A detection and identification method, equipment and storage medium for spores in gynecological microecology

Abstract

The invention relates to a detection and recognition method, device and storage medium for spores in gynecological microecology, comprising the following steps: inputting an image to be detected into an effective target AI detection model, detecting the corresponding target, and then using machine vision technology for a second Determine whether the target is a spore; the generation method of an effective target detection AI model includes: collecting multi-scene microscopic images; conventionally performing manual category labeling; expanding the labeling area of ​​epithelial cell nuclei by introducing hyperparameters; The effective training set with a small gap within the class is based on the backbone network architecture and combined with the deep learning target detection framework to input the effective training set into the convolutional neural network for training to obtain an effective target AI detection model. In the present invention, by introducing hyperparameters to expand the labeling area of ​​epithelial cell nuclei, an effective target detection AI model can be trained; and then the machine vision technology is used to determine whether the target is a spore, effectively reducing the false detection rate and missed detection rate of spores.

Description

technical field [0001] The invention relates to the field of intelligent detection and identification of spores in microscopic images of gynecological reproductive tract microecology, in particular to an intelligent detection and identification method, equipment and storage medium of spores in gynecological microecology. Background technique [0002] Under certain conditions, Candida invades the female lower genital tract and causes inflammation of the vulvar skin and mucous membranes. It is often called vulvovaginal pseudomycosis, also known as fungal vaginitis or fungal vaginitis, and spores are one of them. A common pathogenic fungus of the female genital tract. At present, the Gram staining method for biological specimens of female reproductive tract secretion is the most commonly used test method, and the morphological examination under the microscope is the gold standard for gynecological vaginal microecological diagnosis. It plays a crucial role in diagnosing female ...

AI Technical Summary

Problems solved by technology

For epithelial cell nuclei that are very similar to spores, the currently commonly used labeling rule is to mark according to the smallest circumscribed upright rectangle of the epithelial cell nucleus, but the areas of different epithelial cell nuclei are different in size and have a large gap, among which the area is small and the Gram stain is deeply stained The morphology of epithelial nuclei and spores is very similar. Therefore, in the prior art, marking epithelial nuclei and spores according to the conventional minimum circumscribed upright rectangle may easily cause large intra-class gaps between epithelial cell nuclei and small inter-class gaps between epithelial cell nuclei and spores. It is very good to train an effective target detection model, even if combined with traditional computer vision technology, it is easy to cause false detection of spores

In addition, for the exfoliated white blood cell nuclei and sperm, because the amount of such data is very small, it is difficult to collect a large amount of data. Therefore, in the case of insufficient data, it is impossible to use deep learning technology to perform a good comparison with the spores. distinguish

Images