Medical Image Classification Technology Based On The KAP Directed Graph Model

With the rapid development of computer science and technology and the large number of popularity of medical image acquisition devices,medical images have been widely used in the clinical field,which contains the CT images,MRI images,PET images,and other modes.They provide rich and valuable information for the doctor’s diagnosis and improve the efficiency and accuracy of medical diagnosis,which greatly promote the development of the medical field.At this stage,with the rapid growth of medical image data,how to classify medical image data efficiently and accurately and better assist doctors to diagnose become a hot and difficult research.However,due to the special nature of the field of medicine and the complexity of medical images,have increased the difficulty of medical image classification.Currently,the research on medical image classification technique has just started,there are still many critical issues that must be resolved,continue a in-depth study of medical image classification algorithm has a very important practical significance.In addition,how to show classification results to doctors intuitively and accurately is gradually being taken seriously.Visual information has become an effective mean to communicate and present information.Especially,with the rapid development of D3 visualization technology,it combines the graphics,text,color and other visual elements together,and allow doctors to perform interactive operations,make physicians more quickly access and understand the classification results.Aiming at the existing problems,this paper has studied medical image classification algorithms and implemented visualization of medical image classification,the main work is as follows:1)For the graph model is not accurate in the medical image modeling process,propose the KAP directed graph model(K Nearest Neighbor Texture Angular Points,KAP).KAP directed graph model uses texture features of brain CT image to extract corner as the vertex of the graph model,which ensures the representation of vertex in the KAP directed graph and reduces the complexity of the composition.Meanwhile,through establish directed edges in the KAP directed graph,fully consider the spatial structure relationship between the vertex and the vertex.Combine with knowledge of the medical field and assign to each vertex of thecorresponding movable range,so that the model can truly reflect the characteristics of brain CT images,and it’s more practical application value.2)For the medical image classification process,the similarity matching accuracy is not high between graph models,propose the multi-step matching method of the KAP directed graph model.Which comprises the following three steps: Firstly,put forward the coarse-grained matching algorithm between KAP directed graph models which is used to return the initial matching sequences.Secondly,put forward fine-grained matching algorithm between KAP directed graph models,this algorithm introduces QFV and HFV descriptor which is used to make a further match.Finally,put forward map optimization algorithms,including pseudo-homogeneous structure exclusion method and retrieve strategy.In order to achieve the accurate matching process between KAP directed graph models and obtain the same vertex and common sub-graph.3)Take comparative tests,and evaluate classifier in terms of classification time,accuracy and recall rate.Compared with existing medical image classification method which proves the time complexity and accuracy of the algorithm have achieved good results.Finally,design the visualization classification system of medical image to display the classification process and results.