Breast Cancer Detection Using Local Binary Patterns ISTANBUL

Breast Cancer Detection Using Local Binary Patterns ISTANBUL UNIVERSITY DEPARTMANT OF COMPUTER ENGINEERING BURCU BEKTAS ESPACE VOCATION PARIS HAUSSMAN SAINT-LAZARE RUYA SAMLI 28. 12. 2017

Outline Introduction and Motivation Related Works Image Pre-Processing Feature Extractions Classification Conclusion and Suggestions 28/12/2017 Bektas, Samli 2

Introduction and Motivation Total Number and Percentage Distributions of the Most Common Cancer in Women Cancer Statistics 2014 Year of Turkey, Republic of Turkey, Ministry of Health 28/12/2017 Bektas, Samli 3

Introduction and Motivation CA: A Cancer Journal for Clinicians Volume 67, Issue 1, pages 7 -30, 5 JAN 2017 DOI: 10. 3322/caac. 21387 28/12/2017 Bektas, Samli 4

Related Works LBP FEATURES FOR BREAST CANCER DETECTION Pavel Kr´al, Ladislav Image pre-processing: Otsu method Feature Extraction: Local Binary Patterns Classification: SVM Hybrid Gabor based Local Binary Patterns Texture Features for classification of Breast Mammograms Amal Al. Qoud and M. Arfan Jaffar Image pre-processing: CLAHE Feature Extraction: Gabor Filter and LBP Classification: SVM 28/12/2017 Bektas, Samli 5

Data. Set I used The Mammographic Image Analysis Society (mini-MIAS) database, organized by J Suckling et al. in 1994. The team developed a database of digital mammograms. Films taken in the UK National Breast Screening Programme (NBSP). All images are 1024 x 1024 size. The database 1 consists of 322 digitized mammograms (among which it consist 202 normal and 120 abnormal images). It also includes radiologist’s markings on the locations of abnormalities if present. 1 http: //peipa. essex. ac. uk/info/mias. html 28/12/2017 Bektas, Samli 6

Data. Set 1 http: //peipa. essex. ac. uk/info/mias. html 28/12/2017 Bektas, Samli 7

Method 28/12/2017 Bektas, Samli 8

Image Pre-Processing The aim of pre-processing is to improve the quality of the image, improvement of the image data that suppresses undesired distortions and to enhance the image features for further processing 2. 2 (G. Hemalath, C. P, Sumathi, ¨Preprocessing Techniques of Facial Image with Median and Gabor Filters¨, International Conference On Information Communication And Embedded System, 2016). 28/12/2017 Bektas, Samli 9

Pre-Processing – Gaussian Filter Gaussian filtering is used to blur images and remove noise and detail 3 3 https: //www. cs. auckland. ac. nz/courses/compsci 373 s 1 c/Patrices. Lectures/Gaussian%20 Filtering_1 up. pdf 28/12/2017 Bektas, Samli 10

Pre-Processing – Binary Image was binarized with threshold value 0. 3 A binary image is a digital image that has only two possible values for each pixel 28/12/2017 Bektas, Samli 11

Pre-Processing – Desired Region Undesired background has been deleted and I got the region I wanted. To make the suspicious area segmentation is easier and hidden areas more visible, CLAHE (Contrast Limited Adaptive Histogram Equalization) is implemented. 28/12/2017 Bektas, Samli 12

• In pattern recognition and in image processing, feature extraction is a special form of dimensionality reduction. • The main goal of feature extraction is to obtain the most relevant information from the original data and represent that information in a lower dimensionality space 4. • There are many methods of Future Extracting Methods. • We used the Local Binary Pattern (LBP) method. 4. Gaurav Kumar, Pradeep Kumar Bhatia, ¨A Detailed Review of Feature Extraction in Image Processing Systems¨, Fourth International Conference on Advanced Computing & Communication Technologies, 2014. 28/12/2017 Bektas, Samli 13

LBP • Local Binary Patterns (LBP) is a method for texture feature extraction mostly used for recognition techniques. • The extracted features are useful for classifying breast cancer abnormality in mammograms 5 • We achived 59 attributes for each image and 2 class (normal and abnormal) 5 S. Naresh, S. Vani Kumari, ¨Breast Cancer Detection using Local Binary Patterns¨, International Journal of Computer Applications, 2015 28/12/2017 Bektas, Samli 14

Conclusion and Future Work By pre-processing methods we done noise reduction and image enhancement. By using Local Binary Pattern (LBP) we extracted the texture features. We used Navie Bayes method for classification 64. 9068 % - Accuracy 28/12/2017 Bektas, Samli 15

burcu. bektas@İstanbul. edu. tr ruyasamli@istanbul. edu. tr With love from Istanbul Thank you for your patience