Chronic disease intervention path self-adaptive adjustment method based on intention semantic driving

Abstract

The invention discloses a chronic disease intervention path self-adaptive adjustment method based on intention semantic driving. According to the method, dynamic adjustment and optimization of an intervention path are realized by establishing a patient health intention model, collecting health data in real time, deducing and evaluating candidate intervention schemes, selecting a path most matched with a patient target for execution, and performing loop iteration according to continuous monitoring feedback. According to the method, the intention of the patient is taken as the core, so that chronic disease management is more intelligent and personalized, the accuracy and targeting of intervention are remarkably enhanced, and the treatment satisfaction degree of the patient is improved.

Description

Technical Field

[0001] This invention belongs to the fields of medical information technology and artificial intelligence, and specifically relates to an adaptive adjustment method for chronic disease intervention paths driven by intent semantics in the Data-Information-Knowledge-Wisdom-Intent (DIKWP) model. Background Technology

[0002] The management of chronic diseases typically requires continuous adjustments to intervention plans based on patients' treatment responses and life circumstances. However, most existing disease management processes rely on pre-defined guidelines and lack mechanisms for real-time adjustment based on individual goals, often resulting in over- or under-intervention. Especially when considering patients' diverse cultural backgrounds and personal preferences, incorporating their health intentions into the decision-making process becomes a significant challenge. Traditional linear decision-making models struggle to meet this need because they often adjust treatment solely based on biomedical indicators, failing to understand the patient's underlying intentions and values.

[0003] The DIKWP cognitive model, proposed in recent years, emphasizes the importance of intention-driven cognitive processing by adding a "Purpose" layer. This model allows decision-making systems to use the patient's ultimate goal as a guide when processing health data, achieving deep semantic customization. Professor Duan Yucong's research further points out that introducing intention-driven dynamic semantic feedback can effectively address incomplete and contradictory information in the cognitive process. For example, in medical decision-making, when data is incomplete, a decision that meets the goal can be directly generated from high-level intention (W→P), without waiting for the data to be complete; when information conflicts occur, credible information can be prioritized based on the patient's core goal (P→I), resolving the contradiction. These methods provide new ideas for path adjustment in chronic disease management.

[0004] Furthermore, medical practice shows that both over-intervention and under-intervention can harm patients with chronic diseases, and the best results often come from personalized adjustments to achieve a dynamic balance. The concept of "proactive medicine" advocates a middle ground of "neither excessive nor insufficient," meaning that the intensity of intervention should be carefully measured according to the patient's specific situation, avoiding unnecessary burdens while also preventing risks from escalating. This principle requires sophisticated path control algorithms to implement, but current systems lack effective means to quantify ethical principles into algorithms.

[0005] In light of the aforementioned issues, it is necessary to provide a novel approach: integrating the patient's health intention model into chronic disease intervention decision-making, and achieving adaptive optimization of the intervention path through semantic-level intention-driven mechanisms. This method should not only be able to adjust the sequence and intensity of treatment measures based on physiological feedback, but also ensure that each adjustment aligns with the patient's value preferences and ethical requirements, thereby achieving a balance between medical efficacy and humanistic care. Summary of the Invention

[0006] This invention proposes an adaptive adjustment method for chronic disease intervention pathways based on intent semantics. The core idea is to construct a personalized health intent semantic model for patients during chronic disease management, and then dynamically plan intervention steps based on this model. Through continuous monitoring and feedback, the system enables the intervention pathway to be optimized in real time according to changes in patient status and target deviations. The specific implementation is as follows:

[0007] Health Intent Modeling: A "Purpose Layer" semantic model is established for each chronic disease patient, incorporating their long-term health goals and personal preferences. This model uses structured semantic information to characterize the patient's needs and weights regarding treatment effectiveness, quality of life, and cultural beliefs. For example, a diabetic patient's intent model might define their goal as "controlling blood sugar while maintaining a normal life and work," and their preference as "prioritizing diet and exercise management, and minimizing medication dependence." Through questionnaires, doctor-patient communication, and analysis of historical behavioral data, this intent profile is systematically formed and serves as the highest criterion in subsequent decision-making.

[0008] Multi-objective intervention decision-making: In each decision-making cycle (e.g., weekly or monthly follow-up visits), the system aggregates the patient's current changes in various indicators (blood glucose, blood pressure, weight, etc.) and lifestyle data, inputting them into the data and information layer of the DIKWP model for processing. When an intervention plan needs to be formulated or adjusted, the intelligence layer (W) generates several feasible options, including different combinations of medication adjustments, exercise prescriptions, and dietary plans. Then, the system evaluates these options based on the intent model of the Purpose layer, calculating a comprehensive score for each plan in terms of medical effectiveness and patient preference. The scoring considers the guiding role of intention semantics (i.e., the patient's preferred direction); for example, if a patient explicitly refuses a certain therapy or type of food, the score of the corresponding plan will decrease significantly. Finally, the plan with the highest score is selected as the recommended path.

[0009] Pathway Sequence Planning: This invention not only focuses on single-step decision-making but also plans intervention pathway sequences over a period of time. It employs a rolling optimization approach, determining possible subsequent adjustment steps and conditions based on the expected effects of the current plan and the patient's intentions. For example, for situations requiring progressively stronger treatment, the system pre-plans a tiered approach (e.g., lifestyle intervention first, followed by medication if the target is not met), while setting monitoring milestones to assess whether to proceed to the next step. This planning process fully utilizes medical knowledge and empirical rules at the knowledge layer, combined with the patient's intentions, to ensure that every step of the pathway is justified and ethically sound. The intention layer provides macro-level guidance throughout the process, preventing drastic changes or deviations from the patient's vision.

[0010] Continuous Feedback Optimization: While executing the intervention path, the system continuously collects patient feedback data (including objective changes in health indicators and subjective feedback on comfort and compliance). Whenever a preset assessment node is reached, the system adjusts the path based on the actual results: if the patient responds well and is highly satisfied, the intervention continues as planned or even considers reducing its intensity early; if the effect is poor or side effects are significant, an intention-driven adjustment mechanism is triggered. This adjustment mechanism follows the principle of "intention priority" to find areas for improvement: for example, if blood glucose control is not up to standard and the patient has made every effort to implement lifestyle changes, the system will weigh the patient's resistance to increasing medication against the importance of blood glucose control, updating the target weight at the intention level (increasing health priority), thus allowing for the introduction of a medication-enhanced approach in the next step. Conversely, if the patient experiences adverse reactions or resistance, the intention level may lower the priority of certain measures and seek alternative solutions. Through this closed-loop feedback, the intervention path is not static but recalculates the optimal route in real time, like a navigation system, ensuring progress towards the ultimate health goal even when the target changes.

[0011] Handling Abnormal Situations: During pathway adjustment, in the event of an abnormal situation (such as a sudden deterioration of a patient's indicator or the emergence of new comorbidities), the system will immediately interrupt the original pathway execution and enter emergency decision-making mode. At this time, the Wisdom layer and the Purpose layer work together: the Wisdom layer retrieves emergency treatment plans related to the current abnormality from the knowledge base, while the Purpose layer re-prioritizes decisions based on the principle of prioritizing patient safety (default highest intention). The system may temporarily suspend previous long-term goals and switch to a short-term treatment pathway. After the crisis is resolved, an assessment will be made on how to smoothly transition back to the long-term management pathway. This flexible switching mechanism ensures that the system can both plan for the long term and respond to rapidly changing clinical situations, truly achieving robust adaptive adjustment.

[0012] Through the above methods, this invention enables refined and personalized intervention pathway management for patients with chronic diseases. While ensuring medical efficacy, it fully reflects the patient's personal will and ethical considerations, making the intervention plan "intentional and reasonable." Compared with traditional static plans, this method can detect efficacy trends earlier and proactively adjust them, avoiding disease deterioration due to delayed adjustments; it also prevents resource waste and patient burden caused by overtreatment. Its intention-driven semantic evaluation makes artificial intelligence decision-making more humane, and is expected to improve patient adherence to long-term management plans.

[0013] The method flow of the present invention is as follows: Figure 1 As shown, the main technical steps include the following:

[0014] Step 1: Patient Intent Semantic Initialization. Obtain basic patient information and treatment requirements. Clinicians and patients jointly define health goals and constraints, forming a "Patient Intent Semantic Description Document." Using Ontology or semantic modeling tools, transform this document into a formalized Purpose layer model, including key objectives (e.g., control range of indicators), preferences (e.g., preference for non-pharmacological therapies), and contraindications (e.g., rejection of certain components). The model is stored in a semantic database and can be used by subsequent algorithms.

[0015] Step 2: Status Monitoring and Information Update. The system collects patients' current health status data in real time through a combination of IoT medical devices and manual recording. The data layer performs time alignment, anomaly detection, and missing value imputation on the raw data. The information layer summarizes and generates a current status vector, such as "fasting blood glucose = 8.5 mmol / L (above target), average daily steps in the past week = 6000 (below target), subjective fatigue level = moderate," etc. This information is compared and analyzed with previous data to output a "state deviation set."

[0016] Step 3: Multiple Option Generation and Evaluation. The knowledge layer, based on the state bias set and medical knowledge base, infers several alternative intervention options. Each option consists of a set of specific measures (e.g., adjusting breakfast diet + increasing medication dosage + arranging psychological counseling), along with estimates of expected effects and risks. The Wisdom layer then performs semantic scoring on each option: it calls the Purpose layer model to evaluate the fit between each measure in the option and the patient's intention. For example, option A includes "increasing insulin dosage," but if the patient intends to reduce medication, this item lowers the intention-match score for option A; option B includes "daily meditation relaxation," which aligns with the patient's values ​​and raises the match score. Finally, the Wisdom layer combines the medical effect score and the intention-match score to calculate the total score of the options and selects the optimal option as the intervention decision for the current cycle.

[0017] Step 4: Intervention Pathway Update. The selected plan is linked to the previously implemented path to form a new intervention pathway. If the new plan conflicts with the original pathway (e.g., the original plan was to gradually reduce medication, but the new plan requires increasing medication), the conflict is arbitrated at the Purpose level, prioritizing the patient's core health goals. Pathway updates also consider time factors and dependencies: determining the implementation order, frequency, and duration of each intervention measure. For example, the new pathway might specify: "Implement a diet + exercise plan for the next 4 weeks, and recheck indicators after 2 weeks to determine whether to initiate the medication regimen." The system stores the updated pathway as the "Current Activity Pathway" and notifies the relevant implementation units.

[0018] Step 5: Path Implementation and Tracking. The system initiates the intervention measures according to the plan, prompts the patient via the app, and collects implementation data. The feedback module monitors the changing trends of key indicators and compares them with predicted values. When the actual results deviate significantly from the prediction or the patient experiences difficulty in implementation, the triggering conditions are recorded immediately.

[0019] Step 6: Feedback Evaluation and Next Round Decision. Periodically or when triggered, return to Step 2 to re-enter the next decision cycle. Specifically, when a significant anomaly is detected, the emergency strategy library is invoked for immediate handling before returning to the regular process. In each loop, the knowledge layer continuously learns and updates from new data: this includes adjusting patient intent model parameters (e.g., reducing the priority of a recommendation in future plans if a patient repeatedly fails to follow it), and expanding the medical knowledge base (e.g., updating best practices based on newly published clinical studies). This self-learning mechanism ensures that algorithm performance improves over time.

[0020] Through continuous iteration of the above steps, this method can ensure that the chronic disease intervention pathway is always consistent with the patient's current condition and ultimate goal, and can be continuously optimized based on feedback. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 An example structural diagram of a patient intent model is shown.

[0023] Figure 3 This is a diagram illustrating the dynamic adjustment of the intervention path. Detailed Implementation

[0024] The following example illustrates the specific application of the method of the present invention. Assume a 55-year-old type II diabetic patient, B, who also suffers from obesity and hypertension. Patient B's personal intention model sets their goals as "controlling blood sugar and blood pressure to target levels, and losing 10 kg," their preferences as "avoiding complex medication regimens as much as possible, and being willing to accept exercise and dietary adjustments," and their contraindications as "refusing medications containing porcine pancreatic enzymes."

[0025] In the initial phase, the system collected recent health data from patient B: fasting blood glucose 9.0 mmol / L, glycated hemoglobin 8.5%, systolic blood pressure 150 mmHg, and no significant weight loss compared to the previous month. Information layer analysis revealed the following deviation set: blood glucose 20% higher than target, blood pressure 10% higher, no weight improvement, and insufficient exercise. The knowledge layer generated three plans based on medical guidelines: - Plan 1: Increase the metformin dosage on the existing basis; maintain the original lifestyle recommendations. - Plan 2: Introduce intensive insulin therapy; strengthen dietary control and aerobic exercise programs. - Plan 3: Do not add medication for now, strictly adhere to the Mediterranean diet + 10,000 steps daily; combine with acupuncture to assist in lowering blood sugar (patient prefers traditional Chinese medicine conservative therapy).

[0026] The Wisdom layer evaluates the intervention plans. In terms of medical efficacy, Plan 2 provides the fastest blood sugar control, but requires daily insulin injections; Plan 3 has the lowest drug burden, but its efficacy is uncertain. Regarding intent matching, the patient clearly prefers non-invasive therapy and may have poor adherence to Plan 2, while Plan 3 aligns with their values. Overall scoring results: Plan 3 is slightly better than Plan 1, while Plan 2 is ranked last due to its significant deviation from intent. Therefore, the system selects Plan 3 as the current intervention pathway.

[0027] Over the next four weeks, the system guided Patient B to implement Plan 3: a diet management app and a fitness tracker provided daily feedback. The patient adhered well to the lifestyle modifications, but a follow-up examination after four weeks revealed that fasting blood glucose remained above 8.0 mmol / L, showing only slight improvement, and weight loss was only 2 kg. The intention layer determined that the current plan had not achieved its goals and more aggressive measures were needed. However, considering the patient's diligent adherence to lifestyle interventions, resistance to medication intervention in their intention model may have decreased (due to personal experience that lifestyle interventions alone have limited effectiveness). Therefore, in the next round of decision-making, the system dynamically updated the patient's intention model—appropriately increasing the weighting of medication acceptance.

[0028] In the new decision-making cycle, the knowledge layer refers to the data from the previous stage to generate an improved plan: - Adjust oral medications (add a small dose of sulfonylureas) to assist in lowering blood sugar; - Continue to maintain the diet + exercise plan and add a weekly group exercise class to improve motivation; - Monitor the risk of hypoglycemia and follow up and adjust every two weeks.

[0029] The plan, assessed at the Wisdom level, balanced efficacy (medication could potentially achieve target blood glucose levels) with patient preference (maintaining the lowest possible dosage and increasing social and physical activity to improve acceptance). The Purpose level determined that the plan aligned with the patient's long-term goals and approved its implementation. Over the following months, patient B's blood glucose steadily decreased, and weight continued to decline. The system fine-tuned the exercise intensity and medication dosage based on feedback in each round, ultimately controlling fasting blood glucose below 7.0 mmol / L, with blood pressure and weight reaching target ranges.

[0030] This embodiment demonstrates how the method of the present invention continuously adjusts the intervention path with the patient's intention at its core: first, it tries the patient's most desired approach; if the effect is unsatisfactory, it gradually introduces more proactive interventions; and simultaneously, it updates the patient's intention preferences through feedback, adapting the decision-making strategy to changes in the patient's psychological state. Throughout the process, it ensures medical safety and effectiveness while respecting the patient's value choices, significantly improving treatment adherence and satisfaction. This adaptive path adjustment concept and its effects are highly consistent with the intention-driven semantic processing and feedback loops advocated in existing research.

Claims

1. An intention semantics-driven chronic disease intervention path self-adaptive adjustment method, characterized in that The method comprises the following steps: establishing a patient's individual health intention semantic model as a Purpose layer of cognitive decision-making, recording the patient's long-term health goals, preference weights and ethical constraints; collecting and updating the patient's health data in real time to generate an information description of the current health status; based on pre-constructed medical knowledge rules, deducing multiple candidate intervention schemes at the Knowledge / Wisdom layer, and performing multi-objective evaluation on each scheme in combination with the health intention semantic model to select a scheme that best fits the patient's goals as the current intervention path; executing the selected intervention scheme and continuously monitoring patient feedback, comparing the result data with the expected result; if a deviation in therapeutic effect is detected or an abnormal situation occurs, adjusting the patient's intention model or target priority through the Purpose layer according to the feedback, and returning to the step of regenerating the scheme for cyclic iteration to realize dynamic adjustment and optimization of the intervention path.

2. The method of claim 1, wherein: The multi-objective evaluation is completed by calculating the medical effect score and intention matching score of each candidate scheme, wherein the medical effect score estimates the improvement degree of the scheme on the key health indicators according to the clinical knowledge, and the intention matching score evaluates the compliance of the scheme in meeting the patient's preferences and ethical requirements according to the patient's health intention model, and finally the optimal scheme is selected by comprehensively considering the two types of scores.

3. The method of claim 1, wherein: When the patient's health data is incomplete or the information is in conflict, an intention-driven semantic reasoning strategy is used for decision completion or conflict resolution, including: in the case of data missing, missing information is inferred through semantic association or a decision scheme that meets the patient's goals is directly generated by the Wisdom layer; in the case of information conflict, the core target related to the intention is preferentially selected according to the Purpose layer, and the conflict information unrelated to the intention or secondary is filtered out; in the case of knowledge rule conflict, the reasoning weight is adjusted or the reasoning path is reconstructed through the Wisdom layer to seek a solution that meets the patient's overall goals.

4. The method of claim 1, wherein: A feedback mechanism is introduced to adaptively optimize the intervention path, and when the actual monitoring result deviates from the expected result, the path is re-planned by adjusting the patient's intention model and the knowledge base, wherein the judgment of over-intervention and under-intervention is based on the preset target interval and the doctrine of the mean, so that the new path after adjustment is more in line with the individual needs of the patient while ensuring safety.

5. The method of claim 1, wherein: When an emergency abnormal situation of the patient is detected, the current long-term intervention path is suspended, an emergency plan library is called to perform emergency treatment, and after the abnormality is resolved, the long-term intervention path is reconnected according to the updated health status and intention model of the patient, so that the coordination and unity of long-term health goal management and acute problem handling are considered.

Images