CN115984889B - Medical document integrity analysis method and system based on artificial intelligence - Google Patents
Medical document integrity analysis method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN115984889B CN115984889B CN202310284541.0A CN202310284541A CN115984889B CN 115984889 B CN115984889 B CN 115984889B CN 202310284541 A CN202310284541 A CN 202310284541A CN 115984889 B CN115984889 B CN 115984889B
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- central control
- control unit
- sample data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention relates to the technical field of data processing, in particular to an artificial intelligence-based medical document integrity analysis method and system, which comprises the steps that a data acquisition unit acquires a plurality of medical document scanning images, a data processing unit screens the medical document scanning images with resolution meeting the standard as sample data and randomly divides the sample data into a training set, a verification set and a test set; the modeling unit establishes a neural network model, and inputs a training set into the neural network model for training; the modeling unit tests the neural network model by using the verification set, and the central control unit judges whether to adjust the iteration times and hidden layers of the neural network model according to the test result of the neural network model; the modeling unit uses the test set to verify the neural network model with the test result meeting the standard, and the central control unit judges whether the number of the sample data needs to be increased according to the verification result of the neural network model. The invention improves the analysis accuracy of the integrity of the medical document.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a medical document integrity analysis method and system based on artificial intelligence.
Background
The medical document is a comprehensive record of a medical process, is also a diagnosis and treatment basis of doctors on patients, and the proposal for modifying the medical document is a general name of information data specification documents which are manufactured according to professional technical specification requirements and reflect medical service relations, patient health conditions, medical measures and the like and have preservation values in diagnosis and treatment processes of medical institutions, medical staff, and the like. The medical document represents the medical quality and academic thought level of medical institutions and medical staff, and the important marks for measuring the medical activities and the technical level are provided. The integrity of the medical document plays an important role in improving the medical level, standardizing the medical behavior, guaranteeing the rights and interests of patients and the like.
Chinese patent publication No.: CN113707252B discloses a quality control method and system for intelligent cases, which uses artificial intelligence technology as a core, and finds defect contents and identifies reasons through understanding medical record semantics and evaluating diagnosis and treatment paths.
The integrity analysis of the medical document has important significance for medical work, not only reflects the medical quality and management level of hospitals and reflects the business level of medical staff, but also is an important material for clinical teaching, scientific research, summarized experience and hospital information management. However, in the prior art, quality control and analysis are performed on the integrity of an electronic case, and the medical document not only comprises an electronic medical record, but also comprises paper documents such as an outpatient medical record, an inpatient medical record, an outpatient prescription, a doctor shift report, various application forms, a report form, a three-test form, a nursing shift report, a special nursing record sheet and the like.
Disclosure of Invention
Therefore, the invention provides an artificial intelligence-based medical document integrity analysis method and system, which are used for solving the problems of lower accuracy and lower efficiency of integrity analysis of paper medical documents in the prior art.
To achieve the above object, in one aspect, the present invention provides a medical document integrity analysis method based on artificial intelligence, including:
step S1, a data acquisition unit acquires a plurality of medical document scanning images, a data processing unit screens the medical document scanning images with resolution meeting the standard as sample data and stores the sample data into a sample data set, and the sample data set randomly divides the sample data into a training set, a verification set and a test set;
step S2, a modeling unit establishes a neural network model, sets the iteration times N and the hidden layers A of the neural network model, and inputs the training set into the neural network model for training;
step S3, the modeling unit uses the verification set to verify the neural network model, and the central control unit judges whether to adjust the iteration times and hidden layers of the neural network model according to the verification result of the neural network model;
step S4, the modeling unit tests the neural network model with the test result conforming to the standard by using the test set, and the central control unit judges whether the number of the sample data needs to be increased according to the test result of the neural network model;
and S5, analyzing the integrity of the medical document by using the neural network model.
Further, in the step S1, the data processing unit detects the resolution Q of each of the medical document scanned images, compares Q with a preset resolution Q0, and determines whether the resolution of the medical document scanned image meets a standard according to the comparison result,
if Q is more than or equal to Q0, the data processing unit judges that the resolution of the medical document scanning image corresponding to the resolution Q meets the standard, and the medical document scanning image is used as sample data to be stored in a sample data set;
if Q is smaller than Q0, the data processing unit judges that the resolution of the medical document scanning image corresponding to the resolution Q does not accord with the standard.
Further, in the step S1, the plurality of medical document scanning images acquired by the data acquisition unit include a plurality of medical document scanning images of a first degree of integrity, a plurality of medical document scanning images of a second degree of integrity, a plurality of medical document scanning images of a third degree of integrity, and a plurality of medical document scanning images of a fourth degree of integrity.
Further, in the step S3, the central control unit determines whether to tune the neural network model according to the test result of the neural network model,
when the modeling unit selects sample data A1 and A2. An in the verification set to verify the neural network model, the corresponding preset results are A1 and a2. An, the verification result of the neural network model is b1 and b2.. Bn, the central control unit corresponds the preset result to the numerical value in the verification result, calculates the verification error rate tau, and sets the verification error rate tau
The central control unit compares the verification error rate tau with a preset error rate tau 0,
if tau is less than or equal to tau 0, the central control unit judges that the verification error rate meets the standard, and the verification result of the neural network model meets the standard, so that parameter adjustment on the neural network model is not needed;
if τ > τ0, the central control unit determines that the verification error rate does not meet the standard, and the neural network model verification result does not meet the standard, and parameter adjustment is required to be performed on the neural network model.
Further, in the step S4, when the central control unit determines that the verification result of the neural network model does not meet the standard, that is, τ > τ0, a difference Δτ between the verification error rate τ and the preset error rate τ0 is calculated, and the hidden layer number of the neural network model is adjusted according to Δτ, where Δτ=τ—τ0 is set,
if Δτ is smaller than Δτ1, the central control unit selects α1 to adjust the hidden layer number to a1=a×α1;
if Δτ1 is less than or equal to Δτ2 and Δτ2, the central control unit selects α2 to adjust the hidden layer number to a1=axα2;
if Deltaτ is more than or equal to Deltaτ2, the central control unit selects alpha 3 to adjust the hidden layer number to A1=A×alpha 3;
wherein Δτ1 is a first preset error rate difference, Δτ2 is a second preset error rate difference, Δτ1 is less than Δτ2, α1 is a first hidden layer number adjustment coefficient, α2 is a second hidden layer number adjustment coefficient, α3 is a third hidden layer number adjustment coefficient, values of α1, α2 and α3 are respectively 0.8 < α1 < 1.2,0.6 < α2 < 1.5,0.4 < α3 < 2, and neither α1, α2, nor α3 is 1, and α1, α2, α3 are randomly valued in their corresponding value ranges when the hidden layer number is updated, and when A1 is not a positive integer, the value of A1 is the largest positive integer smaller than A1.
Further, the central control unit is provided with a maximum hidden layer number Amax, the central control unit compares the adjusted hidden layer number A1 with Amax,
if A1 is less than or equal to Amax, the central control unit sets the hidden layer number as A1;
if A1 is larger than Amax, the hidden layer number is set to Amax by the central control unit.
Further, when A1 is larger than Amax, the central control unit calculates the difference value delta A between the hidden layer number A1 and the maximum hidden layer number Amax, adjusts the iteration number of the neural network model according to delta A, sets delta A=A1-Amax,
if Δa is less than Δa1, the central control unit selects β1 to adjust the iteration times to n1=n×β1;
if Δa1 is less than or equal to Δa < Δa2, the central control unit selects β2 to adjust the iteration times to n1=n×β2;
if Δa is greater than or equal to Δa2, the central control unit selects β3 to adjust the iteration times to n1=n×β3;
wherein Δa1 is a first preset hidden layer number difference value, Δa2 is a second preset hidden layer number difference value, Δa1 is less than Δa2, β1 is a first iteration number adjustment coefficient, β2 is a second iteration number adjustment coefficient, β3 is a third iteration number adjustment coefficient, values of β1, β2 and β3 are respectively more than 0.8 and less than β1 and less than 1.2,0.6 and less than β2 and less than 1.5,0.4 and less than β3 and are not 1, each time when the iteration number is updated, β1, β2 and β3 are randomly valued in a corresponding value range, and when N1 is not a positive integer, the value of N1 is the largest positive integer smaller than N1.
Further, when the central control unit determines that the neural network model verification result meets the standard, the modeling unit tests the neural network model by using a test set, and the central control unit determines whether the number of sample data needs to be increased according to the test result, wherein,
when the modeling unit selects sample data C1 and C2. Cm in the test set to test the neural network model, the corresponding preset results are C1 and c2. Cm, the test results of the neural network model are d1 and d2 dm, and the central control unit corresponds the preset results with the numerical values in the test results and calculates a test error rate phi and sets
The central control unit compares the test error rate phi with a preset error rate tau 0,
if phi is less than or equal to tau 0, the central control unit judges that the test error rate meets the standard, the test result of the neural network model meets the standard, and the number of the sample data is not required to be increased;
if phi is more than tau 0, the central control unit judges that the test error rate does not meet the standard, the test result of the neural network model does not meet the standard, and the number of the sample data is required to be increased.
Further, the central control unit calculates a difference delta phi between a test error rate phi and a preset error rate tau 0 when the test result of the neural network model is judged to be not in accordance with the standard, adjusts the number of sample data according to delta phi, sets delta phi = phi-tau 0,
if delta phi is more than or equal to delta phi 2, the central control unit adjusts the number of the sample data to N1, and N1 = 5 x N0 is set;
if Δφ1 is less than or equal to Δφ < Δφ2, the central control unit adjusts the number of sample data to N1, and N1 = 3 x N0 is set;
if Δφ < Δφ1, the central control unit adjusts the number of sample data to N1, and sets N1=2×N0;
wherein, delta phi 1 is the first preset test error rate difference, delta phi 2 is the second preset test error rate difference, and delta phi 1 is less than delta phi 2.
In another aspect, the present invention provides an artificial intelligence based medical document integrity analysis system comprising:
the data acquisition unit is used for acquiring a plurality of medical document scanning images;
the data processing unit is connected with the data acquisition unit and is used for screening the medical document scanning image with the resolution meeting the standard as sample data, storing the sample data into a sample data set and randomly dividing the sample data into a training set, a verification set and a test set;
the modeling unit is connected with the data processing unit and used for establishing a neural network model and training, testing and verifying the neural network model by using sample data;
the central control unit is connected with the modeling unit and used for judging whether the iteration times and hidden layers of the neural network model are adjusted according to the test result of the neural network model, judging whether the number of sample data is required to be increased according to the verification result of the neural network model, and analyzing the integrity of the medical document by using the trained neural network model.
Compared with the prior art, the invention has the beneficial effects that the modeling unit learns sample data with different integrality by establishing the neural network model so as to realize intelligent analysis on the integrality of the medical document, and the invention improves the analysis efficiency of the integrality of the medical document by applying the artificial intelligence technology; if the iteration number and hidden layer number of the neural network model are too large, the model is complex, the operation data are huge, the operation efficiency is low, and even the generalization capability of the model is poor, if the iteration number and hidden layer number of the neural network model are too small, the model is larger in prediction error, the iteration number and hidden layer number are regulated according to the test error rate and the verification error rate, and the neural network model is guaranteed to have better prediction precision.
Furthermore, the medical document scanning image with the resolution meeting the standard is used as the sample data by setting the preset resolution to screen the medical document scanning image, so that the identification degree of the sample data is improved.
Furthermore, the data acquisition unit provides sample data with different integrality by acquiring a plurality of medical document scanning images with different integrality so as to realize that the neural network model analyzes the integrality of the medical document, and improves the accuracy of analyzing the integrality of the medical document by finely dividing the integrality of the sample data.
Furthermore, the invention verifies the trained neural network model by setting the verification set to judge whether the neural network model can be put into use, thereby improving the prediction precision of the neural network model, and limits the error rate to be within 10 percent, preferably to be within 5 percent, and further improving the prediction precision of the neural network model.
Further, when the verification result of the neural network model is judged to be inconsistent with the standard, the central control module adjusts the hidden layer number so as to increase the complexity of the model and further increase the prediction precision of the model.
Furthermore, the complexity of the neural network model is ensured to float in a proper range by setting the maximum hidden layer number, the occurrence of the over-fitting phenomenon is avoided, and meanwhile, the operation rate of the neural network model is ensured by setting the maximum hidden layer number.
Further, when the hidden layer number of the neural network model is adjusted to the maximum value by the central control module, the iteration number of the neural network model is adjusted, so that the prediction accuracy of the model is improved.
Furthermore, the neural network model with the test result meeting the standard is tested by setting the test set so as to judge whether the neural network model can be put into use or not, and the prediction precision of the neural network model is further improved.
Further, when the neural network model test result is judged to be inconsistent with the standard, the central control module adjusts the quantity of the sample data to increase the prediction accuracy of the model.
Drawings
FIG. 1 is a flow chart of a medical document integrity analysis method based on artificial intelligence in accordance with an embodiment of the present invention;
FIG. 2 is a block diagram of an artificial intelligence based medical document integrity analysis system in accordance with an embodiment of the present invention.
Detailed Description
In order that the objects and advantages of the invention will become more apparent, the invention will be further described with reference to the following examples; it should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are merely for explaining the technical principles of the present invention, and are not intended to limit the scope of the present invention.
Furthermore, it should be noted that, in the description of the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be either fixedly connected, detachably connected, or integrally connected, for example; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention can be understood by those skilled in the art according to the specific circumstances.
Referring to fig. 1, a flowchart of an artificial intelligence-based medical document integrity analysis method according to an embodiment of the present invention includes:
step S1, a data acquisition unit acquires a plurality of medical document scanning images, a data processing unit screens the medical document scanning images with resolution meeting the standard as sample data and stores the sample data into a sample data set, and the sample data set randomly divides the sample data into a training set, a verification set and a test set;
the preferred training set, validation set and test set ratio in this embodiment is 7:2:1, a step of;
step S2, a modeling unit establishes a neural network model, sets the iteration times N and the hidden layers A of the neural network model, and inputs the training set into the neural network model for training;
respectively setting input, hiding and output layers of a neural network model, determining the number of neural units of the input layer by using pixels of an image, setting an initial value hidden layer number A and iteration number N, and determining the number of units of the hidden layer according to a formula
Determining, wherein I is the number of input neurons, O is the number of output neurons, a is a constant between 0 and 10, the number of output layer nerve units is the number of target categories, and the expected error is 0.001;
step S3, the modeling unit uses the verification set to verify the neural network model, and the central control unit judges whether to adjust the iteration times and hidden layers of the neural network model according to the verification result of the neural network model;
step S4, the modeling unit tests the neural network model with the test result conforming to the standard by using the test set, and the central control unit judges whether the number of the sample data needs to be increased according to the test result of the neural network model;
and S5, analyzing the integrity of the medical document by using the neural network model.
Specifically, in the step S1, the data processing unit detects the resolution Q of each of the medical document scanned images, compares Q with a preset resolution Q0, determines whether the resolution of the medical document scanned image meets a standard according to the comparison result,
if Q is more than or equal to Q0, the data processing unit judges that the resolution of the medical document scanning image corresponding to the resolution Q meets the standard, and the medical document scanning image is used as sample data to be stored in a sample data set;
if Q is smaller than Q0, the data processing unit judges that the resolution of the medical document scanning image corresponding to the resolution Q does not accord with the standard.
The embodiment of the invention does not specifically limit the resolution of the scanned image of the medical document, and can be set according to the number of sample data or actual needs, in this example, the preset resolution Q0 is set to be more than 100PPI, and preferably the preset resolution Q0 is set to be 150PPI.
According to the medical document scanning method and the medical document scanning device, the preset resolution is set to screen the medical document scanning images, and the medical document scanning images with the resolution meeting the standard are used as sample data, so that the identifiable degree of the sample data is improved.
The present embodiment provides another possibility of performing segmentation and feature extraction after preprocessing such as graying, binarization, sharpening enhancement, etc. on the image in the sample data set. Resolution may be used as a means of screening scanned images, and in another embodiment, the images may be preprocessed by image processing to conform the images to a standard.
Specifically, in the step S1, the plurality of medical document scanning images acquired by the data acquisition unit include a plurality of medical document scanning images of a first degree of integrity, a plurality of medical document scanning images of a second degree of integrity, a plurality of medical document scanning images of a third degree of integrity, and a plurality of medical document scanning images of a fourth degree of integrity.
In this example, it is preferable to set 90% or more of the integrity as the first integrity, 60% or more of the integrity as the second integrity, 60% or more of the integrity as the third integrity, and 30% or less of the integrity as the fourth integrity.
The medical document comprises paper documents such as outpatient medical records, inpatient medical records, outpatient prescriptions, doctor shift reports, various application forms, report forms, three measurement forms, nursing shift reports, special nursing record sheets and the like, and the judgment of the integrity of the scanned image of the medical document can be carried out by the following embodiment modes:
taking an outpatient prescription as an example, a structured template of the outpatient prescription is first created, which includes the following: 1, prescription basic information including clinic numbers, prescription numbers, dates, scientific categories, fees and pharmacies; 2. patient basic information including name, sex, age, clinical diagnosis, long-term medication; 3. medication information including medicine name, usage amount, number of prescribed medicines, and medicine cost; 4. signature information, including physicians, auditors, dispensing pharmacists, checking pharmacists, and dispensing pharmacists. The statistical structured template content item has 20 items in total, and if 10 items of the content are complete, the integrity is 50%. The structural templates of the clinic prescriptions can be adjusted according to actual needs. The embodiment is one of possible implementation manners for determining the integrity of the scanned image of the medical document, and may also be determined by other manners, and the embodiment is not particularly limited, and may be set in a related manner according to actual needs.
According to the invention, the data acquisition unit provides sample data with different integrality by acquiring a plurality of medical document scanning images with different integrality, so that the neural network model can analyze the integrality of the medical document, and the accuracy of analyzing the integrality of the medical document is improved by finely dividing the integrality of the sample data.
In the step S2, the modeling unit sets the number of iterations of the neural network model to N and the hidden layer number to a.
The embodiment of the invention does not limit the iteration times and hidden layers of the neural network model specifically, can be set according to actual requirements, in the example, the iteration times are set to be 1000 < N < 5000, the hidden layers are set to be 6 < A < 32,
the modeling unit learns sample data with different integrality by establishing the neural network model to realize intelligent analysis on the integrality of the medical document, and the invention improves the analysis efficiency of the integrality of the medical document by applying the artificial intelligence technology; if the iteration times and hidden layer numbers of the neural network model are set to be too large, the model is complex, operation data are huge, operation efficiency is low, even generalization capability of the model is poor, if the iteration times and hidden layer numbers of the neural network model are too large or too small, larger prediction errors of the model are caused, an overfitting phenomenon occurs, in the example, the iteration times are set to be 1000 < N < 5000, the hidden layer numbers are set to be 6 < A < 32, good prediction accuracy of the neural network model is guaranteed, and the overfitting phenomenon is avoided.
Specifically, in the step S3, the central control unit determines whether to tune the neural network model according to the test result of the neural network model,
when the modeling unit selects sample data A1 and A2. An in the verification set to verify the neural network model, the corresponding preset results are A1 and a2. An, the verification result of the neural network model is b1 and b2.. Bn, the central control unit corresponds the preset result to the numerical value in the verification result, calculates the verification error rate tau, and sets the verification error rate tau
The central control unit compares the verification error rate tau with a preset error rate tau 0,
if tau is less than or equal to tau 0, the central control unit judges that the verification error rate meets the standard, and the verification result of the neural network model meets the standard, so that parameter adjustment on the neural network model is not needed;
if τ > τ0, the central control unit determines that the verification error rate does not meet the standard, and the neural network model verification result does not meet the standard, and parameter adjustment is required to be performed on the neural network model.
In this example, the preset error rate τ0 is set to 10% or less, and the preset error rate τ0 is preferably set to 10%.
The invention verifies the trained neural network model by setting the verification set to judge whether the neural network model can be put into use, thereby improving the prediction precision of the neural network model, and limits the error rate to be within 10 percent, preferably to be within 5 percent, and further improving the prediction precision of the neural network model.
Specifically, in the step S4, when the central control unit determines that the verification result of the neural network model does not meet the standard, that is, τ > τ0, a difference Δτ between the verification error rate τ and the preset error rate τ0 is calculated, and the hidden layer number of the neural network model is adjusted according to Δτ, where Δτ=τ - τ0 is set,
if Δτ is smaller than Δτ1, the central control unit selects α1 to adjust the hidden layer number to a1=a×α1;
if Δτ1 is less than or equal to Δτ2 and Δτ2, the central control unit selects α2 to adjust the hidden layer number to a1=axα2;
if Deltaτ is more than or equal to Deltaτ2, the central control unit selects alpha 3 to adjust the hidden layer number to A1=A×alpha 3;
wherein Δτ1 is a first preset error rate difference, Δτ2 is a second preset error rate difference, Δτ1 is more than 5% and less than 10% and less than 15% of Δτ2, α1 is a first hidden layer number adjustment coefficient, α2 is a second hidden layer number adjustment coefficient, α3 is a third hidden layer number adjustment coefficient, α1, α2, α3 values are 0.8 and less than α1 and less than 1.2,0.6 and less than α2 and less than 1.5,0.4 and less than α3 and less than 2, and each of α1, α2, α3 is not 1, and each time the hidden layer number is updated, α1, α2, α3 is randomly valued within its corresponding value range, and when A1 is not a positive integer, the value of A1 is a maximum positive integer less than A1.
The hidden layer number can influence the prediction performance of the neural network model to a great extent, and an error becomes larger due to the fact that the hidden layer number is too large or too small. When the verification result of the neural network model is judged to be inconsistent with the standard, the central control module adjusts the hidden layer number so as to increase the complexity of the model and further increase the prediction precision of the model.
In particular to Amax in the central control unit, the central control unit compares the adjusted hidden layer number A1 with Amax,
if A1 is less than or equal to Amax, the central control unit sets the hidden layer number as A1;
if A1 is larger than Amax, the hidden layer number is set to Amax by the central control unit.
In this example, amax is set to 30 < 35, preferably, amax is set to 32.
The invention ensures that the complexity of the neural network model floats in a proper range by setting the maximum hidden layer number, avoids the occurrence of the overfitting phenomenon, and simultaneously ensures the operation rate of the neural network model by setting the maximum hidden layer number.
Specifically, the central control unit calculates the difference value delta A between the hidden layer number A1 and the maximum hidden layer number Amax when A1 is larger than Amax, adjusts the iteration times of the neural network model according to delta A, sets delta A=A1-Amax,
if Δa is less than Δa1, the central control unit selects β1 to adjust the iteration times to n1=n×β1;
if Δa1 is less than or equal to Δa < Δa2, the central control unit selects β2 to adjust the iteration times to n1=n×β2;
if Δa is greater than or equal to Δa2, the central control unit selects β3 to adjust the iteration times to n1=n×β3;
wherein Δa1 is a first preset hidden layer number difference value, Δa2 is a second preset hidden layer number difference value, 1 < Δa1 < 5 < Δa2 < 10, β1 is a first iteration number adjustment coefficient, β2 is a second iteration number adjustment coefficient, β3 is a third iteration number adjustment coefficient, values of β1, β2 and β3 are respectively 0.8 < β1 < 1.2,0.6 < β2 < 1.5,0.4 < β3, and values of β1, β2 and β3 are not 1, and when the iteration number is updated each time, β1, β2 and β3 are randomly valued in the corresponding value ranges, and when N1 is not a positive integer, the value of N1 is the largest positive integer smaller than N1.
When the hidden layer number of the neural network model is regulated to the maximum value by the central control module, the iteration times of the neural network model are regulated, so that the prediction accuracy of the model is improved.
Specifically, when the central control unit determines that the neural network model verification result meets the standard, the modeling unit tests the neural network model by using a test set, and the central control unit determines whether the number of sample data needs to be increased or not through the test result, wherein,
when the modeling unit selects sample data C1 and C2. Cm in the test set to test the neural network model, the corresponding preset results are C1 and c2. Cm, the test results of the neural network model are d1 and d2 dm, and the central control unit corresponds the preset results with the numerical values in the test results and calculates a test error rate phi and sets
The central control unit compares the test error rate phi with a preset error rate tau 0,
if phi is less than or equal to tau 0, the central control unit judges that the test error rate meets the standard, the test result of the neural network model meets the standard, and the number of the sample data is not required to be increased;
if phi is more than tau 0, the central control unit judges that the test error rate does not meet the standard, the test result of the neural network model does not meet the standard, and the number of the sample data is required to be increased.
According to the invention, the neural network model with the test result meeting the standard is tested by setting the test set so as to judge whether the neural network model can be put into use or not, and the prediction precision of the neural network model is further improved.
Specifically, the central control unit calculates a difference delta phi between a test error rate phi and a preset error rate tau 0 when the test result of the neural network model is not in accordance with the standard, adjusts the number of sample data according to delta phi, sets delta phi = phi-tau 0,
if delta phi is more than or equal to delta phi 2, the central control unit adjusts the number of the sample data to N1, and N1 = 5 x N0 is set;
if Δφ1 is less than or equal to Δφ < Δφ2, the central control unit adjusts the number of sample data to N1, and N1 = 3 x N0 is set;
if Δφ < Δφ1, the central control unit adjusts the number of sample data to N1, and sets N1=2×N0;
wherein, delta phi 1 is the first preset test error rate difference, delta phi 2 is the second preset test error rate difference, delta phi 1 is more than 3 percent and less than 6 percent and delta phi 2 is less than 10 percent.
When the neural network model test result is judged to be inconsistent with the standard, the central control module adjusts the number of sample data to increase the prediction precision of the model.
Referring to fig. 2, which is a block diagram of an artificial intelligence-based medical document integrity analysis system according to an embodiment of the present invention, the artificial intelligence-based medical document integrity analysis system includes:
the data acquisition unit is used for acquiring a plurality of medical document scanning images;
the data processing unit is connected with the data acquisition unit and is used for screening the medical document scanning image with the resolution meeting the standard as sample data, storing the sample data into a sample data set and randomly dividing the sample data into a training set, a verification set and a test set;
the modeling unit is connected with the data processing unit and used for establishing a neural network model and training, testing and verifying the neural network model by using sample data;
the central control unit is connected with the modeling unit and used for judging whether the iteration times and hidden layers of the neural network model are adjusted according to the test result of the neural network model, judging whether the number of sample data is required to be increased according to the verification result of the neural network model, and analyzing the integrity of the medical document by using the trained neural network model.
Thus far, the technical solution of the present invention has been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of protection of the present invention is not limited to these specific embodiments. Equivalent modifications and substitutions for related technical features may be made by those skilled in the art without departing from the principles of the present invention, and such modifications and substitutions will be within the scope of the present invention.
The foregoing description is only of the preferred embodiments of the invention and is not intended to limit the invention; various modifications and variations of the present invention will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (5)
1. An artificial intelligence based medical document integrity analysis method, comprising:
step S1, a data acquisition unit acquires a plurality of medical document scanning images, a data processing unit screens the medical document scanning images with resolution meeting the standard as sample data and stores the sample data into a sample data set, and the sample data set randomly divides the sample data into a training set, a verification set and a test set;
step S2, a modeling unit establishes a neural network model, sets the iteration times N and the hidden layers A of the neural network model, and inputs the training set into the neural network model for training;
step S3, the modeling unit uses the verification set to verify the neural network model, and the central control unit judges whether to adjust the iteration times and hidden layers of the neural network model according to the verification result of the neural network model;
step S4, the modeling unit tests the neural network model with the test result conforming to the standard by using the test set, and the central control unit judges whether the number of the sample data needs to be increased according to the test result of the neural network model;
s5, analyzing the integrity of the medical document by using the neural network model;
in the step S1, the plurality of medical document scanning images acquired by the data acquisition unit include a plurality of medical document scanning images of a first degree of integrity, a plurality of medical document scanning images of a second degree of integrity, a plurality of medical document scanning images of a third degree of integrity, and a plurality of medical document scanning images of a fourth degree of integrity;
the medical document comprises an outpatient medical record, an inpatient medical record, an outpatient prescription, a doctor shift report, various application forms, a report form, a three-measurement form, a nursing shift report and a special nursing record form, and the integrity of a scanned image of the medical document is judged by the following modes:
for the determination of the integrity of an outpatient prescription, a structured template of the outpatient prescription is first created, which contains the following: 1. prescription basic information including clinic number, prescription number, date, science, cost, pharmacy; 2. patient basic information including name, sex, age, clinical diagnosis, long-term medication; 3. medication information including medicine name, usage amount, number of prescribed medicines, and medicine cost; 4. signing information, including doctors, auditors, blending pharmacists, checking pharmacists and dispensing pharmacists; counting 20 structured template content items, wherein if 10 content items are complete, the integrity is 50%;
in the step S3, when the central control unit determines whether to tune the neural network model according to the test result of the neural network model,
when the modeling unit selects sample data A1 and A2. An in the verification set to verify the neural network model, the corresponding preset results are A1 and a2. An, the verification result of the neural network model is b1 and b2.. Bn, the central control unit corresponds the preset result to the numerical value in the verification result, calculates the verification error rate tau, and sets the verification error rate tau
The central control unit compares the verification error rate tau with a preset error rate tau 0,
if tau is less than or equal to tau 0, the central control unit judges that the verification error rate meets the standard, and the verification result of the neural network model meets the standard, so that parameter adjustment on the neural network model is not needed;
if τ > τ0, the central control unit determines that the verification error rate does not meet the standard, and the neural network model verification result does not meet the standard, so that parameter adjustment is required to be performed on the neural network model;
in the step S4, when the central control unit determines that the verification result of the neural network model does not meet the standard, that is, τ > τ0, a difference Δτ between the verification error rate τ and the preset error rate τ0 is calculated, the hidden layer number of the neural network model is adjusted according to Δτ, Δτ=τ - τ0 is set,
if Δτ is smaller than Δτ1, the central control unit selects α1 to adjust the hidden layer number to a1=a×α1;
if Δτ1 is less than or equal to Δτ2 and Δτ2, the central control unit selects α2 to adjust the hidden layer number to a1=axα2; if Deltaτ is more than or equal to Deltaτ2, the central control unit selects alpha 3 to adjust the hidden layer number to A1=A×alpha 3;
wherein Δτ1 is a first preset verification error rate difference value, Δτ2 is a second preset verification error rate difference value, Δτ1 is less than Δτ2, α1 is a first hidden layer number adjustment coefficient, α2 is a second hidden layer number adjustment coefficient, α3 is a third hidden layer number adjustment coefficient, values of α1, α2 and α3 are respectively more than 0.8 and less than 1.2,0.6 and less than α2 and less than 1.5,0.4 and less than α3 and are not 1, each time α1, α2 and α3 are updated, the values of α1, α2 and α3 are randomly selected within the corresponding value ranges, and when A1 is not a positive integer, the value of A1 is the largest positive integer smaller than A1;
the central control unit is provided with a maximum hidden layer number Amax, compares the adjusted hidden layer number A1 with Amax,
if A1 is less than or equal to Amax, the central control unit sets the hidden layer number as A1;
if A1 is larger than Amax, the hidden layer number is set as Amax by the central control unit;
when A1 is larger than Amax, the central control unit calculates the difference value delta A between the hidden layer number A1 and the maximum hidden layer number Amax, adjusts the iteration times of the neural network model according to delta A, sets delta A=A1-Amax,
if Δa is less than Δa1, the central control unit selects β1 to adjust the iteration times to n1=n×β1;
if Δa1 is less than or equal to Δa < Δa2, the central control unit selects β2 to adjust the iteration times to n1=n×β2;
if Δa is greater than or equal to Δa2, the central control unit selects β3 to adjust the iteration times to n1=n×β3;
wherein Δa1 is a first preset hidden layer number difference value, Δa2 is a second preset hidden layer number difference value, Δa1 is less than Δa2, β1 is a first iteration number adjustment coefficient, β2 is a second iteration number adjustment coefficient, β3 is a third iteration number adjustment coefficient, values of β1, β2 and β3 are respectively more than 0.8 and less than β1 and less than 1.2,0.6 and less than β2 and less than 1.5,0.4 and less than β3 and are not 1, each time when the iteration number is updated, β1, β2 and β3 are randomly valued in a corresponding value range, and when N1 is not a positive integer, the value of N1 is the largest positive integer smaller than N1.
2. The method for analyzing the integrity of medical documents based on artificial intelligence according to claim 1, wherein in the step S1, the data processing unit detects the resolution Q of each scanned image of the medical document, compares Q with a preset resolution Q0 and determines whether the resolution of the scanned image of the medical document meets the standard according to the comparison result,
if Q is more than or equal to Q0, the data processing unit judges that the resolution of the medical document scanning image corresponding to the resolution Q meets the standard, and the medical document scanning image is used as sample data to be stored in a sample data set;
if Q is smaller than Q0, the data processing unit judges that the resolution of the medical document scanning image corresponding to the resolution Q does not accord with the standard.
3. The method for analyzing the integrity of medical documents based on artificial intelligence according to claim 2, wherein when the central control unit determines that the verification result of the neural network model meets the standard, the modeling unit tests the neural network model by using a test set, and the central control unit determines whether the number of the sample data needs to be increased according to the test result, wherein,
when the modeling unit selects sample data C1 and C2. Cm in the test set to test the neural network model, the corresponding preset results are C1 and c2. Cm, the test results of the neural network model are d1 and d2 dm, and the central control unit corresponds the preset results with the numerical values in the test results and calculates a test error rate phi and sets
The central control unit compares the test error rate phi with a preset error rate tau 0,
if phi is less than or equal to tau 0, the central control unit judges that the test error rate meets the standard, the test result of the neural network model meets the standard, and the number of the sample data is not required to be increased;
if phi is more than tau 0, the central control unit judges that the test error rate does not meet the standard, the test result of the neural network model does not meet the standard, and the number of the sample data is required to be increased.
4. The method for analyzing the integrity of medical documents based on artificial intelligence according to claim 3, wherein the central control unit calculates a difference Δφ between a test error rate φ and a preset error rate τ0 when the test result of the neural network model is determined to be out of standards and adjusts the number of sample data according to Δφ, wherein Δφ=φ - τ0 is set,
if delta phi is more than or equal to delta phi 2, the central control unit adjusts the number of the sample data to N1, and N1 = 5 x N0 is set;
if Δφ1 is less than or equal to Δφ < Δφ2, the central control unit adjusts the number of sample data to N1, and N1 = 3 x N0 is set;
if Δφ < Δφ1, the central control unit adjusts the number of sample data to N1, and sets N1=2×N0;
wherein, delta phi 1 is the first preset test error rate difference, delta phi 2 is the second preset test error rate difference, and delta phi 1 is less than delta phi 2.
5. A system for applying the method of any one of claims 1-4, comprising:
the data acquisition unit is used for acquiring a plurality of medical document scanning images;
the data processing unit is connected with the data acquisition unit and is used for screening the medical document scanning image with the resolution meeting the standard as sample data, storing the sample data into a sample data set and randomly dividing the sample data into a training set, a verification set and a test set;
the modeling unit is connected with the data processing unit and used for establishing a neural network model and training, testing and verifying the neural network model by using sample data;
the central control unit is connected with the modeling unit and used for judging whether the iteration times and hidden layers of the neural network model are adjusted according to the test result of the neural network model, judging whether the number of sample data is required to be increased according to the verification result of the neural network model, and analyzing the integrity of the medical document by using the trained neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310284541.0A CN115984889B (en) | 2023-03-22 | 2023-03-22 | Medical document integrity analysis method and system based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310284541.0A CN115984889B (en) | 2023-03-22 | 2023-03-22 | Medical document integrity analysis method and system based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115984889A CN115984889A (en) | 2023-04-18 |
| CN115984889B true CN115984889B (en) | 2023-06-09 |
Family
ID=85963505
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310284541.0A Active CN115984889B (en) | 2023-03-22 | 2023-03-22 | Medical document integrity analysis method and system based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115984889B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116844733B (en) * | 2023-08-31 | 2023-11-07 | 吉林大学第一医院 | Medical data integrity analysis method based on artificial intelligence |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978947B (en) * | 2015-07-17 | 2018-06-05 | 京东方科技集团股份有限公司 | Adjusting method, dispaly state regulating device and the display device of dispaly state |
| CN109255160B (en) * | 2018-08-17 | 2020-10-16 | 东南大学 | Neural network-based unit delay prediction method and unit delay sensitivity calculation method |
| CN109426813B (en) * | 2018-11-02 | 2022-06-24 | 中电科新型智慧城市研究院有限公司 | Remote sensing image user-defined interest point extraction method based on fuzzy clustering and neural network model |
| CN110533109A (en) * | 2019-09-03 | 2019-12-03 | 内蒙古大学 | A kind of storage spraying production monitoring data and characteristic analysis method and its device |
| CN113792373B (en) * | 2021-11-17 | 2022-02-22 | 中化学建设投资集团北京科贸有限公司 | Personnel behavior monitoring emergency disposal method based on machine vision |
-
2023
- 2023-03-22 CN CN202310284541.0A patent/CN115984889B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115984889A (en) | 2023-04-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chaudhry et al. | Systematic review: impact of health information technology on quality, efficiency, and costs of medical care | |
| Stange | How does provider supply and regulation influence health care markets? Evidence from nurse practitioners and physician assistants | |
| Zou et al. | Statistical evaluation of diagnostic performance: topics in ROC analysis | |
| Kuo et al. | Association of hospitalist care with medical utilization after discharge: evidence of cost shift from a cohort study | |
| CA2456296C (en) | Biometric quality control process | |
| US9251809B2 (en) | Method and apparatus of speech analysis for real-time measurement of stress, fatigue, and uncertainty | |
| Shoukri | Analysis of Correlated Data with SAS and R | |
| Mehrotra et al. | The effect of different attribution rules on individual physician cost profiles | |
| US20110276346A1 (en) | Automated method for medical quality assurance | |
| US20230071400A1 (en) | System and method for assessing medical images | |
| US10269447B2 (en) | Algorithm, data pipeline, and method to detect inaccuracies in comorbidity documentation | |
| US20210098133A1 (en) | Secure Scalable Real-Time Machine Learning Platform for Healthcare | |
| CN115984889B (en) | Medical document integrity analysis method and system based on artificial intelligence | |
| Zhu et al. | A Bayesian approach to measurement bias in networking studies | |
| Kondylakis et al. | Status and recommendations of technological and data-driven innovations in cancer care: Focus group study | |
| Raheja et al. | Data analysis and its importance in health care | |
| Sylolypavan et al. | The impact of inconsistent human annotations on AI driven clinical decision making | |
| Brennan et al. | Patient acuity related to clinical research: Concept clarification and literature review | |
| Mishra et al. | Usage and analysis of big data in E-health domain | |
| Etzioni et al. | Statistics for health data science | |
| US20230274834A1 (en) | Model-based evaluation of assessment questions, assessment answers, and patient data to detect conditions | |
| Rani et al. | The Potential Application of Artificial Intelligence in Healthcare and Hospitals | |
| Stead | Electronic health records | |
| US20190326016A1 (en) | Method of Capturing and Evaluation Uncertainty in Computerized Intelligent Systems for Medical Diagnosis | |
| Courtemanche et al. | Producing comparable cost and quality results from All-Payer Claims Databases |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |