CN-122024986-A - Accurate diagnosis and treatment decision support system for prostate cancer based on knowledge graph

CN122024986ACN 122024986 ACN122024986 ACN 122024986ACN-122024986-A

Abstract

The invention relates to the technical field of medical health data processing and discloses a prostate cancer accurate diagnosis and treatment decision support system based on a knowledge graph, which comprises a multi-mode data acquisition module, a semantic alignment reasoning module and a decision link construction module, wherein the system invokes prostate specific antigen, image scoring, puncture grading and gene data, utilizes a time phase difference between biochemical metabolism transient variation and structural hysteresis response to construct a multi-mode heterogeneous topological graph, generates a phase translation compensation characterization vector through a mapping operator, determines node association weight, and utilizes path searching to generate a deduction subgraph reflecting the disease evolution state.

Inventors

WEI YONG
LIN BOHAN
ZHU ZHONGHUA
KE ZHIBIN
YI ZEYU
Si Shuxin

Assignees

福建医科大学附属第一医院

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (9)

1. The utility model provides a accurate diagnosis and treatment decision support system of prostate cancer based on knowledge graph which characterized in that, the system includes: the multi-mode data acquisition module is used for retrieving medical data comprising a prostate specific antigen sequence, nuclear magnetic resonance image scoring, puncture biopsy grading and gene phenotype data; The semantic alignment reasoning module is connected with the multi-modal data acquisition module and is used for processing medical data through the cross-modal time phase deviation compensation logic, and the semantic alignment reasoning module is used for constructing a multi-modal heterogeneous topological graph by calculating the time phase difference between the transient change of the biochemical metabolic index and the hysteresis response of the anatomical structure characteristic, wherein the multi-modal heterogeneous topological graph takes a diagnosis and treatment entity of a patient as a central node, takes the medical data as an attribute node and utilizes a mapping operator to generate a characterization vector after phase translation compensation in a characteristic space; The decision link construction module is connected with the semantic alignment reasoning module and is used for calculating the association weight among all attribute nodes in the multi-mode heterogeneous topological graph through a cross-mode attention mechanism, extracting characteristic node links with the weight contribution degree meeting a preset weight threshold value of 0.65-0.85 by utilizing a path search algorithm, generating a deduction subgraph reflecting the disease evolution state, and outputting technical parameters representing the prostate cancer metastasis risk.
2. The accurate diagnosis and treatment decision support system for the prostate cancer based on the knowledge graph according to claim 1, wherein the semantic alignment reasoning module introduces knowledge constraints in a clinical diagnosis and treatment ontology library into the characterization dimension of a patient diagnosis and treatment entity through a joint embedding algorithm when constructing a multi-modal heterogeneous topological graph, calculates attenuation factors of all attribute nodes in the multi-modal heterogeneous topological graph according to time stamps of medical data, and reduces the attenuation factors along with the increase of the time interval between the time stamps and the current sampling point so as to lead clinical signs with the sampling frequency not lower than 1 time/month in the medical data to occupy weight dominant positions in phase translation compensation.
3. The accurate diagnosis and treatment decision support system for prostate cancer based on a knowledge graph according to claim 1, wherein the system further comprises a decision security monitoring module, the decision security monitoring module is used for calculating the degree of topological deviation between individual subgraphs of patients in the multi-modal heterogeneous topological graph and a standard medical knowledge graph, and when the degree of topological deviation exceeds a safety boundary value of 0.3, the decision security monitoring module automatically starts a decision feedback cut-off mechanism, switches an output state into an artificial auditing mode, and generates a clinical evidence report containing conflict node pairs.
4. The knowledge-graph-based accurate diagnosis and treatment decision support system for prostate cancer according to claim 1, wherein the semantic alignment reasoning module maps characterization parameters of tumor volume change in an image mode to a real-time calculation space defined by a prostate specific antigen sequence by fitting response time consumption differences of different medical monitoring means through a depth regression network when processing cross-mode time phase deviation compensation logic, and performs time domain registration on feature vectors of different modes by using a phase translation operator.
5. The accurate diagnosis and treatment decision support system for prostate cancer based on a knowledge graph according to claim 1, wherein the decision link construction module introduces a dynamic correction term when calculating the association weight, the dynamic correction term is defined by a time first derivative of a prostate specific antigen sequence, and the decision link construction module improves the priority of a biochemical feature node corresponding to the prostate specific antigen sequence in medical data in a path search algorithm when the first derivative exceeds a rate threshold of 0.2 ng/mL/d.
6. The knowledge-graph-based accurate diagnosis and treatment decision support system for prostate cancer according to claim 1, wherein the multi-modal data acquisition module comprises a laboratory information system interface and an image archiving system interface, and the multi-modal data acquisition module invokes a measurement value of prostate specific antigen of a patient through the laboratory information system interface and matches three-dimensional reconstructed image data in the image archiving system based on a serial number of the measurement value.
7. The knowledge-graph-based accurate diagnosis and treatment decision support system for prostate cancer according to claim 1, further comprising a parameter self-adaptive adjustment module, wherein the parameter self-adaptive adjustment module adjusts the perceived depth of the multi-modal heterogeneous topological graph according to the output value of the technical parameter, and when the technical parameter is greater than 0.75, the parameter self-adaptive adjustment module increases the traversal depth for the genetic phenotype data.
8. The prostate cancer accurate diagnosis and treatment decision support system based on the knowledge graph according to claim 1, wherein the deduction subgraphs generated by the decision link construction module have logic traceability, the system presents logic jump paths among characteristic node links through a front-end interactive interface, and each jump node in the logic jump paths is mapped to a clinical diagnosis and treatment guide item stored in the knowledge graph.
9. The knowledge-graph-based accurate diagnosis and treatment decision support system for prostatic cancer according to claim 1, wherein the system operates in a computing environment supported by a parallel processor array, invokes a desensitization case knowledge base through a data encryption channel to iteratively optimize relation embedding parameters in a multi-mode heterogeneous topological graph, and updates the optimized relation embedding parameters to an embedding matrix of a semantic alignment reasoning module in real time.

Description

Accurate diagnosis and treatment decision support system for prostate cancer based on knowledge graph Technical Field The invention belongs to the technical field of medical health data processing, and particularly relates to a prostate cancer accurate diagnosis and treatment decision support system based on a knowledge graph. Background The accurate diagnosis and treatment of the prostate cancer at present depends on the prostate specific antigen sequence, nuclear magnetic resonance image scoring, puncture biopsy pathology grading and comprehensive judgment of genomics characteristics, a multisource data integration technology is commonly adopted in the medical informatization field, the heterogeneous data are collected and input into an electronic medical record system, a preset static rule or a statistical model is utilized to provide risk early warning for a clinician, and the system has a certain reference value in the standardized flow of processing standard cases, so that the basis of a modern auxiliary decision system is formed. However, the evolution process of the prostate cancer has complex space-time heterogeneity, clinical characterizations of different modes have natural phase differences on physical observation, concentration changes of biochemical indexes have high-frequency transient sensitivity, anatomical morphology changes of whole tissues have physical delay, the prior art generally follows the timestamp synchronization principle, multisource data acquired in the same diagnosis and treatment period are in absolute synchronous disease evolution phases, the logics neglect the asynchronism of biological characterizations, the imaging characteristics tend to dilute high-frequency sensitive biochemical early warning signals, missed diagnosis risks are generated in a critical window period of early tumor metastasis or recurrence, and in addition, due to physical limits or test errors of detection equipment, medical data of different sources often generate semantic conflicts in the same logic space, and the prior art lacks measurement and calibration capability of the semantic tension; the decision system algorithm end fusion logic faces challenges, the prostate cancer evolution has complex space-time heterogeneity, the biochemical index concentration change has high-frequency transient sensitivity, physical hysteresis exists in the whole tissue anatomy form change, for example, chinese patent application publication No. CN120527031A discloses a prostate cancer endocrine therapy drug resistance prediction system based on multi-model artificial intelligence, multiple groups of chemical data are integrated and a prediction model is constructed by using a Stacking integrated learning algorithm, the prior art logic is constructed to have steady-state hypothesis bias, medical data of different modes are defaulted to be in absolute synchronous phase in the disease evolution process, under the actual working condition, the model only completes algorithm dimension feature stack, the underlying biological characterization asynchronous time lag is not decoupled, the image feature dilutes the high-frequency sensitive biochemical early warning signal, the early tumor metastasis or recurrence window period generates diagnosis omission risk, and the heterogeneous index semantic tension lacks logic self-checking capability. Therefore, how to construct a computing architecture capable of decoupling modal asynchronous time lags and calibrating heterogeneous semantic conflicts, and to realize deep logic fusion and causal deduction of prostate cancer multisource clinical evidence, becomes the technical problem to be solved by the invention. Disclosure of Invention The invention provides a prostate cancer accurate diagnosis and treatment decision support system based on a knowledge graph, which comprises: the multi-mode data acquisition module is used for retrieving medical data comprising a prostate specific antigen sequence, nuclear magnetic resonance image scoring, puncture biopsy grading and gene phenotype data; The semantic alignment reasoning module is connected with the multi-modal data acquisition module and is used for processing medical data through the cross-modal time phase deviation compensation logic, and the semantic alignment reasoning module is used for constructing a multi-modal heterogeneous topological graph by calculating the time phase difference between the transient change of the biochemical metabolic index and the hysteresis response of the anatomical structure characteristic, wherein the multi-modal heterogeneous topological graph takes a diagnosis and treatment entity of a patient as a central node, takes the medical data as an attribute node and utilizes a mapping operator to generate a characterization vector after phase translation compensation in a characteristic space; The decision link construction module is connected with the semantic alignment reasoning module and is used for ca