Cell tracking method based on local graph matching and convolutional neural network

Abstract

The invention discloses a cell tracking method based on local graph matching and a convolutional neural network. The method comprises the following steps: S1, segmenting a cell image by adopting a watershed method; S2, constructing and training a convolutional neural network, and extracting the depth similarity of the to-be-matched cell pair by using the trained convolutional neural network; S3, extracting the local triangular map similarity of the to-be-matched cell pair from the cell segmentation image; S4, establishing a similarity matrix by combining and extracting the depth similarity andthe local triangular map similarity of the to-be-matched cell pair, and taking the cell pair corresponding to the maximum value of the similarity matrix as a seed cell; and S5, starting from the obtained seed cells, sequentially matching the adjacent cell pairs by adopting a neighborhood cell diffusion method until all cells are matched. According to the method, the convolutional neural network is introduced to extract the depth similarity of the to-be-matched cell pair, and the cells are tracked by combining the depth similarity and the local triangular map similarity, so that the method hasthe characteristics of wide application range and high tracking accuracy.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a cell tracking method based on local graph matching and convolutional neural network. Background technique [0002] In the research of biomedicine, the tracking of honeycomb closely arranged cells (such as plant meristematic cells, oral epithelial cells) plays a vital role. Currently, many cell tracking methods use a local graph matching model to detect seed cells, and then perform neighborhood cell diffusion growth from the seed cells. [0003] The steps of the existing cell tracking method to automatically track cells are as follows: [0004] a. Cell image segmentation: use the watershed method to segment cell boundaries; [0005] b. Local graph features: In the corresponding local graphs including central cells and adjacent cells, extract angle features, area features, and distance features to form local graph features; [0006] c. Multiple groups of seed cells: u...

AI Technical Summary

Problems solved by technology

First of all, the manual extraction of local image features by the above method is insufficient in discrimination, and is not suitable for image sequences with large time intervals; second, the accuracy of multiple sets of seed cells extracted by the above method is insufficient, which in turn affects the tracking accuracy.

Images