In the background of medical equipm

In the background of medical equipment and image technology innovation, digital medical image data also presents a massive growth trend.in recent years,the rapid development of artificial intelligence is expected to help doctors reduce the workload,improve work efficiency and gradually move towards intelligent 메 디 컬 treatment. Cancer is the 리딩 cause of death in the world,among which liver cancer ranks the fourth. Early diagnosis and treatment is the most effective way to reduce the mortality of liver cancer However.It is a challenge to classify CT images of liver tumors have tfferent pathological changes, and there are many different images of the same disease and different diseases. It is a challenge to classify CT images of liver tumors.and also a research hotspot and difficulty in recent years. Due to the difficulty of data annotation in CT image of liver tumor.there are few effective high-quality annotation data. The early related literatures mainly 포커스 on The research of single-phase and two-dimensional typical layers,and the carefully designed manual 제비 mainly 포커스 on the value 그레이 and texture related aspects. In recent years,the improved model of bovw (bag of visual words) has been widely used in multivase CT images, and has good performance. Due to the high accuracy of medical image classification.the existing feature 스프 methods can not be directly used in clinical practice. Therefore, in this paper, the classification of liver tumors based on multi-phase 3D CT images is studiedfocusing on the key points of feature extraction and selection, as well as the selection of classification model.in order to improve the accuracy and efficiency of classification. the main work is as follows:1) a shallow multi fuature fusion for liver tumor CT image classification is proposed. The traditional extraction algorithm can not adapt to The characteristics of three-dimensionaland multi-phase liver tumor CT image. The design of manual 제비 often needs to be based on The clinical diagnosis experience of radiologists. In order to solve these',this paper focuses on the evolution mode of multi-phase three-dimensional CT images,and proposes a shallow multi fuature fusion tumor CT image classification. According to the bottom-up greedy strategy, this algorithm extracts the manual features in gray value.texture, shape and other aspects,and selects the effective 제비 for the combined 제비 according to Chi 스퀘어 셀 렉. the experimental results show that this feature fusion and-method significantly improves theclassification performance of the manual 제비, and the final classification accuracy can reach about 75%다.