Multi model analysis of mammogram for detection ppts .pptx

Good

1. Multi-model Analysis of Mammograms for Detection of Masses Vijaylaxmi K Kochari (5XC13SPZ19) Under the guidance of Prof. D.A. Kulkarni

2. Contents : Introduction Literature Survey Requirement Analysis & Specification Problem Definition Proposed System System Design Preprocessing Segmentation K Means Clustering Algorithm Feature Extraction Classification Snapshots Conclusion and Future work Summary References

3. Cells in tissues and organs in the human body continue to divide and proliferate. Pace of life has become more rapid. Breast cancer shows no symptoms. Painless masses start to develop. A mammogram is an x-ray of the breast cancer It can also be used to detect and diagnose breast disease in women Introduction :

4. K . Hu, X. Gao and F. Li “ Detection of Suspicious Lesions by Adaptive Thresholding Based on Multiresolution Analysis in Mammograms” R.Nithya , B.Santhi “ Mammogram Classification Using Maximum Difference Feature Selection Method” Literature Survey :

5. Hardware Requirement Minimum PIII machine Minimum 10 GB HDD space Minimum 1 GB RAM Software Requirement Operating system: Windows XP/ Windows 7 platform Coding language: MATLAB 2008 Requirement Analysis & Specification :

6. To design a system that will take less training time and accurately classifies the mammogram as either malignant or benign. Problem Definition :

7. Step 1: Select the mammogram. Step 2: Pre-process the mammogram to enhance it. Step 3: Segment the mammogram using K means algorithm. Step 4: Extract the features of segmented mammogram, store the extracted features for training . Step 5: Build the Support Vector Machine and Feed Forward Back Propagation Neural Network for training and classification of mammogram. Step 6: Once the mammogram is classified, the system shows the type of mammogram. Step 7: Compare the results of both the classifiers and get the best classifier. Step 8: End. Algorithm for the Proposed approach :

8. Fig 1 Block diagram for mammogram classification Proposed Syst em : Mammogram Image Pre-Processing Feature Extraction Classifier Normal Abnormal Segmentation

9. System Design: Database

10. Preprocessing : Pre-processing is necessary to improve the quality of image. Adaptive histogram equalization [1]

11. Preprocessing : Fig 1 Original Mammogram Fig 2 Mammogram after applying Adaptive Histogram Equalization

12. Segmentation : K Means Clustering algorithm is used to implement the segmentation of mammogram . A cluster is a collection of objects which are “similar” between them and are “dissimilar” to the objects belonging to other clusters. [10]

13. K Means Clustering   Step 1: Number of clusters must be known previously to be K Step 2: Select K number of cluster centers such that they are farthest apart from each other μ i =  some value  , i =1,...,k Step 3: Consider each pixel and assign it to the cluster to which it is closest c i ={j : d( x j ,μ i ) ≤ d( x j ,μ l ), l≠i , j=1,...,n} Step 4: Recalculate cluster centers by finding mean of pixels belonging to the same cluster μ i =1/|c i | ∑ j∈ci x j , ∀ i |c|= number of elements in c Step 5: Repeat step 3 and step 4 till shifting of cluster centers are observed. [10] K Means clustering Algorithm :

14. K Means Clustering   2 5 6 8 12 15 18 28 30 K Means clustering Algorithm :

15. K Means Clustering   2 5 6 8 12 15 18 28 30 C1 C2 C3 K Means clustering Algorithm :

16. K Means Clustering   2 5 6 8 12 15 18 28 30 C1 C1 C1 C2 C2 C2 C2 C3 C3 K Means clustering Algorithm :

17. K Means Clustering   2 5 6 8 12 15 18 28 30 C1 C1 C1 C2 C2 C2 C2 C3 C3 4.3(C1) 13.25(C2) 29(C3) K Means clustering Algorithm :

18. K Means Clustering   2 5 6 8 12 15 18 28 30 C1 C1 C1 C2 C2 C2 C2 C3 C3 4.3(C1) 13.25(C2) 29(C3) 2 5 6 8 12 15 18 28 30 C1 C2 C3 K Means clustering Algorithm :

19. Segmentation : Fig 4 Segmented Mammogram Fig 3 Mammogram after applying K Means Clustering

20. Feature Extraction: The transformation of an image into set of features is known as Feature Extraction. Features are fed as input to the classifier that assign them to the class that they represent

21. Classifiers: Two classifiers which are used are Support vector machine Feed-forward back propagation neural network

22. S upport Vector Machine: SVMs are  supervised learning  models Given a set of training examples, each marked for belonging to one of two categories, an SVM training algorithm builds a model that assigns new mammograms into one category or the other.

23. Feed-forward back propagation neural network

24. Snapshots: Figure 8.1 GUI of Multi-model Analysis of Mammograms for Detection of masses.

25. Snapshots: Figure 8.2 Representation of training GUI of SVM and FFBPNN.

26. Snapshots: Figure 8.2 Representation of training GUI of SVM and FFBPNN. Figure 8.9 Feature extraction and training time of SVM and FFBPNN.

27. Snapshots: Figure 8.11 GUI for testing of mammograms using SVM and FFBPNN.

28. Snapshots: Figure 8.12 Mammogram classified as abnormal using classifier SVM

29. Snapshots: Figure 8.13 Mammogram classified as abnormal using classifier FFBPNN

30. Approach which is based on FFBPNN got better results compared to SVM. FFBPNN training time is 2.23% less and accuracy is 10% more than SVM. In future accuracy can be increased by using other classification methods or by extracting other features. Conclusion and Future work :

31. References : [1] K Menaka , S Karpagavalli , “Mammogram Classification using Extreme Learning Machine and Genetic Programming,” International Conference on Computer Communication and Informatics ( ICCCI -2014), Jan. 03 – 05, 2014 [2] K. Hu, X. Gao and F. Li, “Detection of Suspicious Lesions by Adaptive Thresholding Based on Multiresolution Analysis in Mammograms,” IEEE Transactions on Instrumentation and Measurement, vol. 60, pp. 462–472, Feb. 2011. [3] R.Nithya , B.Santhi “Mammogram Classification Using Maximum Difference Feature Selection Method” Journal of Theoretical and Applied Information Technology , Vol. 33, No.2, 2011. [4] M Arfan Jaffar , Nawazish Naveed , Sultan Zia, Bilal Ahmed and Tae-Sun Choi , “ DCT Features Based Malignancy and Abnormality Type Detection Method For Mammograms ” International Journal of Innovative Computing, Information and Control Volume 7, No.9, September 2011 .

32. References : [5] S. Deepa , Dr V Subbiah Bharathi , “Textural Feature Extraction and Classification of Mammogram Images using CCCM and PNN”, IOSR Journal of Computer Engineering (IOSR-JCE), Vol .No. 10, Issue No.6, pages 07-13, June 2013 . [ 6] K. Thangavel , A. Kaja Mohideen , “Semi-Supervised K-Means Clustering for Outlier Detection in Mammogram Classification”, 2010 IEEE [7] S. Shanthi , V. Murali Bhaskaran ,“Computer aided detection and classification of mammogram using self-adaptive resource allocation network classifier ”, Proceedings of the International Conference on Pattern Recognition, Informatics and Medical Engineering , March 21-23, 2012 [ 8] http :// www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-bycancer-type/breast-cancer/mortality#heading-Three [9] http://www.breastcancerindia.net/statistics/stat_global.html [10] http://www.onmyphd.com/? p=k-means.clustering&ckattempt=147

33. References : [11] Leonardo de Oliveira Martins, Geraldo Braz Junior, Aristofanes Correa Silva,Anselmo Cardoso de Paiva, and Marcelo Gattass_, “Detection of Masses in Digital Mammograms using K-means and Support Vector Machine”, Electronic Letters on Computer Vision and Image Analysis 8(2):39-50, 2009 [12] http://peipa.essex.ac.uk/info/mias.html [13] http://homepages.inf.ed.ac.uk/rbf/HIPR2/fourier.htm [14] http://www.frank-dieterle.com/phd/2_7_1.html [15] T.Balakumaran Dr.ILA.Vennila C.Gowri Shankar “Detection of Microcalcification in Mammograms Using Wavelet Transform and Fuzzy Shell Clustering” ( IJCSIS) International Journal of Computer Science and Information Security, Vol. 7, No. 1, 2010 [ 16]Michael Barnathan “ Mammographic Segmentation Using WaveCluster ” Algorithms 2012, 5 ,318-329;doi:10.3390/a5030318, www.mdpi.com/journal/algorithms , ISSN 1999-4893

34. THANK YOU

35. The system takes mammogram, pre-processes it by applying Adaptive Histogram Equalization. The enhanced image is segmented using K Means Clustering algorithm. Statistical features such as mean and standard deviation of a segmented mammogram are extracted. These extracted features are fed as input to the classifier Support Vector Machine. DCT is applied on the segmented mammogram, these extracted features are fed as input to feed-forward back propagation neural network. These classifies the mammogram as benign or malignant. The training time and accuracy of both the classifiers are compared and concluded that FFBPNN is best classifier than SVM. Summary :

36. C onfusion Matrix :

37. Confusion Matrix :