Research And System Implement On Feature Extraction & Feature Selection For Pulmonary Nodules

At present,the mortality rate for lung cancer is higher than that for other kinds of cancers around the world. Conventional chest radiograms have been used to screen lung cancer and computed tomography is widely used to help detect lung cancer, because through this, it is possible to diagnosis correctly. However, this process is exhausting for there are so many images that should be interpreted. In addition, the error of reading could not be avoided. Therefore, computerized automated detectionor diagnosis systems for medical image could be used to help clinicians diagnose, treat, monitor changes, and plan and execute procedures more safely and effectively. Computer-Aided Diagnosis (CAD) techniques provide potential necessaries for the early detection and recognition of pulmonary diseases. By using the digital image processing and pattern recognition techniques, a CAD system has advantages for medical imaging assessment. On the one hand, it can help physicians to reduce the heave workload and improve the working efficiency. On the other hand, it can provide more objective information for physicians to make further diagnosis.In this paper, the medical signs and relative knowledge of pulmonary nodules are summarized. On this basis, the feature extraction method for ROI (Regions of Interest) is studied, and a novel heuristic algorithm is proposed for attribute reduction based on rough set theory. At last, the method of feature extraction and selection is used in a computer- aided detection (CAD) system for lung disease diagnosis.This paper is focus on the research of feature extraction and feature selection for the CAD. The main content is described as follows:First, the medical signs of nodules are named according to their complicated appearances and have their special meanings. And different signs indicate different state to diagnosis. So it is necessary to analyze the important medical signs.Second, a general scheme for nodules feature extraction is proposed after analyzing and the combining the medical signs of nodules in CT images and the expert knowledge. The intuitive features are not only considered also the potential feature.This scheme is analyzed and realized fro the follow aspects, included gray, shape, texture and spatial context, to quantitate the regions of interest (ROI) in CT images.Third, a novel heuristic algorithm is proposed for attribute reduction based on rough set theory. The attribute reduction algorithm based on the core of distinct matrix always has poor efficiency in both time and space. A novel heuristic algorithm is proposed for attribute reduction based on rough set theory. Distinct matrix is not constructed and stored directly and a better core set can be obtained even if the core element is not existent. Obtaining the matrix element and core element are carried out simultaneously and the frequency of attribute set is considered in this method. Experiment results show that by using the reduced attribute set, it not only can decrease the computational complexity but also keep high decision accuracy. The algorithm can find a good attribute subset.Fourth, based on Chongqing foundation project, support diagnosis system develop and design. The system is based on Borland C++ and VTK toolkit. And at present the system performs well especially for the original DICOM format medical images.