An automatic chromosome counting method based on depth learning

Abstract

The invention discloses a chromosome automatic counting method based on depth learning, which comprises the following steps: (1) image collection and preprocessing steps; (2) image classification andregression steps; (3) model training steps; (4) a test counting step, wherein a new sampling strategy is adopted in the step (2), and Faster R-CNN loss function model is improved. The data required bythe invention comes from G-banded chromosomes under a real microscope field of vision, and the method does not need a complex experiment process, has low cost and shorter time consumption, and can automatically and accurately complete the target chromosome counting. The invention uses 1000 examples of annotated chromosome map training model, and then uses 175 examples of annotated chromosome mapfor testing, statistics shows that 175 examples contain 8023 chromosomes, the accuracy of testing is 98.95%, recall rate is 98.67%. Test results show that the time required to complete a chromosome count report using the machine counting and manual correction method is less than 1/3 of the current one.

Description

technical field [0001] The invention relates to the technical fields of computer vision image processing, chromosome counting and the like, in particular to an automatic chromosome counting method. Background technique [0002] At present, karyotype analysis is mainly divided into four steps, namely, chromosome counting, chromosome segmentation, chromosome pairing, sorting correction, and analysis report. The above four steps of the karyotype automatic analysis system (such as Leica CytoVision system) widely used in the market are mainly Rely on manual operation. At present, it takes about 30-50 minutes for a well-trained doctor to complete a case report, and can only complete about ten cases a day, which is inefficient, and the counting step requires the mouse to click on each chromosome, which is easy to cause mouse hands. [0003] Chromosome counting is to count the number of chromosomes within the visible range of the field of view. It is an indispensable link in karyot...

AI Technical Summary

Problems solved by technology

Because the chromosome is long and narrow, but because of its non-rigid properties, it is prone to complex shapes such as bending; secondly, due to the influence of experimental factors, complex states such as adhesion, crossover, and overlap are prone to occur between multiple chromosomes that affect recognition; due to the chromosome system The imaging process is greatly affected by the environment, resulting in different resolutions, that is, the same chromosome may have different b

Images