2000. Active Contour models:
Application to oral Lesion detection in color images‖, IEEE Conference on Systems, Man, and
Cybernetics. IEEE Conference in Systems, Man and Cybernetics, Nashville, TN , USA, 2458 –
2463.
5. Ranjan Rashmi Paul, Anirban Mukherjee, Pranab K. Dutta, Swapna Banerjee, Mousumi, Pal,
Jyotirmoy Chatterjee and Keya Chaudhuri, 2005. A novel wavelet neural network based
pathological stage detection technique for an oral precancerous condition, Journal of Clinical
Pathology, 58 (9): 932 –938.
 REFERENCES
6. Muthu Rama Krishnan.M, Chandran Chakraborthy, Ajoy Kumar Ray, “Wavelet based texture
classification of oral histopathological sections”, International Journal of Microscopy, Science,
Technology, Applications and Education, 897-906.
7. Simon Kent, 1996. Diagnosis of oral cancer using Genetic Programming – A Technical Report
CSTR-96-14 CNES-96-04.
8. Venkatakrishnan.S, Ramalingam.V, Palanivel.S, 2013. Classification of Oral Sub mucous
Fibrosis using SVM. Int.J. of Computer Applications, 78(3), 8-11.
9. Arushi Tetarbe, Tanupriya Choudhury,Teoh Teik Toe, Seema Rawat “Oral Cancer Detection
Using Data Mining tool”. IEEE 2017.
10. D.Padmini Pragna, Sahithi Dandu, Meenakzshi M, C. Jyotsna, Amudha J “Health Alert System
to Detect Oral Cancer” International Conference on Inventive Communication and
Computational Technologies(ICICCT 2017), IEEE 2017.
THANK YoU..