Classification of Lung Diseases Using a Combination of Texture, Shape and Pixel Value by K-NN Classifier

Lung diseases are the disorder, issues that affect the lungs, the organs that permit us to breathe and it is the most frequent medical conditions worldwide especially in India. In this work, the problem of lung diseases like the difficulty encountered while classifying the disease in radiography can be solved. In this work, we propose Features Extraction Techniques for classification of Lung Computed Tomography Images. A Combination of Texture, Shape and Pixel Coefficient Feature are developed for Classifying the CT images of lung disease. The proposed system can classify lung images automatically as Normal Lung, Pleural Effusion, Emphysema and Bronchitis. The proposed System contains four steps. In the initial step, the images are pre-processed. In the second step, the images are segmented by Thresholding and Edge Detection. In the third step, the Texture, Shape and Pixel Coefficient Feature are calculated using the GLCM (Gray Level Co-occurrence Matrix), Moment Invariant and WHT (Walsh Hadamard Transform) and combined to form the single descriptor. In the final step, the K-NN, Multiclass-SVM and Decision Tree classifiers are used for classification of Lung images. The images are the CT scan images. The total datasets contain 400 images, 100 images of each disease like the Normal, Pleural Effusion, Emphysema and Bronchitis. The 280 images are used for Training and 120 images are used for Testing. The classification accuracy of folding method accomplished by the K-NN classifier with Global Thresholding is 97.50% for WHT +GLCM, 97.50% for WHT + MI, 94.45% for GLCM + MI, 97.50% for WHT +GLCM+MI. The K-NN classifier with Global Thresholding reduces the time and also gives better results as compared to other methods and classifiers.