A medical data structuring processing method and system

Abstract

The application discloses a medical data structured processing method, comprising the following steps: obtaining voice record data of a doctor's medical record input; eliminating interference sound of the voice record data and saving the medical record voice data; obtaining the medical record voice data input by the doctor, and converting the medical record voice data into corresponding medical record text information; constructing a medical record voice conversion error checking model, and checking errors of the medical record voice conversion based on the converted medical record text information; constructing a medical record knowledge base, matching the medical record text information with a normal evaluation result with the medical record knowledge base, and outputting matched medical term medical record input information based on the medical record knowledge base. The evaluation index of the evaluation result based on the medical record voice conversion error checking model can further feed back the accuracy of the medical record voice conversion, directly express the medical record information actually converted by the doctor, and improve work efficiency.

Description

Technical Field

[0001] This invention relates to the field of medical information processing technology, specifically to a method and system for structuring medical data. Background Technology

[0002] Traditional medical record keeping methods rely on doctors writing or typing, which consumes a lot of time and effort. At the same time, handwriting or keyboard input is prone to input errors, affecting the accuracy of medical record information.

[0003] Against the backdrop of rapid development in medical information technology, intelligent speech-to-text technology is gradually attracting attention and being applied in the medical field. This technology greatly simplifies the medical record-keeping process and improves the accuracy and usability of medical information by converting doctors' dictation into text.

[0004] In existing technologies, speech-to-text engines are typically used to directly transcribe doctors' spoken information into text. However, when dealing with specialized vocabulary, intelligent error recognition during speech-to-text transcription is lacking, leading to issues such as missing words and sentences. Furthermore, while intelligent transcription improves recording efficiency, the resulting text records are conversational and cannot be compiled into systematic, modular records, which is detrimental to hospital archiving and research.

[0005] Therefore, there is an urgent need to design a medical data structure processing method and system to solve the above-mentioned technical problems. Summary of the Invention

[0006] To address the shortcomings of the existing technology, the present invention aims to overcome these shortcomings and provide a method for structuring medical data, which includes the following steps:

[0007] Acquire the voice recording data of the doctor entering medical records; remove interference from the voice recording data and save the medical record voice data;

[0008] Acquire the voice data of the medical record entered by the doctor, and transcribe the voice data of the medical record into the corresponding text information of the medical record;

[0009] Construct a medical record speech-to-text error detection model to perform speech-to-text error detection based on the transcribed medical record text information:

[0010] If the evaluation result based on the medical record speech-to-text error detection model is normal, the corresponding transcribed medical record text information is output.

[0011] If the evaluation result based on the medical record speech-to-text error detection model is abnormal, the corresponding medical record speech data will be re-transcribed into medical record text information.

[0012] Construct a medical record knowledge base, match the textual information of medical records that are assessed as normal with the medical record knowledge base, and output the matched medical terminology medical record entry information based on the medical record knowledge base.

[0013] As a further optimization of the above solution, the method for removing interference from the voice recording data includes the following:

[0014] A voice feature model for doctors entering medical records is constructed, wherein the influence factor of the model is the medical record entry time and the single-segment impulse ratio corresponding to any unit of time;

[0015] Mark any starting point of medical record entry voice acquisition and record it as a marker point. Based on the marker point, record the voice acquisition frequency. The number of marker points is obtained until any of the recorded medical record voice recordings ends. Then the time for entering medical records for:

[0016] ;

[0017] Mark the total impulse of any medical record recording voice. Based on the marker points, frequencies are obtained according to speech. The cumulative equal divisions are Segment, any equal division The impulse of a single segment of medical record voice input is Any equal division The single-segment impulse ratio of the single-segment medical record recording voice is The single-segment impulse ratio is calculated as follows:

[0018] .

[0019] As a further optimization of the above solution, the method also includes:

[0020] Based on the constructed medical record entry time and single-stage impulse ratio Detect all collected and recorded audio:

[0021] If the detected single-segment impulse ratio first rises to the maximum value and then falls, retain the corresponding recorded sound; otherwise, delete all collected recorded sounds and repeat the above steps.

[0022] Based on the recorded audio that first rises to its maximum value and then falls, the audio corresponding to the maximum value of the medical record entry time is retained as the target audio.

[0023] As a further optimization of the above scheme, the method for error detection in medical record speech-to-text transcription based on transcribed medical record text information specifically includes the following:

[0024] Construct an arbitrary distribution model of medical record text information. As shown below:

[0025] ;

[0026] in, The medical record text information distribution model The computational object is represented as the first Medical record entry terms The distribution of; among which Characterized by the current medical record entry term The former Medical record entry terms;

[0027] Constructing a fault-tolerant model for arbitrary medical record text information ,

[0028] ;

[0029] in, This is represented by the number of times the corresponding medical record entry term appears in the medical record knowledge base;

[0030] Based on the constructed medical record text information distribution model Fault tolerance model for medical record text information Calculate the probability of accurate prediction of the next word in the medical record text information for any entered word: where, For regularization constants, It is the stability coefficient. This indicates any combination of two medical record entry terms;

[0031] ;

[0032] As a further optimization of the above scheme, the method for error detection in medical record speech-to-text transcription based on transcribed medical record text information also includes the following:

[0033] The evaluation index based on the assessment results of the aforementioned medical record speech-to-text error detection model is: Then we have:

[0034] ;

[0035] in, The representation is a description of the current entry probability for any medical record entry term. To describe the weights.

[0036] As a further optimization of the above scheme, the method for matching the medical record text information with the evaluation result being normal with the medical record knowledge base specifically includes the following:

[0037] Generate arbitrary medical record text information that has been checked for errors through speech-to-text transcription, and mark all recognized and detected medical record entries as... Then the set of case text information and medical record entry terms is represented as: ;

[0038] Based on the generated set Construct a deep neural-guided convolution model :

[0039] ;

[0040] Constructing a mapping association between case text information and medical terminology The mapping association is constructed as follows:

[0041]

[0042] ;

[0043] Among them, mapping association As the jump degree First jump, jump range As the correspondence of the second jump, mapping association For jump degree As the first jump, jump degree As a correspondence for the second jump, express Dynamic associations.

[0044] As a further optimization of the above scheme, after completing the mapping association... After construction, the method further includes the following:

[0045] Get mapping association Associating the mapping Input to the medical terminology conversion set:

[0046] ;

[0047] in, yes Type function, It is an input parameter matrix based on medical record input terms. It is a mapping association and The result of the dot product, It is a mapping-related offset vector.

[0048] As a further optimization of the above scheme, the constructed medical record voice transcription error detection model includes the starting point for medical record text information input, relationship conversion, and input endpoint.

[0049] This invention also discloses a medical data structuring processing method system, the system comprising the following:

[0050] The voice filtering module is used to acquire voice recording data of doctors entering medical records; remove interference sounds from the voice recording data, and save the medical record voice data;

[0051] The speech-to-text module is used to acquire medical record speech data entered by doctors and transcribe the medical record speech data into corresponding medical record text information;

[0052] The voice verification module is used to build a medical record voice-to-text error-checking model, and performs error checking on the transcribed medical record text information:

[0053] If the evaluation result based on the medical record speech-to-text error detection model is normal, the corresponding transcribed medical record text information is output.

[0054] If the evaluation result based on the medical record speech-to-text error detection model is abnormal, the corresponding medical record speech data will be re-transcribed into medical record text information.

[0055] The transcription and redirection module is used to build a medical record knowledge base. It matches the medical record text information that is evaluated as normal with the medical record knowledge base and outputs the matched medical terminology medical record entry information based on the medical record knowledge base.

[0056] As a further optimization of the above solution, the voice filtering module includes the following:

[0057] The model building unit is used to build a voice feature model for doctors to enter medical records. The influence factor of the model is the single-segment impulse ratio corresponding to the medical record entry time and any unit of time.

[0058] The voice tagging unit is used to mark the start point of any medical record entry voice acquisition and record the mark point, and then record the voice acquisition frequency based on the mark point. The number of marker points is obtained until any of the recorded medical record voice recordings ends. Then the time for entering medical records for:

[0059] ;

[0060] The impulse calculation unit is used to mark the total impulse of any medical record recording voice. Based on the marker points, frequencies are obtained according to speech. The cumulative equal divisions are Segment, any equal division The impulse of a single segment of medical record voice input is Any equal division The single-segment impulse ratio of the single-segment medical record recording voice is The single-segment impulse ratio is calculated as follows:

[0061] .

[0062] Compared with the prior art, the present invention, employing the above-described technical solution, provides a method and system for structuring medical data, which has the following technical advantages:

[0063] 1. This invention designs a medical data structured processing method, specifically by constructing a speech feature model for medical record entry, calculating the single-segment impulse ratio of any medical record entry time and the total impulse of the medical record entry speech, and quickly and accurately filtering out the required medical record entry speech data information based on the line graph of the single-segment impulse ratio.

[0064] 2. This invention proposes a method for calculating the accuracy probability of the next predicted word for any medical record entry word in medical record text information. When there is only one entry word in the medical record text information, the accuracy probability of the next predicted word for any medical record entry word in the medical record text information is the result calculated by any fault-tolerant model of medical record text information. When there are more than one entry word in the medical record text information, the calculation method based on this invention can quickly obtain the corresponding occurrence probability. This invention is designed to solve the problem of doctors recording patients' conditions through speech-to-text medical record audio. It specifically designs a fault-tolerant mechanism for medical record speech-to-text, that is, by converting medical record entry words from speech to text, and predicting the next medical record entry word based on any already transcribed medical record entry word, a medical record speech-to-text data information with high recognition and more in line with the doctor's actual speech is formed.

[0065] 3. The evaluation index designed in this invention, based on the evaluation results of the medical record speech-to-text error-checking model, can further reflect the accuracy of medical record speech-to-text transcription, intuitively express the actual medical record information that doctors need to transcribe, and improve work efficiency.

[0066] 4. This invention creates a medical data structured processing method, specifically focusing on the actual symptoms and medical records. Based on the established medical record knowledge base, it can quickly help doctors match and associate the attributes and medical terms of the actual symptoms, enabling doctors to quickly and accurately match the actual situation during the transcription of medical records, with good results. Attached Figure Description

[0067] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0068] Figure 1This is a schematic diagram of the process of the present invention;

[0069] Figure 2 This is a schematic diagram of one embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of the transcription process of the present invention. Detailed Implementation

[0071] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0072] like Figure 1-2 As shown in the figure, an embodiment of the present invention discloses a method for structuring medical data, which includes the following:

[0073] Acquire the voice recording data of doctors entering medical records; remove interference from the voice recording data and save the medical record voice data;

[0074] Acquire the voice data of medical records entered by doctors and transcribe the voice data into the corresponding text information of medical records;

[0075] Construct a medical record speech-to-text error detection model to perform speech-to-text error detection based on the transcribed medical record text information:

[0076] If the evaluation result based on the medical record speech-to-text error detection model is normal, the corresponding transcribed medical record text information will be output.

[0077] If the evaluation result based on the medical record voice transcription error detection model is abnormal, the corresponding medical record voice data will be re-transcribed into medical record text information.

[0078] Construct a medical record knowledge base, match the textual information of medical records that are evaluated as normal with the medical record knowledge base, and output the matched medical terminology medical record entry information based on the medical record knowledge base.

[0079] More specifically, this invention outputs medical terminology medical record information transcribed from the voice data of the medical record entered by the doctor after matching with the medical record knowledge base, and transmits it to an interactive electronic touch device. The doctor can match the required medical terminology by clicking or manually modifying it. Since this operation is a conventional technology, it will not be described in detail here. Those skilled in the art can quickly understand and master this operation based on this technical solution.

[0080] Specifically, methods for removing interference from voice recording data include the following:

[0081] A speech feature model for doctors entering medical records is constructed. The influencing factors of this model are the medical record entry time and the single-segment impulse ratio corresponding to any unit of time.

[0082] Mark any starting point of the medical record entry voice acquisition and record it as a marker point. Based on the marker point, record the voice acquisition frequency. The number of marker points is obtained until any medical record recording voice ends. Then the time for entering medical records for:

[0083] ;

[0084] Mark the total impulse of any medical record recording voice. Frequency acquisition based on marker points The cumulative equal divisions are Segment, any equal division The impulse of a single segment of medical record voice input is Any equal division The single-segment impulse ratio of the single-segment medical record recording voice is The single-segment impulse ratio is calculated as follows:

[0085] .

[0086] Specifically, the above methods also include:

[0087] Based on the constructed medical record entry time and single-stage impulse ratio Detect all collected and recorded audio:

[0088] If the detected single-segment impulse ratio first rises to the maximum value and then falls, retain the corresponding recorded sound; otherwise, delete all collected recorded sounds and repeat the above steps.

[0089] Based on the recorded audio that first rises to its maximum value and then falls, the audio corresponding to the maximum value of the medical record entry time is retained as the target audio.

[0090] More specifically, this invention provides an embodiment in which hospital doctors summarize patient symptoms to form medical record information and complete the automated data entry of the system. During this process, the doctor's voice input based on the patient's summarized medical record is stable, meaning that the single-segment impulse ratio has a concentrated time period, and its intensity decays by first rising and then falling. Based on the single-segment impulse ratio calculation method and the line graph of the establishment time versus the single-segment impulse ratio, sounds that conform to the concentrated time period and the intensity decay following a rise-then-fall pattern can be quickly extracted from other interfering background noises, such as those from hospitals.

[0091] It should be noted that during the patient's description of their symptoms, there are localized vocal characteristics that also exhibit a concentrated time period and a trend of initial increase followed by decrease in intensity. Therefore, in order to more accurately isolate the patient's voice, it is necessary to further determine the medical record entry time and obtain the maximum value of the medical record entry time. The corresponding voice is the voice data information of the doctor continuously entering the patient's medical record.

[0092] This invention designs a structured processing method for medical data. Specifically, it constructs a speech feature model for medical record entry, calculates the single-segment impulse ratio based on the total impulse of any medical record entry time and the speech of the medical record entry, and quickly and accurately filters out the required medical record entry speech data information based on the line graph of the single-segment impulse ratio.

[0093] Specifically, the method for error checking in medical record speech-to-text transcription based on transcribed medical record text information includes the following:

[0094] Construct an arbitrary distribution model of medical record text information. As shown below:

[0095] ;

[0096] in, For the distribution model of medical record text information The computational object, representing the first Medical record entry terms The distribution of; among which Characterized by the current medical record entry term The former Medical record entry terms;

[0097] More specifically, the present invention This represents the number of terms entered into the medical record. A larger number of terms indicates a larger distribution model of textual information in the medical record. The likelihood of using it to describe the speech-to-text output of medical records decreases. Furthermore, the number of terms entered into the medical record increases, the number of parameters it contains increases, and the distribution model of medical record text information... The computational complexity increases.

[0098] Constructing a fault-tolerant model for arbitrary medical record text information ,

[0099] ;

[0100] in, This represents the number of times the corresponding medical record entry term appears in the medical record knowledge base;

[0101] More specifically, this invention also provides a fault-tolerant model for arbitrary medical record text information. That is, when the medical record entry information reaches a large scale, it is based on an arbitrary medical record text information distribution model. The calculated results are close to the actual situation; however, due to the increased complexity, there may be medical record entries that have not been recorded before, that is, the theoretical probability of the occurrence of such medical record entries is 0.

[0102] Based on the above description of the generation of unrecorded medical record entry terms, i.e., the actual occurrence of any medical record entry term is not zero, this invention specifically designs a fault-tolerant model for arbitrary medical record text information. The occurrence of any medical record entry word in the speech transcription is designed as a probability.

[0103] Specifically, based on the constructed medical record text information distribution model Fault tolerance model for medical record text information Calculate the probability of accurate prediction of the next word in the medical record text information for any entered word: where, It is a regular constant, and 0 < <1; It is the stability coefficient, which ensures that the total probability of transcription prediction for this case entry remains at 1; This indicates any combination of two medical record entry terms; it should be noted that... It is aimed at For prefix, The number of times all possible word combinations appear, It is aimed at Prefix, contain The number of possible word combinations that appear. It is aimed at The number of times possible compound words with the prefix appear; In order to be in All different combinations of terms under the condition of the preceding prefix. In order to be in Conditions containing as antecedent prefixes Different combinations;

[0104]

[0105] More specifically, this invention proposes a method for calculating the accuracy probability of predicting the next term of any medical record entry word in medical record text information. When there is only one entry word in the medical record text information, the accuracy probability of predicting the next term of any medical record entry word in the medical record text information is the fault-tolerant model for any medical record text information. Calculation results: When there are more than one medical record entry term, the calculation method designed based on this invention can quickly obtain the corresponding occurrence probability. This invention is designed to solve the problem of doctors recording patients' conditions through speech-to-text medical records. It specifically designs a fault-tolerant mechanism for medical record speech-to-text, that is, by converting medical record entry terms from speech to text, and predicting the next medical record entry term based on any already transcribed medical record entry term, it forms medical record speech-to-text data information with high recognition accuracy and more closely matches the doctor's actual speech.

[0106] Specifically, methods for error checking in medical record speech-to-text transcription based on transcribed medical record text information also include the following:

[0107] The evaluation index based on the assessment results of the aforementioned medical record speech-to-text error detection model is: Then we have:

[0108] ;

[0109] in, The representation is a description of the current entry probability for any medical record entry term. To describe the weights, when the probability descriptor value is greater than 0.26, Takes the value 1, otherwise The value is 0.

[0110] More specifically, to more concretely represent the error-checking situation of medical record speech-to-text transcription, this invention, based on the above, specifically designed an evaluation index based on the assessment results of the medical record speech-to-text transcription error-checking model. That is, if the evaluation index of any medical record speech-to-text transcription is greater than 0.65, it indicates that the corresponding medical record speech-to-text transcription is valid and matches the actual description; otherwise, it indicates that the corresponding medical record speech-to-text transcription is invalid, containing transcription and recognition errors, and the corresponding sentence is re-transcribed. The evaluation index designed in this invention, based on the assessment results of the medical record speech-to-text transcription error-checking model, can further reflect the accuracy of medical record speech-to-text transcription, intuitively representing the actual medical record information that doctors need to transcribe, thus improving work efficiency.

[0111] Specifically, the method for matching medical record text information with normal evaluation results with the medical record knowledge base includes the following:

[0112] Generate arbitrary medical record text information that has been checked for errors through speech-to-text transcription, and mark all recognized and detected medical record entries as... Then the set of case text information and medical record entry terms is represented as: ;

[0113] Based on the generated set Construct a deep neural-guided convolution model :

[0114] ;

[0115] Constructing a mapping association between case text information and medical terminology The mapping association is constructed as follows:

[0116] ;

[0117] ;

[0118] Among them, mapping association For jump degree As the first jump, jump degree As the correspondence of the second jump, mapping association For jump degree As the first jump, jump degree As a correspondence for the second jump, Represented as Dynamic association;

[0119] Get mapping association , associate mapping Input to the medical terminology conversion set:

[0120] ;

[0121] in, yes Type function, It is an input parameter matrix based on medical record input terms. It is a mapping association and The result of the dot product, It is a mapping-related offset vector.

[0122] Specifically, the constructed model includes the starting point for entering medical record text information, relationship transformation, and the ending point for entry.

[0123] Specifically, the technical solution disclosed in this invention supports the conversion of transcribed medical terminology medical record entry information into a structured medical record display method, and the specific method is as follows:

[0124] A medical record knowledge graph is established within the medical record knowledge base. Nodes include medical record structure nodes (chief complaint, present illness, past medical history, allergy history, medication history, physical examination, tests, laboratory tests, etc.) and medical terminology nodes. The connection between medical record structure nodes and medical terminology nodes represents the probability that the medical terminology node belongs to that node (a higher probability indicates a higher probability of belonging to that node). The connection between medical terminology nodes represents the relationship between terminology ontologies. Commonly used attributes are defined for medical terminology nodes.

[0125] The above conversion process is shown below:

[0126] 1. Extracted medical terms are matched with terminology nodes in the medical record knowledge graph by name, ignoring terms that cannot be matched;

[0127] 2. The terminology node retrieves all synonyms based on the synonym relationships between nodes, and the medical terms with the highest probability of association with the structure node are those used in structured medical records;

[0128] 3. Obtain the structural nodes associated with medical terms and merge them to form a medical record structural node set; merge term nodes into structural nodes; merge terms that cannot be matched with the previous input term.

[0129] 4. Reorganize medical terms according to the medical record structure node set for interface display. Each medical record structure node constitutes a sub-area of ​​the interface, and each sub-area displays the set of medical terms under that structure. The corresponding transcribed medical terms are in the selected state. Other medical terms are in the unselected state, arranged in descending order of their similarity probability with the selected terms and their association probability with the medical record structure nodes.

[0130] More specifically, the present invention also provides a specific embodiment that creates a medical data structured processing method, specifically focusing on stroke medical records. Based on the established stroke medical record knowledge base, it can quickly help doctors match and associate stroke-related symptoms, contraindications, physical examinations, tests, and other medical terms. This enables doctors to quickly and accurately match the actual situation during the process of transcribing medical records, with good results.

[0131] More specifically, the present invention also provides the following embodiments: When doctors are transcribing medical records, they use core keywords related to stroke to eliminate interference and retain only medical terms that match stroke. Based on a deep neural-guided convolutional model, mapping association, and a medical terminology conversion set, non-standardized medical dialogues are transcribed into standardized medical terms to form medical terminology medical record information, as shown below:

[0132] Input 1: Stroke patient with left lower limb weakness for 3 days, has a history of cerebral infarction, is conscious, and has a heart rate of 71 bpm.

[0133] Transcription 1: [(Left lower limb, weakness, 3 days) (Past medical history, cerebral infarction), (Heart rate, 71), (Conscious)];

[0134] Structure 1: [(Present illness: weakness in the left lower limb for 3 days) (Past medical history: cerebral infarction) (Physical examination: heart rate 71 bpm, alert)]

[0135] It should be noted that in the above embodiment, the starting point for entering medical record text information is stroke, the relationship is converted to symptoms and physical examination results, the entry endpoint is heart rate, and the structured processing results are chief complaint, past history, and physical examination. During the structuring process, the default unit (heart rate unit bmp) is added and commonly used hospital synonyms are converted (conscious - alert). Other cases will not be described in detail here.

[0136] It should be noted that another embodiment of the present invention, a medical data structuring processing method system, uses the same technical means as a medical data structuring processing method, and therefore will not be described in detail here.

[0137] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for structuring medical data, characterized in that, The method includes the following: Acquire the voice recording data of the doctor entering medical records; remove interference from the voice recording data and save the medical record voice data; Acquire the voice data of the medical record entered by the doctor, and transcribe the voice data of the medical record into the corresponding text information of the medical record; Construct a medical record speech-to-text error detection model to perform speech-to-text error detection based on the transcribed medical record text information: If the evaluation result based on the medical record speech-to-text error detection model is normal, the corresponding transcribed medical record text information is output. If the evaluation result based on the medical record speech-to-text error detection model is abnormal, the corresponding medical record speech data will be re-transcribed into medical record text information. The method for error detection in medical record speech-to-text transcription based on transcribed medical record text information specifically includes the following: Construct an arbitrary medical record text information distribution model As shown below: ； in, The medical record text information distribution model The computational object is represented as the first Medical record entry terms The distribution of; among which Characterized by the current medical record entry term The former Medical record entry terms; Constructing a fault-tolerant model for arbitrary medical record text information , ； in, This is represented by the number of times the corresponding medical record entry term appears in the medical record knowledge base; Based on the constructed medical record text information distribution model Fault tolerance model for medical record text information Calculate the probability of accurate prediction of the next word in the medical record text information for any entered word: where, For regularization constants, It is the stability coefficient. This indicates any combination of two medical record entry terms; ； The evaluation index based on the assessment results of the aforementioned medical record speech-to-text error detection model is: Then we have: ； in, The representation is a description of the current entry probability for any medical record entry term. To describe the weights; Construct a medical record knowledge base, match the textual information of medical records that are assessed as normal with the medical record knowledge base, and output the matched medical terminology medical record entry information based on the medical record knowledge base.

2. The medical data structuring processing method according to claim 1, characterized in that, The method for removing interference from the voice recording data includes the following: A voice feature model for doctors entering medical records is constructed, wherein the influence factor of the model is the medical record entry time and the single-segment impulse ratio corresponding to any unit of time; Mark any starting point of medical record entry voice acquisition and record it as a marker point. Based on the marker point, record the voice acquisition frequency. The number of marker points is obtained until any of the recorded medical record voice recordings ends. Then the time for entering medical records for: ； Mark the total impulse of any medical record recording voice. Based on the marker points, frequencies are obtained according to speech. The cumulative equal divisions are Segment, any equal division The impulse of a single segment of medical record voice input is Any equal division The single-segment impulse ratio of the single-segment medical record recording voice is The single-segment impulse ratio is calculated as follows: 。 3. The medical data structuring processing method according to claim 2, characterized in that, The method further includes: Based on the constructed medical record entry time and single-stage impulse ratio Detect all collected and recorded audio: If the detected single-segment impulse ratio first rises to the maximum value and then falls, retain the corresponding recorded sound; otherwise, delete all collected recorded sounds and repeat the above steps. Based on the preserved recorded audio that first rises to its maximum value and then falls, the corresponding medical record entry time is retained. The sound corresponding to the maximum value is the target sound.

4. The medical data structuring processing method according to claim 3, characterized in that, The method for matching the medical record text information with the assessment result being normal with the medical record knowledge base specifically includes the following: Generate arbitrary medical record text information that has been checked for errors through speech-to-text transcription, and mark all recognized and detected medical record entries as... Then the set of case text information and medical record entry terms is represented as: ; Based on the generated set Construct a deep neural-guided convolution model : ； Constructing a mapping association between case text information and medical terminology The mapping association The structure is as follows: ； ； Among them, mapping association For jump degree As the first jump, jump degree As the correspondence of the second jump, mapping association For jump degree As the first jump, jump degree As a correspondence for the second jump, Represented as Dynamic associations.

5. A method for structuring medical data according to claim 4, characterized in that, After completing the mapping association After the construction, the method further includes the following: Get mapping association Associating the mapping Input to the medical terminology conversion set: ； in, yes Type function, It is an input parameter matrix based on medical record input terms. It is a mapping association and The result of the dot product, It is a mapping-related offset vector.

6. A method for structuring medical data according to claims 1-5, characterized in that, The constructed medical record speech-to-text error-checking model includes the starting point for medical record text information input, relationship conversion, and input endpoint.

7. A system employing the medical data structuring processing method as described in any one of claims 1 to 6, characterized in that, The system includes the following: The voice filtering module is used to acquire voice recording data of doctors entering medical records; remove interference sounds from the voice recording data, and save the medical record voice data; The speech-to-text module is used to acquire medical record speech data entered by doctors and transcribe the medical record speech data into corresponding medical record text information; The voice verification module is used to build a medical record voice-to-text error-checking model, and performs error checking on the transcribed medical record text information: If the evaluation result based on the medical record speech-to-text error detection model is normal, the corresponding transcribed medical record text information is output. If the evaluation result based on the medical record speech-to-text error detection model is abnormal, the corresponding medical record speech data will be re-transcribed into medical record text information. The transcription and redirection module is used to build a medical record knowledge base, match the medical record text information that the evaluation result is normal with the medical record knowledge base, and output the matched medical terminology medical record entry information based on the medical record knowledge base.

8. The system according to claim 7, characterized in that, The voice filtering module includes the following: The model building unit is used to build a voice feature model for doctors to enter medical records. The influence factor of the model is the single-segment impulse ratio corresponding to the medical record entry time and any unit of time. The voice tagging unit is used to mark the start point of any medical record entry voice acquisition and record the mark point, and then record the voice acquisition frequency based on the mark point. The number of marker points is obtained until any of the recorded medical record voice recordings ends. Then the time for entering medical records for: ； The impulse calculation unit is used to mark the total impulse of any medical record recording voice. Based on the marker points, frequencies are obtained according to speech. The cumulative equal divisions are Segment, any equal division The impulse of a single segment of medical record voice input is Any equal division The single-segment impulse ratio of the single-segment medical record recording voice is The single-segment impulse ratio is calculated as follows: 。

Images