A method for processing electronic medical record information of patients with patent foramen ovale
By obtaining and correcting the disease characteristic vectors of patients with patent foramen ovale in the electronic medical record information database, the problem of misjudgment of evaluation of PFO patients in the prior art is solved, and the evaluation accuracy and auxiliary diagnosis and treatment effect are improved.
Patent Information
- Application Number
- CN202411027034.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-30
AI Technical Summary
The prior art is prone to misjudgment in evaluating the risk level of patients with patent foramen ovale (PFO), resulting in reduced assessment accuracy, especially confusing with patients with atrial septal defect (ASD).
By obtaining the disorder characteristic vectors of patent foramen ovale and reference diseases in the electronic medical record information database, the discrimination ability coefficient of each detection indicator is calculated, and the risk level is adjusted based on these coefficients. , to reduce misjudgment.
It improves the accuracy of the evaluation of the risk level foramen ovale, reduces misjudgment of PFO patients, and enhances the auxiliary diagnosis and treatment effect of medical staff.
Smart Images

Figure CN118553438B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical data mining and analysis processing, and in particular to a method for processing electronic medical record information of a patent foramen ovale patient. Background Art
[0002] Patent Foramen Ovale (PFO) is a congenital heart structural abnormality, a small hole in the heart's atrial septum that does not close normally during embryonic development. The presence of PFO may lead to a series of health risks and safety hazards such as migraine or stroke, which are related to the risk level of PFO; low-risk patients can usually take appropriate preventive measures to reduce the incidence of complications, but for high-risk patients, certain interventional treatments or surgical treatments are required to ensure the patient's life safety; therefore, the risk level assessment of PFO is crucial.
[0003] At present, the patient's medical records are usually extracted, and then the PFO risk level is evaluated based on the machine learning algorithm, so as to assist the relevant medical staff in formulating prevention or treatment plans for the patients. However, some of the medical records extracted from the patient's medical records for evaluating the PFO risk level may be caused by the patient's atrial septal defect (Atrial Septal Defect, ASD) disease. For example, when evaluating the PFO risk level, the patient is usually required to undergo contrast-enhanced transcranial Doppler ultrasound, also known as the bubble experiment, to detect whether these bubbles enter the left heart system through the heart defect and eventually reach the brain through bubble ultrasound; however, ASD patients will also detect similar bubble results during the bubble examination, which makes it easy for the machine learning algorithm to misjudge ASD patients as PFO patients, thereby reducing the accuracy of the PFO risk level assessment, and further affecting the auxiliary diagnosis and treatment effect of relevant medical staff. Summary of the invention
[0004] In order to solve the technical problem that PFO patients are easily misjudged based on medical record information, resulting in reduced accuracy in PFO risk level assessment, the purpose of the present invention is to provide an electronic medical record information processing method for patent foramen ovale patients, and the technical solution adopted is as follows:
[0005] Obtaining all symptom feature vectors of each medical record sample in the electronic medical record information database, wherein the symptom feature vectors at least include a patent foramen ovale vector and a reference symptom vector, and respectively marking the symptom risk of patent foramen ovale and reference symptom for a preset number of the medical record samples;
[0006] In each of the disease feature vectors, according to the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator, combined with the overall distribution of disease risk labels corresponding to each disease, obtain the discrimination ability coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of each of the disease feature vectors;
[0007] The vector dimensions corresponding to the same detection indicators in different types of disease feature vectors are used as dimensions to be analyzed; according to the distinguishing ability coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of the patent foramen ovale vector, and the difference between the distinguishing ability coefficients of the dimension to be analyzed for distinguishing the disease risk level of the patent foramen ovale vector and the reference disease vector, the correction coefficient of each detection indicator corresponding vector dimension in the patent foramen ovale vector is obtained; the patent foramen ovale vector to be evaluated is adjusted according to the correction coefficient to obtain the risk level of the patent foramen ovale vector to be evaluated.
[0008] Furthermore, the method for obtaining the patent foramen ovale vector and the reference disease vector includes:
[0009] In the medical record sample, the detection parameters of all detection indicators of patent foramen ovale disease are used as vector elements to construct a patent foramen ovale vector, and the detection parameters of all detection indicators of reference diseases are used as vector elements to construct a reference disease vector.
[0010] Furthermore, the reference disease at least includes atrial septal defect disease.
[0011] Furthermore, the method for obtaining the discrimination ability coefficient includes:
[0012] Among all the disease feature vectors under each disease, according to the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator, obtain the first importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector;
[0013] Among all the disease feature vectors under each disease, according to the difference between the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator and the overall distribution of the corresponding disease risk labels, the second importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector is obtained;
[0014] According to the first importance and the second importance, the distinguishing ability coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of each disease feature vector is obtained, and the first importance and the second importance are both positively correlated with the distinguishing ability coefficient.
[0015] Furthermore, the method for obtaining the first importance includes:
[0016] Under each disease, under different vector element values of each detection indicator corresponding vector dimension, according to the imbalance of the number of different disease risk labels in all the disease feature vectors, evaluate the first parameter of each vector element value of each detection indicator corresponding vector dimension for distinguishing the disease risk of each disease feature vector;
[0017] According to the concentrated characteristics of all first parameters of all vector element values of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector, the first importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector is evaluated.
[0018] Furthermore, the method for obtaining the second importance includes:
[0019] Under each disease, under each vector element value of each detection index corresponding vector dimension, a first distribution vector is constructed according to the number of disease risk labels in the disease feature vector; under each disease, a second distribution vector is constructed according to the number of disease risk labels in all the disease feature vectors;
[0020] According to the difference between each of the first distribution vectors and the second distribution vector under different vector element values, the second importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each of the disease feature vectors is obtained.
[0021] Furthermore, obtaining the second importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector according to the difference between each of the first distribution vectors and the second distribution vector under different vector element values includes:
[0022] Obtaining a difference vector between each of the first distribution vectors and the second distribution vector under different vector element values;
[0023] Averaging the infinite norms in each of the difference vectors to obtain a second parameter for distinguishing the disease risk corresponding to the disease feature vector for each detection index corresponding vector dimension;
[0024] According to the similarity between the infinite norms in all the difference vectors, the similarity between all the first distribution vectors under different vector element values is evaluated, and the similarity is negatively correlated and mapped as a third parameter for distinguishing the disease risk corresponding to the disease feature vector for each detection indicator corresponding vector dimension;
[0025] A second importance is obtained according to the second parameter and the third parameter, and both the second parameter and the third parameter are positively correlated with the second importance.
[0026] Furthermore, the method for obtaining the correction coefficient includes:
[0027] Under each dimension to be analyzed, according to the difference between the distinguishing ability coefficients corresponding to the patent foramen ovale vector and the reference disease vector, the dimension weight of each dimension to be analyzed in the patent foramen ovale vector is obtained; the dimension weight under each non-dimension to be analyzed is set to a preset normal number, and the dimension weight of the dimension to be analyzed is greater than the preset normal number;
[0028] According to the dimension weight of the vector dimension corresponding to each detection indicator, the discrimination coefficient under the corresponding dimension in the patent foramen ovale vector is weighted to obtain the correction coefficient under the vector dimension corresponding to each detection indicator in the patent foramen ovale vector.
[0029] Further, the step of adjusting the patent foramen ovale vector to be evaluated according to the correction coefficient to obtain the risk level of the patent foramen ovale vector to be evaluated includes:
[0030] The correction coefficient is used as the correction weight of the vector element on the vector dimension corresponding to the patent foramen ovale vector, and the corresponding vector element is weighted by using the correction weight to obtain all the corrected patent foramen ovale vectors and the patent foramen ovale vector to be evaluated;
[0031] A training sample set was constructed based on all corrected patent foramen ovale vectors and corresponding disease risk level labels, and the risk level of the corrected patent foramen ovale vector to be evaluated was obtained based on the KNN algorithm.
[0032] Furthermore, the preset number is 5% of all case samples in the electronic medical record information database.
[0033] The present invention has the following beneficial effects:
[0034] All symptom feature vectors of each medical record sample are obtained in the electronic medical record information database, and the symptom feature vectors at least include a patent foramen ovale vector and a reference symptom vector, and the symptom risk of patent foramen ovale and reference symptoms are respectively labeled for a preset number of medical record samples; in each symptom feature vector, according to the distribution of symptom risk labels under different vector element values of the vector dimension corresponding to each detection indicator, combined with the overall distribution of the symptom risk labels corresponding to each symptom, the discrimination coefficient of the vector dimension corresponding to each detection indicator for distinguishing the symptom risk level of each symptom feature vector is obtained, and the discrimination coefficient reflects the importance of each detection indicator among many detection indicators and its effect on the corresponding symptom risk. The ability to distinguish can be used for subsequent analysis of the importance of each detection indicator in distinguishing different diseases, so as to adjust and correct the patent foramen ovale vector, thereby reducing misjudgment; the vector dimension corresponding to the same detection indicator in the characteristic vectors of different diseases is used as the dimension to be analyzed; according to the difference between the discriminating ability coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of the patent foramen ovale vector and the discriminating ability coefficient of the dimension to be analyzed for distinguishing the disease risk level of the patent foramen ovale vector and the reference disease vector, the correction coefficient under the dimension of each detection indicator corresponding to the vector in the patent foramen ovale vector is obtained; the patent foramen ovale vector to be evaluated is adjusted according to the correction coefficient to obtain the risk level of the patent foramen ovale vector to be evaluated. The present invention first analyzes the ability of each detection indicator to distinguish the risk level of the corresponding disease under different diseases, and then analyzes and adjusts the interference of similar reference diseases on the evaluation of the patent foramen ovale disease in combination with the difference between the ability of the same detection indicator to distinguish the risk level of the corresponding disease of different diseases, thereby reducing misjudgment, improving the accuracy of the evaluation of the risk level of the patent foramen ovale, and thus improving the auxiliary diagnosis and treatment effect on medical personnel. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0036] Figure 1 A flowchart of a method for processing electronic medical record information of a patient with patent foramen ovale provided by one embodiment of the present invention;
[0037] Figure 2 A flowchart of a method for obtaining a distinguishing capability coefficient provided by an embodiment of the present invention;
[0038] Figure 3 A flowchart of a method for obtaining a correction coefficient provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0039] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the electronic medical record information processing method for a patient with patent foramen ovale proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0040] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0041] The specific scheme of the electronic medical record information processing method for a patent foramen ovale patient provided by the present invention is described in detail below with reference to the accompanying drawings.
[0042] See also Figure 1 , which shows a flow chart of a method for processing electronic medical record information of a patient with patent foramen ovale provided by an embodiment of the present invention, specifically comprising:
[0043] Step S1, obtaining all symptom feature vectors of each medical record sample in the electronic medical record information database, the symptom feature vectors at least including a patent foramen ovale vector and a reference symptom vector, and labeling the symptom risk of patent foramen ovale and reference symptoms for a preset number of medical record samples respectively.
[0044] In one embodiment of the present invention, the medical record file information established for each patient each time they visit the doctor for examination is first extracted from the electronic medical record information database, each medical record file information is used as a medical record sample, and all disease feature vectors are extracted from the medical record sample, wherein the disease feature vector includes at least a patent foramen ovale vector and a reference disease vector, the reference disease is a disease with similar detection parameters or onset characteristics as the patent foramen ovale disease, at least including an atrial septal defect disease; the detection parameter results of the reference disease are likely to interfere with the risk level assessment of the patent foramen ovale disease, thereby facilitating subsequent comparative analysis to accurately assess the risk level of the patent foramen ovale disease.
[0045] The detection parameters of all detection indicators related to patent foramen ovale are used as vector elements to construct a patent foramen ovale vector, wherein the detection indicators related to patent foramen ovale at least include relevant indicators such as bubbling level, pulmonary artery pressure and foramen ovale diameter; similarly, the detection parameters of all detection indicators of atrial septal defect are used as vector elements to construct an atrial septal defect vector, wherein the detection indicators related to atrial septal defect at least include relevant indicators such as bubbling level, cardiac catheter flow and atrial septal defect size.
[0046] It should be noted that in other embodiments of the present invention, the implementer may also obtain other types and quantities of reference disease vectors and detection indicators; considering that the detection parameters under some detection indicators, such as the foaming amount level or the diameter size, have corresponding levels, the detected diameter size parameters are divided into certain levels. For example, an oval foramen with a diameter of 1-2 mm is a small PFO, and the value of the corresponding vector element of its corresponding detection parameter is set to 0. When the diameter is 3-5 mm, the value of the corresponding vector element is set to 1, and when it is greater than 5 mm, the vector element is set to 2; for example, other detection indicators also have situations such as being in line with the normal standard range, being critically high, being critically low, or exceeding the standard range, and they are divided into levels respectively, that is, the corresponding vector elements are label values rather than actual detection parameter values.
[0047] In order to analyze the performance of each detection indicator in the patent foramen ovale vector on the patent foramen ovale disease characteristics, and compare and analyze the distinguishing ability of the detection indicator on distinguishing the patent foramen ovale disease characteristics from the reference disease characteristics, considering that it is labor-intensive and inefficient to label all medical record samples with the risk levels of corresponding diseases, an embodiment of the present invention labels the disease risks of patent foramen ovale and reference diseases for a preset number of medical record samples, specifically, the medical record samples are labeled with the disease risk levels of patent foramen ovale and the risk levels of reference diseases through manual labeling. The greater the severity of the disease, the greater the corresponding disease risk label level; the patent foramen ovale vector in the unlabeled medical record sample is used as the patent foramen ovale vector to be evaluated.
[0048] In one embodiment of the present invention, the preset number is 5% of the total number of all medical record samples; in other embodiments of the present invention, implementers may also set the preset number on their own.
[0049] It should be noted that labeling medical record samples with corresponding disease risk levels is an existing technology well known to those skilled in the art and will not be described in detail here.
[0050] Step S2, in each disease feature vector, according to the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator, combined with the overall distribution of disease risk labels corresponding to each disease, obtain the discrimination ability coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of each disease feature vector.
[0051] Considering that many detection indicators in the disease feature vectors corresponding to different diseases reflect the risk level of the corresponding diseases to varying degrees, for example, the size of the foramen ovale diameter is more intuitive and accurate than other indicators in judging the disease risk of patent foramen ovale, and its importance and ability to distinguish disease risks are relatively higher; and considering that if the distribution of disease risk labels is significantly different under different vector element values of the vector dimension corresponding to each detection indicator in the disease feature vector, it means that the vector dimension corresponding to the detection indicator has a good ability to distinguish disease risks. Otherwise, no matter how the vector element values of the vector dimension corresponding to the detection indicator change, the corresponding disease risk remains unchanged, indicating that the ability of the vector dimension corresponding to the detection indicator to distinguish disease risks is weak.
[0052] At the same time, considering that the distribution of symptom risk labels for each symptom is easily affected by the imbalance of medical record sample data, the distribution of some symptom risk labels is relatively concentrated, which leads to inaccurate assessment of the ability to distinguish symptom risks in the vector dimension corresponding to the detection index; if the distribution of the overall symptom risk labels of the medical record samples is combined with the analysis, the relevant consideration weights for the distinguishing ability analysis can be adjusted in combination with the uneven distribution, thereby reducing the impact of the uneven distribution of samples.
[0053] Therefore, in each disease feature vector, the embodiment of the present invention obtains the discrimination coefficient of the disease risk level of each disease feature vector for each detection indicator corresponding vector dimension according to the distribution of disease risk labels under different vector element values of each detection indicator corresponding vector dimension, combined with the overall distribution of disease risk labels corresponding to each disease. The discrimination coefficient reflects the importance of each detection indicator among many detection indicators and its ability to distinguish the corresponding disease risk, which can be used for subsequent analysis of the importance of each detection indicator in distinguishing different diseases, so as to adjust and correct the patent foramen ovale vector, thereby reducing misjudgment.
[0054] Preferably, in one embodiment of the present invention, the method for obtaining the discrimination capability coefficient includes:
[0055] See also Figure 2 , which shows a flow chart of a method for obtaining a distinguishing ability coefficient provided by an embodiment of the present invention:
[0056] Step S201, among all the disease feature vectors under each disease, according to the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator, obtain the first importance of the vector dimension corresponding to each detection indicator in distinguishing the disease risk of each disease feature vector.
[0057] The first importance evaluates the importance of each detection indicator corresponding vector dimension and its ability to distinguish the corresponding disease risk from the distribution of disease risk labels under different vector element values of each detection indicator corresponding vector dimension. The more unbalanced the distribution, the better the ability of the detection indicator corresponding vector dimension to distinguish the disease risk.
[0058] Preferably, in one embodiment of the present invention, considering that when the disease risk level tends to be concentrated at a certain level, it means that the quantity distribution of different disease risk labels is more unbalanced; the method for obtaining the first importance includes:
[0059] Under each disease, under different vector element values of each detection indicator corresponding vector dimension, according to the imbalance of the number of different disease risk labels in all disease feature vectors, evaluate the first parameter of each vector element value of each detection indicator corresponding vector dimension for distinguishing the disease risk of each disease feature vector;
[0060] According to the concentrated characteristics of all first parameters of all vector element values of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector, the first importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector is evaluated.
[0061] The calculation formula of the first parameter specifically includes:
[0062] ;
[0063] In the formula, The dimension number of the vector dimension corresponding to each detection indicator under each disease; The sequence number of the vector element value in the vector dimension corresponding to each detection indicator under each disease; For each disease The detection index corresponds to the vector dimension The value of the vector element corresponds to the number of the disease risk level with the largest total number of labels among all the disease risk level labels; For the The risk level numbers of other diseases except the risk level number of the disease; The total number of levels of risk for each condition is the total number of levels for each condition; For each disease The detection index corresponds to the vector dimension The value of the vector element is the first parameter for distinguishing the disease risk of the corresponding disease feature vector; is an S-type function, used for normalization; For each disease The detection index corresponds to the vector dimension The value of the vector element corresponds to the risk level labels of all different diseases. The total number of labels under each disease risk level; For each disease The detection index corresponds to the vector dimension The value of the vector element corresponds to the risk level labels of all different diseases. The total number of labels under each disease risk level.
[0064] In the calculation formula of the first parameter, the quantitative imbalance of different disease risk labels is evaluated by the difference between the total number of labels corresponding to the maximum risk level under all different vector element values and the total number of labels of other risk levels. The more unbalanced the quantitative distribution is, the larger the first parameter is, which means that the value of the vector element of the vector dimension corresponding to the detection index has a stronger ability to distinguish the disease risk level of each disease characteristic vector, and the corresponding vector dimension is more likely to be more important or have a greater distinguishing ability.
[0065] In another embodiment of the present invention, the calculation method of the first parameter is: statistically calculate the proportion of each disease risk level label under each vector element value of the vector dimension corresponding to each detection indicator under each disease, evaluate the variance between all proportions, and normalize the variance using the maximum and minimum values as the first parameter; the larger the variance, the more uneven the distribution of different disease risk level labels under the corresponding vector element value; in other embodiments of the present invention, the implementer may also adopt other normalization methods, which are not repeated here.
[0066] In other embodiments of the present invention, the implementer may also use information entropy or the Gini coefficient to evaluate the quantitative distribution balance of each disease risk level label under each vector element value of each detection indicator corresponding to the vector dimension under each disease, and then obtain the first parameter; information entropy or the Gini coefficient are both existing technologies well known to those skilled in the art and will not be elaborated here.
[0067] After obtaining the first coefficient of each vector element value of the vector dimension corresponding to each detection indicator under each disease, in one embodiment of the present invention, the mean of the first coefficients under different vector element values is used as the first importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease characteristic vector. When the first coefficients under different vector element values are all large, the larger the corresponding mean, the greater the importance and distinguishing ability of the corresponding vector dimension for distinguishing the disease risk of each disease characteristic vector. In other embodiments of the present invention, the sum, median or mode may also be used as the first importance, which will not be repeated here.
[0068] Step S202, among all the disease feature vectors under each disease, according to the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator and the difference with the overall distribution of the corresponding disease risk labels, obtain the second importance of the vector dimension corresponding to each detection indicator in distinguishing the disease risk of each disease feature vector.
[0069] The second importance is to analyze the distribution of the overall disease risk labels of the medical record samples, thereby reducing the impact of the uneven distribution of medical record sample data or sampling, so as to accurately analyze the ability of each detection indicator corresponding to the vector dimension to distinguish the disease risk of each disease feature vector.
[0070] Preferably, in one embodiment of the present invention, the method for obtaining the second importance includes:
[0071] Under each disease, under each vector element value of each detection index corresponding vector dimension, a first distribution vector is constructed according to the number of disease risk labels in the disease feature vector; under each disease, a second distribution vector is constructed according to the number of disease risk labels in all disease feature vectors;
[0072] According to the difference between each first distribution vector and the second distribution vector under different vector element values, the second importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector is obtained.
[0073] It should be noted that the number of labels can be directly used as the vector elements in the corresponding first distribution vector and the second distribution vector, or the number can be normalized and used as the vector element, which mainly reflects the proportion of the number of labels; when measuring the difference vector, it is necessary to ensure that the sorting order of the vector dimensions, that is, the disease risk level, between the two distribution vectors is consistent, so as to accurately obtain the difference vector.
[0074] As an example, to help understand the first distribution vector or the second distribution vector: for example, among a total of 100 patent foramen ovale vectors, in the vector dimension corresponding to the foramen ovale diameter detection index, the corresponding vector element values are 0, 1, and 2 respectively; when the vector element value is 0 in this vector dimension, there are 41 patent foramen ovale vectors, and the disease risk level labels corresponding to these 41 patent foramen ovale vectors are 21 for level I, 18 for level II, and 2 for level III. For example, there are a total of five patent foramen ovale disease risk levels, and the corresponding first distribution vector is (21, 18, 2, 0, 0); the method for obtaining the second distribution vector is similar, and it is only necessary to construct a vector according to the number of labels corresponding to the different disease risk levels of all patent foramen ovale vectors, which will not be repeated here.
[0075] Specifically, one embodiment of the present invention obtains a difference vector between each first distribution vector and the second distribution vector under different vector element values;
[0076] The infinite norm in each difference vector is averaged to obtain a second parameter of the disease risk of each detection index corresponding to the vector dimension for distinguishing the corresponding disease feature vector;
[0077] According to the similarity between the infinite norms in all difference vectors, the similarity between all first distribution vectors under different vector element values is evaluated, and the similarity is negatively correlated and mapped as the third parameter for distinguishing the disease risk of the corresponding disease feature vector for each detection indicator corresponding vector dimension;
[0078] The second importance is obtained according to the second parameter and the third parameter, and both the second parameter and the third parameter are positively correlated with the second importance.
[0079] The calculation formula for the second importance includes:
[0080] ;
[0081] In the formula, For each disease The second importance of the vector dimension corresponding to each detection indicator in distinguishing the disease risk of the corresponding disease feature vector; The dimension number of the vector dimension corresponding to each detection indicator under each disease; The sequence number of the vector element value in the vector dimension corresponding to each detection indicator under each disease; To find the mean sign; is the symbol of infinity norm; For the A difference vector between each first distribution vector and the second distribution vector under different vector element values under the vector dimension corresponding to the detection index; For the The second parameter of the disease risk of the corresponding disease feature vector is distinguished by the vector dimension corresponding to the detection index; To find the similarity function; The natural constant The exponential function with base ; For the The third parameter of the vector dimension corresponding to each detection indicator is used to distinguish the disease risk of the corresponding disease feature vector.
[0082] It should be noted that the method for obtaining the third parameter is specifically as follows: Under the vector dimension corresponding to each detection indicator, the difference vector between each first distribution vector and the second distribution vector under different vector element values, after taking the infinite norm of all difference vectors, the infinite norms are subtracted pairwise, all differences are averaged, the smaller the mean of the differences, the greater the similarity, the reciprocal operation of the mean of the differences is performed to obtain the similarity; and because the greater the similarity, the weaker the ability of the corresponding vector dimension of the detection indicator to distinguish the corresponding disease risk level, the similarity is negatively correlated with the exponential function to obtain the third parameter, and the smaller the negative correlation mapping value, the smaller the third parameter; in other embodiments of the present invention, considering that the smaller the mean of the differences, the smaller the third parameter, the implementer can also directly use the mean of the differences as the third parameter; the implementer can also obtain the variance of the infinite norm of all difference vectors, the smaller the variance, the greater the similarity, and the third parameter can also be obtained after the corresponding correlation mapping processing.
[0083] In other embodiments of the present invention, the implementer may compare all difference vectors pairwise, use the DTW distance to evaluate the similarity between the vectors, and then average all similarities to obtain the third parameter; other related similarity measurement methods may also be used, which, together with the infinite norm and the above-mentioned correlation measurement methods, are already prior arts well known to those skilled in the art and will not be described in detail here.
[0084] In the calculation formula of the second importance, the second parameter reflects the distribution of disease risk levels under different vector element values under the vector dimension corresponding to each detection indicator, and the average level of the difference between the corresponding disease risk level distribution. The larger the difference, the larger the second parameter, which indicates that the corresponding vector dimension has a stronger ability to distinguish the disease risk levels of this type of disease; the third parameter combines the difference between the disease risk level distribution under different element values and the overall disease risk level distribution, and evaluates the similarity of the disease risk level distribution under different element values, thereby indirectly evaluating the disease risk level distinction ability of the corresponding vector dimension for this type of disease. The larger the third parameter, the stronger the distinction ability; then the second parameter and the third parameter are combined by multiplication. The larger the second parameter or the third parameter, the second importance, and the corresponding stronger the distinction ability; in other embodiments of the present invention, the implementer may also use other basic mathematical operations such as addition or weighted summation or other related mapping methods, which are not repeated here.
[0085] Step S203, according to the first importance and the second importance, obtain the discrimination coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of each disease feature vector, and the first importance and the second importance are both positively correlated with the discrimination coefficient.
[0086] Both the first importance and the second importance can reflect the importance and distinguishing ability of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of the corresponding disease. Therefore, in one embodiment of the present invention, the two are combined by multiplication to obtain the distinguishing ability coefficient; in other embodiments of the present invention, the implementer may also use other basic mathematical operations such as addition or weighted summation or other related mapping methods, which are not repeated here.
[0087] Step S3, taking the vector dimensions corresponding to the same detection indicators in different disease feature vectors as the dimensions to be analyzed; obtaining the correction coefficients of the vector dimensions corresponding to each detection indicator in the patent foramen ovale vector according to the discrimination coefficients of the disease risk levels of the patent foramen ovale vector corresponding to the vector dimensions of each detection indicator, and the difference between the discrimination coefficients of the disease risk levels of the patent foramen ovale vector and the reference disease vector of the dimension to be analyzed; adjusting the patent foramen ovale vector to be evaluated according to the correction coefficients to obtain the risk level of the patent foramen ovale vector to be evaluated.
[0088] Since the same detection indicators may exist in the disease feature vectors corresponding to different diseases to interfere with the risk level assessment of a certain disease, misjudgment is prone to occur; considering that if a detection indicator has a strong distinguishing ability for assessing patent foramen ovale or other reference diseases such as atrial septal defect, it means that the assessment of the corresponding risk level may be more affected by the interference of other reference diseases, and the accuracy of the risk assessment for patent foramen ovale may be lower.
[0089] Therefore, the embodiment of the present invention first uses the vector dimensions corresponding to the same detection indicators in the characteristic vectors of different diseases as the dimensions to be analyzed; by analyzing the corresponding distinguishing ability coefficients and the differences therebetween, the analysis reference weights of the corresponding dimensions in the patent foramen ovale vector are adjusted and corrected, thereby reducing the possibility of misjudgment of the risk level of the patent foramen ovale vector to be evaluated.
[0090] Preferably, in one embodiment of the present invention, the method for obtaining the correction coefficient includes:
[0091] See also Figure 3 , which shows a flow chart of a method for obtaining a correction coefficient provided by an embodiment of the present invention;
[0092] Step S301, in each dimension to be analyzed, according to the difference between the corresponding distinguishing ability coefficients of the patent foramen ovale vector and the reference disease vector, obtain the dimension weight of each dimension to be analyzed in the patent foramen ovale vector; set the dimension weight of each non-dimension to be analyzed to a preset normal number, and the dimension weight of the dimension to be analyzed is greater than the preset normal number.
[0093] Considering that under the dimension to be analyzed, the smaller the difference between the corresponding discriminative ability coefficient of the patent foramen ovale vector and the corresponding discriminative ability coefficients of other reference disease vectors, the greater the interference effect of the dimension to be analyzed on the assessment of the risk level of the patent foramen ovale disease. At the same time, if the difference is less than 0, it means that the reference value of the dimension to be analyzed for assessing the risk level of the reference disease is higher, and the reference value for assessing the risk level of the patent foramen ovale disease is lower; but if the difference is greater than 0 and the difference is larger, it means that the reference value and accuracy for assessing the risk level of the patent foramen ovale disease are higher.
[0094] In one embodiment of the present invention, the difference between the corresponding discrimination coefficient of the patent foramen ovale vector and the corresponding discrimination coefficient of other reference disease vectors is mapped to The normalization process is performed in the function, and the normalized value is used as the exponent in the exponential function with the natural constant e as the base. The value is amplified for subsequent analysis and operation, and the exponential function value is used as the dimension weight under the dimension to be analyzed; the dimension weight under the dimension not to be analyzed is set to 1, so that the dimension weight of the dimension to be analyzed is greater than the dimension weight under the dimension not to be analyzed, thereby increasing the dimension vector elements with strong distinguishing ability and less interference from the reference disease, so that in the subsequent classification process, the weight of the consideration between the marked patent foramen ovale vector and the patent foramen ovale vector to be evaluated on this dimension is greater, thereby improving the accuracy of the disease risk level assessment.
[0095] Step S302, according to the dimension weight of the vector dimension corresponding to each detection indicator, the discrimination coefficient of the corresponding dimension in the patent foramen ovale vector is weighted to obtain the correction coefficient of the vector dimension corresponding to each detection indicator in the patent foramen ovale vector.
[0096] In one embodiment of the present invention, considering that the distinguishing ability coefficient under the dimension to be analyzed also reflects to a certain extent the distinguishing ability and importance of this dimension to the risk level of patent foramen ovale, the dimensional weight of the vector dimension corresponding to each detection indicator is used to weight the corresponding distinguishing ability coefficient to obtain the correction coefficient under the vector dimension corresponding to each detection indicator.
[0097] At this point, the correction coefficient of each detection index in the patent foramen ovale vector corresponding to the vector dimension is obtained.
[0098] Then, the patent foramen ovale vector to be evaluated is adjusted according to the correction coefficient to reduce the possibility of misjudgment of the risk level of the patent foramen ovale disease by each detection index, so as to accurately obtain the risk level of the patent foramen ovale vector to be evaluated.
[0099] Preferably, in one embodiment of the present invention, adjusting the patent foramen ovale vector to be evaluated according to the correction coefficient to obtain the risk level of the patent foramen ovale vector to be evaluated includes:
[0100] The correction coefficient is used as the correction weight of the vector element on the vector dimension corresponding to the patent foramen ovale vector, and the corresponding vector element is weighted by using the correction weight to obtain all the corrected patent foramen ovale vectors and the patent foramen ovale vector to be evaluated;
[0101] A training sample set was constructed based on all corrected patent foramen ovale vectors and corresponding disease risk level labels, and the risk level of the corrected patent foramen ovale vector to be evaluated was obtained based on the KNN algorithm.
[0102] Specifically, the patent foramen ovale feature vectors in all medical record samples in the electronic medical record information database were corrected and adjusted, and then the patent foramen ovale vectors that had been labeled with disease risk and corrected and adjusted were used as training samples to construct a training sample set, and the K-Nearest Neighbors (KNN) algorithm was used to evaluate the risk level of the corrected patent foramen ovale vector to be evaluated, thereby improving the accuracy and efficiency of the risk level assessment of the patent foramen ovale vector.
[0103] It should be noted that the implementer may also use other deep learning algorithms such as neural network classification models to perform label evaluation and classification. This and the KNN algorithm are already existing technologies well known to those skilled in the art and will not be described in detail here.
[0104] After classifying and evaluating the disease risk level of all patent ovale vectors of each medical record sample, it can assist relevant medical personnel in evaluating the patient's condition and formulating prevention or treatment plans.
[0105] In summary, the embodiment of the present invention first obtains all the disease feature vectors of each medical record sample, the disease feature vectors at least including the patent foramen ovale vector and the reference disease vector, and performs corresponding disease risk annotations on a preset number of medical record samples; further obtains the discrimination coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of each disease feature vector; then uses the vector dimension corresponding to the same detection indicator in different disease feature vectors as the dimension to be analyzed; according to the difference between the discrimination coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of the patent foramen ovale vector and the discrimination coefficient of the dimension to be analyzed for distinguishing the disease risk level between the patent foramen ovale vector and the reference disease vector, obtain the correction coefficient under the vector dimension corresponding to each detection indicator in the patent foramen ovale vector, and then adjust the patent foramen ovale vector to be evaluated to obtain the risk level of the patent foramen ovale vector to be evaluated. The present invention first analyzes the ability of each detection indicator under different symptoms to distinguish the risk level of the corresponding symptoms, and then adjusts the interference effect of similar reference symptoms on the evaluation of patent foramen ovale by combining the difference analysis between the abilities of the same detection indicators of different symptoms to distinguish the risk levels of the corresponding symptoms, thereby reducing misjudgment, improving the accuracy of the assessment of the risk level of patent foramen ovale, and further improving the auxiliary diagnosis and treatment effect for medical personnel.
[0106] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0107] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.
Claims
1. A method for processing electronic medical record information of a patient with patent foramen ovale, characterized in that: The method comprises: Obtaining all symptom feature vectors of each medical record sample in the electronic medical record information database, wherein the symptom feature vectors at least include a patent foramen ovale vector and a reference symptom vector, and respectively marking the symptom risk of patent foramen ovale and reference symptom for a preset number of the medical record samples; In each of the disease feature vectors, according to the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator, combined with the overall distribution of disease risk labels corresponding to each disease, obtain the discrimination ability coefficient of the vector dimension corresponding to each detection indicator for distinguishing the disease risk level of each of the disease feature vectors; The vector dimensions corresponding to the same detection indicators in different types of disease feature vectors are used as dimensions to be analyzed; according to the distinguishing ability coefficient of the dimension corresponding to each detection indicator for distinguishing the disease risk level of the patent foramen ovale vector, and the difference between the distinguishing ability coefficients of the dimension to be analyzed for distinguishing the disease risk level of the patent foramen ovale vector and the reference disease vector, the correction coefficient under the dimension corresponding to each detection indicator in the patent foramen ovale vector is obtained; the patent foramen ovale vector to be evaluated is adjusted according to the correction coefficient to obtain the risk level of the patent foramen ovale vector to be evaluated; The method for obtaining the patent foramen ovale vector and the reference disease vector includes: In the medical record sample, the detection parameters of all detection indicators of the patent foramen ovale disease are used as vector elements to construct a patent foramen ovale vector, and the detection parameters of all detection indicators of the reference disease are used as vector elements to construct a reference disease vector; The reference disease condition at least includes atrial septal defect disease; The method for obtaining the discrimination ability coefficient includes: Among all the disease feature vectors under each disease, according to the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator, obtain the first importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector; Among all the disease feature vectors under each disease, according to the difference between the distribution of disease risk labels under different vector element values of the vector dimension corresponding to each detection indicator and the overall distribution of the corresponding disease risk labels, the second importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector is obtained; According to the first importance and the second importance, obtaining a discrimination coefficient for distinguishing the disease risk level of each disease feature vector corresponding to the vector dimension of each detection indicator, wherein the first importance and the second importance are both positively correlated with the discrimination coefficient; The method for obtaining the correction coefficient includes: Under each dimension to be analyzed, according to the difference between the distinguishing ability coefficients corresponding to the patent foramen ovale vector and the reference disease vector, the dimension weight of each dimension to be analyzed in the patent foramen ovale vector is obtained; the dimension weight under each non-dimension to be analyzed is set to a preset normal number, and the dimension weight of the dimension to be analyzed is greater than the preset normal number; According to the dimension weight of the vector dimension corresponding to each detection indicator, the discrimination coefficient under the corresponding dimension in the patent foramen ovale vector is weighted to obtain a correction coefficient under the dimension of the vector corresponding to each detection indicator in the patent foramen ovale vector; The step of adjusting the patent foramen ovale vector to be evaluated according to the correction coefficient to obtain the risk level of the patent foramen ovale vector to be evaluated includes: The correction coefficient is used as the correction weight of the vector element on the vector dimension corresponding to the patent foramen ovale vector, and the corresponding vector element is weighted by using the correction weight to obtain all the corrected patent foramen ovale vectors and the patent foramen ovale vector to be evaluated; A training sample set is constructed based on all corrected patent foramen ovale vectors and corresponding disease risk level labels, and the risk level of the corrected patent foramen ovale vector to be evaluated is obtained based on the KNN algorithm; The method for obtaining the second importance includes: Under each disease, under each vector element value of each detection index corresponding vector dimension, a first distribution vector is constructed according to the number of disease risk labels in the disease feature vector; under each disease, a second distribution vector is constructed according to the number of disease risk labels in all the disease feature vectors; According to the difference between each of the first distribution vectors and the second distribution vector under different vector element values, the second importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each of the disease feature vectors is obtained.
2. The electronic medical record information processing method for a patent foramen ovale patient according to claim 1, characterized in that: The method for obtaining the first importance includes: Under each disease, under different vector element values of each detection indicator corresponding vector dimension, according to the imbalance of the number of different disease risk labels in all the disease feature vectors, evaluate the first parameter of each vector element value of each detection indicator corresponding vector dimension for distinguishing the disease risk of each disease feature vector; According to the concentrated characteristics of all first parameters of all vector element values of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector, the first importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each disease feature vector is evaluated.
3. The method for processing electronic medical record information of a patent foramen ovale patient according to claim 1, characterized in that: The obtaining, according to the difference between each of the first distribution vectors and the second distribution vector under different vector element values, the second importance of the vector dimension corresponding to each detection indicator for distinguishing the disease risk of each of the disease feature vectors comprises: Obtaining a difference vector between each of the first distribution vectors and the second distribution vector under different vector element values; Averaging the infinite norms in each of the difference vectors to obtain a second parameter for distinguishing the disease risk corresponding to the disease feature vector for each detection index corresponding vector dimension; According to the similarity between the infinite norms in all the difference vectors, the similarity between all the first distribution vectors under different vector element values is evaluated, and the similarity is negatively correlated and mapped as a third parameter for distinguishing the disease risk corresponding to the disease feature vector for each detection indicator corresponding vector dimension; A second importance is obtained according to the second parameter and the third parameter, and both the second parameter and the third parameter are positively correlated with the second importance.
4. The method for processing electronic medical record information of a patent foramen ovale patient according to claim 1, characterized in that: The preset number is 5% of all case samples in the electronic medical record information database.
