| With the development of medical imaging, medical image has become a very important diagnosis and treatment technology. However, extensive use of digital medical equipment such as CT, MR, DSA, DR in clinical diagnosis and treatment, and the rapid spread of computer technology for healthcare, have make a lot of growth in medical image data, and efficient use of medical image resources has become a challenge. Hospitals collect hundreds of imaging data everyday, and automatic image annotation can be an important step when searching for images in huge databases. Automatic medical image annotation not only improves the medical image use efficiency, but also is the prerequisite for medical image processing and analysis.Image mining deals with the extraction of implicit knowledge, image data relationship, or other patterns not explicitly stored in the images, to meet the challenge of huge medical image data sets. Image classification is a basic research task in image mining, and mainly solves the problem of automatic image category annotation. In this thesis, methods of image classification are applied to medical image, and medical image classification algorithms based on multi-features are designed and implemented.The characteristics of medical image are analyzed in this thesis, and the feature extraction methods for medical image are discussed. In order to describe the characteristics of medical image more fully, different features are extracted in this thesis. Image features, such as gray-scale, texture features, shape features, and features extracted in the frequency domain are used. In addition, medical images are blocked to retain their spatial relations. An independent classification framework is proposed and applied to medical image classification task with medical image block and feature extraction methods. Experiments show that proposed method works possesses higher accuracy than some traditional methods for medical image classification. |
PDF Full Text Request