Personalized data source generation method and apparatus for assisting electrostimulation regimen decision making

Abstract

This invention discloses a method and apparatus for generating personalized data sources for assisting in decision-making regarding electrical stimulation protocols. Belonging to the field of biomedical engineering technology, it acquires high-precision EEG signals and medical record text information from patients and preprocesses them to obtain multimodal feature vectors. These multimodal feature vectors are then used for RAG medical knowledge retrieval and multi-agent cross-validation to identify characteristic EEG indicators of neurological / psychiatric diseases, revealing abnormal brain regions. Based on these abnormal brain region results, a personalized data source is generated to assist in decision-making regarding electrical stimulation protocols. This assists decision-makers in making informed decisions, providing objective diagnostic evidence for various neurological / psychiatric diseases and overcoming excessive reliance on doctors' subjective experience.

Description

Technical Field

[0001] This invention relates to the field of biomedical engineering technology, and in particular to a method and apparatus for generating personalized data sources for assisting in decision-making regarding electrical stimulation protocols. Background Technology

[0002] Electroencephalography (EEG), with its high temporal resolution, can reflect the brain's functional state in real time and is an important tool for the auxiliary diagnosis and early screening of neurological / psychiatric diseases such as epilepsy, depression, Alzheimer's disease, and Parkinson's disease. Transcranial electrical stimulation (TCS), as a non-invasive brain modulation technique, modulates neuronal excitability through weak electrical currents and has been widely used in the clinical treatment and rehabilitation of the aforementioned diseases.

[0003] Existing integrated EEG acquisition and transcranial electrical stimulation devices only achieve hardware function superposition, and their level of intelligence as a clinical auxiliary tool is low, especially in terms of decision support, where they have significant technical defects: the devices can only display basic waveforms, lack in-depth feature extraction and state assessment for clinical decision-making, cannot autonomously analyze complex EEG signals, the stimulation parameter setting relies on classical paradigms and physician experience, lacks individual physiological data support, cannot achieve accurate personalized data sources, and are difficult to provide doctors with effective auxiliary diagnostic decision-making basis. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method and apparatus for generating personalized data sources to assist in decision-making regarding electrical stimulation protocols. This method can autonomously analyze electroencephalogram (EEG) data and generate personalized data sources to assist decision-makers in making medical decisions.

[0005] On the one hand, a method for generating personalized data sources to assist in decision-making for electrical stimulation programs is provided, including: Acquire multimodal data from patients, including multichannel EEG signals and medical record text information; Multimodal data is preprocessed to obtain multimodal feature vectors; By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain results of abnormal brain regions. Personalized data sources are generated based on abnormal findings in brain regions to assist decision-makers in making informed decisions about electrical stimulation protocols.

[0006] Furthermore, the multimodal data is preprocessed to obtain multimodal feature vectors, including: Multi-scale convolutional networks are used to extract spatiotemporal features from input multi-channel EEG signals and discretize them into EEG lexical units with continuous temporal features. The patient's medical record text information is converted into text lexical units using a text embedding model; The obtained EEG lexical units and text lexical units are input into a cross-modal alignment network for modal alignment to obtain multimodal feature vectors.

[0007] Furthermore, the multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases through RAG medical knowledge retrieval and multi-agent cross-validation, yielding results of brain region abnormalities, including: Multimodal feature vectors are used as query keys to retrieve highly relevant external prior knowledge from a locally deployed professional medical vector database. The retrieved external prior knowledge is dynamically concatenated with the patient's current multimodal feature vector to obtain enhanced prompts with a deep medical background.

[0008] Furthermore, enhanced prompts with deep medical background are input into a proprietary large-scale multi-agent collaborative network model to perform multi-agent collaborative diagnosis and thought chain interaction. We obtain precise localization of abnormal brain regions and characteristic EEG indicators, as well as highly interpretable quantitative diagnostic results for brain region abnormalities.

[0009] Furthermore, the proprietary large-scale multi-agent collaborative network model includes: an EEG feature analysis agent, a clinical comprehensive diagnosis agent, and a security review agent; Each agent uses thought chain technology to conduct multiple rounds of natural language interaction and cross-validation based on enhanced prompts from retrieved knowledge. Among them, the EEG feature analysis agent focuses on the identification of frequency band abnormal features and the analysis of brain network connectivity; the clinical comprehensive diagnosis agent combines medical history and clinical data to make a comprehensive judgment on the patient's condition; and the security review agent performs threshold verification and risk screening on diagnostic conclusions and stimulation parameters in accordance with clinical safety standards.

[0010] Furthermore, personalized data sources include personalized multi-point transcranial electrical stimulation parameter combinations for the current patient's specific neuropsychiatric disease, accurately defining and outputting parameters such as target electrode locations, stimulation current intensity, and cycle duration.

[0011] On the other hand, an apparatus for generating personalized data sources for assisting in decision-making regarding electrical stimulation protocols is provided, comprising: employing the aforementioned method for generating personalized data sources for assisting in decision-making regarding electrical stimulation protocols. EEG acquisition module: Acquires multimodal data from patients, including multichannel EEG signals and medical record text information; Feature vector acquisition module: preprocesses the multimodal data to obtain multimodal feature vectors; Brain region abnormality result acquisition module: By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain brain region abnormality results; Data source generation module: Generates personalized data sources based on abnormal brain region results to assist decision-makers in making decisions on electrical stimulation protocols.

[0012] Furthermore, an electronic device is also provided, including: Memory, used for non-transitory storage of computer-readable instructions; and Processor, for executing the computer-readable instructions, When the computer-readable instructions are executed by the processor, they perform the method described in the first aspect above.

[0013] In another aspect, a storage medium is also provided for non-transitory storage of computer-readable instructions, wherein when the non-transitory computer-readable instructions are executed by a computer, the method described in the first aspect is performed.

[0014] In another aspect, a computer program product is also provided, including a computer program that, when run on one or more processors, is used to implement the method described in the first aspect above.

[0015] The above technical solution has the following advantages or beneficial effects: This invention discloses a personalized data source generation method and device for assisting in electrical stimulation protocol decision-making. It combines collected electroencephalogram (EEG) signals with medical record text information, and through RAG medical knowledge retrieval and multi-agent cross-validation, autonomously completes symptom assessment and assisted diagnosis. It accurately identifies abnormalities in the patient's EEG signals, such as characteristic EEG indicators of various neurological / psychiatric diseases like depression, epilepsy, and Parkinson's disease, such as reduced low-frequency alpha waves in depression, epileptiform discharges in epilepsy, and abnormal rhythm networks in Parkinson's disease. Based on these abnormal EEG signal results, it generates personalized data sources for assisting in electrical stimulation protocol decision-making, such as electrode stimulation points, current intensity, and stimulation cycle duration. This provides objective auxiliary diagnostic evidence for various neurological / psychiatric diseases, overcoming excessive reliance on doctors' subjective experience. Attached Figure Description

[0016] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0017] Figure 1 The flowchart below shows a personalized data source generation method for assisting in decision-making regarding electrical stimulation protocols, as described in Example 1. Figure 2 This is a block diagram of the EEG signal acquisition unit described in Embodiment 1; Figure 3 The schematic diagram is of the signal acquisition module described in Embodiment 1. Figure 4 This is a spectrogram of an electroencephalogram (EEG) simulation signal as described in Example 1; Figure 5 This is a waveform diagram of an electroencephalogram (EEG) simulation signal as described in Example 1; Figure 6 The circuit diagram is of the half-bridge circuit of the transcranial electrical stimulation unit described in Example 1. Figure 7 This is a schematic diagram of the human-computer interaction interface described in Embodiment 1. Detailed Implementation

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the invention. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] In this embodiment of the invention, "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the description of this invention, "multiple" refers to two or more.

[0021] Furthermore, to facilitate a clear description of the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0022] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0023] All data acquisition in this embodiment is carried out in accordance with laws and regulations and with user consent, and the data is used legally.

[0024] RAG: Retrieval-Augmented Generation.

[0025] Example 1 This embodiment provides a method for generating personalized data sources to assist in decision-making regarding electrical stimulation protocols, such as... Figure 1 As shown, it includes: Acquire multimodal data from patients, including multichannel EEG signals and medical record text information; Multimodal data is preprocessed to obtain multimodal feature vectors; By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain results of abnormal brain regions. Personalized data sources are generated based on abnormal brain region findings to assist decision-makers in making decisions regarding auxiliary electrical stimulation protocols.

[0026] The personalized data source generation method for assisting electrical stimulation program decision-making described in this embodiment includes the following steps: S1: Acquire the patient's multimodal data, which includes multichannel EEG signals and medical record text information.

[0027] S11: Acquire multi-channel EEG signals through the EEG signal acquisition unit.

[0028] This embodiment describes acquiring 8-channel EEG signals through an EEG signal acquisition unit, such as... Figure 2 As shown, the EEG signal acquisition unit mainly includes a signal acquisition module, a power management module, and a main control module. Multi-channel EEG signals are acquired by the signal acquisition module, then pass through a protection circuit before entering the ADS1299 for amplification and digitization processing. The signals are then transmitted to the main control module via the SPI interface. The main control module is responsible for data processing, storage, and communication. The core of the EEG signal acquisition circuit is the ADS1299 chip, a high-precision ADC chip. This chip features high precision and low noise, making it suitable for acquiring EEG signals and preprocessing the acquired signals.

[0029] The signal acquisition module is responsible for acquiring weak EEG signals from the scalp and performing hardware processing such as filtering, amplification, and analog-to-digital conversion on the acquired signals. It mainly consists of acquisition electrodes, an EEG cap, ear clip shielded wires, an ADS1299 chip, and other preprocessing circuits such as a pre-filter circuit.

[0030] The ADS1299-8 is a low-noise 8-channel 24-bit analog-to-digital converter (ADC) suitable for bioelectrical signal measurements. Its main functions are amplification, filtering, and acquisition of weak bioelectrical signals, such as electroencephalogram (EEG) signals. Its 24-bit high resolution and low noise characteristics enable accurate acquisition of EEG signals, ensuring high-quality data. Internally integrated programmable gain amplifiers (PGA) and bias drive amplifiers ensure sufficient amplification of weak signals and effective preprocessing before analog-to-digital conversion.

[0031] To eliminate interference from external noise signals on the system's data acquisition, a pre-amplifier low-pass filter circuit was designed to filter out high-frequency noise from the sampling electrodes, limit the frequency of the signal input to the ADS1299 chip, and simultaneously improve the accuracy and stability of the acquired signal, thereby enhancing signal quality and making the signal cleaner. This filter circuit employs a first-order RC passive low-pass filter. The values ​​of the resistors and capacitors in the circuit determine the filter's cutoff frequency for high-frequency signals. Based on the cutoff frequency formula, two series resistors were selected with a value of 5kΩ each, and the capacitor with a value of 4.7nF.

[0032] The circuit schematic of the signal acquisition module is as follows: Figure 3 As shown, in addition to the pre-processing of the input signal by the low-pass filter circuit, the subsequent signal amplification, filtering, analog-to-digital conversion and other functions are mainly implemented by the ADS1299 chip and its peripheral circuits consisting of current limiting, voltage dividing resistors, filtering and decoupling capacitors.

[0033] In one embodiment, in addition to the core acquisition module, the EEG signal acquisition system is equipped with an integrated EEG cap for EEG acquisition and transcranial electrical stimulation (TCS), EEG signal acquisition electrodes, conductive gel, shielding wires, and silver chloride ear clips, among other auxiliary equipment. The EEG cap features a compatible design where the EEG signal acquisition electrodes and TCS electrodes share the same cap, meeting the requirements for multi-channel EEG signal acquisition and multi-channel TCS output, avoiding electrode displacement and interference, and improving ease of wear and signal stability.

[0034] Multi-channel EEG signal acquisition can obtain quantitative data such as amplitude, frequency, and power spectral density (PSD), generating time-domain waveforms, spectrograms, and EEG topographic maps to intuitively reflect the brain's functional state. For example... Figure 4 , Figure 5 The above are the spectrum and waveform of the EEG simulation signal with a single-channel acquisition period of 200ms and a peak-to-peak value of 8μV in this embodiment.

[0035] The power management module mainly supplies power to the various modules of the EEG signal acquisition unit. It consists of chips such as LM2662, LT3042 and LT3094 and related DC-DC circuits, providing stable, low-noise positive and negative power supply voltages for ADS1299 and Raspberry Pi 5.

[0036] The LM2662 chip converts the +5V voltage input from the Raspberry Pi 5's power pin to -5V, serving as the input voltage for the LT3094 chip. This simplifies the power supply module design, eliminating the need for additional transformers or complex circuitry. The LT3042 chip provides stable positive power supplies of +3.3V and +2.5V for the ADS1299 and other analog circuits. Its extremely low noise and high power supply rejection ratio ensure accurate signal acquisition. The LT3094 chip, used in conjunction with the LM2662 chip, converts the input -5V voltage to -2.5V, providing a stable negative power supply voltage for the ADS1299 and ensuring the stability and low noise characteristics of the negative power supply.

[0037] The ADS1299 chip integrates a serial peripheral interface compatible with the SPI standard. The signal acquisition module and the main control module are connected via pin headers and sockets, using SPI communication to achieve stable transmission and reception of multi-channel raw EEG data. The main control module preprocesses, denoises, and enhances the raw EEG data, employing algorithms such as 0.5–40Hz bandpass filtering and baseline correction to remove environmental interference and power frequency noise, retaining effective EEG components to form high-quality time-series data, resulting in multi-channel EEG signals. The main control module communicates with the host computer via serial port, uploading the preprocessed multi-channel EEG signals to the host computer for intelligent diagnostic analysis and visualization.

[0038] S12: Obtain medical record text information.

[0039] The medical record text information described in this embodiment includes the patient's name, age, gender, chief complaint, present illness, past medical history, drug allergy history, epidemiological history, physical examination results, auxiliary examination conclusions, neuropsychiatric scale scores, and clinical diagnosis.

[0040] S2: Preprocess the multimodal data to obtain multimodal feature vectors.

[0041] S21: To address the continuous and non-stationary nature of EEG signals, a multi-scale convolutional network is used to extract spatiotemporal features from the input multi-channel EEG signals and discretize them into EEG lexical units with continuous temporal features. S22: Transform discrete auxiliary information such as patient medical history and scale scores into text units through a text embedding model; S23: Input the obtained EEG lexical units and text lexical units into the cross-modal alignment network for modal alignment to obtain multimodal feature vectors.

[0042] This embodiment can eliminate modal heterogeneity in a unified high-dimensional feature space through modal alignment, thereby achieving precise anchoring of abnormal neurophysiological bands with clinical medical semantics.

[0043] S3: By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain abnormal brain regions.

[0044] S31: Perform RAG medical knowledge retrieval on the multimodal feature vectors; To ensure the rigor of evidence-based medicine in the diagnostic results, the RAG mechanism was introduced. The obtained multimodal feature vectors are used as query keys to retrieve highly relevant external prior knowledge from a locally deployed professional medical vector database.

[0045] External prior knowledge includes: authoritative clinical guidelines for transcranial electrical stimulation, atlases of electroencephalography (EEG) pathology, and historical successful intervention cases.

[0046] S32: Dynamically concatenate the retrieved external prior knowledge with the patient's current multimodal feature vector to obtain enhanced prompts with deep medical background.

[0047] The enhanced prompts obtained in this embodiment can provide accurate objective knowledge anchors for subsequent reasoning in large language models.

[0048] S33: Then, multi-agent cross-validation is performed to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain abnormal brain regions.

[0049] Enhanced prompts with deep medical backgrounds are input into a proprietary large-scale multi-agent collaborative network model to perform multi-agent collaborative diagnosis and thought chain interaction.

[0050] The proprietary large-scale multi-agent collaborative network model comprises: an EEG feature analysis agent, a clinical comprehensive diagnosis agent, and a security review agent. Each agent utilizes Chain of Thought (CoT) technology for multi-round natural language interaction and cross-validation based on enhanced prompts from retrieved knowledge. The EEG feature analysis agent focuses on identifying frequency band abnormalities and analyzing brain network connectivity, such as abnormal slow-wave activity. The clinical comprehensive diagnosis agent combines clinical data, including medical history, to comprehensively assess the patient's condition. The security review agent performs threshold verification and risk screening on diagnostic conclusions and stimulus parameters according to clinical safety guidelines, ensuring the safety and compliance of the diagnosis and treatment process and output plan.

[0051] By leveraging the debate and consensus-building mechanism among agents in a proprietary large-scale multi-agent collaborative network model, the system outputs precise localization of abnormal brain regions and characteristic EEG indicators, as well as highly interpretable quantitative auxiliary brain region abnormality diagnosis results.

[0052] S4: Generate personalized data sources based on abnormal brain region findings to assist decision-makers in making decisions on electrical stimulation protocols; S41: Generate personalized data sources for decision-making on auxiliary electrical stimulation protocols based on abnormal brain region findings; Based on the comprehensive diagnostic conclusions and identified target sources obtained from a proprietary large-scale multi-agent collaborative network model, abstract medical diagnostic logic is transformed into quantitative control instructions executable by the lower-level machine. Combining stimulation safety thresholds from the RAG knowledge base, personalized data sources are inferred and output to assist in decision-making for electrical stimulation protocols. These personalized data sources include personalized multi-point transcranial electrical stimulation parameter combinations for the patient's specific neuropsychiatric disease, precisely defining and outputting parameters such as target electrode locations, stimulation current intensity, and cycle duration, providing a basis for precise closed-loop physical intervention.

[0053] S42: A personalized data source used to assist decision-makers in making decisions regarding electrical stimulation protocols.

[0054] It assists decision-makers in making decisions based on personalized data sources used to support electrical stimulation protocols.

[0055] The transcranial electrical stimulation unit receives personalized data sources confirmed by the decision-maker, and can also receive control commands sent by the operator through the human-computer interaction interface.

[0056] The transcranial electrical stimulation unit includes: a power supply module, a minimum system module, a DAC digital-to-analog converter module, and a current output module.

[0057] The power module, serving as the core of the transcranial electrical stimulation unit's energy supply, employs a multi-chip collaborative power supply scheme to ensure stable operation of each unit. Three DC-DC chips—MAX1771, AMS1117, and TPS54331—are selected to construct a hierarchical power supply network: the MAX1771 chip is responsible for boosting the input 12V voltage, specifically providing high-voltage power to the current output module. The boost formula is as follows: , This indicates the upper voltage divider resistor connected to the output HV; This indicates the lower voltage divider resistor connected to GND. The boosted voltage is... .

[0058] The TPS54331 chip steps down the 12V input voltage to 5V to power circuit units such as the DAC digital-to-analog converter; the AMS1117 chip further regulates the 5V voltage to 3.3V to meet the power supply requirements of the main control microcontroller.

[0059] This hierarchical power supply design effectively achieves precise voltage conversion and distribution, ensuring the different power supply voltage requirements of different functional modules and improving the stability and reliability of the overall circuit.

[0060] The minimum system module is the control core of the transcranial electrical stimulation unit, encompassing the main control chip, reset circuit, crystal oscillator circuit, download circuit, and power supply unit, forming a complete control infrastructure. The main control chip is the STM32F303VCT6, which features a high-performance ARM Cortex-M4 core, abundant peripheral resources, and powerful computing capabilities. It can efficiently perform tasks such as multi-channel voltage control, data processing, and instruction execution, providing core hardware support for the precise control of the 8-channel transcranial electrical stimulation.

[0061] The DAC (Digital-to-Analog Converter) module is a DAC circuit based on four MCP4728-E / UN digital-to-analog converter chips. When the gain of these chips is set to 2, the output voltage range is 0.000V to 4.096V. Accordingly, the input voltage range of the operational amplifier in the circuit is synchronously set to 0.000V-4.096V. The sampling resistor in the circuit is 680Ω, and two 100Ω current-limiting resistors are configured to achieve circuit protection. Theoretically, the output current range of this circuit is approximately 0-6mA.

[0062] This embodiment uses a DAC digital-to-analog converter module to output a stable and precise control voltage, thereby achieving precise control over the on / off state of the MOSFETs in the H-bridge circuit.

[0063] The current output module adopts an H-bridge topology, and a single complete output circuit consists of two identical half-bridge circuits. A half-bridge circuit is as follows: Figure 6 As shown, this circuit demonstrates a single-channel output half-bridge circuit for transcranial electrical stimulation. A complete single-channel output consists of two identical half-bridge circuits. Core components include a 2N5401 transistor, a BSS123 MOSFET, an ADA4807-4 operational amplifier, and an SMAJ90C transient suppression diode. The transistor and MOSFET form a switching control unit responsible for switching the circuit's on / off state; the operational amplifier amplifies and conditions the signal, ensuring output signal stability; and the transient suppression diode, in conjunction with a current-limiting resistor, provides overvoltage and overcurrent protection to prevent component damage. An eight-channel parallel output interface, numbered LNO0-LNO15, is also designed to meet the application requirements of simultaneous or independent multi-channel stimulation, providing flexible support for diverse applications in clinical and research settings.

[0064] This embodiment outputs an electrical stimulation scheme through a transcranial electrical stimulation unit: control commands are sent to the main control chip STM32F303VCT6 of the transcranial electrical stimulation unit via a serial port. The microcontroller then controls the output of the DAC chip and the on / off state of each MOSFET. The A and C channels of each DAC chip are set to output a high level; the specific output voltage should be determined according to the required current in the electrical stimulation treatment scheme. Simultaneously, the PH pin of the corresponding channel is set to zero. The B and D channels of each DAC chip are set to 0, and the PH pin of the corresponding channel is set to one (output high level), thus ensuring a constant current output. For example, if channels one and two form a DC output loop, then DAC chip VT1 is set to output a high voltage, and VT2 is set to output 0V. At this time, MOSFET Q36 is turned on and Q13 is turned off. Simultaneously, PH1 is controlled to output a low level and PH2 to output a high level, causing transistor Q28 to turn off and Q5 to turn on.

[0065] This embodiment also includes a human-computer interaction interface, such as... Figure 7 As shown, the human-computer interaction interface is deployed on the host computer to display EEG data in a visual manner. At the same time, users can operate and issue commands to the integrated system through this interface, such as stopping EEG signal acquisition or transcranial electrical stimulation, selecting stimulation programs, and adjusting electrical stimulation parameters.

[0066] The human-computer interface is the interaction point between the system and the user. It integrates functions such as data display, system control, parameter configuration, diagnostic result presentation and treatment process management, and realizes integrated management and control of the entire process of EEG acquisition, intelligent diagnosis and transcranial electrical stimulation.

[0067] Key functions include: Intelligent diagnosis and treatment result display: Synchronously displaying the auxiliary diagnostic conclusions of the disease, the results of brain region abnormality localization, and the personalized data source for auxiliary electrical stimulation plan decision-making output by the intelligent diagnosis and treatment control unit, presented clearly in text form; Visualization of brain region abnormalities: Dynamically marking abnormal brain regions, abnormality levels, and corresponding channels using two-dimensional atlases; Data visualization display: Real-time reception of EEG data, electrode impedance information, and device operating status, presented intuitively in the form of waveform diagrams, spectrum diagrams, and parameter tables; System operation control: Supporting operations such as starting and stopping EEG signal acquisition, starting and stopping transcranial electrical stimulation, and emergency stop; Stimulation parameter adjustment: Supporting online adjustment of transcranial electrical stimulation current, stimulation duration, electrode position, and stimulation mode.

[0068] Example 2 This embodiment provides a personalized data source generation device for assisting in electrical stimulation protocol decision-making. The device is characterized by employing a personalized data source generation method for assisting in electrical stimulation protocol decision-making as described in the above embodiment, comprising: Feature vector acquisition module: preprocesses the multimodal data to obtain multimodal feature vectors; Brain region abnormality result acquisition module: By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain brain region abnormality results; Data source generation module: Generates personalized data sources based on abnormal brain region results to assist decision-makers in making decisions on electrical stimulation protocols.

[0069] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0070] The proposed device can be implemented in other ways. For example, the device embodiments described above are merely illustrative, and the division of the modules described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another device, or some features may be ignored or not executed.

[0071] Example 3 This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein, the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform the method described in Embodiment 1.

[0072] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0073] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0074] In the implementation process, each step of the above method can be completed by the integrated logic circuits in the processor hardware or by software instructions.

[0075] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0076] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0077] Example 4 This embodiment also provides a storage medium for storing computer instructions, which, when executed by a processor, complete the method described in Embodiment 1.

[0078] Example 5 This embodiment also provides a computer program product, including a computer program that, when run on one or more processors, implements the method described in Embodiment 1.

[0079] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating personalized data sources to assist in decision-making regarding electrical stimulation protocols, characterized in that, include: Acquire multimodal data from patients, including multichannel EEG signals and medical record text information; Multimodal data is preprocessed to obtain multimodal feature vectors; By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain results of abnormal brain regions. Personalized data sources are generated based on abnormal findings in brain regions to assist decision-makers in making informed decisions about electrical stimulation protocols.

2. The method for generating personalized data sources for assisting in electrical stimulation protocol decision-making according to claim 1, characterized in that, The multimodal data is preprocessed to obtain multimodal feature vectors, including: Multi-scale convolutional networks are used to extract spatiotemporal features from input multi-channel EEG signals and discretize them into EEG lexical units with continuous temporal features. The patient's medical record text information is converted into text lexical units using a text embedding model; The obtained EEG lexical units and text lexical units are input into a cross-modal alignment network for modal alignment to obtain multimodal feature vectors.

3. The method for generating personalized data sources for assisting in electrical stimulation protocol decision-making according to claim 1, characterized in that, By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases, yielding results of abnormal brain regions, including: Multimodal feature vectors are used as query keys to retrieve highly relevant external prior knowledge from a locally deployed professional medical vector database. The retrieved external prior knowledge is dynamically concatenated with the patient's current multimodal feature vector to obtain enhanced prompts with a deep medical background.

4. The method for generating a personalized data source for assisting in electrical stimulation protocol decision-making according to claim 3, characterized in that, Enhanced prompts with deep medical backgrounds are input into a proprietary large-scale multi-agent collaborative network model to perform multi-agent collaborative diagnosis and thought chain interaction. We obtain precise localization of abnormal brain regions and characteristic EEG indicators, as well as highly interpretable quantitative diagnostic results for brain region abnormalities.

5. A method for generating personalized data sources for assisting in electrical stimulation protocol decision-making according to claim 4, characterized in that, The proprietary large-scale multi-agent collaborative network model includes: an EEG feature analysis agent, a clinical comprehensive diagnosis agent, and a security review agent; Each agent uses thought chain technology to conduct multiple rounds of natural language interaction and cross-validation based on enhanced prompts from retrieved knowledge. Among them, the EEG feature analysis agent focuses on the identification of frequency band abnormal features and the analysis of brain network connectivity; the clinical comprehensive diagnosis agent combines medical history and clinical data to make a comprehensive judgment on the patient's condition; and the security review agent performs threshold verification and risk screening on diagnostic conclusions and stimulation parameters in accordance with clinical safety standards.

6. The method for generating a personalized data source for assisting in electrical stimulation protocol decision-making according to claim 1, characterized in that, Personalized data sources include personalized multi-point transcranial electrical stimulation parameter combinations for the current patient's specific neuropsychiatric disease, accurately defining and outputting parameters such as target electrode locations, stimulation current intensity, and cycle duration.

7. A personalized data source generation device for assisting in decision-making regarding electrical stimulation protocols, characterized in that, A method for generating personalized data sources for assisting electrical stimulation protocol decision-making, as described in any one of claims 1-6, includes: Feature vector acquisition module: preprocesses the multimodal data to obtain multimodal feature vectors; Brain region abnormality result acquisition module: By using RAG medical knowledge retrieval and multi-agent cross-validation, multimodal feature vectors are used to identify characteristic EEG indicators of neurological / psychiatric diseases and obtain brain region abnormality results; Data source generation module: Generates personalized data sources based on abnormal brain region results to assist decision-makers in making decisions on electrical stimulation protocols.

8. An electronic device, characterized in that, include: Memory is used to store computer-readable instructions in a non-transitory manner. as well as Processor, for executing the computer-readable instructions, When the computer-readable instructions are executed by the processor, they perform a personalized data source generation method for assisting in decision-making regarding electrical stimulation programs, as described in any one of claims 1-6.

9. A storage medium, characterized in that, Non-transitory storage of computer-readable instructions, wherein, when executed by a computer, the non-transitory computer-readable instructions perform a personalized data source generation method for assisting in electrical stimulation program decision-making as described in any one of claims 1-6.

10. A computer program product, characterized in that, The invention includes a computer program that, when running on one or more processors, implements the personalized data source generation method for assisting in decision-making regarding electrical stimulation protocols as described in any one of claims 1-6.

Images