AI mycobacterium tuberculosis identification model training and detection method

Through the AI ​​M. tuberculosis identification model training method, a digital slide library was established and the AI ​​training module was used to solve the problem of low screening efficiency and accuracy of M. tuberculosis in the existing technology, and efficient and accurate detection results were achieved , and support remote diagnosis, benefiting remote areas.

CN118942643AInactive Publication Date: 2024-11-12JIANGSU COLLEGE OF NURSING
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410981289.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

When screening Mycobacterium tuberculosis, the sample size is large, time-consuming and labor-intensive, and it is prone to missed testing, resulting in low diagnostic efficiency and accuracy.

Method used

Using the AI ​​M. tuberculosis identification model training method, by establishing a digital slide library, cell annotation and AI training module, the Mycobacter tuberculosis identification model is created and detected, which improves detection efficiency and accuracy, reduces missed detection, and supports remote consultation.

Benefits of technology

It improves the detection efficiency and accuracy of Mycobacterium tuberculosis, reduces dependence on local testers, better benefit people in remote areas, and accelerates the establishment of a digital medical system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942643A_ABST
    Figure CN118942643A_ABST
Patent Text Reader

Abstract

The invention discloses an AI mycobacterium tuberculosis identification model training and detection method, and relates to the technical field of medical systems. The method comprises the following steps: S1, establishing a picture library; S2, scanning a tuberculosis sputum smear; S3, carrying out cell labeling and AI identification; S4, carrying out model training of mycobacterium tuberculosis identification; according to the method, tuberculosis sputum smears are collected, a digital slide library is established, cells in the digital slide library are labeled and trained, through an AI training module, establishment training and detection training, training of a mycobacterium tuberculosis recognition model is completed, meanwhile, the model is tested, the detection efficiency and the detection accuracy are improved, the situation of missing detection is avoided, and the detection accuracy of the mycobacterium tuberculosis recognition model is improved. Meanwhile, dynamic zero clearing of tuberculosis is promoted, remote consultation is supported, so that dependence on local inspectors is reduced, people in remote areas can be better benefited, and meanwhile, big data sharing accelerates establishment of a digital medical system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of medical systems, and in particular to an AI Mycobacterium tuberculosis recognition model training and detection method. Background Art

[0002] Mycobacterium tuberculosis, commonly known as tuberculosis bacillus, is the pathogen that causes tuberculosis. The bacteria can invade various organs throughout the body, but most commonly cause pulmonary tuberculosis. Tuberculosis is an ancient disease that is widely distributed around the world and is the leading cause of death among bacterial infectious diseases.

[0003] Mycobacterium tuberculosis is an aerobic bacterium that grows slowly, but it is also very resistant to dryness, cold, acid, and alkali. Under direct sunlight, it takes 2 to 7 hours for the tuberculosis bacilli in sputum to be killed.

[0004] Tuberculosis diagnosis methods generally include chest imaging examination, bacteriological sputum smear examination, molecular biology examination, tuberculosis pathology examination, immunological examination and bronchoscopy. Almost all patients infected with Mycobacterium tuberculosis will undergo sputum smear examination;

[0005] However, the existing technology is time-consuming and labor-intensive when screening Mycobacterium tuberculosis. At the same time, each sputum smear has about 5,000 fields of view, and it takes more than 5 minutes for a microscopist to observe 300 fields of view. In a single working day, the microscopist should read no more than 25 slides, which is time-consuming and labor-intensive, and is prone to missed detection.

[0006] In response to the above problems, the inventors proposed an AI Mycobacterium tuberculosis recognition model training and detection method to solve the above problems. Summary of the invention

[0007] In order to solve the above problems, the purpose of the present invention is to provide an AI Mycobacterium tuberculosis recognition model training and detection method.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solution: an AI Mycobacterium tuberculosis recognition model training method, comprising the following steps:

[0009] S1. Create a picture library

[0010] Collect tuberculosis sputum smears from the clinic, take out the slide rack in the slide scanning imaging system, place the collected tuberculosis sputum smears on the slide, then load the slide, make sure the slide is flat, put it into the slide rack, and close the door;

[0011] S2. Scanning of tuberculosis sputum smear

[0012] Scan the slides one by one through the slide scanning imaging system to observe whether the cell images of tuberculosis sputum smears are clear, scan the tuberculosis sputum smears, name, view and share them, and establish a digital slide library;

[0013] S3, cell labeling and AI recognition

[0014] Check cell information through the digital slide library, select the currently labeled cell category through the AI ​​application set module, add the identification area in the morphology list module, and the slide scanning imaging system automatically counts the number of nucleated cells and preliminarily identifies the cells, then modifies and classifies the cell identifiers, and then reviews the cells in the cell standard library. After the review is passed, the cell labeling and training are completed;

[0015] S4. Model training for Mycobacterium tuberculosis recognition

[0016] The slide is inspected through the scanning viewer, and then the Mycobacterium tuberculosis classification in the AI ​​application set module is used to select Mycobacterium tuberculosis for a screening, and then the morphology list in the morphology list module is used to enter the specimen library, and then the detection library is used for a second screening, and then all areas are reviewed, and then the training is created and tested through the AI ​​training module.

[0017] The present invention also provides a detection method for an AI Mycobacterium tuberculosis recognition model, comprising the following steps:

[0018] S1. Model testing

[0019] Create a training module through the AI ​​training module in the scan viewer, select detection training, then search for tuberculosis bacilli in the labeled data, then select the corresponding image in the training data, and then train;

[0020] S2. Calling model

[0021] Change the training name to be consistent with the slide, wait for the training to finish, and observe the acc value and loss value curves;

[0022] S3. Scan the tuberculosis sputum smear to be tested

[0023] Load the slide on the slide rack, scan the slide by scanning the sample, form a global picture after scanning, view the detailed cell data by zooming in and out, and modify the name of the scanned sample;

[0024] S4. Identification

[0025] Select a sample and select a suitable area on the sample. You can also select the whole film scan, which takes 1 minute and has a true positive rate of 72%.

[0026] S5. Result output

[0027] The AI ​​training module outputs the image results of Mycobacterium tuberculosis, enters the scanning information into the cell library, and then further optimizes the model.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] In the present invention, tuberculosis sputum smears are collected and a digital slide library is established, and the cells in the digital slide library are labeled and trained. Through the AI ​​training module, training is created and tested, and the training of the Mycobacterium tuberculosis recognition model is completed. At the same time, the model is tested to improve the detection efficiency and detection accuracy, avoid missed detection, promote the dynamic clearance of tuberculosis, and support remote consultation, thereby reducing dependence on local inspection personnel, which can better benefit people in remote areas. At the same time, the sharing of big data accelerates the establishment of a digital medical system. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0031] Figure 1 This is a flow chart of an AI Mycobacterium tuberculosis recognition model training method of the present invention;

[0032] Figure 2 The present invention is a flow chart of a detection method of an AI Mycobacterium tuberculosis recognition model. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Example: Figure 1-2 As shown, the present invention provides an AI Mycobacterium tuberculosis recognition model training method, comprising the following steps:

[0035] S1. Create a picture library

[0036] A1. Collect tuberculosis sputum smears from the clinic, use a slide scanning imaging system, turn on the slide scanning imaging system, turn on the computer and power box, and keep the slide scanning imaging system turned on;

[0037] A2. Double-click to open the software Scanner.exe, enter the user name and password, and enter the software working interface;

[0038] A3. Take out the slide rack, load the slide, make sure the slide is flat, put it into the slide rack, and close the door;

[0039] A4. Before scanning, enter the software interface, click the DTENS slide loading button in the upper left corner of the "Scan Sample" area, and enter the Thumbnail window settings;

[0040] A5. Thumbnail window settings:

[0041] 1. Scan to confirm the scanned slide;

[0042] 2. Drag the purple area or adjust the four vertices of the purple area to confirm the scanning range (it cannot exceed the outer green border);

[0043] 3. Automatically distribute focus points (recommended as the preferred option);

[0044] 4. Standardize the focus points. In the purple scanning area, click / left to add focus points (there should be no less than 7 focus points in the area, and the purple scanning area should preferably be within the focus points);

[0045] 5. Clear the focus point and rearrange the points;

[0046] 6. Confirm the scanning lens;

[0047] 7. After the settings are completed, click "Confirm" to enter the Realtime Video window settings;

[0048] A6. Realtime Video window settings: 11 Enter the Realtime Video window (real-time microscope) settings; adjust the B-axis focal length up and down through "Z" and "Z↓", and adjust the clarity under the 2# microscope; use the drop-down menu of the key to set the shooting color (R 3 through "Camara component / B component", saturation, brightness, light source and other parameters to present the cell image in the best state; after the settings are completed, click "OK" to enter the automatic focus state "Rea4.7 Automatic focus check: 1) 2 Enter the automatic focus state;

[0049] A7. Auto focus check:

[0050] 1. Enter the automatic focus state;

[0051] 2. Automatic focus completed:

[0052] When the Selected value is 1111, it indicates good focus. Click "Scan#" to enter the scanning state. When the Selected value is not 1111, click "Cancel" to manually focus.

[0053] Click Mowio in the table to check each focus point one by one to see if the cell image is clear; use the +- keys on the numeric keyboard to modify the piezoelectric value below and adjust the image; click Update in the table to update and save the adjusted data; after checking all focus points, click / "Rescan / Start Scanning" to start scanning;

[0054] S2. Scanning of tuberculosis sputum smear

[0055] A1. Scan slides: enter scanning state;

[0056] A2. Name, view and share scanned samples;

[0057] 1. After scanning, a global picture is formed. The detailed cell data can be viewed through the zoom ruler in the lower left corner;

[0058] 2. Modify the scan sample name:

[0059] 1): In the "Sample List" area on the left, find the scanned sample

[0060] 2): Right-click the sample and select "Change Name" to modify the sample name. The naming method is: examination number-name-sample site-scanning area*Sample site: bone marrow or peripheral blood; Scanning area: whole film or tail);

[0061] 3): Click the "Sample" tab on the menu bar above the homepage to view all scanned digital slide information; you can enter keywords in the name to search for samples;

[0062] 4): Click the "QR Code" button in the menu bar to get the QR code image (right Figure 2 QR code), you can remotely share all local digitized slide information by "copying the address" (reading address) to the computer browser web page or identifying the QR code;

[0063] S3, cell labeling and AI recognition

[0064] A1. Click the "Sample" tab on the homepage to enter the digital slide library and click on the slide to view cell information.

[0065] A2. Select the AI ​​application set module, and manually select the current annotated cell category, such as Xiangya bone marrow; A3. In the morphology list module, click the "+" in the report area to add the recognition area, and click "AI"

[0066] The system then automatically counts the number of nucleated cells and preliminarily identifies the cells;

[0067] A4. Click on the cell to select it (the green box in the menu bar indicates the cell is selected, and the dotted box in the slide view indicates the cell is selected).

[0068] Right-click to modify, classify, delete, and perform other operations on the cell identification;

[0069] A5. If you need to add cells, reselect the cell type (such as medium and young) under the AI ​​application set module.

[0070] Frame the cells and click to complete the selection and identification;

[0071] A6. After the modification is completed, click "Enter all cells into the standard library" in the lower right corner;

[0072] A7. Click "Enter the annotation library" above to enter the annotation review interface, select the cells that need to be reviewed,

[0073] Click "Approved" to complete cell labeling and training;

[0074] S4. Model training for Mycobacterium tuberculosis recognition

[0075] A1. Click Scan Viewer;

[0076] A2. Click on the sample, the page will jump to the slide management, and then select the slide to be tested, eg:

[0077] Sample No. 001 slide;

[0078] A3. Enter the 001 sample interface and click on the Mycobacterium tuberculosis classification in the AI ​​application set;

[0079] A4. Click on tuberculosis bacteria again to conduct a preliminary search for suitable areas;

[0080] A5. Then click on the morphology list, click on Select All, and enter the specimen library;

[0081] A6. Then click on the detection library and perform a secondary screening;

[0082] A7. Select all areas and click Review.

[0083] A8. Then click AI training, create training, and test training;

[0084] A9. Select a category in the labeled data;

[0085] A10. Select reviewed images for AI training.

[0086] The present invention also provides a detection method for an AI Mycobacterium tuberculosis recognition model, comprising the following steps:

[0087] S1. Model testing

[0088] Click "Sample" in the scan viewer, select the "Ai Training" option in the "Ai Training" above, click Create Training, select Detection Training, then find "Tuberculosis Bacillus" in the labeled data, then select the corresponding image in the training data, and click "Train" in the lower right corner;

[0089] S2. Calling model

[0090] A1. Change the training name on the right to be consistent with the slide number, set the width and height to 1024 respectively, change the block size to the default value of 16, and the period to 100000;

[0091] A2. Wait for the training to complete, click the training curve, and observe the acc value and loss value curves;

[0092] S3. Scan the tuberculosis sputum smear to be tested

[0093] A1. Turn on your computer and device.

[0094] A2. Take out the slide rack, load the slides, make sure the slides are flat, put them into the slide rack, and close the door;

[0095] A3. Enter the software interface, click the upper left corner of the "Scan Sample" area, and then click the button to load the slide;

[0096] A4, start scanning the slides;

[0097] A5. After scanning, a global picture is formed. You can view detailed cell data through the zoom scale in the lower left corner, modify the scanned sample name, and click the "Sample" label on the menu bar above the homepage to view all scanned digital slide information;

[0098] S4. Identification

[0099] A1. Click on the sample;

[0100] A2. Enter the slide management page;

[0101] A3. Randomly select a specimen;

[0102] A4. Then click the pattern list column on the right;

[0103] A5. Click the Mycobacterium tuberculosis option, then click the + sign, and then select the rectangular frame selection area. You can select a suitable area on the sample by yourself, or you can choose to scan the whole film. The whole film scan takes 1 minute and the true positive rate is 72%;

[0104] S5. Result output

[0105] A1. Click the AI ​​option, and the page will show AI starts to execute, and then wait for the execution to complete;

[0106] A2. After the execution is completed, the picture result of Mycobacterium tuberculosis will appear;

[0107] A3. Then enter all the image results into the cell library and further optimize the model.

[0108] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for training an AI Mycobacterium tuberculosis recognition model, characterized in that: The following steps are involved: S1. Create a picture library Collect tuberculosis sputum smears from the clinic, take out the slide rack in the slide scanning imaging system, place the collected tuberculosis sputum smears on the slide, then load the slide, make sure the slide is flat, put it into the slide rack, and close the door; S2. Scanning of tuberculosis sputum smear Scan the slides one by one through the slide scanning imaging system to observe whether the cell images of tuberculosis sputum smears are clear, scan the tuberculosis sputum smears, name, view and share them, and establish a digital slide library; S3, cell labeling and AI recognition Check cell information through the digital slide library, select the currently labeled cell category through the AI ​​application set module, add the identification area in the morphology list module, and the slide scanning imaging system automatically counts the number of nucleated cells and preliminarily identifies the cells, then modifies and classifies the cell identifiers, and then reviews the cells in the cell standard library. After the review is passed, the cell labeling and training are completed; S4. Model training for Mycobacterium tuberculosis recognition The slide is inspected through the scanning viewer, and then the Mycobacterium tuberculosis classification in the AI ​​application set module is used to select Mycobacterium tuberculosis for a screening, and then the morphology list in the morphology list module is used to enter the specimen library, and then the detection library is used for a second screening, and then all areas are reviewed, and then the training is created and tested through the AI ​​training module.

2. A detection method applied to the AI ​​Mycobacterium tuberculosis recognition model according to claim 1, characterized in that: The following steps are involved: S1. Model testing Create a training module through the AI ​​training module in the scan viewer, select detection training, then search for tuberculosis bacilli in the labeled data, then select the corresponding image in the training data, and then train; S2. Calling model Change the training name to be consistent with the slide, wait for the training to finish, and observe the acc value and loss value curves; S3. Scan the tuberculosis sputum smear to be tested Load the slide on the slide rack, scan the slide by scanning the sample, form a global picture after scanning, view the detailed cell data by zooming in and out, and modify the name of the scanned sample; S4. Identification Select a sample and select a suitable area on the sample. You can also select the whole film scan, which takes 1 minute and has a true positive rate of 72%. S5. Result output The AI ​​training module outputs the image results of Mycobacterium tuberculosis, enters the scanning information into the cell library, and then further optimizes the model.

Citation Information

Cited By

  • Mycobacterium tuberculosis antigen screening system

    CN122657891A