Nuclear Location Method In H&E Image Based On Distance Estimation

A new method based on the distance estimation is proposed in this paper for locating the nuclei in hematoxylin and eosin(H&E)stains.The H&E image is firstly converted to the image of binary nuclei and then the pixel's distances in binary image is estimated to find the candidate location points of nuclei.To locate the nuclei,the synchronous and adjustable firing pulse coupled neural network(SAFPCNN)model and a new nucleus location method based on distance estimation is proposed for image binarization and seeking nuclei.1)SAFPCNN is a new kind of pulse coupled neural network(PCNN)model with the neurons of bilayer structure.The master part of neurons receives the stimulation signal and generate the pulse sequence of natural firing.In the Slave part,this natural pulse sequence and adjacent neurons' the stimulation is treated to produce the coupled pulse sequence.This would effectively reduce the miss segmentation.2)The nucleus location method based on distance estimation is a rapid,effective and practical methods.This method takes advantage of the objective reality that the center of gravity of a convex area is the furthest from the boundary of this region.Combined with the convolution estimation method of distance,we select the candidate center points of the region to get better marking results.According to the experimental results,a high accuracy and efficiency are achieved by this method in experiments,which the 95.26% in precision and 1.54 second in recognizing a thousand nuclei,the precision and accuracy basically meet the needs of the actual application.