A NOVEL SHAPE BASED FEATURE EXTRACTION TECHNIQUE FOR DIAGNOSIS OF LUNG DISEASES USING EVOLUTIONARY APPROACH

• Published in: ICTACT journal on soft computing Vol. 4; no. 4; pp. 804 - 810
• Main Authors: C, Bhuvaneswari, P, Aruna, D, Loganathan
• Format: Journal Article
• Language: English
• Published: ICT Academy of Tamil Nadu 01.07.2014
• Subjects: Bayes Net; F-Measure; Feature Extraction; Multilayer Perceptron; Neural Networks; Sensitivity; Specificity
• ISSN: 0976-6561; 2229-6956
• DOI: 10.21917/ijsc.2014.0115

Lung diseases are one of the most common diseases that affect the human community worldwide. When the diseases are not diagnosed they may lead to serious problems and may even lead to transience. As an outcome to assist the medical community this study helps in detecting some of the lung diseases specifically bronchitis, pneumonia and normal lung images. In this paper, to detect the lung diseases feature extraction is done by the proposed shape based methods, feature selection through the genetics algorithm and the images are classified by the classifier such as MLP-NN, KNN, Bayes Net classifiers and their performances are listed and compared. The shape features are extracted and selected from the input CT images using the image processing techniques and fed to the classifier for categorization. A total of 300 lung CT images were used, out of which 240 are used for training and 60 images were used for testing. Experimental results show that MLP-NN has an accuracy of 86.75 % KNN Classifier has an accuracy of 85.2 % and Bayes net has an accuracy of 83.4% of classification accuracy. The sensitivity, specificity, F-measures, PPV values for the various classifiers are also computed. This concludes that the MLP-NN outperforms all other classifiers.