Prostate cancer life answer security control method with traceable evidence slices

Abstract

The application discloses a prostate cancer life question and answer safety control method of traceable evidence slices, and comprises the following steps: evidence structuring and traceability management, evidence retrieval enhancement generation, referee evaluation and safety control, and trace archiving and iterative updating. The application can improve the evidence-based consistency, checkability and safety of question and answer contents.

Description

Technical Field

[0001] This invention belongs to the fields of medical and health informatics and artificial intelligence technology, and more specifically, relates to a method for safe control of prostate cancer lifestyle questions based on traceable evidence slices. Background Technology

[0002] Prostate cancer patients often require continuous health education and management advice during treatment and follow-up, covering aspects such as diet and nutrition, exercise, weight management, psychological support, lifestyle habits, and adherence. While general-purpose language models possess strong interactive capabilities in health Q&A, they still exhibit significant limitations in disease-specific lifestyle management scenarios.

[0003] (1) The evidence is not traceable: the model answers often cannot be clearly matched with specific guidelines or document paragraphs, making it difficult for users to verify the source and reliability, and also not conducive to auditing and the delineation of responsibility boundaries.

[0004] (2) Risk of hallucination and inconsistency: When there is a lack of evidence constraints, content that is inconsistent with the evidence-based evidence, overly certain or potentially harmful may be generated, especially in patient self-management scenarios where there are safety risks.

[0005] (3) High-quality data is difficult to scale: Real doctor-patient dialogues are subject to privacy restrictions, expert annotation is costly, and it is difficult to quickly form patient-style question-and-answer training data that covers different topics and has diverse expressions.

[0006] (4) Assessment and risk control are difficult to close the loop: Traditionally, the process relies on manual experts to review each item, which is inefficient and subjective, making it difficult to form a reusable and scalable quality and safety control process.

[0007] Therefore, there is an urgent need for an implementable technical solution to structure evidence-based evidence and enable traceable management, enforce evidence binding during the question-and-answer generation stage, and improve the credibility and security of disease-specific lifestyle management questions and answers by structuring, quantifying and intercepting risks through referee evaluation and security control mechanisms. Summary of the Invention

[0008] The purpose of this invention is to provide a question-and-answer construction method and system for lifestyle management of prostate cancer. It constructs a traceable knowledge base of evidence slices (cleansing and structuring evidence materials, slicing them into the smallest semantic units and storing them quantitatively, and establishing metadata indexes such as source, location, topic, and version to achieve evidence traceability and version management). Based on evidence retrieval enhancement, it constructs patient-oriented question-and-answer systems (generating patient-oriented questions around lifestyle themes, retrieving relevant evidence slices and generating answers, and strongly binding question-and-answer records with evidence slice IDs and source metadata to form structured, reusable data assets). Furthermore, it introduces a referee evaluation and safety control closed loop (evaluating candidate answers from multiple dimensions under evidence constraints, determining unsafe based on rules or thresholds, triggering control strategies such as refusal to answer, prompting medical attention, downgrading output, and transferring to manual review, and archiving the evaluation results). This forms a technical route that ensures traceable evidence, controllable generation, and closed-loop evaluation, improving the evidence-based consistency, verifiability, and security of question-and-answer content, and solving the problems of untraceable evidence, high risk of hallucinations, difficulty in data scaling, and difficulty in closed-loop control of evaluation in existing technologies.

[0009] To address at least one of the aforementioned technical problems, according to one aspect of the present invention, a method for safety control of prostate cancer lifestyle questions based on traceable evidence slices is provided, comprising the following steps:

[0010] Evidence structuring and traceability management;

[0011] Enhanced evidence retrieval generation;

[0012] Referee assessment and safety control;

[0013] Archived records and iteratively updated.

[0014] Furthermore, the evidence structuring and traceability management includes:

[0015] Select relevant index texts on lifestyle management for prostate cancer;

[0016] Unified as a structured text object;

[0017] Semantic slicing is performed by dividing the data into the smallest semantic units, ensuring that each slice independently expresses a fact or conclusion that can be used to support the recommendations.

[0018] Vectorization and indexing: generate an embedding vector for each slice and write it to the vector database; create a slice_id for each slice and write it to metadata synchronously.

[0019] Furthermore, the metadata includes: source_id and original text location.

[0020] Furthermore, the enhanced evidence retrieval generation includes:

[0021] Generate common patient questions and add several rewrites to cover different wording;

[0022] Perform semantic retrieval on the question and return several pieces of evidence most relevant to the question;

[0023] The question and the retrieved evidence slices are input into the generative model, requiring the answers to be evidence-based, patient-oriented, and avoid prescription-like expressions.

[0024] Structured archiving outputs questions and answers as structured records, writes them to traceability fields, and records retrieval sorting information.

[0025] Furthermore, in the semantic retrieval of the question, the number of retrievals is a preset value, which is adjusted according to the results.

[0026] Furthermore, the traceability fields include: slice_id_list, source_id_list, and original text location.

[0027] Furthermore, the referee assessment and safety controls include:

[0028] Referee configuration involves specifying roles, dimensions, scoring anchors, and output JSON field requirements within the evaluation context; evidence slices are then provided to the referee model as evaluation criteria.

[0029] Multidimensional evaluation outputs scores for candidate responses across dimensions of evidence-based consistency, relevance, understandability, feasibility, and risk / safety, along with brief explanations.

[0030] Unsafe determination and action: When there is a situation that seriously conflicts with the evidence or is a high-risk recommendation, mark it as unsafe and trigger control actions;

[0031] Traceability and iteration: judge_scores, unsafe_flag, and control_action are written into the question and answer records for subsequent quality analysis and rule iteration.

[0032] According to another aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the prostate cancer lifestyle question-and-answer security control method for traceable evidence slices of the present invention.

[0033] According to another aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the prostate cancer lifestyle question-and-answer security control method for traceable evidence slices of the present invention.

[0034] Compared with existing technologies, the beneficial effects of the above-described method of the present invention are as follows:

[0035] The evidence in this invention is traceable and verifiable. By binding slice_id with metadata index, any question and answer can locate the source of the evidence and the original text location, thereby improving credibility and auditability.

[0036] This invention can reduce illusions and improve consistency by injecting Top-k evidence slices during the generation stage, making the answers closer to evidence-based evidence and reducing unfounded outputs.

[0037] This invention enables data assetization and iterative processing, binding and archiving questions and answers with evidence in a structured manner, facilitating continuous expansion, version management, and reuse in model training and evaluation.

[0038] This invention provides safe, controllable, and scalable assessment. The referee assessment structures the quality evaluation, and the unsafe judgment triggers control strategies, reducing potential harm and minimizing manual review costs.

[0039] This invention is general and scalable, without limiting specific models or platforms, and its technical approach can be extended to other disease or health management topics. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments will be briefly described below. Obviously, the drawings described below only relate to some embodiments of the present invention and are not intended to limit the present invention.

[0041] Figure 1 This is a schematic diagram of the overall system of a preferred embodiment of the present invention;

[0042] Figure 2 This is a flowchart illustrating the method route and safety closed-loop of a preferred embodiment of the present invention; Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention.

[0044] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0045] Example 1:

[0046] like Figures 1-2As shown, this invention provides a method for safety control of prostate cancer lifestyle questions based on traceable evidence slides, comprising the following steps:

[0047] Evidence structuring and traceability management;

[0048] Enhanced evidence retrieval generation;

[0049] Referee assessment and safety control;

[0050] Archived records and iteratively updated.

[0051] Evidence structuring and traceability management includes:

[0052] Evidence knowledge base construction:

[0053] (1) Sources of evidence: Select textual materials such as guidelines, systematic reviews or clinical studies related to lifestyle management of prostate cancer.

[0054] (2) Cleaning and structuring: Remove irrelevant content such as author and publication information and unify it into structured text objects (e.g., JSON lists in paragraph form).

[0055] (3) Semantic slicing: Slice according to the smallest semantic unit to ensure that each slice can independently express a fact or conclusion that can be used to support the recommendation.

[0056] (4) Vectorization and indexing: Generate an embedding vector for each slice and write it to the vector database; create a slice_id for each slice and synchronously write metadata such as source_id and original text position.

[0057] Enhanced evidence retrieval generation, including:

[0058] Patient-centric question-and-answer construction and evidence binding:

[0059] (1) Question design: Based on the themes of "exercise intensity and frequency", "dietary structure and nutritional supplementation", "weight control", "psychological stress management" and "follow-up and compliance", generate common questions from patients and supplement them with several rewritten versions to cover different wording.

[0060] (2) Evidence retrieval: Perform semantic retrieval on the question and return several pieces of evidence most relevant to the question; the number of retrievals is a preset value and can be adjusted according to the effect.

[0061] (3) Generate answers: Input the questions and retrieved evidence slices into the generation model. The answers should follow the evidence, be patient-oriented, and avoid prescription-like expressions.

[0062] (4) Structured archiving: Output the questions and answers as structured records, write trace fields such as slice_id_list / source_id_list / original text position, and record the search sorting information.

[0063] Archiving and iterative updates include:

[0064] Referee assessment and unsafe security controls;

[0065] (1) Referee configuration: Write the role, dimension, scoring anchor and output JSON field requirements in the evaluation context; provide the evidence slices as the evaluation basis to the referee model.

[0066] (2) Multidimensional evaluation: Output scores for candidate responses in dimensions such as evidence consistency, relevance, understandability, feasibility, and risk / safety, and provide brief explanations.

[0067] (3) Unsafe judgment and action: When there is a serious conflict with the evidence or a high-risk suggestion, mark it as unsafe and trigger control actions; for example, refuse to answer and suggest medical treatment, or downgrade to a general suggestion and provide the source of evidence, or transfer to manual review.

[0068] (4) Traceability and iteration: Write judge_scores, unsafe_flag and control_action into the question and answer record for subsequent quality analysis and rule iteration.

[0069] 4. Optional deformation methods

[0070] (1) Vector database, embedding model, generative model and referee model can all be replaced with functionally equivalent implementations, without limiting the specific names.

[0071] (2) Evaluation dimensions can be added or adjusted according to the theme; differentiated control strategies can also be adopted according to the unsafe category.

[0072] (3) Manual sampling and automatic rule verification can be added to continuously improve the quality of data and output.

[0073] This study emphasizes the importance of data-driven approaches to improve the accuracy of medical imaging-assisted diagnostic systems in identifying rare and novel lesions.

[0074] This invention can be implemented as a software system running on a computer device, and the system includes at least the following functional modules:

[0075] 1. Evidence Acquisition and Cleaning Module: Acquire literature, guidelines, or evidence-based materials related to prostate cancer lifestyle management, clean the text, remove irrelevant content (such as author information, DOI, copyright notices, etc.), and unify it into a structured format (e.g., JSON).

[0076] 2. Semantic Slicing and Vectorized Storage Module: The cleaned text is divided into several evidence slices (chunks) according to the smallest semantic unit. An embedding vector is generated for each slice and written to the vector database to support semantic retrieval.

[0077] 3. Metadata Extraction and Traceability Index Module (Core): Assigns a unique identifier (slice_id) to each evidence slice and records metadata such as source identifier (source_id), document type, subject tag, original text position (chapter / paragraph / sentence or character offset), cleaning version number, and entry time, to achieve evidence traceability and version management.

[0078] 4. Patient-oriented question construction module: Generates basic patient-oriented questions around themes such as diet, exercise, weight management, psychological support, and adherence, and generates several rewritten expressions to enhance data coverage and language diversity.

[0079] 5. Retrieval Enhancement Generation Module: Performs vector retrieval on patient-oriented questions and returns a set of relevant evidence slices; inputs "question + evidence slice content (or its summary)" into the generation model according to a unified prompt template to obtain candidate answers.

[0080] 6. Structured Output and Evidence Binding Archiving Module (Core): Saves the question-and-answer results as structured records (e.g., structured dialogue JSON / parsable structured records), and writes a list of evidence slice IDs, source metadata references, and retrieval sorting into the records, forming auditable and reusable data assets.

[0081] 7. Referee Evaluation and Safety Control Module (Core Framework Protection Point): Constructs the referee evaluation context (including role definition, evaluation dimensions, scoring criteria, and structured output constraints), evaluates candidate answers from multiple dimensions under evidence constraints, and triggers control strategies based on unsafe judgment rules.

[0082] Any generated answer must be able to trace back to the source and location of the evidence through slice_id and source_id; the evaluation phase is not just about scoring, but about triggering output control actions based on the evaluation results, thus forming a closed loop.

[0083] Referee Model Evaluation Framework and Unsafe Security Controls:

[0084] 1. Evaluation input: includes at least question Q, candidate answer A, a set of evidence slices associated with the answer (or the content corresponding to slice_id), and a set of preset evaluation rules.

[0085] 2. Assessment Context Construction: Inject evidence slices into the assessment context and write constraints such as the judge role (e.g., "evidence-based medicine reviewer / patient education reviewer"), dimension definition, scoring anchor, and output format (e.g., JSON field) to reduce the arbitrariness of the assessment output.

[0086] 3. Multidimensional evaluation: This should include at least the following: consistency with evidence (whether it is consistent with the evidence and whether key information is missing), relevance (whether it answers the question), comprehensibility (patient readability), feasibility (whether the recommendations are actionable), and risk / safety (whether there is potential harm or misleading information). Each dimension can be scored from 1 to 5 points or other discrete ratings, with a brief explanation provided.

[0087] 4. Unsafe criteria: including but not limited to: serious inconsistency with evidence; giving advice that may directly harm the patient; inducing the patient to adjust treatment or medication without authorization; giving strong definitive conclusions about high-risk behaviors without providing medical advice; providing diagnostic / prescription content beyond the scope of lifestyle management, etc.

[0088] 5. Output Control Strategy: When an answer is deemed unsafe, execute preset actions, such as: refusing to answer and prompting the user to consult a doctor; downgrading the answer to general lifestyle advice and only providing the source of evidence; triggering manual review; or requesting additional information before answering. Control actions and reasons are recorded in a structured log for easy auditing and iteration.

[0089] Traceability data structure and trace fields:

[0090] Structured output and evidence binding archiving module: Saves the question and answer results as parsable structured records, and binds at least the evidence slice identifier and its source information (such as slice_id / source_id and corresponding position index) to support reverse tracing and verification of the answer content.

[0091] Referee evaluation and safety control module: The multi-dimensional evaluation results, unsafe judgments and triggered control actions are linked and archived with the question and answer record for subsequent auditing, statistics and iterative optimization.

[0092] Example 2:

[0093] The computer-readable storage medium of this embodiment stores a computer program that, when executed by a processor, implements the steps in the prostate cancer lifestyle question-and-answer safety control method for traceable evidence slices of Embodiment 1.

[0094] The computer-readable storage medium in this embodiment can be an internal storage unit of the terminal, such as the terminal's hard disk or memory; the computer-readable storage medium in this embodiment can also be an external storage device of the terminal, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc. equipped on the terminal; furthermore, the computer-readable storage medium can include both the terminal's internal storage unit and external storage devices.

[0095] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0096] Example 3:

[0097] The computer device of this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the prostate cancer lifestyle question-and-answer safety control method of traceable evidence slices in Embodiment 1.

[0098] In this embodiment, the processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The memory can include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0099] Those skilled in the art will understand that the content disclosed in the embodiments can be provided as a method, system, or computer program product. Therefore, this solution can take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this solution can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage) containing computer-usable program code.

[0100] This solution is described with reference to flowchart illustrations and / or block diagrams of methods and computer program products according to embodiments of this solution. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the flowchart illustrations and / or block diagrams.Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0101] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0102] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0103] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0104] The examples described herein are merely preferred embodiments of the invention and are not intended to limit the concept and scope of the invention. Any modifications and improvements made by those skilled in the art to the technical solutions of the invention without departing from the design concept of the invention should fall within the protection scope of the invention.

[0105] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the specific embodiments described above. The specific embodiments and descriptions in the specification are merely for further illustrating the principles of the invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the claims and their equivalents.

Claims

1. A method for safety control of prostate cancer lifestyle questions based on traceable evidence slides, characterized in that, Includes the following steps: Evidence structuring and traceability management; Enhanced evidence retrieval generation; Referee assessment and safety control; Archived records and iteratively updated.

2. The method as described in claim 1, characterized in that, The structured and traceable management of evidence includes: Select relevant index texts on lifestyle management for prostate cancer; Unified as a structured text object; Semantic slicing is performed by dividing the data into the smallest semantic units, ensuring that each slice independently expresses a fact or conclusion that can be used to support the recommendations. Vectorization and indexing: generate an embedding vector for each slice and write it to the vector database; create a slice_id for each slice and write it to metadata synchronously.

3. The method as described in claim 2, characterized in that, The metadata includes: source_id and original text location.

4. The method as described in claim 3, characterized in that, The enhanced evidence retrieval generation includes: Generate common patient questions and add several rewrites to cover different wording; Perform semantic retrieval on the question and return several pieces of evidence most relevant to the question; The question and the retrieved evidence slices are input into the generative model, requiring the answers to be evidence-based, patient-oriented, and avoid prescription-like expressions. Structured archiving outputs questions and answers as structured records, writes them to traceability fields, and records retrieval sorting information.

5. The method as described in claim 4, characterized in that: In the semantic retrieval of the question, the number of searches is a preset value, which is adjusted according to the results.

6. The method as described in claim 5, characterized in that, The traceability fields include: slice_id_list, source_id_list, and original text location.

7. The method as described in claim 6, characterized in that, The referee evaluation and security controls include: Referee configuration involves specifying roles, dimensions, scoring anchors, and output JSON field requirements within the evaluation context; evidence slices are then provided to the referee model as evaluation criteria. Multidimensional evaluation outputs scores for candidate responses across dimensions of evidence-based consistency, relevance, understandability, feasibility, and risk / safety, along with brief explanations. Unsafe determination and action: When there is a situation that seriously conflicts with the evidence or is a high-risk recommendation, mark it as unsafe and trigger control actions; Traceability and iteration: judge_scores, unsafe_flag, and control_action are written into the question and answer records for subsequent quality analysis and rule iteration.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the steps in the prostate cancer lifestyle question-and-answer safety control method for traceable evidence slices as described in any one of claims 1 to 7.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the prostate cancer lifestyle question-and-answer safety control method for traceable evidence slices as described in any one of claims 1 to 7.

Images