CN-122025175-A - Medical meta-space self-care intervention method based on hybrid intelligence

CN122025175ACN 122025175 ACN122025175 ACN 122025175ACN-122025175-A

Abstract

The invention provides a medical meta-space self-care intervention method based on hybrid intelligence, which comprises the following steps of S1, forming a multi-mode data set for self-care intelligent decision, S2, preprocessing various multi-mode data, outputting a multi-dimensional initial characteristic data sequence, S3, introducing a vector sequence continuing a self-attention mechanism, and S4, mapping the vector sequence into different subspaces by a multi-head attention mechanism by adopting mutually independent parameter matrixes, thereby obtaining the medical meta-space self-care intervention method Score vector calculated by individual attention module The method comprises the steps of S5, selecting an adaptive architecture from different self-attention neural network architectures according to the difference condition of the multi-modal medical data characteristic channels, carrying out multi-modal data characteristic fusion calculation, S6, carrying out neural network splicing operation on the extracted multi-modal data deep characteristic information, outputting a prediction result, and S7, training a model constructed by the flow by using a self-care multi-scene task data set, and continuously optimizing model parameters.

Inventors

YAN MIAN
Lai Diandian
YANG QIAOHONG
CHEN YANYA
LI MINGXING
LI WANSHAN

Assignees

暨南大学

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (6)

1. A medical metaspace self-care intervention method based on hybrid intelligence comprises the following steps: s1, integrating human individual intelligence and medical element space intelligent agents through bidirectional information coding to form a multi-mode data set for self-care intelligent decision; s2, cleaning missing values and abnormal values, aligning modes, normalizing data and extracting core features of various multi-mode data in sequence, and outputting a multi-dimensional initial feature data sequence; s3, introducing a vector sequence continuing a self-attention mechanism, wherein the vector sequence comprises Q (Query), K (Key ), V (Value ), Q (Query), K (Key ) and V (Value ) which respectively represent Query embedding, key embedding and Value embedding; s4, the multi-head attention mechanism adopts mutually independent parameter matrixes Sequence of vectors Mapping into different subspaces to obtain Score vector calculated by individual attention module ; S5, selecting an adaptive architecture from different self-attention neural network architectures according to the difference condition of the multi-modal medical data characteristic channels, and carrying out multi-modal data characteristic fusion calculation; S6, outputting a prediction result through neural network splicing operation on the extracted multi-mode data deep characteristic information; and S7, training the constructed model by using the self-care multi-scene task data set, and continuously optimizing model parameters.
2. The medical metaspace self-care intervention method based on hybrid intelligence as claimed in claim 1, wherein in particular, in S1, the method comprises the steps of: s11, constructing a medical metaspace intelligent agent data fusion model driven by Human individual intelligence in a mixed mode based on a Human-in-the-loop (HITL) mode; S12, in the self-care process, user cognition and behavior data capable of realizing feature recognition and knowledge combing are output to a medical metaspace intelligent carrier server through a specific information channel; And the user can acquire the cognition and behavior data of the user through a contextual dialogue, a behavior interaction and the like in the feature recognition and knowledge management mode. S13, based on user cognition and behavior data, the medical meta-space intelligent carrier server completes a closed loop of a man-machine intelligent fusion mode through task, learning, analysis and advice progressive calculation process on the medical meta-space intelligent body data fusion model S14, systematically fusing the multi-mode data generated in the human-computer interaction process through information fusion of the medical meta-space intelligent body and the self-care individual intelligent knowledge base around the self-care intervention method to realize the integration operation of incomplete and unstructured health data to form medical mode data; And S15, realizing fusion integration of different medical modal data in a specific space-time domain by mixing the physical energy code and the concept code of the man-machine information, and finally aligning and outputting the multi-modal data set under medical meta-space man-machine intelligent data fusion.
3. A hybrid intelligence based medical meta-space self-care intervention method as defined in claim 1, wherein: In S4, the specific calculation formula is that the projection matrix The length of the code of the sequence is represented, The position is self-focusing on the head, 。。
4. A hybrid intelligence based medical meta-space self-care intervention method as defined in claim 1, wherein: And the user can acquire the cognition and behavior data of the user through a contextual dialogue, a behavior interaction and the like in the feature recognition and knowledge management mode.
5. A hybrid intelligence based medical meta-space self-care intervention method as defined in claim 1, wherein: The multimodal data includes physiological characteristic data, psychological characteristic data, environmental scene data, physical space data, intelligent system feedback data, and meta-space initial data.
6. A hybrid intelligence based medical meta-space self-care intervention method as defined in claim 1, wherein: the neural network architecture comprises early summation, early serial connection, layered attention, cross attention, serial connection cross attention and the like, and adopts adaptive architectures selected from different self-attention neural network architectures to perform multi-mode data feature fusion calculation, wherein in a paradigm design, the multi-mode data feature fusion calculation is performed by And As a multi-modal input, And An embedded flag representing the modal data, Then it is a sequence of multimodal fusion measures, Representing the processing of the transducer block, For multi-head attention, a specific calculation formula is as follows: (1) Early summation: Wherein, the Is the sum of the elements and, And Is the weight of the sample, and the weight of the sample, , , 。 (2) Early tandem: (3) Hierarchical attention (multi-stream to single stream): (4) Hierarchical attention (single stream to multi stream): (5) Cross attention: (6) Series cross attention: 。

Description

Medical meta-space self-care intervention method based on hybrid intelligence Technical Field The invention relates to the technical field of intelligent medical treatment and health management, in particular to a medical metaspace self-care intervention method based on hybrid intelligence. Background With the continuous development of artificial intelligence technology in the field of medical health, there is an increasing need for health management, especially in self care. The self-care intervention decision-making needs to consider multidimensional characteristics such as physiology, psychology, environment and the like of a user, and particularly user cognition and behavior feedback data generated in the human-computer interaction process is an important component element for assisting the intelligent decision of the self-care intervention method. However, the current self-care intervention method mainly depends on single-mode data, and lack of deep fusion and comprehensive analysis on multi-mode data may lead to insufficient understanding of the health state of the user, and it is difficult to provide accurate intervention advice. In addition, most of the existing self-care methods are based on preset rules, human-computer interaction is mostly in a unidirectional mode, and the lack of an effective user feedback mechanism may result in lower matching degree between care suggestions and actual demands. Therefore, a medical metaspace self-care intervention method based on hybrid intelligence is needed, the defects of the prior art in aspects of data integration, man-machine cooperation, feature fusion and the like are overcome, and more accurate personalized care intervention is realized. Disclosure of Invention In order to solve the technical problems, the invention provides a medical metaspace self-care intervention method based on hybrid intelligence, which comprises the following steps: s1, integrating human individual intelligence and medical element space intelligent agents through bidirectional information coding to form a multi-mode data set for self-care intelligent decision; s2, cleaning missing values and abnormal values, aligning modes, normalizing data and extracting core features of various multi-mode data in sequence, and outputting a multi-dimensional initial feature data sequence; s3, introducing a vector sequence continuing a self-attention mechanism, wherein the vector sequence comprises Q (Query), K (Key ), V (Value ), Q (Query), K (Key ) and V (Value ) which respectively represent Query embedding, key embedding and Value embedding; s4, the multi-head attention mechanism adopts mutually independent parameter matrixes Sequence of vectorsMapping into different subspaces to obtainScore vector calculated by individual attention module; S5, selecting an adaptive architecture from different self-attention neural network architectures according to the difference condition of the multi-modal medical data characteristic channels, and carrying out multi-modal data characteristic fusion calculation; S6, outputting a prediction result through neural network splicing operation on the extracted multi-mode data deep characteristic information; and S7, training the constructed model by using the self-care multi-scene task data set, and continuously optimizing model parameters. Specifically, in S1, the following steps are included: s11, constructing a medical metaspace intelligent agent data fusion model driven by Human individual intelligence in a mixed mode based on a Human-in-the-loop (HITL) mode; S12, in the self-care process, user cognition and behavior data capable of realizing feature recognition and knowledge combing are output to a medical metaspace intelligent carrier server through a specific information channel; And the user can acquire the cognition and behavior data of the user through a contextual dialogue, a behavior interaction and the like in the feature recognition and knowledge management mode. S13, based on user cognition and behavior data, the medical meta-space intelligent carrier server completes a closed loop of a man-machine intelligent fusion mode through task, learning, analysis and advice progressive calculation process on the medical meta-space intelligent body data fusion model S14, systematically fusing the multi-mode data generated in the human-computer interaction process through information fusion of the medical meta-space intelligent body and the self-care individual intelligent knowledge base around the self-care intervention method to realize the integration operation of incomplete and unstructured health data to form medical mode data; And S15, realizing fusion integration of different medical modal data in a specific space-time domain by mixing the physical energy code and the concept code of the man-machine information, and finally aligning and outputting the multi-modal data set under medical meta-space man-machine intelligent data fusion. In S4, the specific