Thyroid fine needle puncture rapid screening method and system based on deep learning

Automatic analysis of fine needle aspiration samples of thyroid gland through deep learning technology solves the problems of subjectivity and inefficiency in traditional diagnostic methods, and achieves efficient and accurate cell recognition and thyroid disease screening, reducing the burden on pathologists.

CN119068478AActive Publication Date: 2024-12-03THE AFFILIATED HOSPITAL OF QINGDAO UNIV +1
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202411108705.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-12-03
Estimated Expiration
2044-08-13

AI Technical Summary

Technical Problem

Traditional diagnosis of fine thyroid aspiration depends on doctors’ visual judgment of cytology characteristics, with high subjectivity and high misdiagnosis rates, and due to insufficient manpower for pathologists, screening efficiency is low.

Method used

Using deep learning-based methods, the digital images of thyroid fine needle puncture samples are automatically analyzed, and deep learning positioning and classification models are used to quickly identify cells and determine whether they are positive, reducing the work burden of pathologists, improving screening efficiency and accuracy.

Benefits of technology

Automatic thyroid fine needle aspiration sample screening has been achieved, which has reduced the work burden of doctors, improved the accuracy and efficiency of diagnosis, and provided important support for the early diagnosis and treatment of thyroid diseases.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068478A_ABST
    Figure CN119068478A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of biomedicine deep learning, and provides a thyroid fine needle puncture rapid screening method and system based on deep learning. The method comprises the following steps: firstly, converting each sub-image into an HSV color space, and then carrying out fast radial symmetric transformation and threshold segmentation to obtain a binary image; generating a cell distribution scatter diagram by taking salient points in the binarized image as the positions of cells; performing kernel density estimation on the obtained cell scatter diagram to obtain a density histogram; performing threshold filtering on the density histogram by adopting a preset scatter distribution density threshold to obtain a target image matrix; inputting a sub-image corresponding to the target image into a deep learning positioning model to obtain a positioning matrix of the target cell; intersecting the positioning matrix of the target cell with the target image matrix to obtain an RoI image; and based on the RoI image, adopting a deep learning classification model to obtain a plurality of categories so as to determine whether the thyroid fine needle puncture digital image is positive or not.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of biomedical deep learning technology, and in particular to a method and system for rapid screening of thyroid fine needle aspiration based on deep learning. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Thyroid nodules are one of the most common endocrine diseases in clinical practice. The incidence rate of thyroid nodules detected by thyroid ultrasound is about 20% to 76%, and malignant nodules account for about 5% to 15%. Fine needle aspiration biopsy (FNAB) is an important method for preoperative evaluation of benign and malignant thyroid nodules and reduction of diagnostic surgery. It can also effectively predict the aggressiveness of malignant tumors. Ultrasound-guided FNA is suitable for the preoperative diagnosis of thyroid nodules suspected of malignancy at the first visit, and for considering recurrence and metastatic thyroid cancer. Due to the large number of thyroid fine needle aspiration examinations and the serious shortage of pathologists, as well as the tens of thousands of cells on the smear that need to be identified one by one by the pathologist, the screening efficiency is low. Summary of the invention

[0004] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a method and system for rapid screening of thyroid fine needle aspiration based on deep learning. The present invention utilizes deep learning related technologies to automatically analyze and screen thyroid fine needle aspiration (FNA) samples, quickly identify cells in thyroid fine needle aspiration images, reduce the workload of pathologists, and greatly improve screening efficiency and accuracy.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] The first aspect of the present invention provides a method for rapid screening of thyroid fine needle aspiration based on deep learning.

[0007] A fast screening method for thyroid fine needle aspiration based on deep learning, including:

[0008] The acquired thyroid fine needle puncture digital image is cropped to obtain several sub-images;

[0009] Each sub-image is first converted to the HSV color space, and then fast radial symmetry transformation and threshold segmentation are performed to obtain a binary image; the significant points in the binary image are used as the location of the cell to generate a cell distribution scatter plot; the obtained cell scatter plot is subjected to kernel density estimation to obtain a density histogram; the density histogram is threshold filtered using a preset scatter distribution density threshold to obtain a target image matrix;

[0010] Based on inputting the sub-image corresponding to the target image into the deep learning positioning model, the positioning matrix of the target cell is obtained; the positioning matrix of the target cell is intersected with the target image matrix, and the point with the result of 1 is selected as the RoI area to obtain the RoI image;

[0011] Based on the RoI image, a deep learning classification model is used to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive.

[0012] Furthermore, the process of converting each sub-image into the HSV color space first includes: converting each sub-image from RGB to the HSV color space, filtering according to preset H, S, and V channel thresholds to obtain an image containing only the cell nucleus.

[0013] Furthermore, the process of performing fast radial symmetry transformation and threshold segmentation to obtain a binary image includes: performing fast radial symmetry transformation on an image containing only cell nuclei, standardizing it, mapping the values ​​in the image to the interval [0, 255], and then performing OTSU threshold segmentation to obtain a binary image.

[0014] Furthermore, the training process of the deep learning localization model includes: dividing the sub-image into n×n regions, if a region contains the target cell, then the region is set to 1, otherwise it is set to 0, to obtain an n×n labeling matrix; using the n×n labeling matrix to train the deep learning localization model.

[0015] Furthermore, the deep learning positioning model includes: N convolutional layers, a maximum pooling layer, a convolutional layer for integrating the number of feature map channels, and a sigmoid layer that maps the output to [0, 1].

[0016] Furthermore, the method comprises: based on the RoI image, using a deep learning classification model to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive, and the method comprises: based on the RoI image, using a deep learning classification model to obtain several categories corresponding to the RoI image; according to the frequency of occurrence of each category, calculating the category probability of the entire thyroid fine needle puncture digital image; if the category probability exceeds the preset probability threshold of each category, outputting the thyroid fine needle puncture digital image as positive.

[0017] The second aspect of the present invention provides a thyroid fine needle aspiration rapid screening system based on deep learning.

[0018] A thyroid fine needle aspiration rapid screening system based on deep learning, including:

[0019] A cropping module is configured to: crop the acquired thyroid fine needle puncture digital image to obtain a plurality of sub-images;

[0020] The processing module is configured to: convert each sub-image into the HSV color space first, and then perform fast radial symmetry transformation and threshold segmentation to obtain a binary image; use the significant points in the binary image as the location of the cell to generate a cell distribution scatter plot; perform kernel density estimation on the obtained cell scatter plot to obtain a density histogram; use a preset scatter distribution density threshold to perform threshold filtering on the density histogram to obtain a target image matrix;

[0021] A positioning intersection module is configured to: input a sub-image corresponding to the target image into a deep learning positioning model to obtain a positioning matrix of the target cell; intersect the positioning matrix of the target cell with the target image matrix, select a point with a result of 1 as the RoI area, and obtain the RoI image;

[0022] The classification module is configured to: based on the RoI image, use a deep learning classification model to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive.

[0023] Furthermore, the processing module is also configured to: convert each sub-image from RGB to HSV color space, filter according to preset H, S, and V channel thresholds to obtain an image containing only the cell nucleus; perform a fast radial symmetry transformation on the image containing only the cell nucleus, and standardize it, map the values ​​in the image to the [0, 255] interval, and then perform OTSU threshold segmentation to obtain a binary image.

[0024] Furthermore, the classification module is also configured to: based on the RoI image, use a deep learning classification model to obtain several categories corresponding to the RoI image; calculate the category probability of the entire thyroid fine needle puncture digital image according to the frequency of occurrence of each category; if the category probability exceeds the preset probability threshold of each category, output the thyroid fine needle puncture digital image as positive.

[0025] A third aspect of the present invention provides a computer-readable storage medium.

[0026] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for rapid screening of thyroid fine needle aspiration based on deep learning as described in the first aspect above.

[0027] A fourth aspect of the present invention provides a computer device.

[0028] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for rapid screening of thyroid fine needle aspiration based on deep learning as described in the first aspect above are implemented.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] Traditional thyroid fine needle aspiration diagnosis relies on the doctor's visual judgment of cytological features, which has the problems of subjectivity and high misdiagnosis rate. To this end, the present application automatically identifies and classifies cytological features by learning a large amount of sample data, thereby improving the accuracy and consistency of diagnosis. The present invention uses the digital image of the fine needle aspiration sample as input, and after analysis and processing by the deep learning model, outputs the judgment result of whether the sample is positive. This automated screening method can not only reduce the workload of doctors, but also improve the efficiency and reliability of screening, providing important support for the early diagnosis and treatment of thyroid diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0032] Figure 1 is a flow chart of a method for rapid screening of thyroid fine needle aspiration based on deep learning shown in the present invention;

[0033] Figure 2 It is a flowchart of background removal shown in the present invention;

[0034] Figure 3 is a structural diagram of a positioning model shown in the present invention;

[0035] Figure 4 It is a framework diagram of the deep learning model shown in the present invention. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0038] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0039] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0040] Embodiment 1

[0041] like Figure 1 As shown, this embodiment provides a method for rapid screening of thyroid fine needle puncture based on deep learning. This embodiment uses the method applied to a server as an example. It can be understood that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application. In this embodiment, the method includes the following steps:

[0042] The acquired thyroid fine needle puncture digital image is cropped to obtain several sub-images;

[0043] Each sub-image is first converted to the HSV color space, and then fast radial symmetry transformation and threshold segmentation are performed to obtain a binary image; the significant points in the binary image are used as the location of the cell to generate a cell distribution scatter plot; the obtained cell scatter plot is subjected to kernel density estimation to obtain a density histogram; the density histogram is threshold filtered using a preset scatter distribution density threshold to obtain a target image matrix;

[0044] Based on inputting the sub-image corresponding to the target image into the deep learning positioning model, the positioning matrix of the target cell is obtained; the positioning matrix of the target cell is intersected with the target image matrix, and the point with the result of 1 is selected as the RoI area to obtain the RoI image;

[0045] Based on the RoI image, a deep learning classification model is used to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive.

[0046] The present invention uses the digital image of the fine needle aspiration sample as input, analyzes and processes it through a deep learning model, and outputs a judgment result on whether the sample is positive. This automated screening method can not only reduce the workload of doctors, but also improve the efficiency and reliability of screening, providing important support for the early diagnosis and treatment of thyroid diseases.

[0047] The following is a method for rapid screening of thyroid fine needle aspiration based on deep learning described in this embodiment, including the following steps:

[0048] Step (1): Background removal. The digital image of thyroid fine needle aspiration (FNAB) is cropped at 5 times magnification, and the cropped sub-image is converted from RGB to HSV color space, and filtered according to the preset H, S, and V channel thresholds to obtain an image containing only the cell nucleus. The obtained image is then subjected to a fast radial symmetry transform (FRST) and standardized, and the values ​​in the image are mapped to the interval [0, 255]. Subsequently, OTSU threshold segmentation is performed to obtain a binary image. The salient points in the binary image are used as the locations of the cells to generate a cell distribution scatter plot. A scatter point distribution density threshold is set, and a kernel density estimation (KDE) is performed on the obtained cell scatter plot to obtain a two-dimensional density histogram. The density histogram is then threshold filtered using the scatter point distribution density threshold to obtain the final result, as shown in FIG. Figure 2 If all the results are 0, it is considered to be a background image, otherwise it is considered to contain target cells and wait for the next step of processing.

[0049] Step (2): Cell localization.

[0050] Step (2-1): Label the cropped sub-image by dividing the sub-image into n×n regions. If a region contains the target cell, the region is set to 1, otherwise it is set to 0. The n×n matrix thus obtained is used as the label.

[0051] Step (2-2): Train the localization model based on the annotations obtained in step (2-1). The localization model consists of N convolutional layers (convolution kernel size 3×3, step size 1) and a maximum pooling layer (2×2, step size 2), as well as a convolutional layer for integrating the number of feature map channels (convolution kernel size 1×1) and a sigmoid layer that maps the output to [0, 1]. Figure 3 As shown, for example, n = 3, the network output is a matrix of 3 × 3 × 1. The network is trained using binary cross entropy loss.

[0052] Step (2-3): The original sub-image corresponding to the non-background image obtained in step (1) is sent to the convolutional neural network in step (2-2), and the network output result is obtained and the result obtained in step (1) is intersected. The point with a value of 1 in the obtained intersection result is taken as the region of interest (RoI), and its coordinates are recorded. The RoI coordinates are converted to the coordinates corresponding to the original slice, and the slice is cropped at 20 times to obtain the RoI image to be classified.

[0053] Step (3): Cell classification. Figure 4 As shown, using steps (1) and (2) to collect RoI images, establish a classification data set, and annotate the cell types in the RoI image according to type priority. For example, if myeloid cancer cells and lymphocytes coexist, they are marked as myeloid cancer cells. Establish a deep learning classification model, train and test the classification data set, so that the model can accurately classify various types of cells to the greatest extent, for example, establish a ResNet-34 classification model.

[0054] Step (4): Result statistics. The slices are processed according to steps (1), (2), and (3) to obtain a set of RoIs and corresponding categories. According to the frequency of occurrence of each category (multinucleated giant cells, benign follicular epithelial cells, lymphocytes, follicular tumor cells, atypical follicular cells of unknown significance, eosinophilic follicular cells, medullary carcinoma cells, and papillary carcinoma cells), the category probability of the entire slice is calculated. Set the probability threshold of each category. If the category probability exceeds the threshold, output the corresponding category.

[0055] Embodiment 2

[0056] This embodiment provides a thyroid fine needle aspiration rapid screening system based on deep learning.

[0057] A thyroid fine needle aspiration rapid screening system based on deep learning, including:

[0058] A cropping module is configured to: crop the acquired thyroid fine needle puncture digital image to obtain a plurality of sub-images;

[0059] The processing module is configured to: convert each sub-image into the HSV color space first, and then perform fast radial symmetry transformation and threshold segmentation to obtain a binary image; use the significant points in the binary image as the location of the cell to generate a cell distribution scatter plot; perform kernel density estimation on the obtained cell scatter plot to obtain a density histogram; use a preset scatter distribution density threshold to perform threshold filtering on the density histogram to obtain a target image matrix;

[0060] A positioning intersection module is configured to: input a sub-image corresponding to the target image into a deep learning positioning model to obtain a positioning matrix of the target cell; intersect the positioning matrix of the target cell with the target image matrix, select a point with a result of 1 as the RoI area, and obtain the RoI image;

[0061] The classification module is configured to: based on the RoI image, use a deep learning classification model to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive.

[0062] In some embodiments, the processing module is further configured to: convert each sub-image from RGB to HSV color space, filter according to preset H, S, and V channel thresholds to obtain an image containing only the cell nucleus; perform a fast radial symmetry transformation on the image containing only the cell nucleus, and standardize it, map the values ​​in the image to the [0, 255] interval, and then perform OTSU threshold segmentation to obtain a binary image.

[0063] In some embodiments, the system also includes a model training module, which is configured to: divide the sub-image into n×n regions, if a region contains the target cell, then set the region to 1, otherwise it is set to 0, to obtain an n×n labeling matrix; use the n×n labeling matrix to train the deep learning positioning model.

[0064] In some embodiments, the deep learning positioning model includes: N convolutional layers, a maximum pooling layer, a convolutional layer for integrating the number of feature map channels, and a sigmoid layer that maps the output to [0, 1].

[0065] In some embodiments, the classification module is further configured to: based on the RoI image, use a deep learning classification model to obtain several categories corresponding to the RoI image; calculate the category probability of the entire thyroid fine needle puncture digital image according to the frequency of occurrence of each category; if the category probability exceeds the preset probability threshold of each category, output the thyroid fine needle puncture digital image as positive.

[0066] Embodiment 3

[0067] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the method for rapid screening of thyroid fine needle aspiration based on deep learning as described in the first embodiment above are implemented.

[0068] Embodiment 4

[0069] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for rapid screening of thyroid fine needle aspiration based on deep learning as described in the first embodiment are implemented.

[0070] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0071] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0074] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for rapid screening of thyroid fine needle aspiration based on deep learning, characterized in that: include: The acquired thyroid fine needle puncture digital image is cropped to obtain several sub-images; Each sub-image is first converted to the HSV color space, and then fast radial symmetry transformation and threshold segmentation are performed to obtain a binary image; the significant points in the binary image are used as the location of the cell to generate a cell distribution scatter plot; the obtained cell scatter plot is subjected to kernel density estimation to obtain a density histogram; the density histogram is threshold filtered using a preset scatter distribution density threshold to obtain a target image matrix; Based on inputting the sub-image corresponding to the target image into the deep learning positioning model, the positioning matrix of the target cell is obtained; the positioning matrix of the target cell is intersected with the target image matrix, and the point with the result of 1 is selected as the RoI area to obtain the RoI image; Based on the RoI image, a deep learning classification model is used to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive.

2. The method for rapid screening of thyroid fine needle aspiration based on deep learning according to claim 1, characterized in that: The process of converting each sub-image to the HSV color space first includes: converting each sub-image from RGB to the HSV color space, filtering according to preset H, S, V channel thresholds, and obtaining an image containing only the cell nucleus.

3. The method for rapid screening of thyroid fine needle aspiration based on deep learning according to claim 2, characterized in that: The process of performing fast radial symmetry transformation and threshold segmentation to obtain a binary image includes: performing fast radial symmetry transformation on an image containing only cell nuclei, standardizing the image, mapping the values ​​in the image to the interval [0, 255], and then performing OTSU threshold segmentation to obtain a binary image.

4. The method for rapid screening of thyroid fine needle aspiration based on deep learning according to claim 1, characterized in that: The training process of the deep learning localization model includes: dividing the sub-image into n×n regions, if a region contains the target cell, then the region is set to 1, otherwise it is set to 0, and an n×n labeling matrix is ​​obtained; the deep learning localization model is trained using the n×n labeling matrix.

5. The method for rapid screening of thyroid fine needle aspiration based on deep learning according to claim 1, characterized in that: The deep learning positioning model includes: N convolutional layers, a maximum pooling layer, a convolutional layer for integrating the number of feature map channels, and a sigmoid layer that maps the output to [0, 1].

6. The method for rapid screening of thyroid fine needle aspiration based on deep learning according to claim 1, characterized in that: The method comprises: based on the RoI image, using the deep learning classification model to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive, and comprising: based on the RoI image, using the deep learning classification model to obtain several categories corresponding to the RoI image; according to the frequency of occurrence of each category, calculating the category probability of the entire thyroid fine needle puncture digital image; if the category probability exceeds the preset probability threshold of each category, outputting that the thyroid fine needle puncture digital image is positive.

7. A thyroid fine needle aspiration rapid screening system based on deep learning, characterized in that: include: A cropping module is configured to: crop the acquired thyroid fine needle puncture digital image to obtain a plurality of sub-images; The processing module is configured to: convert each sub-image into the HSV color space first, and then perform fast radial symmetry transformation and threshold segmentation to obtain a binary image; use the significant points in the binary image as the location of the cell to generate a cell distribution scatter plot; perform kernel density estimation on the obtained cell scatter plot to obtain a density histogram; use a preset scatter distribution density threshold to perform threshold filtering on the density histogram to obtain a target image matrix; A positioning intersection module is configured to: input a sub-image corresponding to the target image into a deep learning positioning model to obtain a positioning matrix of the target cell; intersect the positioning matrix of the target cell with the target image matrix, select a point with a result of 1 as the RoI area, and obtain the RoI image; The classification module is configured to: based on the RoI image, use a deep learning classification model to obtain several categories to determine whether the thyroid fine needle puncture digital image is positive.

8. The thyroid fine needle aspiration rapid screening system based on deep learning according to claim 7, characterized in that: The processing module is further configured to: convert each sub-image from RGB to HSV color space, filter according to preset H, S, V channel thresholds to obtain an image containing only the cell nucleus; perform a fast radial symmetry transformation on the image containing only the cell nucleus, standardize it, map the value in the image to the [0, 255] interval, and then perform OTSU threshold segmentation to obtain a binary image; or, The classification module is further configured to: based on the RoI image, use a deep learning classification model to obtain several categories corresponding to the RoI image; calculate the category probability of the entire thyroid fine needle puncture digital image according to the frequency of occurrence of each category; if the category probability exceeds the preset probability threshold of each category, output the thyroid fine needle puncture digital image as positive.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the deep learning-based thyroid fine needle aspiration rapid screening method as described in any one of claims 1 to 6 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the deep learning-based thyroid fine needle aspiration rapid screening method as described in any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • Image analysis method and device, electronic equipment and computer readable medium

    CN109948429A

  • Gastroscope pathological image classification method based on weak supervised learning

    CN111985536A

  • Automatic analysis system and method for full-slice digital pathological image

    CN112435243A

  • Thyroid cytology multi-type cell detection method based on deep learning

    CN114187277A

  • Method and device for acquiring to-be-interpreted cell image

    CN115223165A