Interstitial Lung Disease Detection Using Template Matching Combined Sparse Coding

Abstract

Interstitial lung disease (ILD), representing a collection of disorders, is considered to be the deadliest one, which increases the mortality rate of humans. In this thesis, an automated scheme for detection and classification of ILD patterns is presented, which eliminates low inter-class feature variation and high intra-class feature variation in patterns, caused by translation and illumination variation effects. Region of interests (ROIs) of four lung tissue patterns (healthy, emphysema, ground glass and fibrosis) selected from a publicly available database are identified and utilized after pre-processing them. A novel and efficient approach for feature extraction is proposed using template matching combined sparse coding, which extracts features invariant to translation and illumination effects, from defined regions of interest (ROIs) within lung parenchyma. The system is tested by a K-nearest neighbor (Knn) classifier, exploiting the sparse coded features. Performance of the proposed scheme was evaluated using various measures such as recall, specificity accuracy, F-score and precision. The proposed scheme shows a better performance for the value K=100 of Knn classifier with an overall accuracy of 81.3%.