Female anemia risk assessment method

By constructing a risk assessment model for women with anemia based on facial and tongue image characteristics, the problem of difficulty in accurately screening and evaluating anemia in the prior art is solved, and a non-invasive and accurate risk assessment of anemia is achieved, which is suitable for a variety of medical scenarios.

CN118942698APending Publication Date: 2024-11-12SHANGHAI UNIV OF T C M
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410983808.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

It is difficult to accurately screen and evaluate the anemia situation in women in areas with scarce medical resources. The traditional method of obtaining blood samples is invasive and inconvenient, and there is a lack of simple, effective and accurate anaemia risk assessment methods.

Method used

By constructing a female anemia risk assessment model based on facial and tongue image characteristics, using color space values ​​and height quality index, the Logit model is used to predict the probability of anemia, and combining LASSO regression and logistic regression analysis to analyze the screening characteristics, an accurate anemia risk assessment model is constructed.

Benefits of technology

It realizes a non-invasive and accurate risk assessment of anemia in women, is suitable for different age groups, reduces the risk of model overfitting, and provides a more reliable auxiliary diagnostic tool for scenarios where medical resources are abundant or scarce.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942698A_ABST
    Figure CN118942698A_ABST
Patent Text Reader

Abstract

The invention discloses a female anemia risk assessment method, and belongs to the technical field of image information processing. The female anemia risk assessment method comprises the following steps: according to a face image of a female individual subject, obtaining a color space value of a selected area in the face of the female individual subject; and obtaining the anemia probability of the female individual subject according to the color space value of the selected area and the height quality index of the female individual subject. According to the female anemia risk assessment method, the female anemia probability is accurately predicted through the face image and the tongue image of the subject.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image information processing, and in particular to a method for assessing the risk of anemia in women. Background Art

[0002] According to the World Health Organization, anemia is defined as a hemoglobin concentration below the standard value, such as hemoglobin concentration of men <130g / L, non-pregnant women <120g / L, pregnant women <110g, etc. In addition, normal hemoglobin levels are also affected by age, race, and altitude. Anemia is a global health problem. According to data from the World Health Organization, anemia affects 1.62 billion people, equivalent to 24.8% of the world's population, especially the health impact on women. Among women worldwide, the prevalence and average severity of anemia are high, with about 29% of non-pregnant women and 38% of pregnant women suffering from anemia.

[0003] Iron deficiency anemia can also affect cognitive ability, motor ability, etc. due to reduced blood oxygen transport and reduced oxidative capacity of iron-deficient cells. Preoperative anemia can also increase the risk of intraoperative blood loss or death, increase the risk of postoperative complications, and prolong recovery time.

[0004] Active intervention in the early stages of anemia can prevent further development of anemia and is also the main way to control the prevalence of anemia. Among them, a complete blood cell test is the main criterion for determining whether anemia is present, but blood samples are obtained through invasive punctures and require the presence of professional medical personnel and specialized testing equipment. For areas where medical and health resources are relatively scarce, how to accurately obtain information on hemoglobin levels is crucial to determining the anemia status of women and reducing the probability of severe anemia in women.

[0005] Therefore, it is necessary to optimize the screening and assessment methods for anemia in women and provide a simpler, more effective and more accurate method for assessing the risk of anemia in women. Summary of the invention

[0006] In order to solve at least one aspect of the above problems and defects in the prior art, an embodiment of the present invention provides a method for assessing the risk of anemia in women, which obtains the anemia probability of individual female subjects by constructing a female anemia risk assessment model based on the features of facial images and tongue images. The technical solution is as follows:

[0007] According to one aspect of the present invention, a method for assessing anemia risk in women is provided.

[0008] The method for assessing anemia risk in women includes the following steps:

[0009] Based on the facial image of the female individual subject, obtaining a color space value of a selected region in the face of the female individual subject;

[0010] The anemia probability of the individual female subject is obtained according to the color space value of the selected region and the body mass index of the individual female subject.

[0011] According to the color space value of the selected area and the body mass index of the female individual subject, the anemia probability of the female individual subject is obtained, including: the color space value of the selected area and the body mass index of the female individual subject are used based on the female anemia risk assessment model to obtain the anemia probability.

[0012] In some embodiments, the female anemia risk assessment model is expressed as:

[0013] Logit(P)=P 0 +k 1 *BMI+k 2 *A 1 +k 3 *A 2 +k 4 *B 1 +k 5 *B 2 +k 6 *C 1 +k 7 *D 1 ;

[0014] In the formula, Logit(P) represents the probability of anemia in female individual subjects, P 0 represents a constant, k 1 , k 2 , k 3 , k 4 , k 5 , k 6 and k 7 denote coefficients, BMI denotes the body mass index of female individual subjects, A 1 represents the first color space value of the first selected region of the female individual subject, A 2 The second color space value representing the first selected area of ​​the female individual subject, B 1 The first color space value of the second selected region representing the female individual subject, B 2 A second color space value representing a second selected region of the female individual subject, C 1 Represents the first color space value of the third selected area of ​​the female individual subject, D 1 A first color space value representing a fourth selected region of the female individual subject.

[0015] In some embodiments, P 0 The value range of k is set to -16 to -12. 1The value range of k is set to 0.09~0.3, 2 The value range of k is set to 0.1~0.4, 3 The value range of k is set to 0.8~0.3, 4 The value range of k is set to 0.1~0.4, 5 The value range of k is set to -0.4~0, 6 The value range of k is set to -0.7 to -0.3, 7 The value range is set to 0.1~0.4.

[0016] In some embodiments, the first selected area is the lips, the second selected area is the cheeks, the third selected area is the cheeks, and the fourth selected area is the nose.

[0017] In some embodiments, when the color space is the Lab color space, P 0 The value of is set to -14.697, k 1 The value of is set to 0.167, k 2 The value of is set to 0.214, k 3 The value of is set to 0.155, k 4 The value of is set to 0.294, k 5 The value of is set to -0.216, k 6 The value of is set to -0.535, k 7 The value of is set to 0.236.

[0018] In some embodiments, the method for constructing a female anemia risk assessment model comprises the following steps:

[0019] According to the facial images of female anemia subjects, the facial anemia color features and the facial anemia texture features of the group are obtained;

[0020] According to the tongue images of the female anemia subjects, the color features and texture features of the anemia tongue of the group are obtained;

[0021] According to the group facial anemia color feature, the group facial anemia texture feature, the group tongue anemia color feature, and the group tongue anemia texture feature, a first screening is performed on the group facial anemia color feature, the group tongue anemia color feature, the group facial anemia texture feature, and the group tongue anemia texture feature to obtain the group facial anemia color feature, the group tongue anemia color feature, and the group tongue anemia texture feature after the first screening;

[0022] Perform a second screening based on the facial anemia color features, tongue anemia color features, and tongue anemia texture features of the group after the first screening to obtain the facial anemia color features of the group after the second screening;

[0023] Based on the facial anemia color characteristics of the group after the second screening, a female anemia risk assessment model was constructed.

[0024] In some embodiments, the first screen is a LASSO regression analysis and the second screen is a logistic regression analysis.

[0025] In some embodiments, the group facial anemia color features obtained based on the facial image include the anemia color features of the entire face of the female group anemia subjects and the anemia color features of each area in the face, and the group facial anemia texture features obtained based on the facial image include the facial anemia texture granularity features, facial anemia texture density features, and facial anemia texture depth features of the female group anemia subjects;

[0026] The group tongue anemia color features obtained based on tongue images include the group tongue coating anemia color features and the group tongue quality anemia color features. The group tongue anemia texture features obtained based on tongue images include the group tongue coating anemia texture granularity features, the group tongue quality anemia texture granularity features, the group tongue coating anemia texture density features, the group tongue quality anemia texture density features, the group tongue coating anemia texture depth features and the group tongue quality anemia texture depth features.

[0027] In some embodiments, based on the facial images and tongue images of the female group anemia subjects, the group facial anemia color features, the group facial anemia texture features, the group tongue color anemia features and the group tongue anemia texture features are obtained, including:

[0028] According to the facial images and tongue images of the female anemia subjects, extracting the individual facial anemia color features, individual facial anemia texture features, individual tongue anemia color features and individual tongue anemia texture features of each female individual subject;

[0029] Divide the female anemic subjects into at least two groups according to uniform criteria;

[0030] By statistically analyzing the anemia color characteristics of each group of subjects in at least two groups of subjects, a distribution of anemia color characteristics of facial anemia, a distribution of anemia color characteristics of tongue coating, and a distribution of anemia color characteristics of tongue quality are obtained;

[0031] By statistically analyzing the anemia texture features of each group of subjects in at least two groups of subjects, a distribution of facial anemia texture features, a distribution of tongue coating anemia texture features, and a distribution of tongue quality anemia texture features were obtained;

[0032] The distribution of group facial anemia color characteristics is the group facial anemia color characteristics, the distribution of group facial anemia texture characteristics is the group facial anemia texture characteristics, the distribution of group tongue coating anemia color characteristics and the distribution of group tongue quality anemia color characteristics are the group tongue anemia color characteristics, the distribution of group tongue coating anemia texture characteristics and the distribution of group tongue body anemia texture characteristics are the group tongue anemia texture characteristics.

[0033] In some embodiments, the group facial anemia texture features include group facial anemia texture features obtained by angular direction second-order matrix analysis, group facial anemia texture features obtained by contrast analysis, group facial anemia texture features obtained by entropy analysis, and group facial anemia texture features obtained by inverse matrix analysis;

[0034] The group tongue anemia texture features include the group tongue body anemia texture features and the group tongue quality anemia texture features obtained through the second-order matrix analysis of the angular direction, the group tongue body anemia texture features and the group tongue quality anemia texture features obtained through entropy analysis, the group tongue body anemia texture features and the group tongue quality anemia texture features obtained through mean value analysis, and the group tongue quality anemia texture features and the group tongue body anemia texture features obtained through contrast analysis.

[0035] The female anemia risk assessment method provided by the embodiments of the present invention has at least one or part of the following advantages:

[0036] (1) The female anemia risk assessment method provided by the embodiment of the present invention can accurately predict the probability of female anemia by acquiring facial features from female facial images;

[0037] (2) The female anemia risk assessment method provided in the embodiment of the present invention screens the collected data of the female anemia population multiple times, removes the redundant features in the collected data that are not highly correlated with female anemia, and retains the features that are highly correlated with female anemia to construct a female anemia risk assessment model, thereby ensuring the accuracy of the female anemia probability predicted by the female anemia risk assessment model, and providing doctors with a more accurate auxiliary reference basis;

[0038] (3) The female anemia risk assessment method provided in the embodiment of the present invention performs statistical analysis on the collected data of female anemia population at different age stages, and then uses the same feature screening criteria for feature screening, thereby retaining the subtle differences in the characteristics of anemia at different age stages, such as the subtle differences in the color of anemia at different age stages on the face and tongue. This makes the female anemia risk assessment model applicable to women of different ages, and when performing risk assessment on women of different age stages, it still ensures the accuracy of predicting the probability of female anemia.

[0039] (4) The female anemia risk assessment method provided in the embodiment of the present invention uses data comparison between female anemia patients and female healthy patients, for example, the color and texture features of facial images and tongue images of the anemia patients and the healthy patients are compared, thereby further ensuring the relevance of the selected features with anemia and improving the accuracy of the prediction data of the female anemia risk assessment model;

[0040] (5) The female anemia risk assessment method provided by the embodiment of the present invention is a non-invasive operation in the process of obtaining characteristic value data, which is superior to the invasive blood sampling and blood test method. It is not only applicable to scenarios with good medical conditions, but also to scenarios with poor medical conditions. At the same time, it also avoids the infection risk brought to the subject by invasive operations;

[0041] (6) The female anemia risk assessment method provided by the embodiment of the present invention can also provide more accurate auxiliary reference data for other occasions of facial diagnosis or tongue diagnosis;

[0042] (7) In the process of constructing a female anemia risk assessment model, the female anemia risk assessment method provided by the embodiment of the present invention reduces the amount of data processing, improves the data processing accuracy, and reduces the risk of model overfitting through multiple screening and analysis of redundant feature values. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] These and / or other aspects and advantages of the present invention will become apparent and readily understood from the following description of preferred embodiments in conjunction with the accompanying drawings, in which:

[0044] Figure 1 A flowchart of a method for assessing anemia risk in women according to one embodiment of the present invention;

[0045] Figure 2 A flowchart of constructing a female anemia risk assessment model according to one embodiment of the present invention;

[0046] Figure 3 Based on Figure 2 The flowchart shown is for obtaining facial anemia characteristic values ​​and tongue anemia characteristic values ​​of a subject group;

[0047] Figure 4 for Figure 2 The schematic diagram of the collinearity analysis results of color features and facial features in the first screening process is shown;

[0048] Figure 5 for Figure 2 The schematic diagram of the first screening using LASSO regression method is shown;

[0049] Figure 6 for Figure 5 The schematic diagram of the convergence process of characteristic variables in the LASSO regression method screening process is shown;

[0050] Figure 7 For Figure 2 Shown is a schematic diagram of the results of mixed matrix analysis of the logistic regression model for the data of the female group of anemic subjects. DETAILED DESCRIPTION

[0051] The technical solution of the present invention is further specifically described below by examples and in conjunction with the accompanying drawings. In the specification, the same or similar reference numerals indicate the same or similar components. The following description of the embodiments of the present invention with reference to the accompanying drawings is intended to explain the overall inventive concept of the present invention and should not be construed as a limitation of the present invention.

[0052] Clinical studies have shown that in addition to blood sampling for complete blood cell examination, non-invasive testing of hemoglobin through the skin, fingertips, nail beds, mucous membranes, etc. of individual subjects can reflect their anemia status to a certain extent. At the same time, non-invasive testing can reduce the impact of invasive blood sampling on individual subjects, such as pain and infection risks.

[0053] Clinical studies have also shown that anemia is associated with the color of certain parts of the body, such as the face, lips, and conjunctiva. Taking the face as an example, as part of the skin, the largest organ in the human body, it will be affected by chromophores such as bilirubin or melanin in hemoglobin detection, but changes in facial color and texture can reflect certain physiological or pathological information of the person. Similarly, the surface of the tongue is covered by mucous membranes and is less affected by confusing chromophores (such as melanin), and its surface color and texture can also serve as reference features for physiological or pathological information. Therefore, changes in these specific features of the face and tongue can serve as clinical markers for early risk screening and assessment of anemia.

[0054] Since the number of female anemia patients in the key monitoring age group (e.g., 29 to 49 years old) is much higher than that of male anemia patients, the subject samples in this embodiment are all from females, and the screening and assessment subjects of the constructed anemia risk assessment model are also females.

[0055] The main clinical data used in the female anemia risk assessment method of this embodiment includes the facial image and tongue image of the subject. The detailed collection process of the facial image and tongue image includes:

[0056] A tongue diagnosis instrument is used to collect facial images and tongue images of the subject. Alternatively, the tongue diagnosis instrument may be, for example, a TFDA-1 tongue diagnosis instrument.

[0057] Before shooting, it is necessary to check the basic conditions of the subject's collection site, including: the subject avoids emotional fluctuations or strenuous exercise; the subject's face is kept clean, without obvious staining or makeup; the subject's tongue surface is free of foreign matter and staining.

[0058] During the shooting, when the facial image is collected, the subject's face is centered and the eyes are slightly closed; when the tongue image is collected, the subject opens his mouth and sticks out his tongue, and the tongue surface is flat and not tense. When the subject's posture is stable, the camera is used to shoot. In order to obtain stable images that meet the basic requirements of the color space and have a higher resolution, the camera shooting parameters can alternatively include: color temperature 5000k, lumen 4800lx, aperture f6.3, ISO200, exposure 1 / 125s, pixel 12 million, image resolution 5568*3712.

[0059] During the shooting process, facial images and tongue images were taken twice for each subject, and the captured images were subjected to preliminary quality screening. For example, images with fogging, underexposure, overexposure, staining, nervousness of the subject, abnormal shapes, etc. were deleted to retain clear and complete facial and tongue images.

[0060] like Figure 1 As shown, the female anemia risk assessment method of this embodiment includes the following steps:

[0061] Step S101: obtaining a color space value of a selected area in the face of a female individual subject according to a facial image of the female individual subject;

[0062] Step S102: Obtaining the anemia probability of the female individual subject according to the color space value of the selected region and the body mass index of the female individual subject.

[0063] In one example, the female individual subject is a single female individual subject.

[0064] Common color spaces in image processing include RGB, CMYK, Lab, HSK NTSC, YCbCr, HSV, etc. Among them, the color space value can be the parameters of the corresponding color space itself, or any parameter in the corresponding color space or any combination thereof. For example, in the RGB color space, the color space value can be R value, G value, B value, or any one of these three values, or any two values, etc.

[0065] In one example, the color space may also use the Lab color space, where L represents lightness, and the larger the L value, the brighter it is; a represents the red-green axis, where a positive value represents red, and the higher the value, the redder it is, and a negative value represents green, and the lower the value, the greener it is; and b represents the yellow-blue axis, where a positive value represents yellow, and the higher the value, the yellower it is, and a negative value represents blue, and the lower the value, the bluer it is.

[0066] In one example, step S102 also includes: obtaining the anemia probability based on the female anemia risk assessment model by combining the color space value of the selected area and the body mass index (BMI) of the female individual subject. The selected area may be the entire face area, a partial face area, the entire tongue area, or a portion of the tongue area. Of course, those skilled in the art may also understand that the selected area may be any combination of the face area and the tongue area, such as a combination of the entire face area and the entire tongue area, a combination of a partial face area and the entire tongue area, a combination of a partial face area and a partial tongue area, or a combination of the entire face area and a partial tongue area. This example is only an illustrative example, and those skilled in the art may determine the selected area based on actual circumstances.

[0067] In one example, the tongue includes two parts: tongue coating and tongue quality. The tongue coating is a layer of moss-like material on the tongue surface. The tongue quality, also known as the tongue body, is the muscle and vascular tissue of the tongue.

[0068] In one example, the expression of the anemia risk assessment model for women is:

[0069] Logit(P)=P 0 +k 1 *BMI+k 2 *A 1 +k 3 *A 2 +k 4 *B 1 +k 5 *B 2 +k 6 *C 1 +k 7 *D 1 ;

[0070] In the formula, Logit(P) represents the probability of anemia in female individual subjects, P 0 represents a constant, k 1 , k 2 , k 3 , k 4 , k 5 , k 6 and k 7denote coefficients, BMI denotes the body mass index of female individual subjects, A 1 represents the first color space value of the first selected region of the female individual subject, A 2 The second color space value representing the first selected area of ​​the female individual subject, B 1 The first color space value of the second selected region representing the female individual subject, B 2 A second color space value representing a second selected region of the female individual subject, C 1 Represents the first color space value of the third selected area of ​​the female individual subject, D 1 A first color space value representing a fourth selected region of the female individual subject.

[0071] In one example, P 0 The value range of k is set to -16 to -12. 1 The value range of k is set to 0.09~0.3, 2 The value range of k is set to 0.1~0.4, 3 The value range of k is set to 0.8~0.3, 4 The value range of k is set to 0.1~0.4, 5 The value range of k is set to -0.4~0, 6 The value range of k is set to -0.7 to -0.3, 7 The value range is set to 0.1~0.4.

[0072] Those skilled in the art will understand that due to different environments (such as differences between the north and the south), individual differences, such as different skin colors and different levels of individual health (for example, whether there are other diseases, such as high blood pressure, heart disease, etc., their skin color and tongue color will be affected by the corresponding diseases). Therefore, although the same color space is used to characterize the faces of female subjects, the specific values ​​of the coefficients are affected by environmental factors, individual differences, etc., and the selected areas will also be different.

[0073] In one example, the first selected area is the lips, the second selected area is the cheeks, the third selected area is the cheeks, and the fourth selected area is the nose. Of course, those skilled in the art will appreciate that the selected area may also be other areas of the face, such as the brow, eyes, jaw, brow, etc.

[0074] In one example, when the color space is the Lab color space, P 0 The value of is set to -14.697, k 1 The value of is set to 0.167, k 2 The value of is set to 0.214, k 3The value of is set to 0.155, k 4 The value of is set to 0.294, k 5 The value of is set to -0.216, k 6 The value of is set to -0.535, k 7 The value of is set to 0.236.

[0075] Of course, those skilled in the art will understand that when the color space is other color spaces, or due to individual differences, environmental differences, etc., the above values ​​will be different, but as long as the values ​​satisfy P 0 The value range of k is set to -16 to -12. 1 The value range of k is set to 0.09~0.3, 2 The value range of k is set to 0.1~0.4, 3 The value range of k is set to 0.8~0.3, 4 The value range of k is set to 0.1~0.4, 5 The value range of k is set to -0.4~0, 6 The value range of k is set to -0.7 to -0.3, 7 The value range of is set to 0.1 to 0.4, thereby ensuring the prediction accuracy of the anemia probability. This example is only an illustrative example, and those skilled in the art should not be construed as a limitation of the present invention.

[0076] In one example, the present invention uses the Lab color space to express color features as an example to illustrate the female anemia risk assessment method for the processing of facial images and tongue images. When the color space is RGB, CMYK, HSK NTSC, YcbCr, HSV, etc., that is, when the color features are expressed in any of the above color spaces, the principle of the female anemia risk assessment method is exactly the same as the principle of expressing color features in the Lab color space, and will not be repeated here.

[0077] In one example, if Figure 2 As shown, the method for constructing a female anemia risk assessment model includes the following steps:

[0078] Step S201: obtaining a group facial anemia color feature and a group facial anemia texture feature according to facial images of female group anemia subjects;

[0079] Step S202: obtaining a group tongue anemia color feature and a group tongue anemia texture feature according to the tongue images of the female group anemia subjects;

[0080] Step S203: performing a first screening on the group facial anemia color features, the group facial anemia texture features, the group tongue anemia color features, and the group tongue anemia texture features according to the group facial anemia color features, the group facial anemia color features, the group tongue anemia texture features, and the group tongue anemia texture features to obtain the group facial anemia color features, the group tongue anemia color features, and the group tongue anemia texture features after the first screening;

[0081] Step S204: performing a second screening based on the facial anemia color features, tongue anemia color features and tongue anemia texture features of the group after the first screening to obtain the facial anemia color features of the group after the second screening;

[0082] Step S205: construct a female anemia risk assessment model based on the facial anemia color characteristics of the group after the second screening.

[0083] In one example, the female group anemia subjects are multiple female individual anemia subjects. In the process of building a female anemia risk assessment model, clinical data can be collected first, and the sample size is selected according to the actual situation. In this example, a total of 533 cases were collected, including 381 females in the healthy group, aged 29 to 49 years old, with an average age of 36.74±7.186 years old; 152 females in the anemia group, aged 29-49 years old, with an average age of 35.76±7.616 years old.

[0084] In one example, if Figure 3 As shown, step S201: according to the facial images and tongue images of the female group anemia subjects, the group facial anemia color features, the group facial anemia texture features, the group tongue color anemia features and the group tongue anemia texture features are obtained, including:

[0085] Step S301: extracting individual facial anemia color features, individual facial anemia texture features, individual tongue anemia color features, and individual tongue anemia texture features of each female individual subject based on the facial images and tongue images of the female anemia subjects in the group;

[0086] Step S302: Divide the female anemic subjects into at least two groups of subjects according to a unified standard;

[0087] Step S303: Analyzing the anemia color characteristics of each group of subjects in at least two groups of subjects through anemia color statistics to obtain a group facial anemia color characteristic distribution, a group tongue coating anemia color characteristic distribution, and a group tongue quality anemia color characteristic distribution;

[0088] Step S304: analyzing the anemia texture features of each group of subjects in at least two groups of subjects through anemia texture statistics to obtain a group facial anemia texture feature distribution, a group tongue coating anemia texture feature distribution, and a group tongue quality anemia texture feature distribution;

[0089] Step S305: The group facial anemia color feature distribution is the group facial anemia color feature, the group facial anemia texture feature distribution is the group facial anemia texture feature, the group tongue coating anemia color feature distribution and the group tongue quality anemia color feature distribution are the group tongue anemia color feature, the group tongue coating anemia texture feature distribution and the group tongue body anemia texture feature distribution are the group tongue anemia texture feature.

[0090] In one example, the collected samples of 533 female anemic subjects can be divided into 3 age gradients according to age stages (as shown in Table 1). Of course, those skilled in the art can also divide according to other unified standards, for example, they can be grouped according to the BMI index, or they can be grouped according to the region where the subjects are located. This example is only an illustrative example, and those skilled in the art should not be understood as a limitation of the present invention.

[0091] For example, grouping is performed again with 10 years old as an age gradient, that is, it is divided into three age gradient groups of 20 to 29 years old, 30 to 39 years old and 40 to 49 years old, and the healthy group and the anemia group are divided into three groups again according to these three age groups, that is, the healthy group is divided into three age gradient groups of 20 to 29 years old, 30 to 39 years old and 40 to 49 years old, and the anemia group is divided into three age gradient groups of 20 to 29 years old, 30 to 39 years old and 40 to 49 years old, a total of six groups. Of course, those skilled in the art can select according to the actual population characteristics and sample size, for example, 5 years old can be an age gradient. This example is only an illustrative example, and those skilled in the art should not be construed as a limitation of the present invention.

[0092] Table 1 Basic information of female subjects

[0093]

[0094] Among them, SBP stands for systolic blood pressure, DBP stands for diastolic blood pressure, and BMI stands for body mass index.

[0095] Then, data annotation is performed: We annotate the data based on the subject's laboratory results, which are determined by hemoglobin concentration (Hb). The definition threshold for anemia in women is hemoglobin below 110g / L. In this example, facial images, tongue images, and blood samples of all subjects were collected on the same day. If the subject's Hb is <110g / L, it is considered to be in the anemia group and the label is set to 1. If the subject's Hb is ≥110g / L, it is considered to be in the healthy control group and the label is set to 0.

[0096] Next, statistical analysis is performed: for normally distributed quantitative data, Indicates its central tendency and dispersion trend. For skewed quantitative data, Median (P25, P75) is used to indicate its central tendency and dispersion trend. If the variables compared between groups meet the normality and variance homogeneity, independent sample t test and one-way analysis of variance are used; if the variables do not meet the requirements, the Mann-Whitney U test is used for both samples, the Kruskal-Wallis H test is used for comparison among the three groups, and then the two-by-two comparisons are performed. Categorical variables are represented by n (%), and the chi-square test is used for comparison among multiple classification data. P ≤ 0.05 is considered statistically significant. All statistical analyses were performed using SPSS27.0.

[0097] Finally, the color and texture features of facial and tongue images are analyzed: the existing facial diagnosis and analysis system (FDAS) is used to extract the color and texture features of different parts of the face; the existing tongue diagnosis and analysis system version 2.0 (TDAS V2.0) is used to extract the color and texture features of the tongue body (TB) and tongue coating (TC).

[0098] Of course, those skilled in the art may also use methods such as deep learning, principal component analysis, etc. to perform feature extraction. This example is only an illustrative example, and those skilled in the art should not understand it as a limitation of the present invention.

[0099] In one example, the group facial anemia color features obtained based on the facial images of the female group anemia subjects include: the anemia color features of the entire face of the female group anemia subjects and the anemia color features of each area of ​​the face (for example, lips, forehead, between the eyebrows, nose, jaw, left cheek, right cheek, left cheek and right cheek); the group facial anemia texture features obtained based on the facial images include the facial anemia texture granularity features, facial anemia texture density features and facial anemia texture depth features of the female group anemia subjects.

[0100] In one example, the group tongue anemia color features obtained based on the tongue image include: the group tongue coating anemia color features and the group tongue quality anemia color features; the group tongue anemia texture features obtained based on the tongue image include the group tongue coating anemia texture granularity features, the group tongue quality anemia texture granularity features, the group tongue coating anemia texture density features, the group tongue quality anemia texture density features, the group tongue coating anemia texture depth features and the group tongue quality anemia texture depth features.

[0101] In one example, the anemia texture granularity features (e.g., clarity features), anemia texture depth features, etc. of the tongue images and facial images of female anemia subjects (including individuals and groups) can be obtained by contrast (CON). Similarly, the texture granularity features and texture depth features of healthy female subjects can also be obtained. The anemia texture granularity features (e.g., coarseness features) and anemia texture density features (e.g., uniformity, etc.) are obtained by the angular direction second-order matrix (ASM). The density features of the texture of the tongue image and facial image (e.g., grayscale distribution complexity, texture change speed, etc.) can be obtained by entropy (ENT) and inverse variance (IDM). Those skilled in the art will understand that the feature analysis method can also use more common statistical analysis methods, such as QR decomposition iteration method, Jacobi iteration method, etc. This example is only an illustrative example, and those skilled in the art should not be understood as a limitation of the present invention. The following content about the eigenvalue algorithm is basically similar to the calculation principle here, and will not be repeated later. For example, when the color space is the Lab color space, the color features of the subject's face include: the L component value, a component value, and b component value of the whole face (total), lips (lip), left cheek (lzyg), right cheek (rzyg), left cheek (lcheek), right cheek (rchee), nose (nose), forehead (foreh), eyebrow (eyeb), and jaw (jaw). And texture feature values ​​can also be obtained from several different angles (exemplarily including 0°, 45°, 90°, 135°) and the specific distance, and the texture feature values ​​include: texture features Asm_0, Asm_1, Asm_2 and Asm_3 corresponding to the second-order matrix in the angular direction at the above four angles; texture features Con_0, Con_1, Con_2 and Con_3 corresponding to the contrast at the above four angles; texture features Idm_0, Idm_1, Idm_2 and Idm_3 corresponding to the inverse matrix at the above four angles; texture feature values ​​Ent_0, Ent_1, Ent_2 and Ent_3 corresponding to the entropy at the above four angles.

[0102] In one example, the tongue diagnosis and analysis system can extract the color features of the subject's tongue quality, the texture features of the tongue quality, the color features of the tongue coating, the texture features of the tongue coating, and the tongue coating index. In the Lab color space, the color features of the tongue quality include the components of the tongue body corresponding to channels L, a, and b in the Lab color space: TB-L, TB-a, and TB-b; and the components of the tongue coating corresponding to channels L, a, and b in the Lab color space: TC-L, TC-a, and TC-b. The tongue quality texture features include: TB-ASM, TB-ENT, TB-CON, TB-IDM, and TB-MEAN, and the tongue coating texture features include: TC-ASM, TC-ENT, TC-CON, TC-IDM, and TC-MEAN. Among them, MEAN represents the average value of the texture.

[0103] The statistical analysis data of facial images are shown in Table II, and the statistical analysis data of tongue images are shown in Table III.

[0104] Table 2 Statistical analysis of facial image data (mean (SD), median (P 25 ,P 75 ))

[0105]

[0106]

[0107] Table 3 Statistical analysis of tongue image data (mean (SD), median (P 25 ,P 75 ))

[0108]

[0109]

[0110] In one example, in the group tongue anemia texture feature, the larger the contrast eigenvalue, entropy eigenvalue and average eigenvalue, and the smaller the second-order matrix eigenvalue in the angular direction, the rougher the tongue texture.

[0111] Furthermore, the tongue coating index can be calculated based on the pixel value of the tongue image, where Per-all is defined as the ratio of the tongue coating area to the total tongue area. The Per-all value has a higher judgment value for thick coating, and the larger the Per-all value, the thicker the tongue coating; Per-part is defined as the ratio of the tongue coating area to the tongue-free area. The Per-part value has a higher judgment value for thin coating, and the larger the Per-part value, the thinner the tongue coating.

[0112] The facial anemia color features, facial anemia texture features, tongue anemia color features, and tongue anemia texture features of the subjects in the anemia group and the healthy group were statistically analyzed. The results showed:

[0113] (1) The anemia color characteristics of the face as a whole, different parts of the face, tongue quality, and tongue coating were the same in all age groups: in the Lab color space, the color value of the L channel of the anemia group was greater than that of the healthy group; the color value of the a channel of the anemia group was less than that of the healthy group; and the color value of the b channel of the anemia group was greater than that of the healthy group.

[0114] (2) There are significant differences in facial anemia texture features in the 30-39 age group: the angular second-order matrix eigenvalue, which reflects the uniformity of image distribution and the coarseness of texture, is larger in the anemia group; the entropy eigenvalue, which reflects the complexity of image distribution, is smaller in the anemia group.

[0115] (3) There were significant differences in the anemia texture features of tongue quality and tongue coating in the age group of 40 to 49 years old: the contrast eigenvalue, entropy eigenvalue and average eigenvalue of the tongue body, as well as the contrast eigenvalue, entropy eigenvalue and average eigenvalue of the tongue coating, in the anemia group were smaller than those in the healthy group; the second-order matrix eigenvalue of the angle direction of the tongue body and tongue coating, in the anemia group were larger than those in the healthy group.

[0116] In one example, the obtained characteristic data were statistically distributed, and each group of characteristic data had normal distribution and skewed distribution. If the inter-group comparison variables all met the normality and variance homogeneity, independent sample test and one-way analysis of variance were used; if the inter-group comparison variables did not meet the normality and variance homogeneity, the Mann-Whitney U test was used, and the comparison between the three groups was conducted by the Kruskal-Wallis H test, and then two-by-two comparisons were conducted. At the same time, the chi-square test was used for the inter-class comparison of multi-classified data. Among all the characteristic data, only when the deviation P≤0.05, the corresponding characteristic value is statistically significant. 66 characteristic values ​​were obtained through the above statistics (as shown in Tables 2 and 3).

[0117] In one example, as shown in step S203, the obtained 66 eigenvalues ​​can be screened down to 26 eigenvalues ​​through a first screening.

[0118] The above 66 eigenvalues ​​include facial image features (including facial anemia color features and facial anemia texture features) and tongue image features (including tongue anemia color features and tongue anemia texture features) of anemic subjects in the female group. All of these eigenvalues ​​have different degrees of influence on the anemia risk assessment model. There are some redundant eigenvalues ​​among them. In order to prevent the model from overfitting due to a large number of eigenvalues, it is necessary to reduce the dimension of the eigenvalues ​​through certain screening principles to screen out eigenvalues ​​that are highly correlated with anemia.

[0119] like Figure 4 As shown in the figure, a part of the collinearity analysis results between independent variables are shown. The size and color of the circle in the figure represent the correlation between any two independent variables. The gradually darkening blue circle represents the increasing correlation between the corresponding two independent variables, and the gradually darkening red circle represents the decreasing correlation between the corresponding two independent variables. Among the 66 eigenvalues, the eigenvalues ​​related to anemia meditation include the following 12 eigenvalues ​​of the L channel in the Lab color space, namely, the overall facial L value (total-L), lip L value (lip-L), forehead L value (foreh-L), glabellar L value (eyeb-L), nose L value (nose-L), mandibular L value (jaw-L), left zygomatic L value (lzyg-L), right zygomatic L value (rzyg-L), left cheek L value (lcheek-L), right cheek L value (rcheek-L), tongue coating L value (TC-L) and tongue quality L value (TB-L).

[0120] The above 12 eigenvalues ​​were subjected to collinearity analysis among independent variables, and the results showed that there was obvious correlation between the eigenvalues, that is, there was obvious multicollinearity among the independent variables in the high-dimensional data set composed of the basic information of the subjects, facial image features (including facial anemia color features and facial anemia texture features) and tongue image features (including tongue anemia color features and tongue anemia texture features).

[0121] In one example, alternatively, LASSO regression analysis is used as the first screening method. In the LASSO regression analysis method, by adding an L1 norm as a penalty constraint in the calculation of minimizing the residual sum of squares (RSS), when the penalty coefficient λ is sufficiently large, the coefficient can be compressed to 0, thereby effectively screening out the variables most relevant to the analysis target, playing a role in reducing the dimension of the independent variable and avoiding the risk of overfitting the model. Those skilled in the art will appreciate that the method of regression analysis can use other conventional methods, such as logistic regression, stepwise regression, ridge regression, etc. This example is only an illustrative example, and those skilled in the art should not be construed as a limitation of the present invention. The following content about the regression algorithm is basically similar to the calculation principle here, and will not be repeated here.

[0122] In one example, specifically, corresponding to a high-dimensional data set consisting of basic information of the subject, facial image features (including facial anemia color features and facial anemia texture features) and tongue image features (including tongue anemia color features and tongue anemia texture features), LASSO regression is used for feature screening, and the complexity of the female anemia risk assessment model is adjusted by the penalty coefficient λ value (λ=0.011 is selected in this embodiment). The larger the penalty coefficient λ value, the greater the degree of compression of the eigenvalue variable, and the more streamlined the model accordingly. Figure 5 It shows that as the parameter Log(λ) increases, the penalty of LASSO regression gradually increases, and the number of eigenvalue variables screened out gradually decreases; the ordinate represents the minimum mean square error, and the upper abscissa represents the number of eigenvalue variables. According to the actual trend characteristics of the regression curve, a vertical dotted line is drawn at the minimum mean square error (λ=0.011) and the standard error of the minimum distance (λ=0.032), and the screening range for the number of eigenvalue variables is obtained. Further, Figure 6 It shows that the coefficient value of the eigenvalue variable will gradually converge with the increase of the parameter Log(λ).

[0123] After the first screening mentioned above, the results showed that after the initial 66 eigenvalues ​​were screened by LASSO regression method, the remaining 26 eigenvalues ​​were determined to be highly correlated with anemia. The 26 eigenvalues ​​included: body mass index (BMI), systolic blood pressure (SBP), total face a value (total-a), total face b value (total-b), lip L value (lip-L), lip b value (lip-b), forehead a value (foreh-a), forehead b value (foreh-b), glabellar b value (eyeb-b), nose a value (nose-a), nose b value (nose-b), mandibular a value (foreh-a), forehead b value (foreh-b), eyebrow b value (eyeb-b), nose a value (nose-a), nose b value (nose-b), mandibular a value (foreh-b), forehead b value (foreh-b), forehead L value (foreh-L), forehead b value (foreh-b), forehead L value (foreh-b), forehead L value (foreh-b), forehead L value (foreh-b), forehead L value (foreh-b), nose L value (foreh-b), mandibular L value (foreh-a), nose L value (foreh-b), mandibular L value (foreh-L), forehead L value (foreh-b), nose L value (foreh-b), mandibular L value (foreh-L), face L value (foreh-b ...a), face L value (foreh-b), face L value (for value (jaw-a), left zygomatic b value (lzyg-b), right zygomatic a value (rzyg-a), left cheek a value (lcheek-a), right cheek a value (rcheek-a), texture features of inverse matrix at 0° (Idm_0), texture features of contrast at 90° (Con_0), ratio of tongue coating area to no coating area (Per-Part), tongue coating a value (TC-a), tongue coating b value (TC-b), tongue angle direction second-order matrix features (TB-ASM), tongue average features (TB-MEAN), tongue L value (TB-L), tongue a value (TB-a) and tongue b value (TB-b).

[0124] In one example, alternatively, logistic regression analysis can be used as a second screening method in step S204. Through the second screening, the 26 eigenvalues ​​obtained in step S203 are further screened to 7 eigenvalues, which include: body mass index (BMI), lip L value (lip-L), lip b value (lip-b), left zygomatic b value (lzyg-b), right zygomatic a value (rzyg-a), left cheek a value (lcheek-a) and nose b value (nose-b).

[0125] After the above multiple screenings, the expression of the female anemia risk assessment model is:

[0126] Logit(P)=P 0 +k 1 *BMI+k 2 *A 1 +k 3 *A 2 +k 4 *B 1 +k 5 *B 2 +k 6 *C 1 +k 7 *D 1 ;

[0127] In the formula, the definitions of various parameters are as follows:

[0128] Logit (P) represents the probability of anemia for individual female subjects;

[0129] P 0 represents a constant;

[0130] k 1 , k 2 , k 3 , k 4 , k 5 , k 6 and k 7 denote coefficients respectively;

[0131] BMI represents the body mass index of individual female subjects, also known as body mass index or body mass index. The specific calculation method of BMI adopts the conventional algorithm, that is, it is obtained by dividing the weight (kg) by the square of the height (m);

[0132] A 1 represents the lip L value (lip-L) of female anemic individual subjects;

[0133] A 2 represents the lip-b value (lip-b) of female anemic individual subjects;

[0134] B 1 It represents the left zygomatic b value (lzyg-b) of an individual female anemic subject;

[0135] B 2 represents the right zygomatic a value (rzyg-a) of an individual female anemic subject;

[0136] C 1 represents the left cheek a value (lcheek-a) of an individual female anemic subject;

[0137] D 1 It represents the nose b value (nose-b) of female anemic individual subjects.

[0138] In one example, the expression of the anemia probability Logit (P) of a female individual subject is assigned a value, and the female anemia risk assessment model obtained by multi-factor logistic regression analysis and fitting is:

[0139] Logit(P)=

[0140] -14.697+0.167BMI+0.214A 1 +0.155A 2 +0.294B 1 -0.216B 2 -0.535C 1 +0.236D 1 .

[0141] In one example, alternatively, a logistic regression algorithm is used, and the sample data is divided into a training set and a test set, wherein the ratio of the training set to the test set is set to 7:3, and the modeling effect of each eigenvalue parameter included in the Logistic regression model in different specific models is analyzed.

[0142] Furthermore, according to the confusion matrix results of the logistic regression model (such as Figure 7 As shown in the figure, accuracy, precision, F1 score, sensitivity, specificity and area under the ROC curve (AUC) were used as evaluation indicators to evaluate the predictive performance of the female anemia risk assessment model through true labels and predicted labels.

[0143] The female anemia risk assessment model obtained in this embodiment was evaluated, and the obtained evaluation indicators are shown in Table 4.

[0144] Table 4 Evaluation indicators of the female anemia risk assessment model

[0145] Evaluation indicators AUC Accuracy Precision Sensitivity Specificity F1 Calculated Value 0.841 0.756 0.630 0.370 0.912 0.466

[0146] The higher the value of the area under the ROC curve (AUC) (the closer it is to 1.0), the better the prediction effect of the model, that is, the higher the accuracy of the model prediction.

[0147] The female anemia risk assessment method provided by the embodiments of the present invention has at least one or part of the following advantages:

[0148] (1) The female anemia risk assessment method provided by the embodiment of the present invention can accurately predict the probability of female anemia by acquiring facial features from female facial images;

[0149] (2) The female anemia risk assessment method provided in the embodiment of the present invention screens the collected data of the female anemia population multiple times, removes the redundant features in the collected data that are not highly correlated with female anemia, and retains the features that are highly correlated with female anemia to construct a female anemia risk assessment model, thereby ensuring the accuracy of the female anemia probability predicted by the female anemia risk assessment model, and providing doctors with a more accurate auxiliary reference basis;

[0150] (3) The female anemia risk assessment method provided in the embodiment of the present invention performs statistical analysis on the collected data of female anemia population at different age stages, and then uses the same feature screening criteria for feature screening, thereby retaining the subtle differences in the characteristics of anemia at different age stages, such as the subtle differences in the color of anemia at different age stages on the face and tongue. This makes the female anemia risk assessment model applicable to women of different ages, and when performing risk assessment on women of different age stages, it still ensures the accuracy of predicting the probability of female anemia.

[0151] (4) The female anemia risk assessment method provided in the embodiment of the present invention uses data comparison between female anemia patients and female healthy patients, for example, the color and texture features of facial images and tongue images of the anemia patients and the healthy patients are compared, thereby further ensuring the relevance of the selected features with anemia and improving the accuracy of the prediction data of the female anemia risk assessment model;

[0152] (5) The female anemia risk assessment method provided by the embodiment of the present invention is a non-invasive operation in the process of obtaining characteristic value data, which is superior to the invasive blood sampling and blood test method. It is not only applicable to scenarios with good medical conditions, but also to scenarios with poor medical conditions. At the same time, it also avoids the infection risk brought to the subject by invasive operations;

[0153] (6) The female anemia risk assessment method provided by the embodiment of the present invention can also provide more accurate auxiliary reference data for other occasions of facial diagnosis or tongue diagnosis;

[0154] (7) In the process of constructing a female anemia risk assessment model, the female anemia risk assessment method provided by the embodiment of the present invention reduces the amount of data processing, improves the data processing accuracy, and reduces the risk of model overfitting through multiple screening and analysis of redundant feature values.

[0155] Although some embodiments of the present general inventive concept have been shown and described, it will be appreciated by those skilled in the art that changes may be made to these embodiments without departing from the principles and spirit of the present general inventive concept, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for assessing anemia risk in women, comprising the following steps: According to the facial image of the female individual subject, obtaining a color space value of a selected area in the face of the female individual subject; The anemia probability of the female individual subject is obtained according to the color space value of the selected area and the body mass index of the female individual subject.

2. The method for assessing anemia risk in women according to claim 1, wherein: Obtaining the anemia probability of the female individual subject according to the color space value of the selected area and the body mass index of the female individual subject, comprising: The anemia probability is obtained by combining the color space value of the selected area and the body mass index of the female individual subject based on a female anemia risk assessment model.

3. The method for assessing anemia risk in women according to claim 2, wherein: The expression of the female anemia risk assessment model is: Logit(P)=P0+k1*BMI+k2*A1+k3*A2+k4*B1+k5*B2+k6*C1+k7*D1; In the formula, Logit(P) represents the probability of anemia of the female individual subject, P0 represents a constant, k1, k2, k3, k4, k5, k6 and k7 represent coefficients respectively, BMI represents the body mass index of the female individual subject, A1 represents the first color space value of the first selected area of ​​the female individual subject, A2 represents the second color space value of the first selected area of ​​the female individual subject, B1 represents the first color space value of the second selected area of ​​the female individual subject, B2 represents the second color space value of the second selected area of ​​the female individual subject, C1 represents the first color space value of the third selected area of ​​the female individual subject, and D1 represents the first color space value of the fourth selected area of ​​the female individual subject.

4. The anemia risk assessment method according to claim 3, wherein: The value range of P0 is set to -16 to -12, the value range of k1 is set to 0.09 to 0.3, the value range of k2 is set to 0.1 to 0.4, the value range of k3 is set to 0.8 to 0.3, the value range of k4 is set to 0.1 to 0.4, the value range of k5 is set to -0.4 to 0, the value range of k6 is set to -0.7 to -0.3, and the value range of k7 is set to 0.1 to 0.

4.

5. The method for assessing anemia risk in women according to claim 4, wherein: The first selected area is the lips, the second selected area is the cheeks, the third selected area is the cheeks and the fourth selected area is the nose.

6. The method for assessing anemia risk in women according to claim 5, wherein: When the color space is the Lab color space, the value of P0 is set to -14.697, the value of k1 is set to 0.167, the value of k2 is set to 0.214, the value of k3 is set to 0.155, the value of k4 is set to 0.294, the value of k5 is set to -0.216, the value of k6 is set to -0.535, and the value of k7 is set to 0.

236.

7. The method for assessing anemia risk in women according to any one of claims 2 to 6, wherein: The method for constructing the female anemia risk assessment model comprises the following steps: According to the facial images of female anemia subjects, the facial anemia color features and the facial anemia texture features of the group are obtained; According to the tongue images of the female anemia subjects, the color features and texture features of the anemia tongue of the group are obtained; According to the group facial anemia color feature, the group facial anemia texture feature, the group tongue anemia color feature, and the group tongue anemia texture feature, a first screening is performed on the group facial anemia color feature, the group tongue anemia color feature, the group facial anemia texture feature, and the group tongue anemia texture feature to obtain the group facial anemia color feature, the group tongue anemia color feature, and the group tongue anemia texture feature after the first screening; Perform a second screening based on the facial anemia color features, tongue anemia color features, and tongue anemia texture features of the group after the first screening to obtain the facial anemia color features of the group after the second screening; The female anemia risk assessment model is constructed based on the facial anemia color characteristics of the group after the second screening.

8. The method for assessing anemia risk in women according to claim 7, wherein: The first screening was LASSO regression analysis, and the second screening was logistic regression analysis.

9. The method for assessing anemia risk in women according to claim 7, wherein: The group facial anemia color features obtained based on the facial images of the female group anemia subjects include the anemia color features of the entire face of the female group anemia subjects and the anemia color features of each area in the face. The group facial anemia texture features obtained based on the facial images include the facial anemia texture granularity features, facial anemia texture density features and facial anemia texture depth features of the female group anemia subjects. The group tongue anemia color features obtained based on tongue images include the group tongue coating anemia color features and the group tongue quality anemia color features. The group tongue anemia texture features obtained based on tongue images include the group tongue coating anemia texture granularity features, the group tongue quality anemia texture granularity features, the group tongue coating anemia texture density features, the group tongue quality anemia texture density features, the group tongue coating anemia texture depth features and the group tongue quality anemia texture depth features.

10. The method for assessing anemia risk in women according to claim 9, wherein: According to the facial images and tongue images of the female group anemia subjects, the group facial anemia color features, the group facial anemia texture features, the group tongue color anemia features and the group tongue anemia texture features are obtained, including: According to the facial images and tongue images of the female anemia subjects, extracting the individual facial anemia color features, individual facial anemia texture features, individual tongue anemia color features and individual tongue anemia texture features of each female individual subject; Dividing the female group of anemic subjects into at least two groups of subjects according to a unified standard; By statistically analyzing the anemia color characteristics of each group of subjects in at least two groups of subjects, a distribution of anemia color characteristics of facial anemia, a distribution of anemia color characteristics of tongue coating, and a distribution of anemia color characteristics of tongue quality are obtained; By statistically analyzing the anemia texture features of each group of subjects in at least two groups of subjects, a distribution of facial anemia texture features, a distribution of tongue coating anemia texture features, and a distribution of tongue quality anemia texture features were obtained; The group facial anemia color feature distribution is the group facial anemia color feature, the group facial anemia texture feature distribution is the group facial anemia texture feature, the group tongue coating anemia color feature distribution and the group tongue quality anemia color feature distribution are the group tongue anemia color feature, the group tongue coating anemia texture feature distribution and the group tongue body anemia texture feature distribution are the group tongue anemia texture feature.

11. The method for assessing anemia risk in women according to claim 10, wherein: The group facial anemia texture features include the group facial anemia texture features obtained by the second-order matrix analysis of the angle direction, the group facial anemia texture features obtained by the contrast analysis, the group facial anemia texture features obtained by the entropy analysis and the group facial anemia texture features obtained by the inverse matrix analysis; The group tongue anemia texture features include the group tongue body anemia texture features and the group tongue quality anemia texture features obtained through the second-order matrix analysis of the angular direction, the group tongue body anemia texture features and the group tongue quality anemia texture features obtained through entropy analysis, the group tongue body anemia texture features and the group tongue quality anemia texture features obtained through mean value analysis, and the group tongue quality anemia texture features and the group tongue body anemia texture features obtained through contrast analysis.

Citation Information

Cited By

  • Method and system for detecting fatty liver patient missed in liver color ultrasound examination

    CN119920453A

  • Anemia detection device, anemia detection model training method and anemia detection model training device

    CN121505676A