Lymph Nodes Detecting For Stomach CT Images Based On Primal Sketch Sequence Tracking And Multi-view Strategy

Because of the convenience and efficiency of the CT image in the process of obtaining, CT image is widely used in medical diagnosis and treatment. Now days, the analysis and extraction of the disease information from CT sequence images are mainly according to the experience of doctors, which not only requires the medical professionalism, but also increases the amount of doctors' work. So with the development of computer technology, it is a need of using computer to process the CT image information in clinical application. In this paper, we proposed a new method for detecting lymph nodes in stomach CT images based on primal sketch sequence tracking and mult-view strategy. It has the advantages of fast processing and low false alarm rate. It converted the usual processing method based on pixel point information of a CT image into a processing method based on sketch lines of a primal sketch, which has the advantages of fast processing and low false alarm rate. This paper mainly completes the following works:(1) The pre process of the primal sketch of the single frame CT image. Due to the noise of the original CT image and other reasons, thus existing two sketch lines that enclose the same area, but are not connected in the primal sketch. Therefore we put forward a sketch line's connection method, and according to the size, position and shape characteristics of lymph nodes, we extracted these features based on sketch lines in order. The first one is length feature of sketch lines. The second one is CT image gray feature depended on sketch lines. The third one is multilateral gather shape feature of sketch lines. They are all mainly for distinguishing the sketch lines covering the large organ and other body organizations from the sketch lines surrounding lymph and blood vessel that is removing the sketch lines that are not likely enclosing the lymph nodes.(2) For the sketch line which maybe enclosed the suspected lymph nodes, In this paper, a method of extracting primary sketch's adaptive window is proposed according to the sketch line matching principle based on the primary sketch sequence. And we also proposed a method of extracting suspected lymph nodes' adaptive window in the original CT image which the position of the adaptive window locate in. The two methods are mainly in order to find the frame of the largest lymph node located in, and adaptive window for the region of suspected lymph nodes in the frame.(3) Based on the suspected lymph nodes' adaptive window, we extract the suspected lymph node sequence image blocks on the primary sketch sequence. Because of the difference of the characteristics of different lymph nodes' size, we use the different Lymph node detection methods for the different sequence image blocks. One method is based on multi-view strategy, which is used for the sequence image block whose size is more than11×11. The other one is based on Low rank decomposition and centroid tracking, which is used for the sequence image block whose size is not more than 11×11.The simulation results also show that the proposed method in this paper can reduce processing time and false alarm rate.