The invention relates to a method for distinguishing differentiation grade of
liver cancer by
convolution neural network. The method comprises the following steps: step S1, preparing data: acquiring MPM images of
liver cancer through a multi-
photon microscope; 2, enlarging the MPM image
data set of the
liver cancer: adjusting the pixel of the original pixel size of 512x512 collected in the step S1, rotating the image horizontally, rotating vertically symmetrically,
cutting and the like, so as to obtain the image number of the
training set to be 16 times of the original image number; S3, designing a
convolution neural network to obtain the differentiation result of the differentiation grade: designing the structure of the
convolution neural network. The convolution neural network consists of eight
layers: the first layer to the fifth layer is called convolution layer, which is used to extract the detailed features from the image; layer 6 to layer 8 are three fully connected
layers, andafter layer 8, the probability of differentiation grade of liver
cancer image is output. The invention can effectively overcome the shortcomings of time-consuming,
missed diagnosis and strong subjectivity in the current clinical
biopsy technology.