Recognition method of fluorescent sputum smear of Mycobacterium tuberculosis based on deep neural network

Abstract

The present invention relates to a method for identifying fluorescent sputum smears of Mycobacterium tuberculosis based on a deep neural network, comprising the following steps: Step S1, input digital bacillus image; image; step S3, matching the optimized image with the preset candidate bacilli; if the matching fails, then proceed to step S5; step S4, automatically identify the bacillus on the optimized image that matches successfully; step S5, output the Bacillus information for digital bacillus images. Its advantage is that it can quickly locate the suspected bacillus target in the bacillus image, effectively reduce redundant information, and reduce the workload of identification; effectively replace the method of manual statistics, assist the manual diagnosis process, improve the efficiency of manual diagnosis, and improve the accuracy of diagnosis; The deep neural network is combined through different structural processing to improve the training results and quickly identify the type of bacillus image.

Description

technical field [0001] The invention relates to the technical field of bacillus picture recognition, in particular to a method for recognizing fluorescent sputum smear of Mycobacterium tuberculosis based on a deep neural network. Background technique [0002] Tuberculosis is a zoonotic chronic infectious disease caused by Mycobacterium tuberculosis infection. Mycobacterium tuberculosis is a single-infection lethal bacterium. In recent years, with the increase of multiple nausea and tumor cases and the emergence of multi-drug resistant strains, it has greatly increased the difficulty of preventing Mycobacterium tuberculosis infection. Rapid diagnosis is a key link in controlling the transmission and spread of tuberculosis. [0003] When diagnosing pulmonary tuberculosis, doctors need to observe slices. Each sample needs 2-3 slices, and the inspection field of view of each slice needs to be greater than 100. Therefore, laboratory physicians need to spend a long time checkin...

AI Technical Summary

Problems solved by technology

[0005] However, there are some defects in this diagnostic process: For example, doctors need to observe all areas of the entire section, which is time-consuming and labor-intensive; when the positive rate is more than 2+, the number of bacilli in a single field of view may exceed 50, and it is difficult to record and count accurately manually; When the distribution is scattered, such as 0 to 9, it is easy to miss the judgment

[0007] Therefore, there is an urgent need for a Mycobacterium tuberculosis digital picture recognition algorithm that can effectively segment digital images, quickly identify and locate suspected bacteria, and improve diagnostic efficiency, but there is no report on this algorithm at present

Images