CN-122025074-A - Intelligent optimization AI diagnosis method and system based on four-diagnosis combined parameter

CN122025074ACN 122025074 ACN122025074 ACN 122025074ACN-122025074-A

Abstract

The invention relates to the field of AI diagnosis, in particular to an intelligent optimization AI diagnosis method and system based on four diagnosis combined parameters. The method comprises the steps of obtaining a diagnosis image, obtaining pulse diagnosis information, generating and constructing a patient customized diagnosis question through AI based on the obtained face diagnosis image, tongue diagnosis image and pulse diagnosis information, re-proportioning diagnosis weights of the face diagnosis image, tongue diagnosis image and pulse diagnosis information according to answer conditions of the diagnosis questions, outputting a preliminary diagnosis result, re-weighting four diagnoses in the preliminary diagnosis result based on the preliminary diagnosis result combined with an AI database, constructing a diagnosis data prediction matrix, obtaining a final diagnosis result through the matrix, and outputting predicted conditions.

Inventors

LING LIN

Assignees

广州德生智能信息技术有限公司

Dates

Publication Date: 20260512
Application Date: 20251218

Claims (7)

1. The intelligent optimization AI diagnosis method based on four diagnosis combined parameters is characterized by comprising the following steps: s100, acquiring diagnostic images, wherein the diagnostic images comprise facial diagnosis images and tongue diagnosis images; S200, generating and constructing a patient customized inquiry question through AI based on the acquired facial diagnosis image, tongue diagnosis image and pulse diagnosis information; S300, re-proportioning diagnosis weights for the facial diagnosis image, the tongue diagnosis image and the pulse diagnosis information according to the question answer condition, and outputting a preliminary diagnosis result; s400, re-weighting four diagnoses in the preliminary diagnosis results based on the preliminary diagnosis results and an AI database, constructing a diagnosis data prediction matrix, obtaining a final diagnosis result through the matrix, and outputting a predicted illness state.
2. The intelligent optimization AI diagnosis method based on four diagnosis combined parameters according to claim 1, wherein the four diagnosis and the parameters are respectively represented as face diagnosis, tongue diagnosis, pulse diagnosis and inquiry diagnosis, an image diagnosis result is obtained by collecting a user face diagnosis image and a tongue diagnosis image according to the face diagnosis and the tongue diagnosis, then a pulse diagnosis result is obtained according to pulse diagnosis information, inquiry questions are generated based on a hospital database AI through the results, re-proportioning weights are carried out on the image diagnosis result and the pulse diagnosis result through user answers, a preliminary diagnosis result is obtained, and the preliminary diagnosis result is output.
3. The four-diagnosis-combination-based intelligent optimization AI diagnosis method according to claim 1, wherein in step S400, based on the preliminary diagnosis result, each item of data in the preliminary diagnosis result is organized into a data set, the data set is established with a diagnosis data prediction matrix, and each item of data of the preliminary diagnosis result collected in the database is computationally analyzed by combining each item of data of the preliminary diagnosis result with a hospital database.
4. The intelligent optimization AI diagnosis method based on four diagnosis parameters according to claim 3, wherein when a diagnosis value of a preliminary diagnosis result is acquired, the diagnosis value is defined as Te, a relevant diagnosis value of the preliminary diagnosis is acquired in a hospital database, the diagnosis data prediction matrix is divided into a plurality of nodes, a node outputting the diagnosis result is defined as a main node, each node subjected to weight judgment is defined as a sub-node, a node state S is correspondingly set for each sub-node, S is expressed as T or F, T is expressed as a diagnosis result accessing and outputting the sub-node, F is expressed as a new diagnosis result outputting by re-proportioning weight for the sub-node, a state initial value of the common node is set as T, and a network connection graph is constructed through the state, when the node state sta is T, the common node is in the state of outputting the diagnosis result and fixes the diagnosis result, and is connected with the main node through a network, the node state is bound with the corresponding node number, and the node state can be queried through the node number; Acquiring a plurality of keywords of the inquiry answers according to semantic analysis and access records, wherein the keywords are main search words for searching the inquiry answers, acquiring the retrieval amount of the keywords in the inquiry answers which are informationized in a root node Mist, constructing a retrieval amount sequence E according to the order of the retrieval amount from large to small, and calculating the weight coefficient of the retrieval amount , = The said And For retrieving the k-th and k+1-th bit elements of the sequence E, according to Assigning the keywords to obtain the weights of the retrieval quantity on the keywords , The d is the total element amount of the retrieval amount sequence Re, and the d-1 is a weight coefficient The weight of the key words The visit condition of the inquiry answers is represented, and the weight ratio of the child nodes where the inquiry answers with more visit conditions are located is changed greatly.
5. The intelligent optimization AI diagnosis method based on four diagnosis parameters according to claim 3, wherein the diagnosis results corresponding to the keywords are integrated into a data stream, the data stream is arranged by the modification sequence recorded in the contract and the sequence data is constructed, data= [ , ......, N is the total number of data streams output at the current moment, and the data streams are processed The flow rate of the data stream in the data stream sequence data is acquired, and a flow rate sequence ve= [ is constructed , ......, A critical value L of the weight sequence Ve is obtained through calculation, L= + , For the value of the i-th bit element in the weight sequence Ve, For the maximum value in the weight sequence Ve, For the minimum value in the weight sequence, exp () is an exponential function, and the threshold value L is set as the lowest weight value when the keyword is searched, if If not less than L, the weight value is normal, and the node state is assigned to the node as T, if If the weight value is lower than L, assigning a node state of F to the node; S505, recording the number of the child node as D, wherein the number D is a set of child node numbers which are arranged in a small-to-large order, connecting the child node with a main node according to the number, obtaining an encryption matrix DEN, constructing a matrix according to the weight value of the child node connected with the main node, adding the array DEN into data streams in each node according to the weight value, constructing a set A, A= [ a1, a2, a3...az ] of the data sets, constructing x data chains for the time period t1, constructing a matrix on each vertex of the D, constructing a GTP model (DEN, x) of the mapping relation network array DEN by using each data chain as a column, and the mathematical expression form of the model G (DEN, x) is as follows: G(DEN,x)= ; The matrix G is ordered according to the sequence of the preceding columns, node sequence storage data sets of encryption weights are determined according to the GTP (DEN, x) matrix, a server of an encryption system is installed in the vertex, a plurality of computer terminals are connected for each common node from the vertex, the data sets in each common node are distributed to S different servers in S shares, secondary encryption is carried out in the servers, and specific diagnosis results of the diagnosis data prediction matrix are output.
6. The four-diagnosis co-parameter based intelligent optimization AI diagnosis method according to claim 3, wherein the diagnosis data prediction matrix is used for outputting a three-dimensional decision output comprising a probability thermodynamic diagram, an anatomical structure association map and a natural language diagnosis suggestion; The natural language diagnosis suggestion is generated by fusing template filling and neural text generation technology and comprises ICD (information-coding) recommendation and a differential diagnosis list; the self-adaptive model optimization engine is also used for setting a multi-level confidence coefficient checking mechanism, and triggering the expert rule engine to perform double checking when the model output confidence coefficient is lower than a threshold value; The expert rule engine is used for integrating the latest clinical guideline and evidence-based medical evidence base by adopting a knowledge graph reasoning technology.
7. The intelligent optimization AI diagnosis system based on four diagnosis and combination is characterized by comprising a diagnosis system memory and a diagnosis system processor, wherein the diagnosis system processor can realize the steps in the intelligent optimization AI diagnosis method based on four diagnosis and combination according to any one of claims 1-5 when executing the computer program.

Description

Intelligent optimization AI diagnosis method and system based on four-diagnosis combined parameter Technical Field The invention relates to the field of AI diagnosis, in particular to an intelligent optimization AI diagnosis method and system based on four diagnosis combined parameters. Background The traditional medical image diagnosis is highly dependent on physician experience, has strong subjectivity, low efficiency and missed diagnosis and misdiagnosis risks, adopts single-mode image analysis in the prior art, is difficult to capture space-time correlation characteristics of disease development, faces challenges such as small sample learning, insufficient model generalization capability and the like, has feature gaps in multi-mode data fusion, has a stiff dynamic feature extraction mechanism, cannot adapt to focus form evolution, and is not capable of effectively updating a model, and diagnosis results are distorted under different diagnosis environments and various conditions through fixed weight proportion, so that the intelligent optimization AI diagnosis method and system based on four diagnosis parameters are needed to solve the problems. Disclosure of Invention In view of the limitations of the prior art, the present invention aims to provide an intelligent optimized AI diagnosis method and system based on four diagnosis parameters, so as to solve one or more technical problems existing in the prior art, and at least provide a beneficial choice or creation condition. An intelligent optimization AI diagnosis method based on four diagnosis combined parameters, which comprises the following steps: s100, acquiring diagnostic images, wherein the diagnostic images comprise facial diagnosis images and tongue diagnosis images; S200, generating and constructing a patient customized inquiry question through AI based on the acquired facial diagnosis image, tongue diagnosis image and pulse diagnosis information; S300, re-proportioning diagnosis weights for the facial diagnosis image, the tongue diagnosis image and the pulse diagnosis information according to the question answer condition, and outputting a preliminary diagnosis result; s400, re-weighting four diagnoses in the preliminary diagnosis results based on the preliminary diagnosis results and an AI database, constructing a diagnosis data prediction matrix, obtaining a final diagnosis result through the matrix, and outputting a predicted illness state. Further, the four diagnosis and the parameter are respectively expressed as facial diagnosis, tongue diagnosis, pulse diagnosis and inquiry diagnosis, image diagnosis results are obtained by collecting facial diagnosis images and tongue diagnosis images of a user according to the facial diagnosis and the tongue diagnosis, pulse diagnosis results are obtained according to pulse diagnosis information, inquiry diagnosis questions are generated based on a hospital database AI through the results, the weight of the image diagnosis results and the pulse diagnosis results is re-proportioned through user answers, a preliminary diagnosis result is obtained, and the preliminary diagnosis result is output. Further, in step S400, based on the preliminary diagnosis result, each item of data in the preliminary diagnosis result is organized into a data set, the data set is established into a diagnosis data prediction matrix, and each item of data of the preliminary diagnosis result collected in the database is calculated and analyzed by combining each item of data of the preliminary diagnosis result with a hospital database. Further, when a diagnosis value of a preliminary diagnosis result is acquired, the diagnosis value is defined as Te, a relevant diagnosis value of the preliminary diagnosis is acquired in a hospital database, the diagnosis data prediction matrix is divided into a plurality of nodes, the node outputting the diagnosis result is defined as a main node, each node of which the data is subjected to weight judgment is defined as a sub-node, each sub-node is correspondingly provided with a node state S, S is expressed as T or F, the T is expressed as the diagnosis result of the sub-node which is accessed and output, the F is expressed as the new diagnosis result which is output by the sub-node through the weight matching again, the state initial value of the common node is set as T, and a network connection diagram is constructed through the state of the network connection diagram, wherein when the node state sta is T, the common node is in the state of outputting the diagnosis result and fixing the diagnosis result, the network connection is carried out with the main node, the node state is bound with the corresponding node number, and the node state can be inquired through the node number; Acquiring a plurality of keywords of the inquiry answers according to semantic analysis and access records, wherein the keywords are main search words for searching the inquiry answers, acquiring the retrieval amount