A portable eye medical image intelligent auxiliary diagnosis device

Abstract

The application relates to a portable eye medical image intelligent auxiliary diagnosis device, which comprises a portable eye medical image acquisition device and a portable eye medical image intelligent auxiliary diagnosis box, the latter comprising an embedded device and a display device connected thereto, the embedded device being pre-burned with an intelligent auxiliary diagnosis program configured to perform the following steps: reading an eye medical image acquired by the acquisition device; segmenting the eye medical image by using an image segmentation algorithm to extract a sclera region image; inputting the sclera region image into a pre-trained neural network model for identification analysis to obtain auxiliary diagnosis information; and controlling the display device to display the auxiliary diagnosis information. The application can be applied to various scenes, reduces the dependence of medical image identification on the professional knowledge and experience of doctors, significantly reduces the misjudgment of diagnosis caused by human subjective factors, has high portability, can display auxiliary diagnosis information in real time, and has high real-time performance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical intelligent auxiliary diagnosis, in particular to a portable eye medical image intelligent auxiliary diagnosis device. BACKGROUND

[0002] Eye medical images are important basis for medical diagnosis of eye diseases and hepatobiliary diseases, but the existing medical diagnosis based on eye medical images needs to rely on the medical knowledge and experience of doctors, and is highly subjective, for example:

[0003] (1) Eye disease (such as dry eye disease) medical image auxiliary diagnosis

[0004] With the popularity of electronic products, the increasing burden of eye use, etc., dry eye disease has become the most common ophthalmic disease other than refractive error, and there are about 400 million dry eye patients in China, and the dry eye detection rate is also increasing year by year. At present, dry eye diagnosis in China needs to be carried out in the dark room of professional diagnosis and treatment institutions, and large UHR-OCT or Placido disc equipment is used to check the eye conditions of patients. Doctors diagnose eye diseases according to medical images based on medical knowledge and experience.

[0005] (2) Other diseases (such as hepatobiliary diseases) are diagnosed by eye images

[0006] With the acceleration of social life rhythm, the population size of hepatobiliary diseases is increasing year by year, and there are nearly 100 million patients at present, and the disease has the characteristics of infectious foci, which brings many inconveniences to diagnosis and treatment. At present, hepatobiliary doctors mainly judge the disease condition by checking the degree of scleral jaundice of patients through non-invasive examination method. This method is highly subjective and cannot be realized.

[0007] Through the above analysis, it can be seen that the doctor's judgment of the degree of scleral jaundice and the diagnosis of dry eye disease has a strong subjectivity, and a large amount of experience accumulation is needed. The degree of scleral jaundice is not suitable for quantification, and the accuracy of manual interpretation is closely related to the experience of doctors, and the objectivity is insufficient.

[0008] The prior art also proposes a method of using image recognition algorithm to interpret medical images to assist doctors in diagnosis. This method needs to use large special equipment to collect eye data of patients, obtain eye images, tear film break-up time, etc. Then the data is transmitted to the special PC / server end for processing and analysis to obtain eye red analysis, tear film profile, etc. to assist doctors in diagnosis. But this auxiliary diagnosis method needs to be matched with a computer / server equipped with intelligent recognition software, and also has certain requirements for the environment, loses portability, and cannot cope with the scene of carrying medical image collection equipment, such as providing mobile detection for patients with difficulty in movement, home self-checking of patients, medical support, rescue, etc. SUMMARY

[0009] To solve the problems in the prior art that the medical diagnosis assistance method based on eye medical images needs to rely on the medical professional knowledge and experience of doctors, is subjective, and adopts the form of a server carrying intelligent recognition software to assist in diagnosis, does not have portability and cannot adapt to the needs of various scenes, a portable eye medical image intelligent assistance diagnosis device is provided, which adopts an embedded device with information display and edge computing capability, carries an image recognition and information processing module, can be flexibly deployed to a medical image acquisition terminal, and supports medical support, rescue and patient home self-test and various portable use scenes.

[0010] To solve the above problems, the present application adopts the following technical solutions:

[0011] A portable eye medical image intelligent assistance diagnosis device, comprising a portable eye medical image acquisition device and a portable eye medical image intelligent assistance diagnosis box, the portable eye medical image intelligent assistance diagnosis box comprising an embedded device and a display device connected with the embedded device, the embedded device having a pre-burned intelligent assistance diagnosis program, the intelligent assistance diagnosis program being configured to perform the following steps:

[0012] reading an eye medical image acquired by the portable eye medical image acquisition device;

[0013] segmenting the eye medical image by using an image segmentation algorithm to extract a sclera region image;

[0014] inputting the sclera region image into a pre-trained neural network model, the neural network model performing identification analysis on the sclera region image to obtain assistance diagnosis information;

[0015] controlling the display device to display the assistance diagnosis information.

[0016] Compared with the prior art, the present application has the following beneficial effects:

[0017] The portable eye medical image intelligent assistance diagnosis device provided by the present application can make patient examination and post-treatment review more convenient compared with the current domestic large-scale detection equipment, can cope with patient home self-test, provide inpatient examination in the ward, medical support, rescue and various scenes, and can reduce the dependence of medical image identification on the professional knowledge and experience of doctors, significantly reduce the diagnosis errors caused by human subjective factors, provide intelligent diagnosis assistance, have high portability, can display assistance diagnosis information in real time, and have strong real-time performance. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1A principle schematic view of a portable eye medical image intelligent auxiliary diagnosis device according to the present application;

[0019] Figure 2 An application scenario schematic view of a portable eye medical image intelligent auxiliary diagnosis device according to the present application. DETAILED DESCRIPTION

[0020] The technical solutions of the present application will be described in detail below with reference to the accompanying drawings and preferred embodiments.

[0021] The present application provides a portable eye medical image intelligent auxiliary diagnosis device, as shown in the figure, which comprises a portable eye medical image acquisition device and a portable eye medical image intelligent auxiliary diagnosis box, wherein the portable eye medical image intelligent auxiliary diagnosis box specifically comprises an embedded device and a display device, and the display device is connected with the embedded device. Figure 1 The intelligent auxiliary diagnosis device of the present application collects the ocular surface image of the patient through the portable eye medical image acquisition device, obtains the eye medical image of the patient, then transmits the eye medical image of the patient to the embedded device, pre-burns the intelligent auxiliary diagnosis program in the embedded device, processes and analyzes the image through the intelligent auxiliary diagnosis program, thereby obtaining the corresponding auxiliary diagnosis information, and finally displays the auxiliary diagnosis information through the display device, so as to facilitate the user to quickly obtain the auxiliary diagnosis information.

[0022] At present, deep learning technology is mainly used for image recognition, and deep learning has high computational overhead and high requirements for hardware devices. The edge computing device can provide the computing capability to meet the real-time image recognition of the deep learning algorithm, has low cost, is convenient to deploy in the medical image acquisition terminal, has high response performance, and can add the intelligent auxiliary diagnosis function to the medical image acquisition terminal, so that the intelligent auxiliary diagnosis based on the edge computing technology is used in the present application. The function of the embedded device is the auxiliary diagnosis based on the medical image, including but not limited to the judgment of dry eye, liver and gallbladder diseases, etc. The embedded device in the present application generally refers to all portable embedded devices to meet the requirement of convenience, and the embedded device can be optionally an image recognition chip loaded with an image recognition and information processing module, wherein the image recognition and information processing module is loaded with various image segmentation, image recognition, machine learning regression or traditional regression algorithms, so as to realize the portable auxiliary diagnosis, real-time data acquisition and image discrimination result.

[0023] Code of image segmentation, image recognition, etc. And combined into an intelligent auxiliary diagnosis program including dry eye syndrome auxiliary diagnosis, liver and gallbladder disease auxiliary diagnosis and other function modules, and then the intelligent auxiliary diagnosis program is pre-burned into the embedded device to complete the function preparation of the portable medical image intelligent auxiliary diagnosis box. Specifically, the intelligent auxiliary diagnosis program pre-burned in the embedded device is configured to perform the following steps:

[0024] Step one: reading the eye medical image collected by the portable eye medical image collection device.

[0025] The eye medical image is an important basis for medical diagnosis of eye diseases and liver and gallbladder diseases. The present application collects the eye medical image of the patient to be diagnosed or other images used for medical diagnosis or research by using the portable eye medical image collection device, which can be but not limited to a special camera and other devices.

[0026] Step two: segmenting the eye medical image by using an image segmentation algorithm to extract the sclera region image.

[0027] This step segments the eye medical image of the patient to be diagnosed read in step one to extract the image of the part to be analyzed (i.e. the sclera part) to obtain the corresponding sclera region image.

[0028] When segmenting the eye medical image by using the image segmentation algorithm, a variety of algorithms can be used for segmentation, such as RGB color threshold segmentation, HSV color threshold segmentation, gray threshold segmentation, and expansion and corrosion operation fine tuning on the eye medical image, so as to obtain the complete sclera region image as much as possible.

[0029] Step three: inputting the sclera region image into a pre-trained neural network model, and the neural network model recognizing and analyzing the sclera region image to obtain auxiliary diagnosis information.

[0030] With the rapid development and wide application of artificial intelligence, image processing, image recognition and information analysis technology are increasingly mature, and can replace manual image interpretation to a certain extent. The intelligent medical image discrimination technology used in the present application mainly includes image segmentation technology such as watershed algorithm, Canny operator image segmentation, and intelligent image segmentation algorithm such as neural network feature extraction algorithm. The image recognition technology generally uses machine learning method, uses neural network model for supervised learning, so as to achieve the effect of picture recognition. Information processing and analysis mainly dig the relationship between data through data mining, and the traditional method is least square method, and the intelligent algorithm generally uses machine learning method, and uses neural network model for regression.

[0031] The present application utilizes the built training data set to train the neural network by the method of supervised learning, and the structure of the neural network model is as follows:

[0032] (1) taking a full connection layer as an input layer, and the number of nodes is the number of image pixels;

[0033] (2) taking 4 continuous convolution layers + a Relu activation function as a hidden layer of the neural network, and the number of nodes is 128;

[0034] (3) adding a full connection layer after the hidden layer, and the number of nodes is 128;

[0035] (4) taking a full connection layer + a softmax function as an output layer, and the number of nodes is the number of possible results.

[0036] It should be pointed out that the neural network model used for image recognition in the present application includes but is not limited to the above model structure, and for those skilled in the art, a number of deformations can be made on the basis of the above neural network model.

[0037] The sclera region image is input into the pre-trained neural network model, the neural network model recognizes and analyzes the sclera region image, for example, taking dry eye as an example, when the eye medical image is the eye image of a dry eye patient, the sclera region image is input into the pre-trained neural network model, after model recognition and analysis, the output is the probability of the existence of dry eye disease and the degree of disease, and other auxiliary diagnosis information; taking the judgment of sclera yellowing degree as an example, when the eye medical image is the eye image of a patient with liver and gallbladder disease, the sclera region image is input into the pre-trained neural network model, after model recognition and colorimetric analysis, the output is the predicted serum bilirubin content and other auxiliary diagnosis information.

[0038] In order to obtain the pre-trained neural network model and realize the intelligent recognition function of the portable eye medical image intelligent auxiliary diagnosis box, the training data set is pre-prepared, and the pre-prepared training data set is trained by the method of supervised learning. According to the different diagnosis of the portable eye medical image intelligent auxiliary diagnosis device, different training data sets are established, for example: taking dry eye as an example, the training data set is obtained by the following steps:

[0039] (1) collecting a large number of eye images of dry eye patients;

[0040] (2) segmenting all the collected eye images to extract the eye sclera image, the segmentation method used in this step can be an image segmentation algorithm in the prior art;

[0041] (3) using expert scoring method, the field experts score the probability of the eye sclera image having dry eye disease and the degree of the disease, then record and store the eye sclera image and the corresponding score, finally obtain the training data set, train the neural network model using the training data set, and obtain the trained neural network model, which can be used for auxiliary diagnosis of dry eye disease.

[0042] In addition, taking the judgment of the degree of scleral yellowing as an example, the training data set is obtained by the following steps:

[0043] (1) Collect a large number of eye images of patients with hepatobiliary diseases and healthy people;

[0044] (2) Segment all collected eye images to extract eye sclera images. Similarly, the segmentation method used in this step can be an image segmentation algorithm in the prior art;

[0045] (3) Detect the serum bilirubin content of the human body corresponding to the collected eye image, record and store the eye sclera image and the corresponding serum bilirubin content, obtain the training data set, train the neural network model using the training data set, and obtain the trained neural network model, which can be used for auxiliary diagnosis of hepatobiliary diseases.

[0046] Step four: controlling the display device to display the auxiliary diagnosis information.

[0047] The present application displays the auxiliary diagnosis information on the display screen connected to the embedded device, which is convenient for users to quickly obtain the auxiliary diagnosis information. Taking dry eye disease as an example, the display device displays the probability of the presence or absence of dry eye disease and the degree of the disease. Taking hepatobiliary disease as an example, the display device displays the predicted serum bilirubin content.

[0048] The portable eye medical image intelligent auxiliary diagnosis device solves the problems of the prior art that medical image interpretation relies on the professional knowledge and experience of doctors, which is subjective, and that medical image intelligent auxiliary diagnosis mostly adopts the form of servers carrying intelligent recognition software, which is not portable. It can be applied to patient home self-checking, providing inpatient examination and medical support in the ward for patients with difficulty in movement, and various scenes such as medical rescue, as shown in Figure 2 The implementation method of patient home self-checking or doctor auxiliary diagnosis is as follows:

[0049] (1) Use a portable eye medical image acquisition device to acquire eye medical images;

[0050] (2) The eye medical image data is transmitted into the portable eye medical image intelligent auxiliary diagnosis box, image recognition is carried out by the embedded device (such as an image recognition chip) carrying a pre-trained neural network model, and the auxiliary diagnosis information is displayed by the display device (such as an information display screen);

[0051] (3) The doctor or the patient who carries out self-checking at home can quickly obtain the auxiliary diagnosis result by reading the information display screen of the portable eye medical image intelligent auxiliary diagnosis box.

[0052] Compared with the large domestic detection equipment, the portable eye medical image intelligent auxiliary diagnosis device of the application can make the patient examination and postoperative review more convenient, and can cope with patient self-checking at home, provide ward examination for patients who are inconvenient to move, and provide medical support, rescue and other scenes.

[0053] The intelligent recognition technology of medical images can reduce the dependence of medical image identification on the professional knowledge and experience of doctors, and significantly reduce the misjudgment of diagnosis caused by human subjective factors. The existing medical image intelligent recognition technology mostly adopts the form of servers carrying intelligent recognition software. When used, the image data of the medical image acquisition device is connected to the server for intelligent recognition. This way cannot support the portable use of the medical image acquisition device. In order to provide intelligent diagnosis assistance for portable medical image acquisition devices, portability needs to be considered, and the image identification result needs to be displayed in real time, that is, the auxiliary diagnosis information needs to be displayed in real time, and the real-time performance is strong.

[0054] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above-mentioned embodiments are not described, but as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0055] The above-mentioned embodiments only express several embodiments of the application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, some modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A portable intelligent auxiliary diagnosis device for eye medical image, characterized in that, The application relates to a portable eye medical image acquisition device and a portable eye medical image intelligent auxiliary diagnosis box, wherein the portable eye medical image intelligent auxiliary diagnosis box comprises an embedded device and a display device connected with the embedded device; the embedded device is provided with an image recognition chip loaded with an image recognition and information processing module; an intelligent auxiliary diagnosis program is pre-burned in the embedded device; the intelligent auxiliary diagnosis program is configured to perform the following steps: reading an eye medical image acquired by the portable eye medical image acquisition device; segmenting the eye medical image by using an image segmentation algorithm to extract a sclera region image; the segmentation of the eye medical image by using the image segmentation algorithm comprises the following processes: sequentially performing RGB color threshold segmentation, HSV color threshold segmentation, grayscale threshold segmentation and inflation corrosion operation fine adjustment on the eye medical image; inputting the sclera region image into a pre-trained neural network model; the neural network model performs recognition analysis on the sclera region image to obtain auxiliary diagnosis information; the structure of the neural network model is as follows: (1) taking a full connection layer as an input layer, and the number of nodes is the number of image pixels; (2) taking four continuous convolution layers and a Relu activation function as hidden layers of the neural network, and the number of nodes is 128; (3) adding one full connection layer after the hidden layers, and the number of nodes is 128; (4) taking a full connection layer and a softmax function as an output layer, and the number of nodes is the number of possible results; controlling the display device to display the auxiliary diagnosis information.

2. The portable eye medical image intelligent assistant diagnostic device according to claim 1, wherein, The eye medical image is an eye image of a dry eye patient, and the auxiliary diagnosis information comprises the presence or absence of a dry eye disease and the probability of the disease degree.

3. The portable intelligent auxiliary diagnostic apparatus for eye medical imaging according to claim 2, wherein, The neural network model is trained by using a training data set through a supervised learning method; the training data set is obtained through the following steps: collecting a large number of eye images of dry eye patients; segmenting all the collected eye images to extract eye sclera images; adopting an expert scoring method to score the probability of the dry eye disease and the disease degree of the eye sclera images, recording and storing the eye sclera images and the corresponding scores to obtain the training data set.

4. The portable intelligent auxiliary diagnostic apparatus for eye medical imaging according to claim 1, wherein, The eye medical image is an eye image of a liver and gall disease patient, and the auxiliary diagnosis information comprises serum bilirubin content.

5. The portable intelligent auxiliary diagnostic apparatus for eye medical imaging according to claim 4, wherein, The neural network model is trained by using a training data set through a supervised learning method; the training data set is obtained through the following steps: collecting a large number of eye images of liver and gall disease patients and healthy people; segmenting all the collected eye images to extract eye sclera images; detecting the serum bilirubin content of the human body corresponding to the collected eye images, recording and storing the eye sclera images and the corresponding serum bilirubin content to obtain the training data set.

Images