Research And Implementation Of Automatic Detection Algorithm In Tissue Pathology Image

Digital pathology represents one of the major evolutions modern medicines.Pathological examinations constitute the gold standard in many medical protocols,and also play a critical and legal role in the diagnosis process.With the development of computer information technology,the introduction of fast digital slide scanners provide whole slide images has led to a revival of interest in image analysis applications in pathology.The intelligent digital pathology image analysis using computer can objectively,high efficiency,high accuracy to assist doctors in the diagnosis of disease,it is also a trend of medical diagnosis digital analysis in the future.Pathological image automatic analysis mainly has been divided into three portions: the extraction of interesting area in image,feature extraction and classification learning.This paper will research and exploration around the above questions.The extraction of interest area is the key to how to realize the accurate image segmentation,which directly affect the subsequent image analysis.This paper introduced an improved watershed segmentation approach that works with hematoxylin and eosin(H&E)stained breast histopathology image,acquiring precise segmentation result of nucleus.The procedure can be split into three steps :(1)pre-processing with color deconvolution,morphological operators and fast radial symmetry transform,(2)marker-controlled watershed segmentation at multiple scales and with different markers,(3)merging of the results from multiple scales.The simulation results revealed that the proposed method has good performance in both detection and segmentation accuracy,and simultaneously,improves the over-segmentation of traditional algorithm.The extraction of feature is the core of accurate diagnosis,at which texture feature has been widely applied in the cell classification field.Image textures are inherent and complex visual patterns that reflect the information of gray level statistics,spatial distribution,synthetic structure and so on.This paper bring in a descriptor that encodes image microscopic configuration by a linear configuration model,so the final local configuration pattern(LCP)feature integrates both the microscopic feature represented by optimal model parameters,which is obtained using this descriptor,and local feature represented by pattern occurrences.The optical model parameters are estimate by an efficient least squares estimator.To achieve rotation invariance,which is a desired property for texture features,Fourier transform is applied to the estimated parameter vectors.Finally,the transforms vectors are concatenated with local pattern occurrences to construct LCPs.In order to realize the automatic diagnosis and analysis of pathological image,through the study of classification algorithms,to classify the extracted features and annotations,this paper utilizes extreme learning machine(ELM),which is an emerging learning algorithm for the generalized single hidden layer feedforward neural networks,of which the hidden node parameters are randomly generated and the output weights are analytically computed.Discussing the extreme learning machine theory,to analyze its origin,advantage and disadvantage,one of the open problems in ELM research is whether the number of hidden nodes can be further reduced without affecting learning effectiveness.This paper puts forward an improved extreme learning machine,in which hidden nodes are calculated by feedback residual error.The algorithm has universal approximation ability with theories analysis and experiments.In theory,this algorithm tends to reduce network output error to 0 at an extremely early stage.Furthermore,a relationship between the network output error and the network output weights in the improved ELM is shown.Simulation results demonstrate that the proposed method can be tens to hundreds of times faster than other incremental ELM algorithms.The consequence of test compared to others algorithm shows that high speed and accuracy of the new algorithm.