Screening method for cell slide smear specimen for thyroid fine needle examination

Abstract

The present invention discloses a screening method for cell slide smear specimens for thyroid fine needle examination, the method comprising a diagnostic model generation step (S110); a specimen smearing step (S120), a tomographic image acquisition step (S130), a classification step (S140), and a diagnostic stage determination step (S150), in which first image data (A), which provides optimal information suitable for diagnosing papillary thyroid carcinoma from the same specimen image, and second image data (B), which is unsuitable for diagnosis, and a dataset on results of papillary thyroid carcinoma diagnosis stage determination based on feature analysis between the first image data (A) and the second image data (B) are pre-trained, whereby both clear and blurred tomographic images can be utilized in the diagnostic determination.

Description

Screening method for thyroid fine needle aspiration cytology slide smear specimens

[0001] The present invention relates to a screening method for a thyroid fine needle aspiration cytology slide smear specimen, which utilizes a clear image and a blurred image of a tomographic image based on a microscope for diagnosis.

[0002] Typically, papillary cancer, the most common type of thyroid cancer, has a good prognosis, with a 20-year survival rate of 99%. However, if left untreated for a long period of time, it can develop into a tumor with a poor prognosis, such as anaplastic carcinoma, or invade surrounding organs, such as the esophagus and trachea, making treatment difficult.

[0003] In addition, after performing a fine needle biopsy, which is a screening test for thyroid cancer, the possibility of malignancy is evaluated based on the Bethesda classification system, and surgical treatment is performed based on the size and shape of the nodule. In the case of ultrasound examination and fine needle biopsy, they can be performed using a mobile ultrasound and syringe, but accessibility to a pathologist and pathology testing facility to analyze the results can be an issue.

[0004] In cases where access to such a medical system is not readily available, thyroid screening can be performed cost-effectively if primary screening on glass slides can be used to initially assess the possibility of malignancy and determine whether referral to a higher-level medical institution is necessary.

[0005] Meanwhile, in the case of thyroid fine needle aspiration biopsy slides, cells are spread on the slide in multiple layers, so it is difficult to focus on the location necessary for diagnosis and make a diagnosis compared to other types of slide pathology tests.

[0006] Accordingly, clear photographs selected by pathologists or expensive full-layer digital slide scanners are being utilized, but full-layer digital slide scanners are only available at higher-level medical institutions, so a technology that can perform primary examinations cost-effectively by replacing multilayer scanning is required.

[0007] Prior literature includes Korean Patent Publication No. 10-2280764 (Method and device for rapid label-free blood cancer diagnosis using 3D quantitative phase imaging and deep learning, July 22, 2021) and Korean Publication of Patent Publication No. 10-2022-0119669 (Method and system for digital staining of microscopic examination images using deep learning, August 30, 2022).

[0008] The technical problem to be achieved by the idea of ​​the present invention is to provide a screening method for a thyroid fine needle aspiration cytology slide smear specimen, which can select a clear image and a blurry image from a single image of a smear specimen obtained by tomography by a microscope without multilayer imaging by an expensive multilayer scanner, and can apply the blurry image to the basic diagnosis and discrimination of thyroid papillary cancer by utilizing a diagnostic model learned through not only the clear image but also the blurry image.

[0009] In order to achieve the above-mentioned object, an embodiment of the present invention comprises: a diagnostic model generation step of generating an artificial intelligence-based diagnostic model by learning in advance a dataset of first image data, which derives optimal information suitable for diagnosing papillary thyroid cancer of the same specimen image, second image data which are unsuitable for diagnosis, and a thyroid papillary cancer diagnosis stage determination result according to feature analysis of the first and second image data, which have been built as big data in advance; a specimen smear step of smearing a multilayered specimen collected through a fine needle aspiration of a subject for diagnosing papillary thyroid cancer on a slide; a tomographic image acquisition step of acquiring and storing a specimen image by taking a tomographic image of a specific tomographic layer of the specimen using a microscope equipped with an autofocus function; a classification step of selecting and classifying the first image data and the second image data from the specimen image using an artificial intelligence-based selection model; And, through the above diagnostic model, a diagnosis stage determination step for outputting the thyroid papillary cancer diagnosis stage determination result by inputting the first image data and the second image data classified in the classification step; The present invention provides a screening method for a thyroid fine needle aspiration biopsy cell slide smear specimen.

[0010] Here, the above selection model can be constructed by learning in advance the selection results of the first image data and the second image data according to the pre-set clarity of the specimen image and the specimen image constructed in advance as big data.

[0011] Additionally, the second image data may be a low-resolution image, an image containing noise, or an out-of-focus image.

[0012] In addition, in the above diagnosis stage determination step, a feature corresponding to a malignant stage of thyroid papillary cancer or a feature corresponding to a malignant tumor stage can be extracted from the first image data and the second image data, and the thyroid papillary cancer diagnosis stage determination result can be screened.

[0013] Additionally, features corresponding to the above malignant stage or features corresponding to the above malignant tumor stage may include papillary formation, cell nuclear overlapping, psammoma body due to calcification, or nuclear groove.

[0014] In addition, the microscope can obtain the specimen image by taking a picture while being fixed in the Z-axis direction with respect to the slide.

[0015] According to the present invention, it is possible to select a clear image and a blurry image from a single image of a smear specimen obtained by tomography based on a microscope without multilayer imaging using an expensive multilayer scanner, and to apply the blurry image to the basic diagnosis and discrimination of papillary thyroid cancer by utilizing a diagnostic model learned through not only the clear image but also the blurry image.

[0016] FIG. 1 is a schematic flowchart illustrating a screening method for a thyroid fine needle aspiration biopsy cell slide smear specimen according to an embodiment of the present invention.

[0017] FIG. 2 illustrates a system configuration diagram for implementing a screening method for a thyroid fine needle aspiration biopsy cell slide smear specimen of FIG. 1.

[0018] Figure 3 illustrates the process of constructing a diagnostic model for a screening method for a thyroid fine needle aspiration biopsy cell slide smear specimen of Figure 1.

[0019] Figure 4 illustrates the Bethesda step applied to the screening method for the thyroid fine needle aspiration biopsy cell slide smear specimen of Figure 1.

[0020] <Explanation of symbols>

[0021] S110: Diagnostic model creation stage

[0022] S120: Specimen smear stage

[0023] S130: Single-layer image acquisition step

[0024] S140: Classification stage

[0025] S150: Diagnosis stage determination stage

[0026] 10: Slide 20: Microscope

[0027] 110: Diagnostic Model 120: DB

[0028] 130: Selection model

[0029] Hereinafter, an embodiment of the present invention having the above-described features will be described in more detail with reference to the attached drawings.

[0030] A screening method for a thyroid fine needle aspiration cytology slide smear specimen according to an embodiment of the present invention comprises: a diagnostic model generation step (S110) in which an artificial intelligence-based diagnostic model (110) is generated by pre-learning a dataset of first image data (A) from which optimal information suitable for diagnosing thyroid papillary cancer of the same specimen image is derived, second image data (B) unsuitable for diagnosis, and thyroid papillary cancer diagnosis stage determination results according to feature analysis of the first image data (A) and the second image data (B); a specimen smear step (S120) in which a multilayer specimen collected through fine needle aspiration is smeared on a slide (10); a tomographic image acquisition step (S130) in which a specific single layer of the specimen is photographed using a microscope (20) to acquire and store a specimen image; a classification step (S140) in which the first image data (A) and the second image data (B) are selected and classified from the specimen image by the artificial intelligence-based selection model (130); The gist of the invention is to utilize clear and blurry images of tomographic images for diagnosis and determination, including a diagnosis stage determination step (S150) that inputs the first image data (A) and the second image data (B) classified in the classification step through a diagnosis model (110) and outputs the results of the diagnosis stage determination of thyroid papillary cancer.

[0031] Hereinafter, with reference to the drawings, a screening method for a thyroid fine needle aspiration biopsy cell slide smear specimen of the above-described configuration is specifically described as follows.

[0032] First, the diagnostic model creation step (S110) creates an artificial intelligence-based diagnostic model (110) that derives a thyroid papillary cancer diagnosis stage determination result by using a clean data image and a blurred data image of the same specimen image taken for thyroid papillary cancer diagnosis.

[0033] For example, referring to FIG. 3, a diagnostic model (110) can be created by extracting and learning in advance a first image data (clear image) (A) that derives optimal information suitable for diagnosing thyroid papillary cancer of the same specimen image from a DB (120) in which fine needle smear slide photographs for thyroid papillary cancer screening are stored as big data in advance, a second image data (blurry image) (B) that is not suitable for diagnosis, and a dataset of thyroid papillary cancer diagnosis stage determination results based on feature analysis of the first image data (A) and the second image data (B).

[0034] Additionally, CNN (Convolutional Neural Network), GAN (Generative Adversarial Networks), etc. can be applied as a diagnostic model (110), but are not particularly limited thereto.

[0035] In addition, the diagnostic model (110) defines a correlation between the second image data (B) corresponding to the first image data (A) and the thyroid papillary cancer diagnosis stage determination result according to the correlation between the first image data (A) and the thyroid papillary cancer diagnosis stage determination result, so that the second image data (B) can be used to determine the thyroid papillary cancer diagnosis stage.

[0036] Afterwards, in the specimen smear step (S120), a multilayered specimen collected through a fine needle biopsy of a subject seeking to be diagnosed with thyroid papillary cancer is prepared by smearing it on a slide (10).

[0037] Thereafter, in the step of obtaining a single-layer image (S130), a single image of a sample can be obtained and stored by taking a single-layer photograph of a specific single-layer of the sample using a microscope (20) equipped with an autofocus function, and then stored in the DB (120).

[0038] Meanwhile, the microscope (20) can search for and focus on the photographic layer of the area required for diagnosis while being fixed in the Z-axis direction with respect to the slide (10), and obtain a single specimen image by taking a tomographic image.

[0039] In addition, without a multi-layer scanning process for the specimen, it is possible to focus on the photographing layer of the area required for diagnosis and take a tomographic scan. The search for the photographing layer is performed by analyzing the statistical correlation between the photographing layer of the specimen image stored in the DB (120) and the diagnosis stage determination, and the optimal photographing layer is preset, and the height of the microscope (20) with respect to the Z axis can also be set.

[0040] Thereafter, in the classification step (S140), the first image data (A) and the second image data (B) are selected and classified from the previously acquired sample images by an artificial intelligence-based selection model (130).

[0041] Here, the second image data (B) may be a low-resolution image, an image containing noise, or an out-of-focus image.

[0042] That is, when a smear specimen formed in multiple layers is scanned in a tomographic manner, images that derive optimal information suitable for diagnosis and images that are inappropriate for deriving optimal information are mixed, so the selection model (130) can be constructed by learning in advance the selection results of the first image data (A) and the second image data (B) according to the specimen image constructed in advance as big data and the preset clarity of the specimen image.

[0043] The selection model (130) constructed in this manner can be used to select the first image data (A) and the second image data (B), and the second image data (B) can be converted into fuzzy data so that not only the clear first image data (A) but also the blurry second image data (B) can be applied to determine the stage of diagnosis of thyroid papillary cancer.

[0044] Meanwhile, when converting the second image data (B) into fuzzy data, the blurry image can be converted into a clear image by an image generation artificial intelligence model, or the blurry image can be converted into a clear image by removing noise from the image by a noise filter.

[0045] Thereafter, in the diagnosis stage determination step (S150), the first image data (A) and the second image data (B) classified in the preceding classification step (S140) are input through the previously constructed diagnosis model (110), respectively, and the thyroid papillary cancer diagnosis stage determination result is output, thereby performing basic screening for thyroid papillary cancer, so that cases with a high possibility of thyroid papillary cancer can be determined in advance before a precise diagnosis using expensive equipment.

[0046] That is, in the diagnosis stage determination step (S150), according to the Bethesda system, features corresponding to the malignant stage of papillary thyroid cancer or features corresponding to the malignant tumor stage are extracted from the first image data (A) and the second image data (B), and the thyroid papillary cancer diagnosis stage determination result is distinguished from stages 1 to 4 (insufficient specimen, benign, atypical cell, follicular tumor volume) and screened primarily.

[0047] Here, features corresponding to the stage of malignancy suspicion or the stage of malignancy include, for example, papillary formations that structurally resemble fingers, nuclear overlapping in which the nuclei of cells appear to overlap, psammoma bodies due to calcification, or nuclear grooves in which the nuclei of cells appear to resemble coffee beans.

[0048] Therefore, by configuring a screening method for thyroid fine needle aspiration cytology slide smear specimens as described above, clear and blurry images can be selected from a single image of a smear specimen obtained by tomography based on a microscope without multilayer photography, and the blurry image can be utilized for diagnosis determination by a diagnostic model learned through not only clear images but also blurry images.

[0049] The embodiments described in this specification and the configurations illustrated in the drawings are only the most preferred embodiments of the present invention and do not represent all of the technical ideas of the present invention. Therefore, it should be understood that there may be various equivalents and modified examples that can replace them at the time of filing this application.

[0050]

Claims

1. A diagnostic model generation step of generating an artificial intelligence-based diagnostic model by learning in advance a dataset of first image data, which is constructed as big data in advance and provides optimal information suitable for diagnosing papillary thyroid cancer of the same specimen image, second image data that is not suitable for diagnosis, and the results of determining the thyroid papillary cancer diagnosis stage based on feature analysis of the first image data and the second image data; A specimen smear step in which a multilayered specimen collected through a fine needle aspiration biopsy of a subject for the diagnosis of papillary thyroid cancer is smeared on a slide; A tomographic image acquisition step of acquiring and storing a specimen image by taking a tomographic photograph of a specific section of the specimen using a microscope equipped with an autofocus function; A classification step for selecting and classifying the first image data and the second image data from the sample image using an artificial intelligence-based selection model; and A diagnosis stage determination step for outputting the thyroid papillary cancer diagnosis stage determination result by inputting the first image data and the second image data classified in the classification step through the above diagnosis model; including; Screening methods for thyroid fine needle aspiration cytology slide smear specimens.

2. In paragraph 1, The above selection model is characterized in that it is constructed by learning in advance the selection results of the first image data and the second image data according to the pre-set clarity of the specimen image and the specimen image constructed in advance as big data. Screening methods for thyroid fine needle aspiration cytology slide smear specimens.

3. In paragraph 1, The above second image data is characterized by being a low-resolution image, an image containing noise, or an image that is out of focus. Screening methods for thyroid fine needle aspiration cytology slide smear specimens.

4. In paragraph 1, In the above diagnosis stage determination step, the feature corresponding to the malignant stage of thyroid papillary cancer or the feature corresponding to the malignant tumor stage is extracted from the first image data and the second image data, and the result of determining the diagnosis stage of thyroid papillary cancer is screened. Screening methods for thyroid fine needle aspiration cytology slide smear specimens.

5. In paragraph 4, The features corresponding to the above malignant stage or the features corresponding to the above malignant tumor stage are characterized by including papillary formation, cell nuclear overlapping, psammoma body due to calcification, or nuclear groove. Screening methods for thyroid fine needle aspiration cytology slide smear specimens.

6. In paragraph 1, The above microscope is characterized in that it obtains the specimen image by taking a photograph while fixed in the Z-axis direction with respect to the slide. Screening methods for thyroid fine needle aspiration cytology slide smear specimens.

Images