An Instance Segmentation Method of Crossed Chromosome Image Based on Chromosome Trisection Feature Point Location

Abstract

The invention discloses a cross-chromosome image instance segmentation method based on the positioning of chromosome trisection feature points, which divides an image containing chromosomes into S×S grid units, and according to the position of the trisection feature points of the chromosome instance, Chromosome instances are assigned to multiple grid units, each grid unit is responsible for predicting the instance class of an instance. Each output channel is responsible for predicting the instance mask map of the chromosome object to which the grid cell belongs. Compared with the instance location segmentation network which only uses the instance center point to complete the instance segmentation, the method proposed in the present invention uses the chromosome trisection feature points to complete the chromosome instance segmentation, and has a higher segmentation accuracy for cross-overlapped chromosomes, solving the problem of chromosome The segmentation effect of the image is still not ideal, and the small-volume chromosome is easy to be missed, which greatly improves the robustness of the chromosome instance segmentation algorithm and speeds up the segmentation speed of the chromosome instance.

Description

technical field [0001] The present application relates to the field of medical images, in particular to a method for segmenting instances of cross-chromosome images based on the location of chromosome trisection feature points. Background technique [0002] Currently, in a typical karyotype analysis process, it is necessary to perform image processing such as segmentation and classification on chromosome images. Some automatic chromosome analysis systems have been developed in China, but the analysis effect still needs to be improved when the chromosome images overlap and stick together. Traditional segmentation methods include threshold-based on the image histogram, based on morphological erosion and dilation, and based on morphological contour analysis. However, these methods have their limited scope of use, for the shape of the target chromosome and the pattern of overlapping regions. Due to the characteristics of small objects, complex shapes, many adhesions, and diffic...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a cross-chromosome image instance segmentation method based on chromosome trisection feature point positioning, the purpose is to solve the problem that the segmentation effect of chromosome image is still not ideal, and small-volume chromosomes are easy to be missed, and greatly improve the chromosome instance segmentation algorithm Robustness, which speeds up the segmentation of chromosome instances

Images