A Novel Feature Selection based Ensemble Decision Tree Classification Model for Predicting Severity Level of COPD Disease

In the current era, research on automated knowledge extraction from Chronic Obstructive Pulmonary Disease (COPD) images is growing rapidly. COPD becomes a highly prevalent disease that impacts both patients and healthcare system. In various medical applications, image classification algorithms are used to predict the disease severity that can help in early diagnosis and decision-making process. Also, for large scale and complex medical images, machine learning techniques are more efficient,accuracy and reliable. Traditional image classification models such as Naïve Bayesian, Neural Networks, SVM, Regression models. etc are used to classify the image using the annotated ROI and image texture features. These models are used as a diagnostic tool in analyzing the COPD and disease prediction. These models are not applicable to classify the COPD using the disease severity level. Also, the accuracy and false positive rate of existing classification models is still far from satisfactory, due to lack of feature extraction and noise handling methods. Therefore, developing an effective classification model for predicting the severity of the COPD using features derived from CT images is a challenge task.In this paper, an ensemble feature selection based classification model was developed, using images features extracted from COPD patients’ CT scan images, to classify disease into “Severity level ” and “Normal level” categories, representing their riskof suffering a COPD disease. We applied five different classifier methods and three state-of-the-art ensemble classifiers to the COPD dataset and validated their performance in terms of F-measure and false positive rate. We found that proposed feature selection based ensemble classifier (F-measure 0.957) had the highest average accuracy for COPD classification.