CN-122020263-A - Intelligent grading system for eye bursting severity of integrated image histology

CN122020263ACN 122020263 ACN122020263 ACN 122020263ACN-122020263-A

Abstract

The invention relates to the field of medical image information, in particular to an intelligent grading system for the severity of a sudden eye for integrating image histology. The system comprises a data acquisition module, a data processing module, a fusion analysis matrix, a construction stage fusion feature matrix, a final fusion feature matrix and a data classification module, wherein the data acquisition module is used for acquiring clinical features and medical image features, the data processing module is used for determining dimension reduction clinical features and dimension reduction medical image features, constructing the fusion analysis matrix, calculating standardized obstructivity of each dimension reduction medical image feature, determining main component clinical features, constructing the fusion feature matrix, calculating main component standard obstructivity, acquiring the final fusion feature matrix, the data classification module is used for determining a reference vector according to the final fusion feature matrix, combining the dimension reduction medical image features and the main component clinical features of each patient into a personal feature vector, inputting the reference vector and the personal feature vector into a trained intelligent classification model, and outputting classification grades of each patient through the intelligent classification model. The invention can improve the grading effect on the sudden eye disease.

Inventors

LIU PEI
DONG XIAOHUI

Assignees

西安市第一医院

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. The utility model provides an intelligent grading system of outstanding eye degree of integration image group is said, which is characterized in that includes: the data acquisition module is used for acquiring clinical characteristics and medical image characteristics of each patient; the data processing module is used for determining dimension-reducing clinical features and dimension-reducing medical image features from clinical features and medical image features for each patient; According to the fusion analysis matrix corresponding to each dimension-reduction medical image feature, calculating the standardized obstruction of each dimension-reduction medical image feature; According to the clinical features of the main components and the medical image features of the dimension reduction, constructing a phase fusion feature matrix, and calculating the main component standard obstruction of the phase fusion feature matrix; If the standard obstructive of the principal component is smaller than a preset obstructive threshold, determining the phase fusion feature matrix as a final fusion feature matrix, otherwise, standardizing the clinical features of the principal component, and constructing the final fusion feature matrix according to the standardized clinical features of the principal component; And the data classification module is used for determining a reference vector according to the final fusion feature matrix, combining the dimensionality reduction medical image features and the main component clinical features of each patient into a personal feature vector, inputting the reference vector and the personal feature vector into a trained intelligent classification model, and outputting the classification grade of each patient through the intelligent classification model.
2. The intelligent grading system for the severity of the salient eye of integrated image histology according to claim 1, wherein the data processing module determines, for each patient, a dimension-reduced clinical feature and a dimension-reduced medical image feature from the clinical features and the medical image features, specifically comprising: constructing an association standard feature matrix according to clinical features and medical image features of each patient, and respectively determining reference clinical features and reference medical image features from the clinical features and the medical image features according to the association standard feature matrix of each patient; Calculating a first association degree of each reference clinical feature and a second association degree of each reference medical image feature, and determining a dimension-reduction clinical feature and a dimension-reduction medical image feature from the reference clinical feature and the reference medical image feature according to a first preset association threshold and a second preset association threshold respectively, wherein the value range of the first preset association threshold and the second preset association threshold is (0, 1), the first preset association threshold is related to the reference clinical feature, and the second preset association threshold is related to the reference medical image feature.
3. The intelligent grading system for the severity of the salient eye of integrated image histology according to claim 2, wherein the data processing module constructs a correlation standard feature matrix according to clinical features and medical image features for each patient, and determines reference clinical features and reference medical image features from the clinical features and the medical image features respectively according to the correlation standard feature matrix of each patient, specifically comprising: Combining each clinical feature with the medical image feature as a data pair for each patient, constructing an initial feature matrix by taking the data pair of each clinical feature as elements in each row of the matrix, and carrying out standardized processing on the initial feature matrix to obtain an initial standard feature matrix; Calculating the pearson correlation coefficient of each clinical feature and the medical image feature respectively for each patient, and carrying out position adjustment on elements in the initial standard feature matrix according to the pearson correlation coefficient of each clinical feature to obtain an association standard feature matrix, wherein the number of lines of data pairs of the clinical feature in the association standard feature matrix and the pearson correlation coefficient of the clinical feature are in a direct proportional relation; and determining clinical features with the pearson correlation coefficient larger than a preset first correlation threshold in the correlation standard feature matrix of each patient as reference clinical features, and acquiring reference medical image features, wherein the preset first correlation threshold is determined by the clinical features of the first 70%.
4. The intelligent grading system for the severity of the salient eye of integrated image histology according to claim 3, wherein the data processing module calculates pearson correlation coefficients for each clinical feature and each medical image feature, comprising: Taking each medical image feature as an element in the data sequence to obtain an image feature sequence; taking the a clinical feature as an element in the data sequence to obtain a clinical feature sequence of the a clinical feature, wherein the image feature sequence and the clinical feature sequence have the same length; acquiring a clinical feature sequence of each clinical feature; calculating the pearson correlation coefficient of the clinical feature sequence of each clinical feature and the image feature sequence respectively to obtain the pearson correlation coefficient of each clinical feature; And (3) carrying out position adjustment on elements in the initial standard feature matrix according to the Pearson correlation coefficient of each clinical feature to obtain an associated standard feature matrix.
5. The intelligent grading system for integrated image histology according to claim 3, wherein the data processing module obtains reference medical image features, comprising: Combining each medical image feature and clinical feature into a data pair aiming at each patient, constructing an image feature matrix by taking the data pair of each medical image feature as elements in each row of the matrix, and carrying out standardized processing on the image feature matrix to obtain an image standard feature matrix; calculating the pearson correlation coefficient of each medical image feature and the clinical feature respectively for each patient, and carrying out position adjustment on elements in the image feature matrix according to the pearson correlation coefficient of each medical image feature to obtain an associated image feature matrix, wherein the number of lines of data pairs of the medical image feature and the pearson correlation coefficient of the medical image feature in the associated image feature matrix are in a direct proportion relation; And determining the associated image features with the pearson correlation coefficient larger than a preset second correlation threshold value in the associated image feature matrix of each patient as reference medical image features, wherein the preset second correlation threshold value is determined through the first 70% of medical image features.
6. The intelligent grading system for the severity of the salient eye of integrated image histology according to claim 2, wherein the data processing module calculates a first degree of association for each reference clinical feature and a second degree of association for each reference medical image feature, comprising: Acquiring the number of patients comprising the b-th reference clinical feature, and acquiring the row number of the corresponding data pair in the association standard feature matrix of the b-th reference clinical feature in the c-th patient; Acquiring the b-th reference clinical feature in an associated standard feature matrix of each patient, summing the row numbers of corresponding data pairs, and obtaining the priority degree of the b-th reference clinical feature; determining a ratio of the number of patients with the b-th reference clinical feature to the priority of the b-th reference clinical feature as a first degree of association of the b-th reference clinical feature; Acquiring a first association degree of each reference clinical feature; Acquiring the number of patients comprising the d-th reference medical image feature, and acquiring the row serial number of the corresponding data pair in the associated image feature matrix of the e-th patient by the d-th reference medical image feature; acquiring the d-th reference medical image feature in an associated image feature matrix of each patient, summing the row serial numbers of corresponding data pairs, and obtaining the priority of the d-th reference medical image feature; Determining the ratio of the number of patients with the d-th reference medical image feature to the priority of the d-th reference medical image feature as a second association degree of the d-th reference medical image feature; A second degree of association of each reference medical image feature is obtained.
7. The intelligent grading system for the eye bursting severity of integrated image histology according to claim 1, wherein the data processing module constructs a fusion analysis matrix corresponding to each dimension-reduced medical image feature according to the dimension-reduced clinical feature and the dimension-reduced medical image feature, and the system specifically comprises: Aiming at the f-th dimension-reduction medical image feature, acquiring the vector feature of the f-th dimension-reduction medical image feature in the g-th reference patient, wherein the dimension-reduction medical image feature of the reference patient comprises the f-th dimension-reduction medical image feature; Combining the vector features of the f-th dimension-reduced medical image feature in the g-th reference patient with each dimension-reduced clinical feature to form a data pair, and constructing a fusion analysis matrix corresponding to the f-th dimension-reduced medical image feature by taking the f-th dimension-reduced medical image feature in the data pair of each reference patient as an element in each row of the matrix; And obtaining a fusion analysis matrix corresponding to each dimension-reduced medical image feature.
8. The intelligent grading system for the eye bursting severity of integrated image histology according to claim 1, wherein the data processing module calculates the standardized obstructive of each dimension-reduced medical image feature according to the fusion analysis matrix corresponding to each dimension-reduced medical image feature, and the system specifically comprises: acquiring the difference value of the rows and the columns in a fusion analysis matrix corresponding to the feature of the f dimension-reduced medical image, and obtaining the feature difference of the feature of the f dimension-reduced medical image; Acquiring differences between each data pair and a preset experience data pair in a fusion analysis matrix corresponding to the f-th dimension-reduction medical image feature, and averaging the differences to obtain experience difference values of each data pair in the fusion analysis matrix corresponding to the f-th dimension-reduction medical image feature, wherein the preset experience data pair is the extremely poor of the data pair formed according to the clinical feature and the medical image feature; Averaging the empirical difference values of the data pairs to obtain an empirical difference value of the f-th dimension-reduction medical image characteristic; determining the product of the feature difference of the f-th dimension-reduction medical image feature and the empirical difference value of the f-th dimension-reduction medical image feature as the standardized obstacle of the f-th dimension-reduction medical image feature; and obtaining the standardized obstacle of each dimension-reduced medical image characteristic.
9. The intelligent grading system for the severity of the salient eye of integrated image histology according to claim 1, wherein the data processing module determines principal component clinical features from the reduced dimension clinical features according to the standardized obstructive of each reduced dimension medical image feature, specifically comprising: Determining the dimension-reduced medical image features with standardized obstruction less than a preset obstruction threshold as first medical image features, and determining the dimension-reduced medical image features with standardized obstruction greater than or equal to the preset obstruction threshold as second medical image features; When the number of the first medical image features is larger than a preset number threshold, normalizing the fusion analysis matrix of the second medical image features, and analyzing the normalization result by using a principal component analysis method to obtain principal component features; determining the dimensionality reduction clinical characteristics in the main component characteristics as the main component clinical characteristics; When the number of the first medical image features is smaller than or equal to a preset number threshold, normalizing the fusion analysis matrix of each dimension-reduced medical image feature, and analyzing the normalization result by using a principal component analysis method to obtain principal component features; And determining the dimension reduction clinical characteristic in the main component characteristics as the main component clinical characteristic.
10. The intelligent grading system for the severity of the salient eye of the integrated image histology according to claim 1, wherein the data processing module constructs a phase fusion feature matrix according to the clinical features of the principal components and the features of the reduced-dimension medical images, and calculates the standard obstruction of the principal components of the phase fusion feature matrix, and the system specifically comprises: the median of the clinical characteristics of the h main component of each patient is obtained, so that the median of the clinical characteristics of the h main component is obtained; obtaining the median value of the clinical characteristics of each main component; Solving the median of the ith dimension-reduction medical image feature of each patient to obtain the median of the ith dimension-reduction medical image feature; obtaining the median value of each dimension-reduced medical image feature; respectively combining the median value of each main component clinical feature and the median value of each dimension-reduction medical image feature in pairs to obtain data pairs, and constructing a phase fusion feature matrix by taking the data pairs as elements in the matrix; the clinical characteristics of the h main component of each patient are averaged to obtain the average value of the clinical characteristics of the h main component; acquiring the average value of clinical characteristics of each main component, and solving the average value to obtain the average value of the main components of the phase fusion characteristic matrix; solving standard deviation of the median value of each main component clinical feature to obtain a main component representative value of the phase fusion feature matrix; normalizing the product of the principal component mean value of the phase fusion feature matrix and the principal component representative value of the phase fusion feature matrix, and determining the normalization result as the principal component standard obstruction of the phase fusion feature matrix.

Description

Intelligent grading system for eye bursting severity of integrated image histology Technical Field The invention relates to the field of medical image information, in particular to an intelligent grading system for the severity of a sudden eye for integrating image histology. Background Thyroid eye disease is an autoimmune disease associated with thyroid dysfunction (most commonly hyperthyroidism, especially graves 'disease), whose central mechanism is that the immune system produces antibodies (mainly TSH receptor antibodies) that attack the thyroid and orbital tissues, resulting in autoimmune inflammation of the intraorbital tissues and muscles, pushing the patient's eyes to bulge outward, thus forming a sudden eye symptom. For thyroid eye diseases, the symptoms of the protruding eye are mainly divided into an active period and a non-active period, and the treatment means and the difficulty of the intervention in the different periods of the disease progress are different, so in order to effectively treat patients suffering from the thyroid eye diseases, the patients are usually required to be subjected to eye examination, and the severity of the protruding eye of the patients is graded by using image histology. The traditional image histology is to convert standard medical images (such as CT, MRI, PET) into high-dimensional and excavateable quantitative data, and then analyze the quantitative data to support decision making, and essentially translate the images into data which can be statistically analyzed and modeled, and in the actual process of using the quantitative data, the whole essence of thyroid eye diseases cannot be completely represented by extracting information from the medical images, and the biological meaning of the same texture feature change in the medical images is blurred. In the data fusion process, a large number of characteristics exist in clinical data and medical images, redundant noise is easily introduced in direct fusion, so that 'dimension disaster' is caused, the risk of model overfitting is increased, and the interpretation and robustness of the model are reduced. Disclosure of Invention The invention provides an intelligent grading system for the severity of the salient eyes of an integrated image histology, which aims to solve the existing problems. The intelligent grading system for the eye bursting severity of the integrated image histology adopts the following technical scheme: an embodiment of the present invention provides an intelligent grading system for the severity of a salient eye for integrated image histology, comprising: the data acquisition module is used for acquiring clinical characteristics and medical image characteristics of each patient; the data processing module is used for determining dimension-reducing clinical features and dimension-reducing medical image features from clinical features and medical image features for each patient; According to the fusion analysis matrix corresponding to each dimension-reduction medical image feature, calculating the standardized obstruction of each dimension-reduction medical image feature; According to the clinical features of the main components and the medical image features of the dimension reduction, constructing a phase fusion feature matrix, and calculating the main component standard obstruction of the phase fusion feature matrix; If the standard obstructive of the principal component is smaller than a preset obstructive threshold, determining the phase fusion feature matrix as a final fusion feature matrix, otherwise, standardizing the clinical features of the principal component, and constructing the final fusion feature matrix according to the standardized clinical features of the principal component; And the data classification module is used for determining a reference vector according to the final fusion feature matrix, combining the dimensionality reduction medical image features and the main component clinical features of each patient into a personal feature vector, inputting the reference vector and the personal feature vector into a trained intelligent classification model, and outputting the classification grade of each patient through the intelligent classification model. Optionally, in the data processing module, for each patient, the dimension-reducing clinical feature and the dimension-reducing medical image feature are determined from the clinical feature and the medical image feature, and specifically include: constructing an association standard feature matrix according to clinical features and medical image features of each patient, and respectively determining reference clinical features and reference medical image features from the clinical features and the medical image features according to the association standard feature matrix of each patient; Calculating a first association degree of each reference clinical feature and a second association degree of each reference medical image