Medical diagnosis optimization method and system based on big data

Abstract

The invention discloses a medical diagnosis optimization method and system based on big data, and relates to the technical field of diagnosis optimization. According to the symptom intensity transition criterion and the collapse mechanism, the target path with the highest diagnosis priority can be quickly focused in the multiple possible diagnosis paths, and the accuracy of preliminary judgment is improved; interference correction is carried out on inconsistent parts between diagnosis paths and medical rules by adopting structural comparison and conflict degree calculation means, so that misjudgment caused by logic deviation or local abnormality is avoided; besides, through construction of a consistency verification graph and a forward and reverse path consistency scoring mechanism, bidirectional verification and screening of the overall structure of the diagnosis path are realized, so that the output path has logic rationality and clinical feasibility.

Description

Technical Field

[0001] The present invention relates to the field of diagnosis optimization technology, and in particular to a medical diagnosis optimization method and system based on big data. Background Art

[0002] With the rapid development of medical informatization and data intelligence, intelligent assisted diagnosis methods based on big data have gradually become a key technical means to improve the efficiency and quality of medical services. Modern medical systems have accumulated a vast amount of diagnostic pathways, symptom evolution processes, and intervention records through long-term practice. How to effectively explore the causal relationships and diagnostic evolution patterns of these processes, and on this basis, achieve individualized and dynamic diagnostic pathway optimization, has become one of the important directions of current medical artificial intelligence research. Traditional decision-making assistance methods are mostly based on symptom sets or symptom matching, which makes it difficult to accurately describe the structural and causal nature of disease evolution, and is not conducive to pathway deduction and timely adjustment of intervention strategies in complex conditions.

[0003] For example, CN118629633B discloses a decision-making support method and system for medical big data, which obtains patient medical information and combines it with the disease type and decision parameter table preset in the database to output an auxiliary diagnosis plan. This method emphasizes the mapping relationship between individual diagnosis and treatment data and the disease knowledge base, and has certain advantages in improving decision-making speed and accuracy. However, the diagnostic decision-making process relies more on static parameter comparison and rule matching, fails to fully model the jumping characteristics of symptoms evolving over time, and lacks an interference and reconciliation mechanism for multi-path diagnostic conflicts, making it difficult to support high-complexity diagnosis and treatment scenarios under multi-source information fusion.

[0004] CN118824507A discloses an intelligent medical auxiliary diagnosis system for multi-source collaborative mining of big data. It uses a multi-core collaborative learning strategy to fuse heterogeneous medical data and achieves good results in prediction and classification. The advantage is that it can integrate multi-dimensional data in key scenarios such as the ICU to achieve a higher-dimensional characterization of the disease. However, this method focuses on the extraction of underlying features and the construction of machine learning models, ignoring the structural evolution characteristics of the clinical pathway itself, lacking the ability to model the causal chain of the disease process, and does not involve high-level logical mechanisms such as jump-triggered diagnostic pathway updates, diagnostic rule conflict reconciliation, and path consistency verification. There are problems with insufficient explanatory power and missing diagnostic logic. Summary of the Invention

[0005] In view of the problems existing in the existing medical big data auxiliary diagnosis technology, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to achieve dynamic updating and consistency verification of diagnostic pathways, and effectively improve the adaptability and reasoning logic of the diagnostic system to complex disease processes.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, the present invention provides a medical diagnosis optimization method based on big data, which includes, based on the historical diagnostic paths contained in the medical big data, extracting the symptom evolution trajectory and intervention record sequence of the target individual, forming a disease course causal chain, and constructing a state superposition set; when the symptom intensity index in the symptom evolution trajectory transitions and exceeds a preset jump threshold, triggering the collapse operation of the state superposition set, converging to the target diagnostic path with the highest diagnostic priority; performing item-by-item structural comparison on the target diagnostic path and the medical rule set, calculating the conflict degree parameter, if the conflict degree parameter is greater than the conflict degree tolerance threshold, then correcting the target diagnostic path according to the interference superposition of the candidate diagnostic path to form a path interference state, and obtaining a corrected diagnostic path; re-embedding the corrected diagnostic path into the disease course causal chain to construct a consistency verification graph, and using the consistency score between the forward evolution chain and the reverse tracing chain as the screening basis, extracting the diagnostic path with the highest score as the final diagnostic recommendation path.

[0009] As a preferred solution of the medical diagnosis optimization method based on big data described in the present invention, the construction of the state superposition set includes: extracting the disease course causal chain in chronological order based on the historical diagnosis and treatment data of the target individual, extracting the parameters of each node in the disease course causal chain, and forming an intervention factor tensor through structural splicing; using the disease course causal chain and the intervention factor tensor as retrieval keys, matching the historical diagnostic path set with similar structure in the big data, and calculating the matching score for each diagnostic path, the matching score is given in combination with the overlap of the disease course causal chain sequence and the similarity of the intervention factor space projection; according to the matching score corresponding to each diagnostic path, constructing the superposition probability amplitude in the target state space, based on the diagnostic path set, and encoding it to the state superposition set.

[0010] As a preferred solution of the big data-based medical diagnosis optimization method described in the present invention, the parameters of each node include the intervention intensity, time span and symptom impact level of each node.

[0011] As a preferred solution of the big data-based medical diagnosis optimization method described in the present invention, the triggering operation of collapsing the state superposition set and converging to the target diagnostic path with the highest diagnostic priority includes: reverse mapping the path feature vectors in the state superposition set, calculating the component of the indicator that triggers the transition in each diagnostic path, and extracting a subset of diagnostic paths whose component is greater than the component threshold; normalizing the probability amplitude of each diagnostic path in the diagnostic path subset to form a collapsed probability distribution; based on the normalized probability amplitude, selecting the diagnostic path corresponding to the maximum probability amplitude as the candidate diagnostic path, and using the number of symptom dimensions covered by the candidate diagnostic path and the consistency of the symptom evolution trajectory as the diagnostic priority determination factor; selecting the diagnostic path with the largest diagnostic priority factor from the candidate diagnostic paths as the final target diagnostic path.

[0012] As a preferred solution of the big data-based medical diagnosis optimization method described in the present invention, the calculation of the conflict degree parameter includes: constructing a diagnostic path structure label sequence and a medical rule label sequence based on the diagnostic node sequence contained in the target diagnostic path and the structural labels in the medical rule set, and establishing a structural comparison matrix for the two sequences; and using the proportion of mismatched units in the structural comparison matrix to characterize the structural conflict degree between the target diagnostic path and the medical rule.

[0013] As a preferred solution of the big data-based medical diagnosis optimization method described in the present invention, the modified diagnostic path includes: screening a set of similar diagnostic paths whose similarity to the symptom evolution trajectory of the target diagnostic path is higher than a similarity threshold from the state superposition set, extracting characteristic substructures and similarity weights, and constructing a weighted interference superposition vector; applying the weighted interference superposition vector to the conflicting section position of the target diagnostic path, performing structural adjustment according to position alignment and semantic adaptation rules, and obtaining a modified diagnostic path.

[0014] As a preferred solution of the big data-based medical diagnosis optimization method described in the present invention, the method of using the consistency score between the forward evolution chain and the reverse tracing chain as the screening basis includes: the consistency verification graph includes a forward reasoning edge set and a reverse tracing edge set; two logical sub-chains are separated from the consistency verification graph, namely the forward evolution chain and the reverse tracing chain; a consistency matching matrix is ​​constructed based on the node pairing of the forward evolution chain and the reverse tracing chain, and a weighted score is performed on the node content, causal depth and trigger timing difference to form a path consistency score.

[0015] In a second aspect, the present invention provides a medical diagnosis optimization system based on big data, which includes:

[0016] The disease course extraction module is used to extract the target individual's symptom evolution trajectory and intervention record sequence based on the historical diagnostic path contained in medical big data, form a disease course causal chain, and construct a state superposition set;

[0017] a jump triggering module configured to trigger a collapse operation on the state superposition set when a jump occurs in the symptom intensity indicator in the symptom evolution track and exceeds a preset jump threshold, and converge into a target diagnostic path with the highest diagnostic priority;

[0018] a path correction module configured to compare the target diagnostic path and the medical rule set item by item in structure, calculate a conflict degree parameter, and if the conflict degree parameter is greater than a conflict degree tolerance threshold, correct the target diagnostic path according to the candidate diagnostic path interference superposition to form a path interference state, and obtain a corrected diagnostic path;

[0019] a consistency verification module configured to re-embed the corrected diagnostic path into the disease course causal chain to construct a consistency verification graph, and extract the diagnostic path with the highest score as the final diagnostic suggestion path as a screening basis for the consistency score between the forward evolution chain and the backward tracing chain.

[0020] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program instructions are executed by the processor, the steps of the medical diagnosis optimization method based on big data according to the first aspect of the present application are implemented.

[0021] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein when the computer program instructions are executed by the processor, the steps of the medical diagnosis optimization method based on big data according to the first aspect of the present application are implemented.

[0022] The present application has the following advantages: according to the symptom intensity jump criterion and the collapse mechanism, the present application can quickly focus on the target path with the highest diagnostic priority among multiple possible diagnostic paths, improving the accuracy of preliminary judgment; by using structural comparison and conflict degree calculation means, the inconsistent parts between the diagnostic path and the medical rules are corrected, thereby avoiding misjudgment caused by logical deviation or local abnormality; in addition, by constructing the consistency verification graph and the forward and backward path consistency scoring mechanism, the present application realizes bidirectional verification and screening of the overall structure of the diagnostic path, making the output path more logically reasonable and clinically feasible. In summary, the present application effectively integrates the diagnosis path deduction and structure optimization strategy driven by big data, improving the precision, robustness and interpretability of the diagnosis process under complex conditions. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0024] Figure 1 A flow chart of a medical diagnosis optimization method based on big data.

[0025] Figure 2 A structural diagram of a medical diagnosis optimization system based on big data. DETAILED DESCRIPTION

[0026] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0027] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from the description, and those skilled in the art can make similar generalizations without departing from the concept of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0028] Secondly, "one embodiment" or "embodiment" referred to herein means that a specific feature, structure or characteristic can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is separate or alternative to other embodiments.

[0029] As described in the above background, with the rapid development of medical informatization and data intelligence, intelligent auxiliary diagnosis methods based on big data have gradually become a key technical means to improve the efficiency of medical services and the quality of diagnosis and treatment. Modern medical systems have accumulated a large amount of diagnosis paths, symptom evolution processes and intervention records in a long period of practice. How to effectively mine the causal relationship of the disease course and the diagnosis evolution law, and on this basis to realize the individualization and dynamic optimization of the diagnosis path, has become one of the important directions of current medical artificial intelligence research. Traditional auxiliary decision-making methods are mostly based on symptom set or disease matching, which is difficult to accurately describe the structure and causality of the disease evolution process, and is not conducive to the path deduction and timely adjustment of intervention strategies under complex conditions.

[0030] Figure 1 A flow chart of a medical diagnosis optimization method based on big data according to an embodiment of the present application. As shown in Figure 1 In the medical diagnosis optimization method based on big data, it includes:

[0031] S1: Based on the historical diagnosis path contained in the medical big data, the symptom evolution track and the intervention record sequence of the target individual are extracted, the disease course causal chain is formed, and the state superposition set is constructed, wherein each superposition state is associated with the probability amplitude of a specific diagnosis path.

[0032] The symptom evolution trajectory is a set of symptom event nodes arranged in chronological order in the causal chain of the disease course. Specifically, symptom events usually include subjective description symptoms (such as pain, dizziness, etc.), objective physical sign symptoms (such as fever, abnormal blood pressure, etc.) and imaging indicator symptoms (such as expansion of lesion range, abnormal density, etc.). Due to the heterogeneity of the structure and granularity of different symptom types in the records, a unified standardized label system is used in the present invention to map various symptom events into time series nodes in a unified format, and at the same time mark their first occurrence time, duration and clinical weight. After the node construction is completed, the nodes are arranged in ascending order according to their time labels to form a symptom evolution trajectory.

[0033] The intervention record sequence is to extract the intervention behavior information received by the target individual from the clinical pathway execution log, prescription instruction set and treatment operation record.

[0034] The construction of the state superposition set includes the following steps:

[0035] First, based on the historical diagnosis and treatment data of the target individual, the causal chain of the disease course is extracted in chronological order, and the intervention intensity, time span and symptom impact level of each node in the causal chain of the disease course are extracted, and the intervention factor tensor is formed through structural splicing.

[0036] In the process of forming the causal chain of the disease course, the present invention adopts the dual rules of structural splicing and temporal mapping to integrate the aforementioned symptom evolution trajectory and intervention record sequence into a unified directed causal chain structure. Specifically, the diagnosis node serves as the observation node in the disease course chain, and the intervention node serves as the action node, and a directed edge is established between the two through the influence path. Each edge represents the delay, reversal or strengthening effect of a specific intervention operation on a certain symptom node. In the process of chain construction, the causal strength value of each intervention-symptom connection is calculated based on the frequency of the effect of similar interventions on similar symptoms in historical samples, and stored as the weight of the edge.

[0037] In an embodiment of the present invention, the causal strength value is calculated by: screening out a set of historical paths with the same symptom node structure as the current individual in medical big data to form a basic sample pool; dividing the basic sample pool into an intervention group and a non-intervention group based on whether it contains the intervention behavior to be evaluated. The two groups of samples must control other key variables to be consistent, such as individual age group, number of comorbidities and intervention time period window, to ensure that only the intervention behavior is the only source of change. Statistically determine whether the target symptoms in the intervention group and the non-intervention group become more severe in the subsequent stage (for example, from a mild cough to a persistent high fever), so as to construct the difference in the probability of occurrence of the intervention effect; the symptom node pairs with a higher abnormal evolution rate in the intervention group than in the non-intervention group are regarded as causal edges, and the frequency of occurrence in similar structures is recorded, and the frequency is used as the basic value of the edge weight for normalization, and used in the causal chain of the disease course.

[0038] The final disease course causal chain can be used to describe the phased change characteristics of the target individual's condition in time evolution, intervention intervention nodes and reaction modes.

[0039] Secondly, with the disease course causal chain and the intervention factor tensor as the retrieval key, the structurally similar historical diagnosis path set in the big data is matched, and the matching degree score of each diagnosis path is calculated, which combines the disease course causal chain sequence overlap and the intervention factor space projection similarity to give.

[0040] Specifically, the matching process is mainly in the form of structural template fitting, and the historical sample path with similar causal structure and intervention distribution is preferentially matched. For each matched diagnosis path, the matching degree score is calculated, which is composed of two parts: one is the disease course causal chain sequence overlap, and the other is the projection similarity of the intervention factor space. The disease course causal chain overlap is measured by the longest common subsequence comparison, and the intervention factor projection similarity is obtained by the inner product and normalization processing of the tensor space. That is, in the double-index structure of the score, both the time logic consistency of the path structure and the application similarity of the intervention behavior are considered, so as to ensure that the selected historical path is consistent with the target individual state in structure and semantics.

[0041] Finally, according to the matching degree score corresponding to each diagnosis path, an overlay probability amplitude in the target state space is constructed, and based on the diagnosis path set, it is coded to the state superposition set.

[0042] Among them, the probability amplitude includes the following operation process:

[0043] All candidate diagnosis paths are sorted according to the structural matching score to generate a one-dimensional score distribution curve; the score curve is statistically divided into several level intervals (for example, five segments), and the number of paths in each level interval and the total score percentage are counted; each path is assigned a relative proportion in the score interval, and the assignment values of all paths are normalized to ensure that the total weight value can be used for subsequent overlay probability normalization; according to the normalized weight of the path, it is used as the contribution proportion in the target state space participating in the collapse evolution, and finally an information state set with a probability amplitude value in structure is formed.

[0044] Among them, each diagnosis path is mapped to a state ground state, and the probability amplitude reflects the adaptation degree in the current state space. In this way, the current diagnosis state of the target individual is represented as a linear superposition structure of historical multiple paths, with the characteristics of the quantum analogy mechanism of "multiple possibilities-single convergence".

[0045] S2: When the symptom intensity index in the symptom evolution trajectory jumps and exceeds the preset jump threshold, trigger the path collapse operation of the state superposition set, and converge to the target path with the highest diagnosis priority.

[0046] First, symptom intensity indicators are quantitative values ​​used to measure the severity of a target individual's physiological or pathological condition, such as acute pain scores, blood oxygen saturation fluctuations, and abnormal body temperature. In a continuous time series symptom node, when the offset between the current observed value of a symptom and its historical mean exceeds a set transition threshold, the symptom is considered to have transitioned.

[0047] The jump threshold is a fixed scalar or a dynamic threshold that is adaptively set according to the disease type and is used to control the trigger sensitivity of path collapse.

[0048] Furthermore, triggering the collapse operation of the state superposition set and converging to the target diagnosis path with the highest diagnosis priority includes:

[0049] S2.1: After the transition detection is confirmed, the path feature vectors in the state superposition set are reversely mapped, the component of the indicator that triggers the transition in each diagnostic path is calculated, and the subset of diagnostic paths whose component is greater than the component threshold is extracted.

[0050] Specifically, in this process, a reverse mapping operation is performed on the vector components in each diagnostic path for the specific symptom indicator where the transition occurs.

[0051] The specific approach is: use the transition index as an index, read each vector component in the diagnostic path, and calculate the proportion of the transition index in the total vector components; when the proportion is higher than the component threshold, it is considered that the path is significantly sensitive to the transition index and is included in the diagnostic path subset.

[0052] S2.2: Normalize the probability amplitude of each diagnostic path in the diagnostic path subset to form a collapsed probability distribution.

[0053] S2.3: Based on the normalized probability amplitude, the diagnostic path corresponding to the maximum probability amplitude is selected as the candidate diagnostic path, and the number of symptom dimensions covered by the candidate diagnostic path and the consistency of the symptom evolution trajectory are used as diagnostic priority determination factors.

[0054] Specifically, the diagnostic node structure in the candidate path is analyzed for the types of overlap with the nodes in the causal chain of the target individual's disease course, including three types of matches: same symptoms (same name), same system (same organ system), and same logical position (same evolutionary stage); it is verified whether the intervention nodes in the candidate path are in an acceptable time window with the intervention records in the target trajectory in terms of time sequence. The time window is calculated by the window value before and after the standard treatment of the disease in the medical big data (such as the 3 days before and after antibiotic intervention is an adaptable segment); the node sequence corresponding to the transition segment in the target trajectory is extracted, and the structure is matched with the corresponding segment in the candidate path. If there are more than three consecutive overlaps, it is marked as a high-consistency path.

[0055] Ultimately, the priority factor was determined by the comprehensive grade combination of the above three types of indicators, and pathways with medium and high fitness grades in both structure and timing were preferentially selected as target pathway candidates, which strengthened the clinical reasoning basis of the results.

[0056] S2.4: Select the diagnostic path with the largest diagnostic priority factor from the candidate diagnostic paths as the final target diagnostic path.

[0057] The target pathway ultimately selected not only best fits the current transition indicators at the structural level, but also has the highest explanatory power and clinical rationality, thus providing a stable foundation for subsequent diagnostic deduction, treatment recommendations, and disease course simulation.

[0058] S3: Perform a structural comparison of the target path and the medical rule set item by item, calculate the conflict degree parameter, and if the conflict degree parameter is greater than the conflict degree tolerance threshold, then the target path is corrected based on the interference superposition of the candidate paths to form a path interference state, obtain the corrected path, and feed it back to the disease course causal chain to correct the weights of early nodes.

[0059] S3.1: Calculate the conflict degree parameter.

[0060] Based on the diagnostic node sequence contained in the target diagnostic pathway and the structural labels in the medical rule set, a diagnostic pathway structure label sequence and a medical rule label sequence were constructed, and a structural comparison matrix was established for the two sequences. The proportion of mismatched units in the structural comparison matrix was used to represent the structural conflict between the target diagnostic pathway and the medical rules.

[0061] Specifically, each cell in the structural comparison matrix corresponds to the matching result between the node in the target path and the node in the rule path, where a match is recorded as “1” and a mismatch or missing is recorded as “0”. If the diagnosis node and the rule node belong to the same drug / symptom category, it is defined as a partial match.

[0062] The structural conflict parameter is calculated as the ratio of the number of mismatched cells in the structural comparison matrix—specifically, the ratio of "0" cells to the total number of cells in the matrix. This conflict parameter serves as a core indicator for measuring the degree of structural discrepancy between the target diagnostic pathway and the medical rules, and is used to determine whether there is a potential for significant guideline violations. When this ratio exceeds the set conflict tolerance threshold, the pathway is deemed to be at risk of structural breach, and the pathway interference correction mechanism must be activated to prevent misdiagnosis and misjudgment.

[0063] In this design, the conflict tolerance threshold is set based on the empirical comparison results between historical paths and standard rules. It is adjustable and adaptable, and differentiated tolerance standards can be selected for different diseases or diagnostic stages.

[0064] S3.2: Modify the diagnostic path.

[0065] A set of similar diagnostic paths whose similarity to the symptom evolution trajectory of the target diagnostic path is higher than the similarity threshold is screened from the state superposition set, and the characteristic substructure and similarity weight are extracted to construct a weighted interference superposition vector. The weighted interference superposition vector is applied to the conflicting section position of the target diagnostic path, and the structure is adjusted according to the position alignment and semantic adaptation rules to obtain the corrected diagnostic path.

[0066] Specifically, similar path screening is based on the local pre-diagnosis feature chain matching mechanism, and does not use the final diagnosis code as the main filtering condition. Specifically, the node matching degree of the evolutionary trajectory (calculated by dividing the longest common subsequence by the average node length), the intervention time sequence overlap (calculating the minimum distance between the timestamps of the two intervention sequences) and the diagnosis result homology (if the final diagnosis code is consistent, it is recorded as 1, otherwise it is the structural similarity value on the disease classification tree) are used as criteria. The similarity threshold (such as 85%, which is set according to actual operations) is set through weighted calculation to limit the paths included in the collection.

[0067] After obtaining a set of similar paths, frequently occurring triplets in the structure are extracted and their frequency of occurrence in the set is analyzed using support statistics (such as Apriori or FP-Growth). The frequencies are normalized to similarity weights in the range [0, 1], which serve as weighting factors when constructing the intervention vector to control the degree of dominance of the triplet in subsequent structural adjustments. A higher frequency indicates that the structure is more frequently verified in similar situations, thus increasing the credibility of the intervention.

[0068] Based on this, a weighted interference superposition vector is constructed to guide the structural adjustment of the conflicting section of the target path.

[0069] During the structural adjustment operation, the weighted interference superposition vector is first applied to the conflicting segments of the target path. Specifically, based on the aforementioned structural comparison matrix, the segment position indices with all zero cell values ​​are extracted and aggregated to form a continuous set of conflicting segments. For each conflicting segment, its start and end positions in the target path are identified. The sequence of feature triples with the largest weight in the weighted interference superposition vector is selected to determine whether it is logically replaceable for the conflicting segment (e.g., identical intervention type, consistent symptom indicator category, and similar structural depth). Structural insertion is then performed based on "semantic adaptation rules." These adaptation rules include two types of constraints: indicator dimension consistency, which requires that all node indicator attributes of the replacement segment match those of the replaced segment; and causal convertibility, which requires that the logical chain before and after the inserted segment does not break, such as preventing the preceding intervention from occurring later than the subsequent symptom. After structural substitution is completed, the replacement segment overwrites the original conflicting segment, and the edge weights and sequence indexes in the causal chain of the disease process are updated to form a new corrected path. This operation not only changes the internal structure of the conflicting segment but also adjusts its logical connection within the overall chain, thus completing the path interference correction.

[0070] This correction mechanism is different from the traditional path deletion and modification mode. By using historical similar paths for "structural projection", it retains the original characteristics of the diagnostic path, while reducing the degree of conflict and enhancing logical coherence.

[0071] After the path correction is completed, based on the structural differences between the corrected segment and the original path, the changed diagnostic nodes are adjusted back to the earlier nodes in the disease causal chain. Specifically, the weight values ​​of these earlier nodes are adjusted to reflect the changes in their role in the final path.

[0072] Weight adjustments are based on factors such as whether a node is retained, replaced, or whether its function has shifted due to logical changes in the corrected path. This feedback mechanism not only maintains the logical consistency of the causal chain before and after path correction, but also creates a closed-loop structure, effectively preventing the risk of local corrections leading to global logical collapse.

[0073] S4: The corrected path is re-embedded into the causal chain of the disease process to construct a consistency verification diagram, and the path consistency score between the forward evolution chain and the reverse tracing chain is used as the screening basis, and the path with the highest score is extracted as the final diagnostic recommendation path.

[0074] First, the node sequence in the revised path is mapped to the corresponding intervention points and clinical event nodes in the original chain. After mapping, the revised path is re-implanted into the causal chain of the disease process as a set of updated nodes, constructing a composite graph structure called a consistency verification graph. This consistency verification graph structure has a bidirectional logical edge set, including forward reasoning edges and backward tracing edges. That is, the consistency verification graph includes a forward reasoning edge set from the starting node to the current diagnosis node, and a backward tracing edge set from the current diagnosis node to the earlier traceable node.

[0075] Two logical sub-chains are separated from the consistency verification diagram, namely the forward evolution chain and the reverse tracing chain. The forward evolution chain refers to the structural path that gradually advances from the starting node of the disease course to the terminal diagnosis node, which is mainly reflected in the evolution process of the target individual's condition from initial onset to high-level diagnosis; the reverse tracing chain traces forward from the current revised diagnosis node, and gradually traces back the intervention factors, previous judgments and potential antecedent diagnosis nodes through the path logic relationship, which is reflected in the reverse reasoning process of the diagnostic logic.

[0076] It should be noted that when constructing a reverse traceability chain structure, it is not used for generative abduction, but rather to verify causal symmetry with the forward chain. If the target path is a forward chain, the corresponding reverse high-frequency path in the historical sample can be obtained. The sequence similarity, causal direction consistency, and frequency support of the node pairs can be compared to verify the rationality of the structure.

[0077] This method uses the abductive patterns in existing data to complete a structural closed-loop assessment without performing direct reasoning operations, which can avoid the logical errors of "post hoc attribution" in medicine.

[0078] A consistency matching matrix is ​​constructed based on the node pairing of the forward evolution chain and the reverse tracing chain, and a weighted score is given to the node content, causal depth, and triggering timing differences to form a path consistency score.

[0079] Among them, the consistency matching matrix uses the node pairs of the forward chain and the reverse chain as the basic unit, and determines three key indicators for each paired unit: node content consistency (content consistency is calculated using cosine similarity), causal depth difference (measures whether the logical level distances of the two nodes in their respective causal chains are consistent) and trigger timing difference (evaluates the difference in occurrence time of the two nodes in actual clinical data).

[0080] Among them, node content consistency evaluates whether the two nodes match in diagnostic semantics, such as whether they belong to the same disease category and whether they are in the same intervention type; causal depth difference measures the depth span of the node in the cause-effect chain to evaluate the structural consistency of the path logic; trigger timing difference reflects whether the time sequence of the two nodes in the actual course of the disease conforms to the normal physiological progression.

[0081] Preferably, the calculation of the causal depth difference includes:

[0082] Let define the forward node The depth in the forward chain is Reverse Node The depth in the reverse chain is Assume the maximum allowable level difference is D max ,but:

[0083]

[0084] If two nodes are at the same or close depth, the score approaches 1; if the depth difference is large, it approaches 0, for example D max =6, then If one is the first visit and the other is the last visit, and the difference is 5, then The structural logic is inconsistent.

[0085] The three indicators are weighted and summed by setting weight factors to form a unit scoring matrix. The process uses a sliding window method to handle structurally misaligned node pairs to ensure the robustness of the scoring results.

[0086] The trigger timing difference may be calculated using a nonlinear attenuation function.

[0087] Ultimately, the consistency score is calculated by summing and averaging the scores of each unit in the node pairing matrix to form an overall path consistency score. A higher score indicates a greater degree of logical overlap and structural synergy between the forward evolution chain and the backward tracing chain, meaning that the path is not only logically coherent but also verifiable from a posteriori perspective.

[0088] By setting scoring and sorting rules, the highest-scoring path is selected from the set of corrected paths as the final diagnostic recommendation. This path undergoes multiple rounds of structural evaluation, graph structure matching, and bidirectional chain logic verification. Its recommendation value is superior to that of a single-way scoring path, embodying the integrated design concept of big data, structural evolution, multi-scale matching, and bidirectional verification.

[0089] Further, such as Figure 2 As shown, this embodiment also provides a medical diagnosis optimization system based on big data, including:

[0090] The disease course extraction module 100 is used to extract the symptom evolution trajectory and intervention record sequence of the target individual based on the historical diagnostic path contained in the medical big data, form a disease course causal chain, and construct a state superposition set;

[0091] A transition trigger module 200 is configured to trigger a collapse operation on the state superposition set when a symptom intensity indicator in the symptom evolution trajectory transitions and exceeds a preset transition threshold, so as to converge to a target diagnostic path with the highest diagnostic priority;

[0092] The path correction module 300 is used to perform a structural comparison between the target diagnostic path and the medical rule set item by item, calculate the conflict degree parameter, and if the conflict degree parameter is greater than the conflict degree tolerance threshold, correct the target diagnostic path based on the interference superposition of the candidate diagnostic paths to form a path interference state, thereby obtaining a corrected diagnostic path;

[0093] The consistency verification module 400 is used to re-embed the modified diagnostic path into the disease causal chain to construct a consistency verification diagram, and use the consistency score between the forward evolution chain and the reverse tracing chain as the screening basis to extract the diagnostic path with the highest score as the final diagnostic recommendation path.

[0094] This embodiment also provides a computer device suitable for the case of a medical diagnosis optimization method based on big data, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the medical diagnosis optimization method based on big data proposed in the above embodiment.

[0095] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0096] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for optimizing medical diagnosis based on big data proposed in the above embodiment is implemented.

[0097] In summary, the present invention can quickly focus on the target path with the highest diagnostic priority among multiple possible diagnostic paths based on the symptom intensity transition criteria and collapse mechanism, thereby improving the accuracy of the preliminary judgment; adopting structural comparison and conflict degree calculation means, the inconsistent parts between the diagnostic path and the medical rules are interfered and corrected, thereby avoiding misjudgments caused by logical deviations or local abnormalities; in addition, the present invention realizes two-way verification and screening of the overall structure of the diagnostic path through the construction of a consistency verification diagram and a forward and reverse path consistency scoring mechanism, making the output path more logically reasonable and clinically feasible. In summary, the present invention effectively integrates the diagnostic path deduction and structural optimization strategy driven by big data, and improves the accuracy, robustness and interpretability of the diagnostic process under complex conditions.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A medical diagnosis optimization method based on big data, characterized by: include: Based on the historical diagnostic paths contained in medical big data, the target individual's symptom evolution trajectory and intervention record sequence are extracted to form a causal chain of the disease process and construct a state superposition set; When the symptom intensity indicator in the symptom evolution trajectory transitions and exceeds the preset transition threshold, the collapse operation of the state superposition set is triggered, converging to the target diagnostic path with the highest diagnostic priority; The target diagnostic path and the medical rule set are structurally compared item by item, and the conflict degree parameter is calculated. If the conflict degree parameter is greater than the conflict degree tolerance threshold, the target diagnostic path is corrected based on the interference superposition of the candidate diagnostic paths to form a path interference state, and the corrected diagnostic path is obtained; The revised diagnostic pathway is re-embedded into the disease causal chain to construct a consistency verification diagram. The consistency score between the forward evolution chain and the reverse tracing chain is used as the screening basis, and the diagnostic pathway with the highest score is extracted as the final diagnostic recommendation pathway.

2. The method for optimizing medical diagnosis based on big data according to claim 1, wherein: The construction of the state superposition set includes: Based on the historical diagnosis and treatment data of the target individual, the causal chain of the disease course is extracted in chronological order, and the parameters of each node in the causal chain are extracted to form an intervention factor tensor through structural splicing; Using the disease course causal chain and intervention factor tensor as search keys, we match historical diagnostic pathway sets with similar structures in big data and calculate a matching score for each diagnostic pathway. The matching score is given by combining the overlap of the disease course causal chain sequence and the similarity of the intervention factor space projection. According to the matching scores corresponding to each diagnostic path, the superposition probability amplitude in the target state space is constructed, based on the diagnostic path set, and encoded into the state superposition set.

3. The big data-based medical diagnosis optimization method according to claim 2, wherein: The parameters of each node include the intervention intensity, time span and symptom impact level of each node.

4. The method for optimizing medical diagnosis based on big data according to claim 1, wherein: The triggering of the collapse operation on the state superposition set to converge to the target diagnosis path with the highest diagnosis priority includes: Perform reverse mapping on the path feature vectors in the state superposition set, calculate the proportion of the indicator that triggers the transition in each diagnostic path, and extract the subset of diagnostic paths whose proportion is greater than the component threshold; Normalize the probability amplitude of each diagnostic path in the diagnostic path subset to form a collapsed probability distribution; According to the normalized probability amplitude, the diagnostic path corresponding to the maximum probability amplitude is selected as the candidate diagnostic path, and the number of symptom dimensions covered by the candidate diagnostic path and the consistency of the symptom evolution trajectory are used as diagnostic priority determination factors; The diagnostic path with the largest diagnostic priority factor is selected from the candidate diagnostic paths as the final target diagnostic path.

5. The method for optimizing medical diagnosis based on big data according to claim 1, wherein: The calculation of the conflict degree parameter includes: Based on the diagnostic node sequence contained in the target diagnostic pathway and the structural labels in the medical rule set, a diagnostic pathway structural label sequence and a medical rule label sequence are constructed, and a structural comparison matrix is ​​established for the two sequences. The proportion of mismatched units in the structural comparison matrix is ​​used to represent the degree of structural conflict between the target diagnostic pathway and the medical rules.

6. The method for optimizing medical diagnosis based on big data according to claim 5, wherein: The modified diagnostic path includes: A similar diagnostic path set whose similarity to the symptom evolution trajectory of the target diagnostic path is higher than a similarity threshold is screened from the state superposition set, characteristic substructures and similarity weights are extracted, and a weighted interference superposition vector is constructed; The weighted interference superposition vector is applied to the conflicting section position of the target diagnostic path, and the structure is adjusted according to the position alignment and semantic adaptation rules to obtain the corrected diagnostic path.

7. The method for optimizing medical diagnosis based on big data according to claim 1, wherein: The screening criteria based on the consistency score between the forward evolution chain and the reverse tracing chain include: The consistency verification graph includes a forward reasoning edge set and a backward tracing edge set; Separate two logical sub-chains from the consistency verification graph, namely the forward evolution chain and the reverse tracing chain; A consistency matching matrix is ​​constructed based on the node pairing of the forward evolution chain and the reverse tracing chain, and a weighted score is given to the node content, causal depth, and triggering timing differences to form a path consistency score.

8. A big data-based medical diagnosis optimization system, based on the big data-based medical diagnosis optimization method according to any one of claims 1 to 7, characterized in that: Also includes: The disease course extraction module is used to extract the target individual's symptom evolution trajectory and intervention record sequence based on the historical diagnostic path contained in medical big data, form a disease course causal chain, and construct a state superposition set; A transition trigger module is used to trigger the collapse of the state superposition set when the symptom intensity indicator in the symptom evolution trajectory transitions and exceeds the preset transition threshold, converging to the target diagnostic path with the highest diagnostic priority; The path correction module is used to perform a structural comparison between the target diagnostic path and the medical rule set item by item, calculate the conflict degree parameter, and if the conflict degree parameter is greater than the conflict degree tolerance threshold, correct the target diagnostic path based on the interference superposition of the candidate diagnostic paths to form a path interference state, thereby obtaining a corrected diagnostic path; The consistency verification module is used to re-embed the revised diagnostic pathway into the disease causal chain to construct a consistency verification diagram, and use the consistency score between the forward evolution chain and the reverse tracing chain as the screening basis to extract the diagnostic pathway with the highest score as the final diagnostic recommendation pathway.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the big data-based medical diagnosis optimization method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the big data-based medical diagnosis optimization method according to any one of claims 1 to 7 are implemented.

Images