Blob based segmentation for lung CT image to improving CAD performance

Abstract

Computer Aided Diagnosis (CAD) acts as a primary tool for the radiologists to have a second opinion for identifying whether the lung is affected by any abnormalities or not. Lung image segmentation and classification plays a vital role in CAD system. Despite many ongoing researches, lung image segmentation has still scope for improvement in terms of accuracy and automation. The proposed blob based segmentation aims to improve the segmentation of the lung image from chest CT in terms of sensitivity and accuracy. Blob based segmentation consists of three important processing stages, in preprocessing stage an automatic thresholding method is used to separate the lung image from background image; in second stage, segmentation of left and right lungs are carried out based on intensity value. Finally, Region of Interest (ROI) is identified from lung image and results are classified using a Neuro Fuzzy Classifier. On comparing with existing methods, the proposed method achieves good result in terms of accuracy and sensitivity.