A double chromosome image cutting method based on a Compact SegUnet self-learning model

Abstract

The invention discloses a double chromosome cutting method based on a Compact SegUnet self-learning model. The method comprises steps of by learning a double-chromosome overlapping data set generatedbased on real chromosome picture simulation, extracting high-dimensional features of different areas of the image by the model, predicting the probability that each pixel of the image belongs to the overlapping area and each chromosome according to the difference between the chromosome overlapping area and the non-overlapping area and the difference between different chromosomes, and finally selecting the classification with the maximum probability to complete the segmentation on the overlapping chromosome pixel level. Compared with a traditional manual observation distinguishing method, the method has the advantages that the efficiency is greatly improved, and the working time and cost are saved; Compared with an existing geometric segmentation method, the problem that partial overlappingis not easy to segment is solved, and practicability is high; Compared with an existing deep learning model, the segmentation accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a double chromosome image cutting method based on a Compact SegUnet self-learning model. Background technique [0002] With the development of electronic computer technology, the automatic image recognition and cutting and separation realized by software algorithm can be preliminarily realized, and the computer can also find out the frame area where the specified category is located in an image. In medicine, image analysis systems are generally used to detect and separate objects in medical images. However, when there is overlap between objects, the object segmentation effect at the pixel level is not ideal. [0003] Human genetic information is carried on chromosomes, so karyotype analysis is the basic method of cytogenetics research, an important means to study the relationship between chromosome shape and structure and its function, and to explore the rela...

AI Technical Summary

Problems solved by technology

This method requires that the chromosomes must be overlapped in a crossover manner, and it does not perform well for partially overlapping chromosomes. At the same time, the first step of the method requires manual marking, which is not fully automated.

[0004] For example, the document "Double Chromosome Cutting Method Based on U-net Deep Learning Model" R.L.Hu, J.Karnowski, R.Fadely, J.-P.Pommier, Image segmenta-tion to distinguish between overlapping human chromosomes, in: 2017Machine Learning for Health (NIPS), Long Beach, CA, 2017 revealed that the accuracy achieved in overlapping areas and two non-overlapping areas, through IoU (Intersection overUnion) The scores indicate that they are 94.7%, 88.2% and 94.4%, respectively, and in 1500 test images, through expert identification, the two-chromosome cutting method based on the U-net deep learning model cannot correctly segment overlapping and non-overlapping images. 225 pieces, far from meeting the actual needs

Images