Emphysema medical image classification using fuzzy decision tree with fuzzy particle swarm optimization clustering

Abstract

Automated classification of medical images using machine learning methods has portrayed a vital role in the field of medical diagnosis. In the research work presented in this paper, the process of classifying the medical image is done in twofold, feature extraction and classification using fuzzy decision tree (FDT) with evolutionary clustering. The feature descriptors of the images are extracted using local diagonal extrema pattern (LDEP). The extracted features are passed to fuzzy particle swarm optimization (FPSO) clustering algorithm to obtain optimal fuzzy partition space for each attribute, which are then later used for inducing FDT. The proposed method of applying FPSO to develop input fuzzy space for Fuzzy ID3 is tested on emphysema CT image to classify the patient’s lung tissue into normal, centribulor emphysema, and paraseptal emphysema. From the results obtained, we observe that our proposed framework improves the classification accuracy of Fuzzy ID3 compared to the other frameworks considered.