Classification of Lobar Pneumonia by Two Different Classifiers in Lung CT Images

摘要

Pneumonia is a lung disorder caused due to microbial infections. The study is an attempt to identify and classify pneumonia into several stages as mild, moderate and severe using Gray Level occurrence Matrix (GLCM) texture features. The study involves 30 subjects (10 mild, 10 moderate and 10 severe) suffering from mild, moderate and severe lobar pneumonia. The CT examination of lung was performed using Philips MX8000 IDT 16 slice CT scanner for all subjects and saved in DICOM format. The obtained images were subjected to median filtering for noise removal and normalisation was done using contrast stretching. Subsequently GLCM was applied to extract various texture features namely contrast, energy, maximum probability, variance, mean, skewness, entropy, standard deviation, autocorrelation, median, mode, cluster prominence, cluster shade, homogeneity and kurtosis. A neural network classification scheme was adopted to identify the various diseased groups. The results of post hoc test (tukey HSD) revealed that there exist statistically significant differences for the texture features namely contrast, energy, mean, standard deviation, autocorrelation and median at the level . The results of the study infer that the proposed method could be effectively used in diagnosis of pneumonia.