Automatic Computer-Aided Diagnosis of Liver Disease Based on Multi-Cascade and Multi-Featured Classifier

With the increasing availability of medical imaging and popularity of routine medical examination, more and more patients have liver disease. Currently, the diagnosis of liver disease relies heavily on doctor's rich clinical experience. However, it is very difficult to locate the lesions from hundreds of computed tomography images, and even more difficult to provide correct diagnosis. Thus, automatic diagnosis of liver disease with the aid of computer is highly promising. In this paper, we proposed an automatic computer-aided diagnosis method based on multi-cascade and multi-featured classifier. The automatic lesion extraction was used as data source of diagnose firstly in this method. The designed multi-cascade and multi-featured classifier makes accuracy rate of each cascade best for liver disease. With this method, Liver cyst, liver hemangioma and liver cancer can be diagnosed successfully from the original multi-phase computed tomography images. The accuracy rate of normal patient or abnormal patient reaches 99.49 percent; as to liver disease, the diagnostic accuracy can reaches more than 93 percent.