Image recognition method and system for adenoid hypertrophy based on visual nasal endoscope

Abstract

The embodiment of the invention provides an adenoid hypertrophy image recognition method and system based on a visual nasal endoscope. The method comprises the steps that a nasal endoscope image of a patient is acquired; wherein the nasal endoscope image comprises at least one of a nasal cavity image, a nasal meatus image and a nasopharynx image; determining an image type corresponding to the nasal endoscope image; selecting a corresponding sub-adenoid hypertrophy image recognition model from the adenoid hypertrophy image recognition models based on the corresponding relationship between the image types and the image models, recognizing the nasal endoscope image through the sub-adenoid hypertrophy image recognition model, and determining a diagnosis result of adenoid hypertrophy; wherein the adenoid hypertrophy image recognition model is obtained by training a plurality of different nasal endoscope sample images; the image model corresponding relation is a mapping relation between a preset image type and the subadenoid hypertrophy image recognition model. According to the scheme, the image recognition accuracy of adenoid hypertrophy can be improved.

Description

Technical Field

[0001] The present application relates to the field of medical technology, and in particular to a method and system for image recognition of adenoid hypertrophy based on a visual nasal endoscope. Background Art

[0002] Adenoid hypertrophy is a common medical condition in children. Traditional nasal endoscopy relies primarily on visual observation and empirical judgment, which is highly subjective and can lead to misjudgment of subtle pathological features. For example, individual physicians may have inconsistent criteria for adenoid size and shape, resulting in discrepancies in diagnostic results.

[0003] Although existing diagnostic methods also use neural network models to identify and process the collected nasal endoscope images, due to factors such as the collection equipment and collection environment, they cannot guarantee that the collected image quality can adapt to the corresponding neural network model, resulting in unstable recognition accuracy. Summary of the Invention

[0004] The embodiments of the present application aim to provide a method and system for image recognition of adenoids hypertrophy based on a visual nasal endoscope, which can improve the accuracy of image recognition of adenoids hypertrophy.

[0005] The technical solution of this application is achieved as follows:

[0006] In a first aspect, an embodiment of the present application provides an image recognition method for adenoid hypertrophy based on a visual nasal endoscope, the method comprising:

[0007] Acquire a nasal endoscopic image of the patient; wherein the nasal endoscopic image includes at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image;

[0008] Determining an image type corresponding to the nasal endoscopic image;

[0009] Based on the correspondence between the image type and the image model, selecting a corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, and recognizing the nasal endoscopy image using the sub-adenoids hypertrophy image recognition model to determine a diagnosis result of adenoids hypertrophy;

[0010] The adenoids hypertrophy image recognition model is obtained by training with a plurality of different nasal endoscopy sample images; and the image model correspondence is a mapping relationship between a pre-set image type and a sub-adenoids hypertrophy image recognition model.

[0011] In the above solution, the sub-adenoids hypertrophy image recognition model includes a first adenoids hypertrophy image recognition model, a second adenoids hypertrophy image recognition model and a third adenoids hypertrophy image recognition model;

[0012] The method further comprises: selecting a corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the image type and the image model, and recognizing the nasal endoscopy image by using the sub-adenoids hypertrophy image recognition model to determine a diagnosis result of adenoids hypertrophy.

[0013] If the image type is the first image type, then based on the image model correspondence, selecting the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy; wherein the first image type indicates that the nasal endoscopic image is any one of the nasal cavity image, the nasal passage image, and the nasopharynx image;

[0014] If the image type is the second image type, then based on the image model correspondence, selecting the second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy; wherein the second image type indicates that the nasal endoscopic image is any two images of the nasal cavity image, the nasal passage image, and the nasopharynx image;

[0015] If the image type is the third image type, based on the image model correspondence, the third adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to identify the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy; wherein, the third image type represents that the nasal endoscopic image is the nasal cavity image, the nasal passage image and the nasopharynx image.

[0016] In the above solution, if the image type is the first image type, then based on the image model correspondence, selecting the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopy image and determine the diagnosis result of the adenoids hypertrophy includes:

[0017] If the image type is the first image type, cropping the nasal endoscopic image to determine a target area image containing adenoids;

[0018] Performing data enhancement processing on the target area image to determine a target enhanced image;

[0019] Based on the correspondence between the first image type and the image model, selecting the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to perform multi-dimensional feature extraction on the target enhanced image to obtain texture features and color features corresponding to the target enhanced image;

[0020] The texture feature and the color feature are fused to obtain a fused feature; and the fused feature is identified to determine a diagnosis result of adenoid hypertrophy.

[0021] In the above solution, if the image type is the second image type, then based on the image model correspondence, selecting the second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopy image and determine the diagnosis result of the adenoids hypertrophy includes:

[0022] If the image type is the second image type, performing cropping and data enhancement on any two of the nasal cavity image, the nasal passage image, and the nasopharynx image to determine a first region image and a second region image;

[0023] Based on the correspondence between the second image type and the image model, selecting the second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, performing multi-dimensional feature extraction and information fusion on the first region image and the second region image, respectively, to determine a first fusion feature and a second fusion feature;

[0024] Based on the first fusion feature and the second fusion feature, identification is performed to determine a first diagnosis result and a second diagnosis result;

[0025] If the first diagnostic result and / or the second diagnostic result indicates adenoids hypertrophy, determining the diagnostic result of adenoids hypertrophy as adenoids hypertrophy;

[0026] If the first diagnostic result and the second diagnostic result indicate that the adenoids are normal, the diagnostic result of adenoid hypertrophy is determined to be normal.

[0027] In the above solution, if the image type is the third image type, then based on the image model correspondence, selecting the third adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopy image and determine the diagnosis result of the adenoids hypertrophy includes:

[0028] If the image type is the third image type, selecting the third adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the third image type and the image model, performing feature extraction on the nasal endoscopy image to determine nasal cavity feature information, nasal passage feature information, and nasopharyngeal feature information;

[0029] Obtaining weight values ​​corresponding to the nasal cavity feature information, the nasal passage feature information, and the nasopharyngeal feature information;

[0030] performing weighted processing on the nasal cavity feature information, the nasal passage feature information, and the nasopharynx feature information based on the weight values ​​corresponding to each of the nasal cavity feature information, the nasal passage feature information, and the nasopharynx feature information to determine a diagnosis score;

[0031] A diagnosis result of adenoid hypertrophy is determined based on the diagnosis score.

[0032] In the above solution, the nasal cavity feature information corresponds to a first weight value, the nasal passage feature information corresponds to a second weight value, and the nasopharyngeal feature information corresponds to a third weight value; the third weight value is greater than the second weight value; the second weight value is greater than the first weight value; the sum of the third weight value, the second weight value, and the first weight value is 1;

[0033] The step of performing weighted processing on the nasal cavity feature information, the nasal passage feature information, and the nasopharyngeal feature information based on respective corresponding weight values ​​of the nasal cavity feature information, the nasal passage feature information, and the nasopharyngeal feature information to determine a diagnosis score includes:

[0034] Performing a multiplication operation based on the first weight value and the nasal cavity feature information to obtain a first score;

[0035] Performing a multiplication operation based on the second weight value and the nasal passage feature information to obtain a second score;

[0036] Performing a multiplication operation based on the third weight value and the nasopharyngeal feature information to obtain a third score;

[0037] performing an AND operation on the first score, the second score, and the third score to determine the diagnosis score;

[0038] and / or,

[0039] Determining the diagnosis result of adenoids hypertrophy based on the diagnosis score includes:

[0040] If the diagnostic score is less than a first threshold, determining that the diagnosis result of adenoids hypertrophy is normal;

[0041] If the diagnostic score is greater than or equal to the first threshold and less than the second threshold, the diagnosis result of the adenoids hypertrophy is determined to be mild adenoids hypertrophy;

[0042] If the diagnostic score is greater than or equal to the second threshold and less than the third threshold, the diagnosis result of the adenoids hypertrophy is determined to be moderate adenoids hypertrophy;

[0043] If the diagnosis score is greater than or equal to the third threshold, the diagnosis result of adenoids hypertrophy is determined to be severe adenoids hypertrophy.

[0044] In the above solution, before selecting the corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the image type and the image model, the method further includes:

[0045] Acquire a nasal endoscopy dataset; wherein the nasal endoscopy dataset includes normal adenoids images and adenoids hypertrophy images; the adenoids hypertrophy images include images of patients of different age groups, genders, and disease stages;

[0046] Dividing the nasal endoscopy dataset into a test set and a training set; wherein the test set and the training set respectively include nasal endoscopy data of different image types;

[0047] Training an initial adenoids hypertrophy sub-image recognition model using nasal endoscopy data of different image types in the training set, and determining a trained adenoids hypertrophy sub-image recognition model;

[0048] Based on the test set, the trained sub-adenoids hypertrophy image recognition model is tested to obtain a test result; if the test result meets the preset conditions, the sub-adenoids hypertrophy image recognition model is determined, thereby determining the adenoids hypertrophy image recognition model.

[0049] In a second aspect, an embodiment of the present application provides an image recognition system for adenoids hypertrophy based on a visual nasal endoscope, wherein the image recognition system for adenoids hypertrophy based on a visual nasal endoscope comprises: a visual nasal endoscope; the visual nasal endoscope comprises a mirror body and a detection probe; the mirror body has a built-in image diagnostic device; the detection probe is connected to the image diagnostic device via a data transmission line; the mirror body has a display screen, wherein,

[0050] The detection probe is used to obtain a nasal endoscopic image of the patient; wherein the nasal endoscopic image includes at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image; and transmit the nasal endoscopic image to the image diagnostic device;

[0051] The image diagnostic device is configured to determine an image type corresponding to the nasal endoscopic image; based on the correspondence between the image type and the image model, select a corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, identify the nasal endoscopic image using the sub-adenoids hypertrophy image recognition model, and determine a diagnosis of adenoids hypertrophy; wherein the adenoids hypertrophy image recognition model is trained using a plurality of different nasal endoscopic sample images; and the image model correspondence is a mapping relationship between a pre-set image type and a sub-adenoids hypertrophy image recognition model;

[0052] The display screen is used to display the diagnosis result of adenoids hypertrophy.

[0053] In a third aspect, an embodiment of the present application provides an image diagnostic device, comprising: a processor and a memory; wherein:

[0054] The memory is used to store computer programs;

[0055] The processor is configured to call and run the computer program from the memory to execute the method according to the first aspect.

[0056] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing executable instructions for causing a processor to execute the method described in the first aspect.

[0057] An embodiment of the present application provides an image recognition method and system for adenoids hypertrophy based on a visual nasal endoscope, the method comprising: obtaining a nasal endoscopic image of a patient; wherein the nasal endoscopic image comprises at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image; determining an image type corresponding to the nasal endoscopic image; based on the correspondence between the image type and the image model, selecting a corresponding sub-adenoids hypertrophy image recognition model from an adenoids hypertrophy image recognition model, and recognizing the nasal endoscopic image using the sub-adenoids hypertrophy image recognition model to determine a diagnosis result of adenoids hypertrophy; wherein the adenoids hypertrophy image recognition model is obtained by training with a plurality of different nasal endoscopic sample images; and the image model correspondence is a mapping relationship between a pre-set image type and a sub-adenoids hypertrophy image recognition model. In the above scheme, the nasal endoscopic image includes at least one of the nasal cavity image, nasal passage image and nasopharynx image. This is because the actual process of collecting nasal endoscopic images is affected by many different factors, resulting in the possibility of multiple types of collected nasal endoscopic images. Therefore, it is necessary to determine the image type corresponding to the nasal endoscopic image; after determining the image type corresponding to the nasal endoscopic image, according to the mapping relationship between the image type and the sub-adenoids hypertrophy image recognition model, the corresponding sub-adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to identify the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy. Since different sub-adenoids hypertrophy image recognition models are provided for different types of nasal endoscopic images to identify the nasal endoscopic image, targeted identification can be achieved, thereby improving the accuracy of adenoids hypertrophy image recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings herein are incorporated into and constitute a part of this specification. These drawings illustrate embodiments consistent with the present application and, together with the specification, serve to illustrate the technical solutions of the present application. Obviously, the drawings described below are merely some embodiments of the present application. Those skilled in the art can, without inventive effort, derive other drawings from these drawings.

[0059] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0060] Figure 1 An optional process diagram of a method for image recognition of adenoid hypertrophy based on visual nasal endoscopy is provided for the embodiment of the present application. Figure 1 ;

[0061] Figure 2 An optional process diagram of a method for image recognition of adenoid hypertrophy based on visual nasal endoscopy is provided for the embodiment of the present application. Figure 2

[0062] Figure 3 An optional process diagram of a method for image recognition of adenoid hypertrophy based on visual nasal endoscopy is provided for the embodiment of the present application. Figure 3 ;

[0063] Figure 4 A schematic structural diagram of an image recognition system for adenoid hypertrophy based on a visual nasal endoscope is provided for an embodiment of the present application;

[0064] Figure 5 A structural diagram of a graphic diagnostic device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0065] To make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the specific technical solutions of the present application will be further described in detail below in conjunction with the drawings in the embodiments of the present application. The following embodiments are used to illustrate the present application but are not intended to limit the scope of the present application.

[0066] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0067] In the following description, references to “some embodiments,” “this embodiment,” “embodiments of the present application,” and examples, etc., describe a subset of all possible embodiments. However, it can be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

[0068] If similar descriptions of "first / second" appear in the application documents, the following explanation is added. In the following description, the terms "first\second\third" involved are merely used to distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0069] Based on this, the embodiment of the present application provides an image recognition method for adenoids hypertrophy based on visual nasal endoscopy. Figure 1 An optional process diagram of a method for image recognition of adenoid hypertrophy based on visual nasal endoscopy is provided for the embodiment of the present application. Figure 1 , will combine Figure 1 The steps shown are explained.

[0070] S101. Obtain a nasal endoscopic image of a patient; wherein the nasal endoscopic image includes at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image.

[0071] In some embodiments of the present application, the nasal endoscopic image includes at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image. When acquiring a nasal endoscopic image of a patient, one, two, or three images may be acquired.

[0072] In some embodiments of the present application, the execution subject of the adenoids hypertrophy image recognition method based on visual nasal endoscope is a visual nasal endoscope.

[0073] In some embodiments of the present application, the image recognition method for adenoids hypertrophy based on visual nasal endoscopy is applicable to the diagnosis scenario of adenoids hypertrophy.

[0074] In some embodiments of the present application, the visual nasal endoscope includes a mirror body and a detection probe. The mirror body has a display screen for displaying the detected nasal endoscopic images and diagnostic results. The mirror body has a built-in image diagnostic device, and the detection probe is connected to the image diagnostic device via a data transmission line. The detection probe is used to obtain nasal endoscopic images in the patient's nasal cavity and transmit them to the image diagnostic device. The design of the mirror body conforms to ergonomic principles and is convenient for doctors to operate. Its shell is made of medical-grade materials with good durability and safety. The display screen uses a high-resolution screen that can clearly display images and diagnostic results. The detection probe has a suitable size and shape and can smoothly enter the nasal cavity for image acquisition. The probe head is made of soft material to avoid damage to the nasal cavity. At the same time, an optical imaging system is integrated inside the probe to obtain high-quality nasal endoscopic images.

[0075] In some embodiments of the present application, the image diagnostic device includes a construction module, a training module, a verification module and a diagnosis module. The construction module is used to select nasal endoscopic images from a hospital case database to construct a nasal endoscopic data set. The training module is used to load a pre-trained neural network model and train the neural network model. The verification module is used to test the trained neural network model. The diagnosis module is used to perform intelligent classification diagnosis on the nasal endoscopic images of the patient detected by the detection probe using the optimal neural network model obtained by the verification module, and output the classification diagnosis result of the nasal endoscopic image (normal adenoids or adenoids hypertrophy).

[0076] Exemplarily, the visual nasal endoscope collects nasal endoscopic images of the patient through a detection probe, specifically by placing the detection probe at different positions inside the patient's nose to collect data, obtaining at least one of the nasal cavity image, nasal passage image and nasopharyngeal image, and transmitting the collected nasal endoscopic image to the image diagnostic device through a data transmission line.

[0077] S102: Determine the image type corresponding to the nasal endoscope image.

[0078] In some embodiments of the present application, the image types corresponding to the nasal endoscopic image can be divided into three types: a first image type, a second image type, and a third image type. Each image type includes different image data of the nasal endoscopic image.

[0079] In some embodiments of the present application, the first image type represents that the nasal endoscopic image is any one of the nasal cavity image, nasal passage image and nasopharynx image; the second image type represents that the nasal endoscopic image is any two of the nasal cavity image, nasal passage image and nasopharynx image; the third image type represents that the nasal endoscopic image is the nasal cavity image, nasal passage image and nasopharynx image.

[0080] In some embodiments of the present application, after receiving the nasal endoscopic image, the image diagnostic device performs initial recognition on the nasal endoscopic image to determine the number of images and image acquisition position information of the nasal endoscopic image, and thus determines the image type corresponding to the nasal endoscopic image based on the number of images and image acquisition position information of the nasal endoscopic image.

[0081] Exemplarily, after acquiring the nasal endoscopic image, the nasal endoscopic image is initially identified to determine that the nasal endoscopic image includes a nasal cavity image and a nasal passage image, and based on the nasal cavity image and the nasal passage image, the nasal endoscopic image is determined to be a second image type.

[0082] S103. Based on the correspondence between the image type and the image model, a corresponding sub-adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model, and the nasal endoscopy image is recognized by the sub-adenoids hypertrophy image recognition model to determine the diagnosis result of adenoids hypertrophy; wherein the adenoids hypertrophy image recognition model is obtained by training with multiple different nasal endoscopy sample images; the image model correspondence is a mapping relationship between the pre-set image type and the sub-adenoids hypertrophy image recognition model.

[0083] In some embodiments of the present application, the sub-adenoids hypertrophy image recognition model includes a first adenoids hypertrophy image recognition model, a second adenoids hypertrophy image recognition model and a third adenoids hypertrophy image recognition model.

[0084] In some embodiments of the present application, the diagnostic criteria for adenoid hypertrophy are as follows:

[0085] Normal adenoids: The adenoids' shape and size within the nasal cavity are within the normal physiological range, with no abnormalities such as hyperplasia or congestion. Specifically, the ratio of the adenoids' volume to the nasal cavity space is within the normal range, and the adenoids' surface is smooth, with a clear boundary between them and the surrounding tissue.

[0086] Adenoid hypertrophy: Enlarged adenoids can block the posterior nasal passages, leading to symptoms such as nasal congestion, runny nose, and snoring. Radiological findings include an imbalance in the proportion of the adenoids to the surrounding tissues, a rough surface, and possible signs of hyperplasia or inflammation. In severe cases, the adenoids may almost completely block the posterior nasal passages, causing symptoms such as difficulty breathing.

[0087] In some embodiments of the present application, the adenoids hypertrophy image recognition model is obtained by training with a plurality of different nasal endoscopy sample images; the image model correspondence is a mapping relationship between a pre-set image type and a sub-adenoids hypertrophy image recognition model.

[0088] Exemplarily, the image-model correspondence is that the first image type corresponds to the first adenoids hypertrophy image recognition model, the second image type corresponds to the second adenoids hypertrophy image recognition model, and the third image type corresponds to the third adenoids hypertrophy image recognition model.

[0089] In some embodiments of the present application, the adenoids hypertrophy image recognition model is a trained neural network model, which includes a convolutional layer, a pooling layer, an activation function and a fully connected layer.

[0090] Convolutional layers play a key role in neural network models. For example, the InceptionV3 model contains multiple convolution kernels of different sizes. These kernels slide over the image to perform convolution operations, automatically extracting local features. For endoscopic images, the convolution layer can extract features such as the shape and texture of the adenoids. Convolution kernels of different sizes can capture features at different scales. For example, a smaller kernel can capture subtle textures on the surface of the adenoids, while a larger kernel can capture the overall shape of the adenoids.

[0091] Pooling layer: The pooling layer usually follows the convolution layer. Its function is to downsample the feature map output by the convolution layer, reducing the amount of data while retaining key features. Commonly used pooling methods include maximum pooling and average pooling. In the present invention, for example, the maximum pooling layer is used, which selects the maximum value of each area in the feature map as the output, so as to highlight the significant features in the image. For nasal endoscopic images, the pooling layer helps to highlight the main features of the adenoids, such as the general shape of the adenoids and the most obvious texture features, while reducing the amount of data processing and improving the computational efficiency of the model.

[0092] Activation function: The activation function is used to introduce nonlinear factors into neurons, so that the neural network can learn and fit complex nonlinear relationships. In this application, the ReLU (Rectified Linear Unit) activation function can be used. When the input of the ReLU function is greater than 0, the output is equal to the input; when the input is less than or equal to 0, the output is 0. This activation function has the advantages of simple calculation and fast convergence speed. For the classification task of nasal endoscopic images, the ReLU activation function can help the model better learn the nonlinear relationship between normal and hypertrophic adenoids, and improve the classification accuracy of the model.

[0093] Fully connected layer: When fine-tuning the pre-trained model, the last fully connected layer needs to be replaced. The fully connected layer integrates the features extracted by the previous layer and outputs the classification results. In this application, it is replaced with a fully connected layer with an output number of 2 (for normal and hypertrophic adenoids). Each neuron in the fully connected layer is connected to all neurons in the previous layer, and information is transferred and calculated through the weight matrix. During the training process, the weights are continuously adjusted to optimize the classification performance of the model.

[0094] In some embodiments of the present application, if the image type is a first image type, then based on the image model correspondence, a first adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy; wherein the first image type indicates that the nasal endoscopic image is any one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image. If the image type is a second image type, then based on the image model correspondence, a second adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy; wherein the second image type indicates that the nasal endoscopic image is any two of a nasal cavity image, a nasal passage image, and a nasopharyngeal image. If the image type is a third image type, then based on the image model correspondence, a third adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy; wherein the third image type indicates that the nasal endoscopic image is any one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image.

[0095] For example, the adenoids hypertrophy sub-image recognition model performs intelligent classification diagnosis on the patient's nasal endoscopic image detected by the detection probe, and outputs the classification diagnosis result of the nasal endoscopic image (normal adenoids or adenoids hypertrophy). The patient's nasal endoscopic image and the diagnosis result of the nasal endoscopic image are displayed on the display screen, thereby realizing visual detection and intelligent diagnosis of nasal endoscopy. When displaying the diagnosis results, different colors or icons can be used to distinguish different diagnosis results, so that doctors can quickly identify them.

[0096] It can be understood that the nasal endoscopic image includes at least one of the nasal cavity image, nasal passage image and nasopharynx image. This is because the actual process of collecting nasal endoscopic images is affected by many different factors, resulting in the possibility of multiple types of collected nasal endoscopic images. Therefore, it is necessary to determine the image type corresponding to the nasal endoscopic image; after determining the image type corresponding to the nasal endoscopic image, according to the mapping relationship between the image type and the sub-adenoids hypertrophy image recognition model, the corresponding sub-adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to identify the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy. Since different sub-adenoids hypertrophy image recognition models are provided for different types of nasal endoscopic images to identify the nasal endoscopic image, targeted identification can be achieved, thereby improving the accuracy of image recognition of adenoids hypertrophy.

[0097] In some embodiments of the present application, Figure 2 As shown, S103 can be implemented through S201 or S202, or 203, as follows:

[0098] S201. If the image type is the first image type, based on the image model correspondence, select the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy; wherein the first image type represents that the nasal endoscopic image is any one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image.

[0099] In some embodiments of the present application, if the image type is the first image type, the nasal endoscopy image is cropped to determine the target area image containing the adenoids; the target area image is data enhanced to determine the target enhanced image; the target enhanced image is subjected to multi-dimensional feature extraction through the first adenoids hypertrophy image recognition model to obtain texture features and color features corresponding to the target enhanced image; the texture features and color features are feature fused to obtain fused features; and the fused features are identified to determine the diagnostic result of adenoids hypertrophy.

[0100] In some embodiments of the present application, if the image type is the first image type, that is, the nasal endoscope image is any one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image, then any one of the nasal cavity image, nasal passage image, and nasopharyngeal image is cropped to determine a target region image including the adenoids. The target region image is then subjected to processing such as translation, scaling, rotation, left-right flipping, and left-right stretching to obtain a target enhanced image.

[0101] In some embodiments of the present application, based on the correspondence between the first image type and the image model, a first adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition models to perform multi-dimensional feature extraction on the target enhanced image, thereby obtaining texture features and color features corresponding to the target enhanced image; wherein the texture features can reflect the morphological characteristics of the adenoids, and the color information can reflect the color characteristics of the adenoids. Feature fusion is performed on the texture features and color features to obtain a fused feature; the fused feature is then identified to determine a diagnosis of adenoids hypertrophy.

[0102] In some embodiments of the present application, the diagnosis results of adenoids hypertrophy include normal and adenoids hypertrophy, and adenoids hypertrophy further includes mild adenoids hypertrophy, moderate adenoids hypertrophy and severe adenoids hypertrophy.

[0103] For example, data augmentation processing includes translation, scaling, rotation, left-right flipping, and left-right stretching. Translation can simulate observation from different angles, scaling can simulate observation from different distances, rotation can simulate different probe orientations within the nasal cavity, and left-right flipping and stretching can increase image diversity. These data augmentation operations can improve the model's generalization capabilities, enabling it to better handle a variety of image conditions in real-world applications.

[0104] It is understandable that since the first image type only includes one type of nasal endoscopic image, and the acquired nasal endoscopic image may not be clear, the nasal endoscopic image is cropped to determine the target area image containing the adenoids; this can effectively extract the area to be identified, thereby reducing the identification of invalid areas and improving the recognition efficiency. The target area image is subjected to data enhancement processing to determine the target enhanced image, which can improve the clarity of the target enhanced image and help improve the accuracy of image recognition. Multi-dimensional feature extraction is performed on the target enhanced image to obtain texture features and color features corresponding to the target enhanced image; feature fusion is performed on the texture features and color features to obtain fusion features; since texture features can reflect the morphological features of the adenoids, and color information can reflect the color features of the adenoids, the fusion features have two information features, and the diagnosis result of adenoids hypertrophy determined by identification through fusion features will be more accurate.

[0105] S202. If the image type is the second image type, based on the image model correspondence, select a second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy; wherein the second image type represents that the nasal endoscopic image is any two images among the nasal cavity image, the nasal passage image and the nasopharynx image.

[0106] In some embodiments of the present application, if the image type is the second image type, any two images among the nasal cavity image, the nasal passage image and the nasopharynx image are cropped and data enhanced respectively to determine the first region image and the second region image; based on the correspondence between the second image type and the image model, a second adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to perform multi-dimensional feature extraction and information fusion on the first region image and the second region image respectively to determine the first fusion feature and the second fusion feature; based on the first fusion feature and the second fusion feature, identification is performed respectively to determine the first diagnostic result and the second diagnostic result; if the first diagnostic result and / or the second diagnostic result characterize adenoids hypertrophy, the diagnostic result of the adenoids hypertrophy is determined to be adenoids hypertrophy; if the first diagnostic result and the second diagnostic result characterize normal adenoids, the diagnostic result of the adenoids hypertrophy is determined to be normal.

[0107] In some embodiments of the present application, if the image type is the second image type, it means that the nasal endoscopy image is any two images among the nasal cavity image, the nasal passage image and the nasopharynx image, then any two images among the nasal cavity image, the nasal passage image and the nasopharynx image are cropped respectively, the cropped target area is determined, and the target area is data enhanced to obtain the first area image and the second area image.

[0108] In some embodiments of the present application, multi-dimensional feature extraction is performed on the first region image and the second region image respectively through the second adenoids hypertrophy image recognition model to obtain texture information and sample color information corresponding to the first region image and the second region image respectively, and information fusion is performed on the texture information and sample color information corresponding to the first region image and the second region image respectively to determine the first fusion feature corresponding to the first region image and the second fusion feature corresponding to the second region image; based on the first fusion feature and the second fusion feature, identification is performed respectively to determine the first diagnostic result and the second diagnostic result.

[0109] In some embodiments of the present application, if at least one of the first diagnostic result and the second diagnostic result indicates adenoids hypertrophy, the diagnostic result of adenoids hypertrophy is determined to be adenoids hypertrophy; if both the first diagnostic result and the second diagnostic result indicate that the adenoids are normal, the diagnostic result of adenoids hypertrophy is determined to be normal.

[0110] It should be noted that both the first diagnostic result and the second diagnostic result represent adenoids hypertrophy. It is necessary to select the diagnostic result with a higher degree of adenoids hypertrophy as the diagnostic result of adenoids hypertrophy based on the degree of adenoids hypertrophy.

[0111] Exemplarily, the first diagnostic result is mild adenoids hypertrophy, and the second diagnostic result is normal, and the diagnostic result of adenoids hypertrophy is determined to be mild adenoids hypertrophy.

[0112] It is understandable that the nasal endoscope image is any two images among the nasal cavity image, the nasal passage image, and the nasopharynx image. Cropping and data enhancement are performed on any two images among the nasal cavity image, the nasal passage image, and the nasopharynx image to determine the first region image and the second region image. This can reduce the recognition of invalid areas, improve recognition efficiency, and improve the clarity of the first region image and the second region image, which is conducive to improving the accuracy of image recognition. Based on the correspondence between the second image type and the image model, a second adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to perform multi-dimensional feature extraction and information fusion on the first region image and the second region image, respectively, to determine the first fusion feature and the second fusion feature; based on the first fusion feature and the second fusion feature, recognition is performed to determine the first diagnosis result and the second diagnosis result; the first fusion feature and the second fusion feature each have two information features, and then the first diagnosis result and the second diagnosis result are determined more accurately by the first fusion feature and the second fusion feature. If the first diagnostic result and / or the second diagnostic result characterizes adenoids hypertrophy, the diagnostic result of adenoids hypertrophy is determined to be adenoids hypertrophy; if the first diagnostic result and the second diagnostic result characterize normal adenoids, the diagnostic result of adenoids hypertrophy is determined to be normal. Based on the two diagnostic results, a comprehensive evaluation is performed to determine whether the final diagnostic result of adenoids hypertrophy is more accurate, thereby improving the accuracy of image recognition of adenoids hypertrophy.

[0113] S203. If the image type is the third image type, based on the image model correspondence, select a third adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of adenoids hypertrophy; wherein the third image type represents that the nasal endoscopic image is a nasal cavity image, a nasal passage image, and a nasopharyngeal image.

[0114] In some embodiments of the present application, if the image type is a third image type, based on the correspondence between the third image type and the image model, a third adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to perform feature extraction on the nasal endoscopy image, and determine the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information; obtain the weight values ​​corresponding to the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information; based on the weight values ​​corresponding to the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information, perform weighted processing on the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information to determine the diagnostic score; based on the diagnostic score, determine the diagnostic result of adenoids hypertrophy.

[0115] In some embodiments of the present application, the nasal cavity feature information corresponds to a first weight value, the nasal passage feature information corresponds to a second weight value, and the nasopharyngeal feature information corresponds to a third weight value; the third weight value is greater than the second weight value; the second weight value is greater than the first weight value; the sum of the third weight value, the second weight value and the first weight value is 1.

[0116] In some embodiments of the present application, a multiplication operation is performed based on the first weight value and the nasal cavity feature information to obtain a first score; a multiplication operation is performed based on the second weight value and the nasal passage feature information to obtain a second score; a multiplication operation is performed based on the third weight value and the nasopharyngeal feature information to obtain a third score; and a sum operation is performed on the first score, the second score, and the third score to determine the diagnostic score.

[0117] In some embodiments of the present application, if the diagnostic score is less than the first threshold, the diagnosis result of adenoids hypertrophy is determined to be normal; if the diagnostic score is greater than or equal to the first threshold and less than the second threshold, the diagnosis result of adenoids hypertrophy is determined to be mild adenoids hypertrophy; if the diagnostic score is greater than or equal to the second threshold and less than the third threshold, the diagnosis result of adenoids hypertrophy is determined to be moderate adenoids hypertrophy; if the diagnostic score is greater than or equal to the third threshold, the diagnosis result of adenoids hypertrophy is determined to be severe adenoids hypertrophy.

[0118] For example, the first threshold value may be 3, the second threshold value may be 5, and the third threshold value may be 7. The first weight value is 0.2, the second weight value is 0.35, and the third weight value is 0.45. The initial first score determined based on the nasal cavity feature information is 7, the initial second score determined based on the nasal passage feature information is 6, and the initial third score determined based on the nasopharyngeal feature information is 5. Based on the first weight value and the initial first score, a multiplication operation is performed to obtain the first score, i.e., 7*0.2=1.4; based on the second weight value and the nasal passage feature information, a multiplication operation is performed to obtain the second score, i.e., 6*0.35=2.1; based on the third weight value and the nasopharyngeal feature information, a multiplication operation is performed to obtain the third score, i.e., 5*0.45=2.25; the diagnostic score is 1.4+2.1+2.25=5.75. Since 5.75 is greater than the second threshold value of 5 and less than the third threshold value of 7, the diagnosis result of adenoid hypertrophy is determined to be moderate adenoid hypertrophy.

[0119] It can be understood that the third adenoids hypertrophy image recognition model is used to extract features from the nasal endoscopy image to determine the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information; the weight values ​​corresponding to the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information are obtained; based on the weight values ​​corresponding to the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information, the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information are weighted to determine the diagnostic score; based on the diagnostic score, the diagnosis result of adenoids hypertrophy is determined. Since the feature information of the nasopharyngeal image has a significant impact on the diagnosis of adenoids hypertrophy, by increasing the weight of the nasopharyngeal feature information corresponding to the nasopharyngeal image, image recognition based on the nasal cavity image, nasal passage image and nasopharyngeal image can make the diagnosis result of adenoids hypertrophy more accurate.

[0120] In some embodiments of the present application, Figure 3 As shown, before executing S103, S104-S107 are also executed as follows:

[0121] S104. Acquire a nasal endoscopy dataset; wherein the nasal endoscopy dataset includes normal adenoids images and adenoids hypertrophy images; the adenoids hypertrophy images include images of patients of different age groups, genders, and disease stages.

[0122] In some embodiments of the present application, the nasal endoscopy dataset includes normal adenoids images and adenoids hypertrophy images. The adenoids hypertrophy images include images of patients of different age groups, genders, and disease stages.

[0123] For example, a nasal endoscopy dataset was constructed by selecting nasal endoscopy images from a hospital case database. These images include normal adenoids and images of patients with varying degrees of adenoid hypertrophy. The image selection process ensured quality and representativeness, encompassing patients of different ages, genders, and disease stages.

[0124] It should be noted that cases in the nasal endoscopy dataset were randomly selected and divided into multiple test and training sets using a 50-fold cross-validation method. During the division process, it was ensured that multiple images of the same case did not appear in both the training and test sets. In other words, all images of the case were either in the training set or in the test set.

[0125] S105. Divide the nasal endoscopy dataset into a test set and a training set; wherein the test set and the training set respectively include nasal endoscopy data of different image types.

[0126] For example, the dataset is divided into a test set and a training set. During the division process, it is important to ensure that multiple images from the same case do not appear in both the training and test sets. That is, all images from the case must be in either the training set or the test set. Scientific division methods, such as stratified sampling or random sampling, are used to ensure the rationality and effectiveness of the training and test sets.

[0127] S106 , training the initial sub-adenoids hypertrophy image recognition model using nasal endoscopy data of different image types in the training set, and determining a trained sub-adenoids hypertrophy image recognition model.

[0128] In some embodiments of the present application, endoscopic data of different image types include endoscopic data of a first type, endoscopic data of a second type, and endoscopic data of a third type. The initial sub-adenoids hypertrophy image recognition model includes a first initial adenoids hypertrophy image recognition model, a second initial adenoids hypertrophy image recognition model, and a third initial adenoids hypertrophy image recognition model. The first type of endoscopic data is used to compare the first initial adenoids hypertrophy image recognition model to determine the first adenoids hypertrophy image recognition model after training; the second type of endoscopic data is used to compare the second initial adenoids hypertrophy image recognition model to determine the second adenoids hypertrophy image recognition model after training; the third type of endoscopic data is used to compare the third initial adenoids hypertrophy image recognition model to determine the third adenoids hypertrophy image recognition model after training.

[0129] Exemplarily, the initial sub-adenoid hypertrophy image recognition model is a neural network model, such as the InceptionV3 model or the ResNet50 model. These neural network models have been preliminarily trained on large-scale image datasets and possess certain feature extraction capabilities. Fine-tuning the neural network model involves removing the last fully connected layer from the model and replacing it with a fully connected layer whose output number is equal to the number of types (here, for normal and hypertrophic adenoids, this can be set to 2). The weights of this fully connected layer, whose output number is equal to the number of types, are randomly initialized to obtain a new neural network model for nasal endoscopic image classification (for adenoids). Fine-tuning can make the model better suited to the characteristics of nasal endoscopic images and diagnostic tasks. The new neural network model is trained on each constructed training set using methods such as stochastic gradient descent. Simultaneously, data augmentation operations are performed on the input image, such as translation, scaling, rotation, left-right flipping, and left-right stretching. Translation simulates observation from different angles, scaling simulates observation from different distances, rotation simulates different probe orientations within the nasal cavity, and left-right flipping and left-right stretching increase image diversity. These data augmentation operations can improve the generalization ability of the model, enabling it to better handle various image situations in real applications.

[0130] S107. Based on the test set, the trained sub-adenoids hypertrophy image recognition model is tested to obtain a test result; if the test result meets a preset condition, the sub-adenoids hypertrophy image recognition model is determined, thereby determining the adenoids hypertrophy image recognition model.

[0131] In some embodiments of the present application, the test set also includes endoscopic test data of different image types, namely, first type endoscopic test data, second type endoscopic test data, and third type endoscopic test data. The preset condition is that the recognition rate of the test results reaches a preset recognition threshold.

[0132] In some embodiments of the present application, the trained first adenoids hypertrophy image recognition model is tested using a first type of nasal endoscopy test data to obtain a first test result. If the first test result meets a preset condition, the first adenoids hypertrophy image recognition model is determined.

[0133] In some embodiments of the present application, the trained second adenoids hypertrophy image recognition model is tested using a second type of nasal endoscopy test data to obtain a second test result. If the second test result meets a preset condition, the second adenoids hypertrophy image recognition model is determined.

[0134] In some embodiments of the present application, the trained third adenoids hypertrophy image recognition model is tested using a third type of nasal endoscopy test data to obtain a third test result. If the third test result meets a preset condition, the third adenoids hypertrophy image recognition model is determined.

[0135] In some embodiments of the present application, the first adenoids hypertrophy image recognition model, the second adenoids hypertrophy image recognition model, and the third adenoids hypertrophy image recognition model are determined as adenoids hypertrophy image recognition models.

[0136] For example, the performance of the neural network model obtained by training is evaluated using accuracy in multi-fold cross validation, and the neural network model with the highest accuracy is the optimal neural network model. Multi-fold cross validation is a commonly used evaluation method. By dividing the data set into different training sets and test sets multiple times, the model is trained and evaluated multiple times, and the average value is taken as the final evaluation result. To verify the performance of the optimal neural network model, for example, between normal adenoids and adenoid hypertrophy, the ROC curve of the optimal neural network model is drawn, and the true positive rate, true negative rate, false positive rate, and false negative rate of the doctors participating in the verification are plotted at the corresponding positions in the ROC curve. By comparing the ROC curve of the model with the doctor's diagnosis results, it can be intuitively judged whether the model can reach or exceed the performance of human experts. If the ROC curve of the optimal neural network model surrounds the doctor's result point, it means that the optimal neural network model can reach or exceed the performance of human experts and can be used for intelligent diagnosis of nasal endoscopic images of actual patients.

[0137] In some embodiments of the present application, determining the adenoids hypertrophy image recognition model may be achieved by following the steps below:

[0138] A. Dataset Construction

[0139] Using the nasal endoscopy database of a hospital's otolaryngology department, we collected nasal endoscopy images to form a nasal endoscopy dataset. When acquiring images, we should ensure the quality and integrity of the images to avoid blurry, damaged, or incomplete images.

[0140] Nasal endoscopic images are divided into two types according to the condition of adenoids (normal and hypertrophic) and annotated. The annotation process should be performed by professional doctors or medical experts to ensure the accuracy and consistency of the annotation.

[0141] Remove nasal endoscopic images that are blurry or do not capture the adenoids. These images may affect the training effect of the model, so they need to be filtered out.

[0142] Cases from the nasal endoscopy dataset were randomly selected and divided into multiple test and training sets using a 50-fold cross-validation method. During the division process, it was important to ensure that multiple images of the same case did not appear in both the training and test sets. In other words, all images of the case were either in the training set or in the test set.

[0143] B. Neural Network Model Training

[0144] Load pre-trained neural network models, such as the InceptionV3 model. When loading a model, ensure the integrity and accuracy of the model to avoid loading a damaged or incorrect model.

[0145] Fine-tune the pre-trained neural network model by removing the final fully connected layer and replacing it with a fully connected layer with two outputs, and randomly initialize the weights. The fine-tuning process should be tailored to the characteristics of the nasal endoscopic images and the diagnostic task to ensure that the model is better adapted to the new task.

[0146] A new neural network model is trained on each constructed training set, with the batch size set to an appropriate value (e.g., 72). The initial learning rate is 0.001, and the learning rate is decreased every 8 cycles, with the new learning rate set to 0.1 times the original learning rate. At the same time, data augmentation operations are performed on the input images, such as translation, scaling, rotation, left-right flipping, and left-right stretching. The specific parameters can be set according to actual conditions, such as random translation and left-right stretching with a maximum value of 0.1 times the image width, random scaling from 0.9 to 1.1 times, random rotation between 0 and 30 degrees, and left-right flipping with a 50% probability. During the training process, the model's training progress and performance indicators, such as the loss function value and accuracy, should be closely monitored, and training parameters should be adjusted in a timely manner to ensure that the model achieves optimal performance.

[0147] C. Model Validation

[0148] Use accuracy to evaluate the performance of the trained neural network model in a five-fold cross-validation. Select the neural network model with the highest accuracy as the optimal neural network model. During the evaluation process, ensure the accuracy and reliability of the evaluation method to avoid erroneous conclusions due to inappropriate evaluation methods.

[0149] The performance of the optimal neural network model was verified using AUC. A receiver operating characteristic (ROC) curve was plotted for the optimal neural network model, comparing normal adenoids with adenoid hypertrophy. The true positive rate, true negative rate, false positive rate, and false negative rate of the participating physicians were plotted at corresponding positions on the ROC curve. By comparing the model's ROC curve with the physicians' diagnostic results, it is possible to intuitively determine whether the model can match or exceed the performance of human experts. If the ROC curve of the optimal neural network model encompasses the physician's diagnostic point, it indicates that the optimal neural network model can match or exceed the performance of human experts and can be used for intelligent diagnosis of nasal endoscopic images of actual patients.

[0150] D. Image diagnosis

[0151] The patient's nasal endoscopic images are fed into the optimal neural network model, which then outputs a classification diagnosis of normal adenoids or adenoid hypertrophy. When inputting images, ensure that the format and size meet the model's requirements to avoid diagnostic errors due to incorrect image formatting.

[0152] At the same time, based on the aforementioned auxiliary illustration generation method, auxiliary illustrations can be provided for the classification diagnosis results to help doctors better understand the basis for the diagnosis. The auxiliary illustrations can intuitively display the model's analysis and judgment results for each pixel in the image, helping doctors further understand the model's working principles and diagnostic process.

[0153] It's understandable that using deep learning technology for intelligent imaging diagnosis of adenoid hypertrophy improves diagnostic accuracy and objectivity. By using deep learning algorithms to learn and analyze large amounts of image data, human interference is reduced. This reduces the workload of doctors, improves diagnostic efficiency, and provides patients with faster and more accurate diagnoses. This allows doctors to more efficiently handle the diagnostic needs of a large number of patients and provide them with more timely treatment recommendations.

[0154] Based on the above-mentioned method for recognizing adenoids hypertrophy by visual nasal endoscope, the present invention also provides an image recognition system for recognizing adenoids hypertrophy by visual nasal endoscope. Figure 4 As shown, Figure 4 The present invention provides a schematic structural diagram of an image recognition system for adenoids hypertrophy based on a visual nasal endoscope. The image recognition system 4 for adenoids hypertrophy based on a visual nasal endoscope includes: a visual nasal endoscope 40; the visual nasal endoscope 40 includes a mirror body 401, a detection probe 402 and a data transmission line 403; the mirror body 401 has a built-in image diagnostic device 4011; the detection probe 401 is connected to the image diagnostic device 4011 via a data transmission line 403; the mirror body 401 has a display screen 4012, wherein,

[0155] The detection probe 402 is used to obtain a nasal endoscopic image of the patient; wherein the nasal endoscopic image includes at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image; and transmit the nasal endoscopic image to the image diagnostic device;

[0156] The image diagnostic device 4011 is configured to determine an image type corresponding to the nasal endoscopic image; based on the correspondence between the image type and the image model, select a corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, identify the nasal endoscopic image using the sub-adenoids hypertrophy image recognition model, and determine a diagnosis of adenoids hypertrophy; wherein the adenoids hypertrophy image recognition model is trained using a plurality of different nasal endoscopic sample images; and the image model correspondence is a mapping relationship between a pre-set image type and a sub-adenoids hypertrophy image recognition model.

[0157] The display screen 4012 is used to display the diagnosis result of adenoids hypertrophy.

[0158] In some embodiments of the present application, the sub-adenoids hypertrophy image recognition model includes a first adenoids hypertrophy image recognition model, a second adenoids hypertrophy image recognition model, and a third adenoids hypertrophy image recognition model;

[0159] The image diagnostic device 4011 is further configured to, if the image type is the first image type, select the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the image model correspondence to identify the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy; wherein, the first image type represents that the nasal endoscopic image is any one of the nasal cavity image, the nasal passage image and the nasopharyngeal image; if the image type is the second image type, select the second adenoids hypertrophy image from the adenoids hypertrophy image recognition model based on the image model correspondence The recognition model identifies the nasal endoscopic image to determine the diagnosis result of the adenoids hypertrophy; wherein, the second image type represents that the nasal endoscopic image is any two images of the nasal cavity image, the nasal passage image and the nasopharynx image; if the image type is the third image type, then based on the image model correspondence, the third adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to identify the nasal endoscopic image to determine the diagnosis result of the adenoids hypertrophy; wherein, the third image type represents that the nasal endoscopic image is the nasal cavity image, the nasal passage image and the nasopharynx image.

[0160] In some embodiments of the present application, the image diagnostic device 4011 is also used to, if the image type is the first image type, crop the nasal endoscope image to determine the target area image containing the adenoids; perform data enhancement processing on the target area image to determine the target enhanced image; based on the correspondence between the first image type and the image model, select the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, perform multi-dimensional feature extraction on the target enhanced image to obtain texture features and color features corresponding to the target enhanced image; perform feature fusion on the texture features and the color features to obtain fusion features; and identify the fusion features to determine the diagnosis result of the adenoids hypertrophy.

[0161] In some embodiments of the present application, the image diagnostic device 4011 is also used to, if the image type is the second image type, crop and data enhance any two images of the nasal cavity image, the nasal passage image and the nasopharynx image respectively to determine a first region image and a second region image; based on the correspondence between the second image type and the image model, select the second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, perform multi-dimensional feature extraction and information fusion on the first region image and the second region image respectively to determine a first fusion feature and a second fusion feature; based on the first fusion feature and the second fusion feature, perform identification respectively to determine a first diagnostic result and a second diagnostic result; if the first diagnostic result and / or the second diagnostic result characterize adenoids hypertrophy, determine that the diagnostic result of the adenoids hypertrophy is adenoids hypertrophy; if the first diagnostic result and the second diagnostic result characterize normal adenoids, determine that the diagnostic result of the adenoids hypertrophy is normal.

[0162] In some embodiments of the present application, the image diagnostic device 4011 is also used to, if the image type is a third image type, select the third adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the third image type and the image model to perform feature extraction on the nasal endoscopy image, and determine the nasal cavity feature information, nasal passage feature information and nasopharyngeal feature information; obtain the weight values ​​corresponding to the nasal cavity feature information, the nasal passage feature information and the nasopharyngeal feature information; based on the weight values ​​corresponding to the nasal cavity feature information, the nasal passage feature information and the nasopharyngeal feature information, perform weighted processing on the nasal cavity feature information, the nasal passage feature information and the nasopharyngeal feature information to determine a diagnostic score; and determine the diagnostic result of the adenoids hypertrophy based on the diagnostic score.

[0163] In some embodiments of the present application, the nasal cavity feature information corresponds to a first weight value, the nasal passage feature information corresponds to a second weight value, and the nasopharyngeal feature information corresponds to a third weight value; the third weight value is greater than the second weight value; the second weight value is greater than the first weight value; the sum of the third weight value, the second weight value, and the first weight value is 1;

[0164] In some embodiments of the present application, the image diagnostic device 4011 is further configured to perform a multiplication operation based on the first weight value and the nasal cavity feature information to obtain a first score; perform a multiplication operation based on the second weight value and the nasal passage feature information to obtain a second score; perform a multiplication operation based on the third weight value and the nasopharyngeal feature information to obtain a third score; and perform a sum operation on the first score, the second score, and the third score to determine the diagnostic score;

[0165] and / or,

[0166] In some embodiments of the present application, the image diagnostic device 4011 is also used to determine that the diagnosis result of the adenoids hypertrophy is normal if the diagnostic score is less than a first threshold value; determine that the diagnosis result of the adenoids hypertrophy is mild adenoids hypertrophy if the diagnostic score is greater than or equal to the first threshold value and less than the second threshold value; determine that the diagnosis result of the adenoids hypertrophy is moderate adenoids hypertrophy if the diagnostic score is greater than or equal to the second threshold value and less than a third threshold value; and determine that the diagnosis result of the adenoids hypertrophy is severe adenoids hypertrophy if the diagnostic score is greater than or equal to the third threshold value.

[0167] In some embodiments of the present application, the image diagnostic device 4011 is also used to obtain a nasal endoscopy dataset before selecting the corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the image type and the image model; wherein the nasal endoscopy dataset includes normal adenoids images and adenoids hypertrophy images; the adenoids hypertrophy images include images of patients of different age groups, genders and disease stages; the nasal endoscopy dataset is divided into a test set and a training set; wherein the test set and the training set respectively include nasal endoscopy data of different image types; the initial sub-adenoids hypertrophy image recognition model is trained by the nasal endoscopy data of different image types in the training set to determine the trained sub-adenoids hypertrophy image recognition model; based on the test set, the trained sub-adenoids hypertrophy image recognition model is tested to obtain a test result; if the test result meets the preset condition, the sub-adenoids hypertrophy image recognition model is determined, thereby determining the adenoids hypertrophy image recognition model.

[0168] Based on the above-mentioned embodiment of the image recognition method of adenoids hypertrophy based on visual nasal endoscope, the embodiment of the present application further provides an image diagnostic device, such as Figure 5 As shown, Figure 5 This is a schematic diagram of the structure of an image diagnostic device provided in an embodiment of the present application. The image diagnostic device 4011 includes a processor 501 and a memory 502. The memory 502 is used to store a computer program; the processor 501 is used to load and execute the computer program from the memory to implement the image recognition method for adenoid hypertrophy based on visual nasal endoscopy as described in the above embodiment.

[0169] In the embodiment of the present application, the processor 501 may be at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, and a microprocessor. It is understood that for different devices, the electronic device used to implement the above-mentioned processor function may also be other, and the embodiment of the present application does not specifically limit this.

[0170] An embodiment of the present application provides a computer-readable storage medium storing a computer program for implementing the image recognition method for adenoid hypertrophy based on visual nasal endoscopy as described in any of the above embodiments when executed by a processor.

[0171] Illustratively, the program instructions corresponding to the image recognition method for adenoids hypertrophy based on visual nasal endoscopy in this embodiment can be stored on a storage medium such as a CD, a hard disk, or a USB flash drive. When the program instructions corresponding to the image recognition method for adenoids hypertrophy based on visual nasal endoscopy in the storage medium are read or executed by an electronic device, the image recognition method for adenoids hypertrophy based on visual nasal endoscopy as described in any of the above embodiments can be implemented.

[0172] In addition, the functional modules in the embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional modules.

[0173] If the integrated unit is implemented in the form of a software functional module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method of this embodiment. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0174] It should be understood that "one embodiment" or "an embodiment" or "some embodiments" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" or "in some embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The above-mentioned serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments. The above description of the various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other. For the sake of brevity, they will not be repeated here.

[0175] The modules described above as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules; they may be located in one place or distributed across multiple network units; some or all of the modules may be selected according to actual needs to achieve the purpose of this embodiment.

[0176] In addition, all functional modules in the embodiments of the present application can be integrated into one processing unit, or each module can be a separate unit, or two or more modules can be integrated into one unit; the above-mentioned integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0177] Those skilled in the art will understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM), magnetic disks or optical disks, and other media that can store program codes.

[0178] The methods disclosed in the several method embodiments provided in the embodiments of this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0179] The features disclosed in several product embodiments provided in the embodiments of this application can be arbitrarily combined to obtain new product embodiments without conflict.

[0180] The features disclosed in several method or device embodiments provided in the embodiments of this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0181] The above is merely an implementation of the embodiments of the present application, but the scope of protection of the embodiments of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the embodiments of the present application. Therefore, the scope of protection of the embodiments of the present application should be based on the scope of protection of the claims.

Claims

1. A method for image recognition of adenoid hypertrophy based on visual nasal endoscopy, characterized in that: Applied to visual nasal endoscopy, the method comprises: Acquire a nasal endoscopic image of the patient; wherein the nasal endoscopic image includes at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image; Determining an image type corresponding to the nasal endoscopic image; Based on the correspondence between the image type and the image model, selecting a corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, and recognizing the nasal endoscopy image using the sub-adenoids hypertrophy image recognition model to determine a diagnosis result of adenoids hypertrophy; The adenoids hypertrophy image recognition model is obtained by training with a plurality of different nasal endoscopy sample images; and the image model correspondence is a mapping relationship between a pre-set image type and a sub-adenoids hypertrophy image recognition model.

2. The method according to claim 1, characterized in that The sub-adenoids hypertrophy image recognition model includes a first adenoids hypertrophy image recognition model, a second adenoids hypertrophy image recognition model and a third adenoids hypertrophy image recognition model; The method further comprises: selecting a corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the image type and the image model, and recognizing the nasal endoscopy image by using the sub-adenoids hypertrophy image recognition model to determine a diagnosis result of adenoids hypertrophy. If the image type is the first image type, then based on the image model correspondence, selecting the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy; wherein the first image type indicates that the nasal endoscopic image is any one of the nasal cavity image, the nasal passage image, and the nasopharynx image; If the image type is the second image type, then based on the image model correspondence, selecting the second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy; wherein the second image type indicates that the nasal endoscopic image is any two images of the nasal cavity image, the nasal passage image, and the nasopharynx image; If the image type is the third image type, based on the image model correspondence, the third adenoids hypertrophy image recognition model is selected from the adenoids hypertrophy image recognition model to identify the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy; wherein, the third image type represents that the nasal endoscopic image is the nasal cavity image, the nasal passage image and the nasopharynx image.

3. The method according to claim 2, characterized in that If the image type is the first image type, then based on the image model correspondence, selecting the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy, including: If the image type is the first image type, cropping the nasal endoscopic image to determine a target area image containing adenoids; Performing data enhancement processing on the target area image to determine a target enhanced image; Based on the correspondence between the first image type and the image model, selecting the first adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to perform multi-dimensional feature extraction on the target enhanced image to obtain texture features and color features corresponding to the target enhanced image; The texture feature and the color feature are fused to obtain a fused feature; and the fused feature is identified to determine a diagnosis result of adenoid hypertrophy.

4. The method according to claim 2, characterized in that If the image type is the second image type, then based on the image model correspondence, selecting the second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy, including: If the image type is the second image type, performing cropping and data enhancement on any two of the nasal cavity image, the nasal passage image, and the nasopharynx image to determine a first region image and a second region image; Based on the correspondence between the second image type and the image model, selecting the second adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, performing multi-dimensional feature extraction and information fusion on the first region image and the second region image, respectively, to determine a first fusion feature and a second fusion feature; Based on the first fusion feature and the second fusion feature, identification is performed to determine a first diagnosis result and a second diagnosis result; If the first diagnostic result and / or the second diagnostic result indicates adenoids hypertrophy, determining the diagnostic result of adenoids hypertrophy as adenoids hypertrophy; If the first diagnostic result and the second diagnostic result indicate that the adenoids are normal, the diagnostic result of adenoid hypertrophy is determined to be normal.

5. The method according to claim 2, characterized in that If the image type is the third image type, then based on the image model correspondence, selecting the third adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model to recognize the nasal endoscopic image and determine the diagnosis result of the adenoids hypertrophy, including: If the image type is the third image type, selecting the third adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the third image type and the image model, performing feature extraction on the nasal endoscopy image to determine nasal cavity feature information, nasal passage feature information, and nasopharyngeal feature information; Obtaining weight values ​​corresponding to the nasal cavity feature information, the nasal passage feature information, and the nasopharyngeal feature information; performing weighted processing on the nasal cavity feature information, the nasal passage feature information, and the nasopharynx feature information based on the weight values ​​corresponding to each of the nasal cavity feature information, the nasal passage feature information, and the nasopharynx feature information to determine a diagnosis score; A diagnosis result of adenoid hypertrophy is determined based on the diagnosis score.

6. The method according to claim 5, characterized in that The nasal cavity feature information corresponds to a first weight value, the nasal passage feature information corresponds to a second weight value, and the nasopharyngeal feature information corresponds to a third weight value; the third weight value is greater than the second weight value; the second weight value is greater than the first weight value; the sum of the third weight value, the second weight value, and the first weight value is 1; The step of performing weighted processing on the nasal cavity feature information, the nasal passage feature information, and the nasopharyngeal feature information based on respective corresponding weight values ​​of the nasal cavity feature information, the nasal passage feature information, and the nasopharyngeal feature information to determine a diagnosis score includes: Performing a multiplication operation based on the first weight value and the nasal cavity feature information to obtain a first score; Performing a multiplication operation based on the second weight value and the nasal passage feature information to obtain a second score; Performing a multiplication operation based on the third weight value and the nasopharyngeal feature information to obtain a third score; performing an AND operation on the first score, the second score, and the third score to determine the diagnosis score; and / or, Determining the diagnosis result of adenoids hypertrophy based on the diagnosis score includes: If the diagnostic score is less than a first threshold, determining that the diagnosis result of adenoids hypertrophy is normal; If the diagnostic score is greater than or equal to the first threshold and less than the second threshold, the diagnosis result of the adenoids hypertrophy is determined to be mild adenoids hypertrophy; If the diagnostic score is greater than or equal to the second threshold and less than the third threshold, the diagnosis result of the adenoids hypertrophy is determined to be moderate adenoids hypertrophy; If the diagnosis score is greater than or equal to the third threshold, the diagnosis result of adenoids hypertrophy is determined to be severe adenoids hypertrophy.

7. The method according to any one of claims 1 to 6, characterized in that Before selecting the corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model based on the correspondence between the image type and the image model, the method further includes: Acquire a nasal endoscopy dataset; wherein the nasal endoscopy dataset includes normal adenoids images and adenoids hypertrophy images; the adenoids hypertrophy images include images of patients of different age groups, genders, and disease stages; Dividing the nasal endoscopy dataset into a test set and a training set; wherein the test set and the training set respectively include nasal endoscopy data of different image types; Training an initial adenoids hypertrophy sub-image recognition model using nasal endoscopy data of different image types in the training set, and determining a trained adenoids hypertrophy sub-image recognition model; Based on the test set, the trained sub-adenoids hypertrophy image recognition model is tested to obtain a test result; if the test result meets the preset conditions, the sub-adenoids hypertrophy image recognition model is determined, thereby determining the adenoids hypertrophy image recognition model.

8. An image recognition system for adenoid hypertrophy based on visual nasal endoscopy, characterized in that: The image recognition system for adenoids hypertrophy based on visual nasal endoscope comprises: a visual nasal endoscope; the visual nasal endoscope comprises a mirror body and a detection probe; the mirror body has a built-in image diagnostic device; the detection probe is connected to the image diagnostic device via a data transmission line; the mirror body has a display screen, wherein, The detection probe is used to obtain a nasal endoscopic image of the patient; wherein the nasal endoscopic image includes at least one of a nasal cavity image, a nasal passage image, and a nasopharyngeal image; and transmit the nasal endoscopic image to the image diagnostic device; The image diagnostic device is configured to determine an image type corresponding to the nasal endoscopic image; based on the correspondence between the image type and the image model, select a corresponding sub-adenoids hypertrophy image recognition model from the adenoids hypertrophy image recognition model, identify the nasal endoscopic image using the sub-adenoids hypertrophy image recognition model, and determine a diagnosis of adenoids hypertrophy; wherein the adenoids hypertrophy image recognition model is trained using a plurality of different nasal endoscopic sample images; and the image model correspondence is a mapping relationship between a pre-set image type and a sub-adenoids hypertrophy image recognition model; The display screen is used to display the diagnosis result of adenoids hypertrophy.

9. The system according to claim 8, characterized in that The image diagnostic device includes: a processor and a memory, wherein: The memory is used to store computer programs; The processor is configured to call and run the computer program from the memory to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Executable instructions are stored, which are used to cause a processor to execute and implement the method according to any one of claims 1 to 7.

Images