Research And Implementation On Liver Cancer Identification Methods Based On Texture Feature And ELM

Liver cancer is a serious malignant tumor.China is one of countries in the world with high incidence of liver cancer.Early diagnosis and treatment is the important measures to reduce mortality in patients with liver cancer.Liver tissue pathology image diagnosis is the gold standard for clinical diagnosis of liver cancer,traditional methods of liver cancer pathology image diagnosis is dependent on the doctor's experience,a large number of pathological image diagnosis to the doctor brought huge workload,it's easy to cause the doctor making fault diagnosis because fatigue,so using computer image recognition technology to assist in the diagnosis of liver cancer pathology image to improve the efficiency and accuracy of digital medical pathology have important research value.Based on the extensive reference at home and abroad about image feature extraction and classification of research results.I take further study of image classification method based on liver cancer cell image texture feature of the liver tissue pathology image,using fractal dimension and multifractal spectrum and more direction fractal dimension on wavelet domain as a texture characteristics,calculation method of fractal dimension was improved,so it can more accurately describe the image texture feature.In pathological image recognition,this paper studies the extreme machine learning algorithm,support vector machine algoritnm is studied at the same time,and through the contrast experiment of the two analyses the advantages and disadvantages of extreme learning machine.In order to solve the problem of the extreme learning machine that classification accuracy is not high on small sample.this paper proposes the KFCM method is combined with ELM algorithm.ELM based on differential evolution algorithm is proposed,using the global optimization ability of differential evolution algorithm,to select extreme learning machine network weights and threshold.and optimize the network model.In accordance with the above algorithm,this article has carried on the experiment,the experimental results show that,in this paper,the improved ELM method to improve the performance of classification.and verify the effectiveness of the extracted texture features.