Study On Computer-aided Diagnosis Of Liver Cancer

In recent years, the liver cancer incidences rate in China has exceeded 25 cases per million people. CT is one of the effective diagnostic tools to detect liver cancer. However, the diagnosis depends on the doctors'experience in medical image knowledge and disease diagnosis, and so the results lack of quantitative analysis and objectivity. A large number of medical image data bring heavy work burden to the doctor. The computer-aided medical diagnostic techniques to do the quantitative analysis of medical images and to provide a reference to the doctors is urgent. Liver cancer CAD based on CT images is one of hot issues and difficult problems in international study.The research of liver cancer CAD can reduce the rate of the misdiagnosis and missed diagnosis. It has an important study meaning and application value to improve the rate of clinical diagnosis of liver cancer.At present, needle examination is an effective method to diagnosis of liver cancer. However, this method will give greater damage to patients. Along with the computer, medical image processing and pattern recognition technology development, we can use pattern recognition technology to detect liver cancer by analysising the statistical features,texture features and chaos features of CT images. This algorithm extract the statistical features, texture features and chaos features from the CT images based on feature extraction technique, and the use the BPNN to detect the liver cancer, which can provide the necessary diagnostic aids to doctors. Among them, the statistical features include the rate of region of in interest (ROI) and the average gray level of the whole liver (contrast) and approximate entropy, texture features include angular second moment, entropy, moment of inertia and homogeneity, chaos features include the fractal dimension and Kc complexity.A large number of experiments have been carried out by using the above-mentioned methods and actual abdominal CT images.The recognition rate of liver cancer based on the texture features is 70%, and the statistical features'is 77%, and chaotic charateristics'is 80%.The experimental results show that chaos features can be a better expression of the characteristics of abdominal CT images and it reflects the chaotic degree of liver tissue. Chaos features are sensitive to a very slight change from the the liver tissue of patients, so it can be an important indicator to liver cancer diagnosis. After fusing the statistical features and chaotic features, and then input to the BPNN, we get the recognize rate which is 83%.The experimental show reveal the efficacy of the chaotic features and statistics features and the BPNN has better image recognition performance. Combined with the clinical experience of physicians, this proposed method can improve malignant diagnostic accuracy.