Research On Automatic Recognition Of Gallbladder Ultrasound Image

Gallbladder is an important organ in human body.Gallbladder disease is a common clinical multiple disease,which greatly affects the physical and mental health of patients.With the rapid development of China's economy,people's living conditions and living standards continue to improve,the number of people suffering from gallbladder diseases is growing.Therefore,early diagnosis and treatment of gallbladder diseases is of great significance.Ultrasound imaging technology is characterized by simple operation,short examination time,real-time dynamic,safe,non-radiation and low examination cost.Therefore,it is mainly used in clinical diagnosis of gallbladder diseases.Using computer technology to recognize gallbladder image automatically can improve the scientificity and accuracy of disease diagnosis,reduce subjective factors and improve the efficiency of diagnosis.According to the characteristics of gallbladder B-mode ultrasound image,the selected region of interest(ROI)includes the sound and shadow areas formed by gallbladder and gallbladder.Then the selected ROI is extracted and optimized.The gallbladder B-mode ultrasound image is automatically recognized by support vector machine(SVM).The innovation and main work of this paper are as follows:Firstly.The gallbladder B-mode ultrasonography image database was established.In t the thesis,105 cases of gallbladder B-mode ultrasound images were selected from a large number of B-mode ultrasound images collected,and a gallbladder B-mode ultrasound image database was preliminarily established,which provided a certain quantity and quality samples for this experiment and the follow-up experimental study.Secondly.In the thesis,we proposed a the method of ROI for the first time,which includes the formation of gallbladder and gallbladder.The pretreatment methods of histogram equalization enhancement and median filter denoising are determined by experiments.Two shape features,five gray difference statistical features,five gray level co-occurrence matrix features,three spectrum analysis features and directional gradient histogram(HOG)features are extracted.The eigenvalues were normalized.Thirdly.In the thesis,we proposed to recognize gallbladder B-mode ultrasound images with Support Vector Machine(SVM).The RBF kernel function of support vector machine(SVM)classifier is optimized through experiments.The penalty parameter C and the kernel parameter gamma are optimized by grid search method.A SVM classifier based on the combination of gray level co-occurrence matrix and directional gradient histogram(HOG)is trained to recognizegallbladder B-mode ultrasound image.The recognition rate of normal gallbladder,The recognition rate of normal gallbladder by classifier was 100%,The recognition rate of gallstones by classifier was87.5%.The recognition rate of cholecystitis by classifier was 90.9%.This classifier has certain reference value for the diagnosis of gallbladder diseases.Fourth.A gallbladder B-mode ultrasound image recognition system was designed and developed.The system has the characteristics of simple operation interface and strong practicability,and can assist clinicians in the diagnosis of gallbladder diseases.