B-mode Ultrasound Texture Recognition Algorithm of Liver Based on Random Forests

Abstract

With the advantages of non ionizing radiation, real-time imaging, multi-directional tomography and dynamic observation of blood flow, B-mode ultrasound has become the preferred method for imaging examination of some organs such as liver, gallbladder, pancreas and spleen. On the one hand, the B-ultrasound diagnostic doctor fixed his eyes on the screen of B- ultrasound monitor, on the other hand, he operated the probe to move on the patient's position to be examined. With dozens of patients are examined every day, so B-mode ultrasound doctors often work at full-load every day and lead to eye fatigue. Eye fatigue easily causes erroneous diagnosis or missed diagnosis. With the development of artificial intelligence technology and computer technology, more and more work done by human can be completed by computer instead. The computer will not feel fatigue for long working hours, and its analysis has objectivity and consistency. Therefore, computer-aided diagnosis is an urgent need in the field of B-mode ultrasound. Random forests algorithm is a machine learning algorithm based on decision tree, which can be used for classification. In this study, a B-mode ultrasound texture recognition algorithm for liver based on random forests is established. Compared with CART decision tree algorithm with sufficient samples, it is found that random forests is superior to CART decision tree in the accuracy of texture recognition, so random forests has a good application prospect in the analysis of B-mode ultrasound texture and computer aided diagnosis of diseases.