Artificial intelligence-based voice-to-text real-time nursing record system

Abstract

This invention relates to the field of nursing technology, specifically disclosing an AI-based real-time speech-to-text nursing record system. The speech acquisition module utilizes wearable devices and an adaptive noise suppression algorithm to achieve high signal-to-noise ratio speech acquisition in a hospital environment. The AI ​​real-time transcription module employs a Transformer architecture model, fine-tuned with nursing professional corpus, and combines accent adaptation and real-time semantic error correction to improve the accuracy of professional terminology and number transcription. The nursing standardization adaptation module automatically formats and generates compliant initial draft records based on a template library and standard dictionary. Data storage employs AES-256 encryption and local plus cloud backup, coupled with hierarchical permissions and dual login verification to ensure medical data security. This invention enables simultaneous generation of nursing operations and records, reducing nurses' paperwork burden, improving the standardization, accuracy, and timeliness of nursing records, and is applicable to multi-departmental clinical nursing scenarios, possessing strong practicality and promotional value.

Description

Technical Field

[0001] This invention relates to the field of nursing technology, specifically to an artificial intelligence-based real-time speech-to-text nursing record system. Background Technology

[0002] In clinical nursing practice, nursing records are core documents reflecting changes in a patient's condition and the implementation of nursing interventions, and also serve as important evidence for medical quality control and the resolution of medical disputes. Currently, clinical nursing records are primarily entered manually by nurses or supplemented later, which presents the following prominent problems: 1. Low work efficiency: After completing patient care procedures, nurses need to spend a lot of time recalling the details of the procedure, organizing patient information, and manually entering nursing records, which takes up a lot of the time nurses spend directly caring for patients, resulting in a waste of nursing human resources. 2. Delayed and inaccurate records: During manual record-keeping, nurses are prone to omitting key information about the patient's condition or misrecording details of nursing procedures due to factors such as memory bias or busy work schedules. This leads to a disconnect between the nursing records and the actual nursing process, affecting the authenticity and completeness of the records, which may in turn affect the scientific nature of subsequent diagnosis and treatment decisions. 3. Difficulty in ensuring standardization: Nursing records have strict industry standards and format requirements. When entering data manually, nurses are prone to problems such as non-standard formatting and non-standard use of terminology, which increases the difficulty of medical quality control. 4. Limitations of existing technology: Most voice-to-text tools on the market are general-purpose and not optimized for nursing scenarios. They cannot recognize nursing terminology, cannot adapt to the standardized format of nursing records, and have high transcription delays. They cannot achieve "synchronous recording of nursing operations" and still require nurses to make modifications and improvements later, failing to fundamentally solve the problem of rewriting.

[0003] To address the aforementioned issues, there is an urgent need to design a nursing record system specifically adapted to clinical nursing scenarios, capable of real-time speech-to-text conversion, and eliminating the need for nurses to rewrite records later. This system should incorporate artificial intelligence technology to optimize transcription accuracy and standardization, thereby freeing up nursing manpower and improving the quality of nursing records. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an artificial intelligence-based real-time voice-to-text nursing record system, which realizes real-time voice collection, real-time transcription, and real-time standardized input of patient information and nursing measures during nursing operations, eliminating the need for nurses to rewrite later, greatly reducing nurses' paperwork workload, and ensuring the authenticity, completeness, and standardization of nursing records.

[0005] To solve the above-mentioned technical problems, the technical solution provided by the present invention is: a real-time speech-to-text nursing record system based on artificial intelligence, including a speech acquisition module, an AI real-time transcription module, a nursing standard adaptation module, a data storage module, an interactive display module, and a permission management module; The voice acquisition module is used to acquire the voice signals of nurses during nursing operations in real time, convert them into digital signals after noise filtering, and transmit them to the AI ​​real-time transcription module in real time. The AI ​​real-time transcription module adopts a transcription model based on the Transformer architecture, and performs real-time transcription of nursing professional terms and numbers through fine-tuning of nursing professional corpus. The nursing standard adaptation module has a built-in nursing record template library and standard dictionary, which automatically organizes and formats the transcribed text, fills in the necessary elements, and generates a first draft of nursing records that conforms to clinical nursing standards. The data storage module uses a combination of local encrypted storage and cloud backup to store voice data, transcribed text, nursing records, and operation logs in real time. The interactive display module supports multi-terminal operation and realizes functions such as voice acquisition control, transcription viewing, record correction, submission, and historical query. The permission management module adopts a hierarchical permission mechanism to control system access and operation permissions.

[0006] Furthermore, the voice acquisition module includes a portable microphone and a noise filtering unit; the portable microphone is a wearable microphone or a handheld microphone terminal; the noise filtering unit adopts an adaptive noise suppression algorithm to specifically filter hospital power frequency noise, instrument noise, and environmental conversation noise, suppressing noise interference through spectral subtraction, enhancing the human voice signal, and ensuring that the signal-to-noise ratio of the acquired voice signal is ≥35dB.

[0007] Furthermore, the transcription model adopts a two-layer architecture of a basic transcription model + nursing professional fine-tuning; the basic transcription model is trained through massive general speech data; the nursing professional fine-tuning stage inputs massive nursing professional speech data and performs fine-tuning in combination with clinical nursing record standards.

[0008] Furthermore, the nursing record template library covers various nursing scenarios, including general wards, intensive care units, and postoperative care. Each scenario corresponds to a preset record template, which includes basic patient information, condition observation, nursing measures, and vital sign records. The standardized dictionary includes nursing professional terms, common abbreviations, and contraindicated expressions, automatically replacing non-standard expressions with standard nursing terms, while also supporting hospital-customized templates and standards.

[0009] Furthermore, the data storage module uses the AES-256 encryption algorithm to encrypt the data, the local encrypted storage is deployed on the hospital's internal server, and the cloud backup uses a dedicated hospital cloud server for regular automatic backups; the operation log records all data access, modification, and deletion operations.

[0010] Furthermore, the interactive display module supports multi-terminal display on nurse station computers, mobile nursing terminals, and tablet computers. It supports voice or manual correction of transcription errors, and automatically synchronizes nursing records to the hospital's electronic medical record system after submission. It also supports querying historical nursing records and corresponding voice data by patient name, hospital number, and record time.

[0011] Furthermore, the permission management module divides users into three levels: administrator, head nurse, and nurse. The administrator is responsible for system configuration, permission allocation, and data maintenance. The head nurse is responsible for viewing nursing records in their department, reviewing nursing records, and viewing operation logs. The nurse can only view and edit the nursing records of the patients under their care, and has the permissions to transcribe and correct records and submit records, but no permissions to delete data or modify templates. User login uses a dual authentication method of account password + facial recognition.

[0012] Furthermore, the AI ​​real-time transcription module integrates a real-time error correction unit and an accent adaptive learning unit. The accent adaptive learning unit records the accent characteristics of different nurses and adapts to the nurses' accents through short-term learning. The real-time error correction unit uses a contextual semantic analysis algorithm to correct homophones, similar shapes, and word order errors.

[0013] The advantages of this invention compared to the prior art are: This invention achieves real-time voice acquisition, real-time transcription, and real-time standardization adaptation, enabling synchronous generation of nursing operations and records. Nurses no longer need to recall and rewrite after completing nursing operations, significantly reducing their paperwork workload and freeing them from tedious paperwork, thus increasing their time for direct patient care. The AI ​​transcription model of this invention is optimized for nursing scenarios, with high accuracy in recognizing professional terms. Combined with a real-time error correction unit and a nursing standard adaptation module, it effectively avoids problems such as memory bias, terminology errors, and non-standard formats, ensuring that nursing records are consistent with the actual nursing process, improving record quality, and providing a reliable basis for diagnosis and treatment decisions. The noise filtering function of the voice acquisition module of this invention is adapted to the complex environment of the hospital, the accent adaptive learning adapts to different nurses' accents, and the multi-terminal support adapts to different nursing scenarios, making it highly practical. This invention employs hierarchical access control, encrypted data storage, and operation log traceability, which complies with medical data security standards, ensures the security and confidentiality of nursing records, and meets industry standards for nursing records. This invention supports custom templates and specifications to adapt to the personalized needs of different hospitals and departments. In the future, it can be optimized through model iteration to increase the adaptability to more nursing scenarios and improve the applicability of the system. Attached Figure Description

[0014] Figure 1 This is a system block diagram of the artificial intelligence-based real-time speech-to-text nursing record system of the present invention. Detailed Implementation

[0015] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0016] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0017] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0018] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0019] The following is a detailed description of the artificial intelligence-based real-time speech-to-text nursing record system of the present invention, with reference to the accompanying drawings.

[0020] Combined with appendix Figure 1 This invention will be described in detail below.

[0021] An AI-based real-time speech-to-text nursing record system includes a speech acquisition module, an AI real-time transcription module, a nursing standard adaptation module, a data storage module, an interactive display module, and an access control module. These modules work together to automate the entire process from speech acquisition to nursing record generation in real time. The specific structure is as follows: The voice acquisition module is used to collect voice commands issued by nurses during nursing operations in real time, including patient condition descriptions, nursing operation details, vital sign data, etc. It adopts multi-channel sound pickup technology to adapt to complex clinical environments, such as noise from ward instruments and conversations, to achieve clear sound pickup.

[0022] The voice acquisition module includes a portable microphone (wearable microphone, handheld microphone terminal) and a noise filtering unit: the portable microphone supports wireless connection, is lightweight and easy for nurses to carry, and has an adjustable pickup distance to adapt to different nursing scenarios, such as bedside care and ward rounds; the noise filtering unit uses an adaptive noise suppression algorithm to specifically filter common noises in the hospital environment, such as power frequency noise, instrument operation noise, and environmental conversation noise. It suppresses noise interference through spectral subtraction while enhancing the human voice signal to ensure that the acquired voice signal is clear and free of noise, thus ensuring the accuracy of subsequent transcription.

[0023] The voice acquisition module converts the acquired voice signals into digital signals in real time and transmits them to the AI ​​real-time transcription module. The transmission delay is ≤300ms, ensuring that the transcription is synchronized with the nursing operation.

[0024] The AI ​​real-time transcription module is the core module of the system, which is used to convert the digital voice signal transmitted by the voice acquisition module into text information in real time. The core adopts a dedicated transcription model based on deep learning, which is different from the general voice transcription model. It is specially optimized and trained for nursing scenarios to achieve high accuracy and low latency transcription.

[0025] The transcription model adopts a two-layer architecture of a basic transcription model and nursing specialty fine-tuning: the basic transcription model is built on the Transformer architecture and trained with massive amounts of general speech data to ensure basic transcription capabilities; in the nursing specialty fine-tuning stage, massive amounts of nursing specialty speech data are input, including nursing terminology, patient descriptions, and nursing operation specifications, and combined with clinical nursing record specifications, the model is fine-tuned to optimize the recognition accuracy of nursing specialty terms, numbers, and abbreviations. The accuracy rate for recognizing nursing specialty terms is ≥98.5%, and the accuracy rate for recognizing numbers is ≥99.8%, avoiding the problems of misjudgment of specialty terms and confusion of numbers in the general transcription model.

[0026] Meanwhile, the module integrates a real-time error correction unit, which uses a contextual semantic analysis algorithm to correct errors in the transcribed text in real time, focusing on correcting homophone errors, similar-looking errors, and word order errors to ensure the accuracy of the transcribed text; the transcription delay is ≤500ms, realizing "the text is generated synchronously after the voice is finished", which is synchronized with the nursing operation in real time and does not require nurses to make supplementary modifications later.

[0027] In addition, to address the transcription error caused by differences in nurses' accents, this module supports accent adaptive learning. It can record the accent characteristics of different nurses and adapt to their accents through short-term learning, thereby further improving the transcription accuracy and solving the transcription pain points for nurses with different accents.

[0028] The nursing standard adaptation module is used to automatically organize and format the text information generated by the AI ​​real-time transcription module according to the clinical nursing record standard and hospital nursing record standard, so as to generate nursing records that meet the requirements without nurses having to manually adjust the format or supplement the standard content.

[0029] The nursing standard adaptation module has a built-in nursing record template library and standard dictionary: The template library covers various clinical nursing scenarios, such as general ward nursing, intensive care nursing, and postoperative nursing. Each scenario corresponds to a preset record template, which includes record items such as basic patient information, condition observation, nursing measures, vital signs, health guidance, format requirements, and terminology standards. The standard dictionary contains nursing professional terms, commonly used abbreviations, and prohibited expressions to ensure that the transcribed text conforms to nursing standards.

[0030] The specific adaptation process is as follows: The system automatically identifies key information in the transcribed text, such as vital signs data, nursing operation names, and keywords describing the patient's condition, and fills them into the corresponding positions in the preset template. It automatically adjusts the text format, such as paragraph spacing, bullet points, and numerical standardization, automatically replaces non-standard expressions with standard nursing terminology, automatically supplements the necessary elements required for the nursing record, such as the record time and nurse's name, and automatically associates the system time and the currently logged-in nurse's information to generate a complete and standardized draft of the nursing record.

[0031] Meanwhile, the nursing standards adaptation module supports a custom template function, allowing hospitals to modify preset templates and add exclusive standards according to their own nursing work requirements, adapting to the personalized needs of different hospitals and departments.

[0032] The data storage module is used to store voice data from the voice acquisition module, transcribed text from the AI ​​real-time transcription module, nursing records generated by the nursing standard adaptation module, as well as system operation logs and user operation records in real time, ensuring the security, integrity and traceability of the data.

[0033] This module uses a combination of local storage and cloud backup: local storage uses an encrypted database deployed on an internal hospital server to store real-time generated nursing records and voice data, ensuring data security and low access latency; cloud backup uses encrypted backup technology to periodically back up local data to a dedicated hospital cloud server to prevent local data loss and support off-site data recovery.

[0034] Data storage follows medical data security standards, and all data is encrypted using the AES-256 encryption algorithm. Only authorized personnel can access the data. All access, modification, and deletion operations are recorded to form a complete operation log, which facilitates medical quality control and traceability. Voice data is stored in association with corresponding nursing records, enabling "text to voice" verification, which facilitates subsequent verification of the authenticity of nursing records.

[0035] The interactive display module is used to enable nurses to interact with the system, including voice acquisition and control, viewing transcribed text, editing nursing records (allowing only minor corrections and eliminating the need for extensive rewriting), record submission, and historical record query. The interface is simple and easy to use, making it suitable for nurses' busy work scenarios.

[0036] This module supports multi-terminal display, including nurse station computers, mobile nursing terminals, and tablets. Nurses can view the transcribed text and generated nursing records in real time via mobile terminals. If a few transcription errors are found, they can be quickly corrected via voice or manual input. After correction, the system automatically updates and stores the data. It supports real-time submission of nursing records, which are automatically synchronized to the hospital's electronic medical record system after submission, eliminating the need for nurses to re-enter the data. It also supports searching historical nursing records and corresponding voice data by keywords such as patient name, hospital number, and record time, facilitating nurses' review and verification.

[0037] In addition, the interactive display module has a reminder function. When the transcription accuracy rate is lower than a preset threshold, it will automatically issue a slight reminder to prompt nurses to pay attention to and check the transcribed content, thereby further ensuring the accuracy of the records.

[0038] The access control module is used to manage system access and operation permissions, ensuring the security and confidentiality of nursing records and complying with medical data management standards.

[0039] The access control module adopts a hierarchical access control mechanism, dividing users into three levels: administrator, head nurse, and nurse. Administrators are responsible for system configuration, template modification, permission allocation, and data maintenance. Head nurses are responsible for viewing all nursing records in their department, reviewing nursing records, and viewing operation logs. Nurses can only view and edit the nursing records of the patients they are responsible for, and cannot access the data of patients under the responsibility of other nurses. They only have the permissions to transcribe and correct records and submit records, but do not have the permissions to delete data or modify templates.

[0040] At the same time, the system records all operations of each user, such as login, voice acquisition, transcription correction, record submission, and query, forming an operation log. This allows for the traceability of the generation and modification process of each nursing record, ensuring that the responsibility for nursing records is traceable.

[0041] The specific implementation process of the artificial intelligence-based real-time speech-to-text nursing record system of this invention is as follows: The workflow of this system is as follows, and it is fully real-time and automated, eliminating the need for nurses to manually rewrite data later: The nurse logs into the system, selects the patient under her care, and activates the voice capture module. While performing nursing procedures, nurses simultaneously describe the patient's condition and the details of the nursing procedure using voice. The voice acquisition module acquires voice signals in real time, and after noise filtering, transmits the digital voice signals to the AI ​​real-time transcription module. The AI ​​real-time transcription module transcribes speech signals into text in real time, and after real-time error correction, transmits the text to the nursing standard adaptation module. The nursing standard adaptation module automatically organizes and formats the transcribed text based on preset templates and nursing standards, fills in the necessary elements, and generates a standard nursing record draft. Nurses can view the transcribed text and draft nursing records in real time through the interactive display module, and quickly correct any rare errors they find. After the nurse confirms that everything is correct, she submits the nursing record. The system stores the nursing record and voice data in real time through the data storage module and synchronizes them to the hospital's electronic medical record system. Head nurses can review nursing records through the system, and administrators can perform system maintenance and access management. All operations are recorded in the operation log, enabling full traceability.

[0042] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A real-time speech-to-text nursing record system based on artificial intelligence, characterized in that: It includes a voice acquisition module, an AI real-time transcription module, a nursing standard adaptation module, a data storage module, an interactive display module, and a permission management module; The voice acquisition module is used to acquire the voice signals of nurses during nursing operations in real time, convert them into digital signals after noise filtering, and transmit them to the AI ​​real-time transcription module in real time. The AI ​​real-time transcription module adopts a transcription model based on the Transformer architecture, and performs real-time transcription of nursing professional terms and numbers through fine-tuning of nursing professional corpus. The nursing standard adaptation module has a built-in nursing record template library and standard dictionary, which automatically organizes and formats the transcribed text, fills in the necessary elements, and generates a first draft of nursing records that conforms to clinical nursing standards. The data storage module uses a combination of local encrypted storage and cloud backup to store voice data, transcribed text, nursing records, and operation logs in real time. The interactive display module supports multi-terminal operation and realizes functions such as voice acquisition control, transcription viewing, record correction, submission, and historical query. The permission management module adopts a hierarchical permission mechanism to control system access and operation permissions.

2. The AI-based real-time speech-to-text nursing record system according to claim 1, characterized in that: The voice acquisition module includes a portable microphone and a noise filtering unit; the portable microphone is a wearable microphone or a handheld microphone terminal; the noise filtering unit adopts an adaptive noise suppression algorithm to filter hospital power frequency noise, instrument noise, and environmental conversation noise, suppresses noise interference through spectral subtraction, enhances human voice signals, and ensures that the signal-to-noise ratio of the acquired voice signal is ≥35dB.

3. The AI-based real-time speech-to-text nursing record system according to claim 2, characterized in that: The transcription model adopts a two-layer architecture of basic transcription model + nursing professional fine-tuning; The basic transcription model is trained using massive amounts of general speech data; the nursing professional fine-tuning stage inputs massive amounts of nursing professional speech data and performs fine-tuning in conjunction with clinical nursing record standards.

4. The AI-based real-time speech-to-text nursing record system according to claim 3, characterized in that: The nursing record template library covers various nursing scenarios, including general wards, intensive care units, and postoperative care. Each scenario corresponds to a preset record template, which includes basic patient information, condition observation, nursing measures, and vital signs records. The standardized dictionary includes nursing professional terms, common abbreviations, and contraindicated expressions, and automatically replaces non-standard expressions with standard nursing terms. It also supports hospital-customized templates and standards.

5. The AI-based real-time speech-to-text nursing record system according to claim 4, characterized in that: The data storage module uses the AES-256 encryption algorithm to encrypt the data. The local encrypted storage is deployed on the hospital's internal server, and the cloud backup uses a dedicated cloud server for regular automatic backups. The operation log records all data access, modification, and deletion operations.

6. The AI-based real-time speech-to-text nursing record system according to claim 5, characterized in that: The interactive display module supports multi-terminal display on nurse station computers, mobile nursing terminals, and tablet computers. It supports voice or manual correction of transcription errors, and automatically synchronizes nursing records to the hospital's electronic medical record system after submission. It also supports querying historical nursing records and corresponding voice data by patient name, hospital number, and record time.

7. The AI-based real-time speech-to-text nursing record system according to claim 6, characterized in that: The access control module categorizes users into three levels: administrator, head nurse, and nurse. Administrators are responsible for system configuration, access control, and data maintenance. Head nurses are responsible for viewing and reviewing nursing records in their department and viewing operation logs. Nurses can only view and edit nursing records for patients under their care, and have the authority to transcribe and correct records and submit records, but not the authority to delete data or modify templates. User login uses a dual authentication method of account password + facial recognition.

8. The artificial intelligence-based real-time speech-to-text nursing record system according to claim 7, characterized in that: The AI ​​real-time transcription module integrates a real-time error correction unit and an accent adaptive learning unit. The accent adaptive learning unit records the accent characteristics of different nurses and adapts to the nurses' accents through short-term learning. The real-time error correction unit uses a contextual semantic analysis algorithm to correct homophones, similar shapes, and word order errors.

Images