Intelligent tongue diagnosis feature extraction method for insomnia diagnosis and insomnia diagnosis system

Abstract

The present application belongs to the technical field of image processing, and particularly relates to an intelligent tongue diagnosis feature extraction method for insomnia diagnosis and an insomnia diagnosis system. The method of the present application comprises the following steps: step 1, insomnia patients are enrolled, and a tongue map database is constructed based on tongue images of the insomnia patients; step 2, a tongue region is segmented from the tongue images; step 3, imageomic features of the tongue region are quantified and extracted; and step 4, the imageomic features are screened through a machine learning algorithm to obtain tongue diagnosis features for constructing an insomnia diagnosis tongue diagnosis model. The features obtained by the above method can be used to further construct an insomnia diagnosis system. The present application has good diagnosis accuracy and good application prospect.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of image processing, and particularly relates to an intelligent tongue diagnosis feature extraction method for insomnia diagnosis and an insomnia diagnosis system. BACKGROUND

[0002] Insomnia is a serious public health problem, which is defined as difficulty falling asleep, difficulty maintaining sleep, or poor sleep quality despite having adequate sleep opportunities, and is accompanied by some form of daytime dysfunction. The prevalence of insomnia is about 10%-20%, of which chronic insomnia accounts for about 50%. The causes and physiology of insomnia are closely related to genetics, environment, behavior and physiological factors. In the prior art, the sleep quality of insomnia patients is evaluated by the Insomnia Severity Index (ISI) as the gold standard.

[0003] In modern medicine, we scientifically view insomnia from different angles, and at the same time, we can also learn from traditional medicine and quantitatively and scientifically design empirical knowledge. It can be found that in the traditional medicine of many countries and nations, there are their own opinions on the diagnosis of insomnia. Traditional Chinese medicine believes that the heart is the key organ of insomnia, and finally manifests as the balance or imbalance of the functional state of yin and yang. Therefore, traditional Chinese medicine can observe and judge the symptoms of insomnia of patients through various manifestations of yin and yang imbalance, such as changes in tongue state.

[0004] In the traditional Chinese medicine hospital, the traditional Chinese medicine doctor can diagnose insomnia through tongue diagnosis. From the perspective of traditional Chinese medicine, the state of the tongue is closely related to the physiology and pathology of the human body. In many past studies, tongue imaging has shown valuable information about the disease state for the diagnosis of various diseases. These experimental results show that in future research, new technologies may be used to quantify tongue information to achieve more accurate clinical diagnosis, especially for insomnia.

[0005] Artificial intelligence has been applied to public health and can bring unprecedented progress to medical practice. It is a very potential idea to use artificial intelligence to analyze tongue imaging to achieve insomnia severity analysis. However, so far, no intelligent classification tool for insomnia based on tongue images has been reported. For tongue images, how to extract image features and what kind of input features can achieve accurate insomnia diagnosis are still problems that need to be solved when constructing the intelligent classification tool for insomnia. SUMMARY

[0006] In view of the defects of the prior art, the application provides an intelligent tongue diagnosis feature extraction method for insomnia diagnosis and an insomnia diagnosis system, aiming to screen out features with good diagnostic performance for insomnia, and to construct an artificial intelligence system that can accurately diagnose the severity of insomnia.

[0007] An intelligent tongue diagnosis feature extraction method for insomnia diagnosis, comprising the following steps:

[0008] Step 1, insomnia patients are enrolled, and a tongue map database is constructed based on tongue images of the insomnia patients;

[0009] Step 2, a tongue region is segmented from the tongue image;

[0010] Step 3, imageomic features of the tongue region are quantified and extracted;

[0011] Step 4, the imageomic features are screened by a machine learning algorithm to obtain tongue diagnosis features for constructing an insomnia diagnosis tongue diagnosis model.

[0012] Preferably, in step 2, the tongue image is segmented by using an nnU-Net algorithm.

[0013] Preferably, in step 3, the imageomic features are quantified and extracted by using a Pyradiomics tool.

[0014] Preferably, in step 3, the obtained imageomic features include 14 shape features, 18 first-order statistical features and 73 texture features.

[0015] Preferably, in step 4, the machine learning algorithm is a LASSO algorithm.

[0016] Preferably, in step 4, the obtained tongue diagnosis features for constructing the insomnia diagnosis tongue diagnosis model include correlation information measurement, maximum two-dimensional diameter, elongation rate, energy, clustered shadow, gray variance, skewness.

[0017] Preferably, in step 4, when the training data of the insomnia diagnosis tongue diagnosis model is labeled, the insomnia severity rating scale is used to determine the insomnia degree.

[0018] The application also provides an intelligent tongue diagnosis system for insomnia diagnosis, comprising:

[0019] A tongue map database management module for acquiring, format converting and storing managing tongue image data;

[0020] An automatic segmentation module for automatically segmenting tongue images;

[0021] A feature quantization module for extracting and quantizing imageomic features;

[0022] An intelligent insomnia tongue diagnosis module for quantifying and diagnosing the insomnia degree of patients by an insomnia diagnosis tongue diagnosis model;

[0023] The tongue diagnosis features for constructing the insomnia diagnosis tongue diagnosis model are obtained by using the above intelligent tongue diagnosis feature extraction method.

[0024] Preferably, the insomnia diagnosis tongue diagnosis model is constructed based on a logistic regression algorithm, and the expression is: 0.085 x correlation information measurement + 0.044 x maximum two-dimensional diameter - 0.033 x elongation rate - 0.030 x energy + 0.014 x cluster shadow + 0.006 x gray scale variance - 0.003 x skewness.

[0025] The application further provides a computer readable storage medium, which stores a computer program for implementing the intelligent tongue diagnosis feature extraction method or a computer program for implementing the intelligent tongue diagnosis system.

[0026] The application preferably utilizes the feature types and model types for diagnosing insomnia by using tongue images, and by utilizing these preferred features, the application can construct an insomnia diagnosis tongue diagnosis model with high accuracy. Experiments show that the insomnia diagnosis tongue diagnosis model of the application has higher accuracy than the artificial tongue diagnosis results of some doctors. The application can meet the application requirements of clinics, can help to solve the diagnosis differences between different hospitals and doctors, can reduce the diagnosis cost of patients, and can provide technical support for early diagnosis and early treatment of the disease. Considering that insomnia has become an important disease affecting global health, the application has good application prospects.

[0027] Obviously, according to the above content of the application, according to the ordinary technical knowledge and common means in the art, other various forms of modifications, replacements or changes can be made without departing from the above basic technical ideas of the application.

[0028] The above content of the application will be further described in detail through the specific embodiments in the form of examples. However, this should not be understood as limiting the scope of the above subject matter of the application to the following examples. Any technology realized based on the above content of the application belongs to the scope of the application. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 The figure is a flowchart of embodiment 1 of the application.

[0030] Figure 2 The figure is a flowchart of embodiment 2 of the application. DETAILED DESCRIPTION

[0031] It should be particularly noted that the algorithms of the steps of data acquisition, transmission, storage and processing not specifically described in the embodiments, and the hardware structure, circuit connection and the like not specifically described can be realized through the existing technology.

[0032] Embodiment 1: Intelligent tongue diagnosis feature extraction method and model construction method for insomnia diagnosis

[0033] As Figure 1As shown, the embodiment provides an intelligent tongue diagnosis feature extraction method for insomnia diagnosis and a method of constructing a model using the obtained features, comprising:

[0034] S1, insomnia patients are enrolled, the degree of insomnia is determined, and tongue images are taken and collected;

[0035] Specifically, 399 insomnia patients were enrolled from two hospital centers. The degree of insomnia was determined according to the ISI standard, and tongue images were taken and collected. The data of one center was used as the training set, and the data of the other center was used as the test set. The degree of insomnia is divided into: no insomnia, mild, moderate and severe.

[0036] S2, convert the tongue image to a DICOM format file, and record the clinical information of the patient, manage the tongue map, and construct a tongue map database;

[0037] Specifically, the following steps are included:

[0038] S21, the format conversion of the tongue image is realized by a Python program;

[0039] S22, the DICOM format tongue image can record the information of the patient's gender, age, etc., for tongue map management, and construct a tongue map database. The database size of the embodiment is expected to reach thousands of cases.

[0040] S3, train a tongue map segmentation model based on the nnU-Net algorithm to automatically segment the tongue region;

[0041] Specifically, the nnU-Net algorithm is built using the Pytorch framework, supervised learning is performed through the artificially annotated tongue map database, the tongue map segmentation model is trained, and is used for automatic segmentation of other tongue images.

[0042] S4, quantize and extract the imageomic features of the tongue region in the tongue image;

[0043] Specifically, the following steps are included:

[0044] S41, the quantization and extraction of imageomic features is realized using the Pyradiomics tool;

[0045] S42, the quantized and extracted imageomic features include 14 shape features, 18 first-order statistical features and 73 texture features.

[0046] S5, screen valuable tongue features by a machine learning algorithm, and construct an insomnia diagnosis tongue diagnosis model;

[0047] Specifically, the following steps are included:

[0048] S51, the machine learning algorithm is a LASSO algorithm, and is implemented by using Scikit-Learn;

[0049] S52, the insomnia diagnosis tongue diagnosis model is constructed based on valuable features screened by the LASSO algorithm and feature weights, and a model calculation formula is: 0.085 x correlation information measurement + 0.044 x maximum two-dimensional diameter - 0.033 x elongation rate - 0.030 x energy + 0.014 x clustered shadow + 0.006 x gray level variance - 0.003 x skewness.

[0050] Embodiment 2: An intelligent tongue diagnosis system for insomnia degree classification

[0051] As shown in Figure 2 , the embodiment provides an intelligent tongue diagnosis system for insomnia degree diagnosis, which comprises:

[0052] a tongue image database management module for acquiring, format converting and storing managing tongue image data;

[0053] an automatic segmentation module for automatically segmenting tongue images;

[0054] a feature quantization module for extracting and quantitatively processing radiomics features;

[0055] an intelligent insomnia tongue diagnosis module for quantitatively and diagnosing insomnia degrees of patients by an insomnia diagnosis tongue diagnosis model;

[0056] wherein the insomnia diagnosis tongue diagnosis model is obtained by the method in Embodiment 1, and an expression of the final model is: 0.085 x correlation information measurement + 0.044 x maximum two-dimensional diameter - 0.033 x elongation rate - 0.030 x energy + 0.014 x clustered shadow + 0.006 x gray level variance - 0.003 x skewness.

[0057] The technical solutions of the present application are further described below through experiments.

[0058] Experimental Example 1: Performance analysis of an insomnia diagnosis tongue diagnosis model

[0059] In Embodiment 1, the feature data set screened by the LASSO algorithm is finally constructed as:

[0060]

[0061]

[0062] The calculation formula of the insomnia diagnosis tongue diagnosis model is: 0.085 x correlation information measurement + 0.044 x maximum two-dimensional diameter - 0.033 x elongation rate - 0.030 x energy + 0.014 x clustered shadow + 0.006 x gray level variance - 0.003 x skewness.

[0063] The diagnostic performance of the above model in the test set was tested, and the diagnostic performance of the insomnia diagnosis tongue diagnosis model finally constructed was:

[0064]

[0065] It can be seen that the insomnia diagnosis tongue diagnosis model obtained by the present application performs very well in accuracy, precision, sensitivity, specificity and F1 index.

[0066] Experimental Example 2: Comparison of intelligent tongue diagnosis system for insomnia degree diagnosis with TCM doctor diagnosis

[0067] I. Experimental method

[0068] In order to highlight the advantages of the tongue diagnosis of the present application compared with the tongue diagnosis of the clinical TCM doctor, the insomnia diagnosis tongue diagnosis model of Example 1 was used to compare the diagnostic performance with three TCM doctors with different seniority on the same tongue image data set, and the results are shown in the following table:

[0069]

[0070]

[0071] Among them, doctors 1-3 are resident doctors, attending doctors and deputy chief physicians. The results show that the intelligent tongue diagnosis system for insomnia diagnosis has certain advantages compared with the artificial diagnosis of doctors, and can be used for clinical application.

[0072] From the above examples and experimental examples, it can be seen that the present application constructs an insomnia diagnosis method and system based on tongue image, which has good diagnostic accuracy and good application prospect.

Claims

1. An intelligent tongue diagnosis feature extraction method for insomnia diagnosis, characterized in that, The method comprises the following steps: Step 1, recruit insomnia patients, and build a tongue atlas database based on tongue images of the insomnia patients; Step 2, segment a tongue region from the tongue image; Step 3, quantify and extract imageomic features of the tongue region; Step 4, screen the imageomic features by a machine learning algorithm to obtain tongue diagnosis features for building an insomnia diagnosis tongue diagnosis model; In step 3, the obtained imageomic features include 14 shape features, 18 first-order statistical features and 73 texture features; In step 4, the machine learning algorithm is a LASSO algorithm; and the tongue diagnosis features for building the insomnia diagnosis tongue diagnosis model include correlation information measurement, maximum two-dimensional diameter, elongation rate, energy, clustered shadow, gray variance and skewness.

2. The intelligent tongue diagnosis feature extraction method according to claim 1, characterized in that: In step 2, the tongue image is segmented by using an nnU-Net algorithm.

3. The intelligent tongue diagnosis feature extraction method according to claim 1, wherein: In step 3, the imageomic features are quantified and extracted by using a Pyradiomics tool.

4. The intelligent tongue diagnosis feature extraction method according to claim 1, characterized in that: In step 4, when the training data of the insomnia diagnosis tongue diagnosis model is labeled, an insomnia severity index is used to determine the insomnia degree.

5. An intelligent tongue diagnosis system for insomnia diagnosis, characterized in that, The system comprises: a tongue atlas database management module for acquiring, format converting and storing and managing tongue image data; an automatic segmentation module for automatically segmenting the tongue image; a feature quantification module for extracting and quantifying imageomic features; an intelligent insomnia tongue diagnosis module for quantifying and diagnosing the insomnia degree of a patient by using the insomnia diagnosis tongue diagnosis model; The tongue diagnosis features for building the insomnia diagnosis tongue diagnosis model are obtained by using the intelligent tongue diagnosis feature extraction method according to any one of claims 1-4.

6. The intelligent tongue diagnosis system according to claim 5, characterized in that: The insomnia diagnosis tongue diagnosis model is built based on a logistic regression algorithm, and its expression is: 0.085 x correlation information measurement + 0.044 x maximum two-dimensional diameter - 0.033 x elongation rate - 0.030 x energy + 0.014 x clustered shadow + 0.006 x gray variance - 0.003 x skewness.

7. A computer-readable storage medium, characterized in that: The computer program for implementing the intelligent tongue diagnosis feature extraction method according to any one of claims 1-4, or the computer program for implementing the intelligent tongue diagnosis system according to claim 5 or 6 is stored on the computer readable medium.

Images