Tuberculosis diagnosis method based on deep-learning

Abstract

A tuberculosis diagnosis method based on deep-learning of the present invention comprises: a step of acquiring an image from a sputum smear slide made for training by using image capturing equipment;a learning step of learning a deep-learning model to be used to determine tuberculosis by using the acquired image; a verification step of verifying accuracy in the deep-learning model learned throughthe learning step; a step of determining negative or positive for tuberculosis by using a weighted value of the deep-learning model learned through the learning step and the verification step; and adisplaying step of providing, to a user, a test result made through the tuberculosis determination step by displaying the test result on a monitor.

Description

technical field [0001] The present invention relates to a tuberculosis inspection method. In more detail, the present invention relates to a tuberculosis inspection method based on deep learning that can reduce inspection errors through the automation of sputum smear tuberculosis inspection. Background technique [0002] Although the standard of living and medical care has been improved, the number of tuberculosis patients continues to increase. According to the World Health Organization (WHO), the number of tuberculosis infections in 2013 was 9 million, of which 1.5 million South Korea also showed a trend of increasing tuberculosis patients and ranked first among OECD countries. [0003] Diagnosis methods for tuberculosis include sputum smear examination, sputum culture examination, drug sensitivity test, chest radiography and tuberculin skin test (Tuberculin Skin Test, TST). [0004] Among them, sputum smear examination and culture method are the most widely used, especia...

AI Technical Summary

Problems solved by technology

[0006] This kind of sputum smear test is widely used because the experimental operation is relatively simple and the result can be judged in a short time. However, the proficiency of the inspector will affect the accuracy of the test result. Therefore, whether there is tuberculosis and the treatment process encountered difficulties in predicting

[0007] Moreover, with respect to the microscope usually used for sputum smear examination, the examiner needs to place a glass slide (slide) on the stage every time an examination is performed, while changing the position and focus of the slide glass, It is troublesome to confirm the presence of tuberculosis one by one with the naked eye, and it is necessary to replace other slides after the inspection

[0008] In order to improve this problem, an automated scheme has been studied, which uses image processing to extract features and machine learning to learn the extracted features, but it is difficult to apply because it cannot exhibit sensitivity and specificity sufficient to replace actual inspections. in actual products or services

[0009] In addition, since the focus of the imaging device used for automatic inspection changes according to the position of the slide glass or the shape of the smeared sputum, it may be difficult to distinguish the form of tuberculosis even if the automatic imaging is performed. question

[0010] Furthermore, since this is a method of confirming nodules one by one with the naked eye, there is a problem of fatigue of the inspector's eyes

Images