A model based segmentation with hybrid feature selection for computer aided diagnosis

Abstract

Detection of lung diseases and disorders is a challenging task in radiology and oncology. Computer Aided Diagnosis (CAD) systems read the inputs from a medical imaging modality such as Computer Tomography (CT) and Magnetic Resonance Imaging (MRI) scans and tends to highlight the suspicious or abnormal patterns which is otherwise referred as Region of Interest (ROI). The ROI is evaluated on the attributes like, shape, size, current newlinepattern, and the growth pattern. Segmentation in a medical image is complicated due to the factors like scanning environment conditions, frequent changes in the intensity of light beam from the scanners, and the noise variations. Segmentation of lung components becomes increasingly difficult due to minor difference between normal and pathological lung tissues. Hence, a method is required to extract the lung model with greater accuracy. A model-based segmentation method to extract the lung shape and corelating with a reference model was presented. The proposed model constructs healthy reference models and uses the shape features as a correlation metric against the input slices. Then numerical analysis indicated that the proposed segmentation approach achieved better results when compared with the other widely used techniques. Feature selection plays an important role in the classification of the datasets. Classification of input data can become a time consuming and newlinetedious task if, there multiple features available to be examined. The major challenge in feature selection is selecting the most important features from a set of abundant features that have a vital impact on classification newline newline