Method and computing device for generating treatment plans based on transdermal drug delivery

By combining a large language model and a medical knowledge base, TCM diagnosis and treatment plans are generated, which solves the problem of relying on physician experience in existing technologies, realizes full-process automation and precision of transdermal drug delivery technology, and improves the standardization and replicability of treatment.

CN123120188APending Publication Date: 2026-11-17HUAXIA CHENGHUANG (BEIJING) HEALTH TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611339427.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing transdermal drug delivery technologies rely on physician experience, with subjective acupoint selection, lack of quantitative parameters, and lack of objective indicators for efficacy evaluation, making it difficult to achieve a seamless process from patient information to an executable plan for applying the treatment patch.

Method used

By using a large language model to perform semantic understanding and field information extraction on consultation information and examination reports, structured disease feature description information is generated. Combined with a pre-built medical knowledge base, TCM diagnosis information is generated, and the acupoints to be applied, drug type and dosage, and application plate operating parameters are determined, realizing the fully automated generation of the entire process from patient information to application treatment plate.

Benefits of technology

It achieves fully automated generation of the entire process from patient information input to application of the treatment plate, reducing reliance on the experience of highly skilled physicians, improving the transparency of acupoint selection and the safety of drug combinations, and ensuring the standardization and precision of treatment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120188A_ABST
    Figure CN123120188A_ABST
Patent Text Reader

Abstract

This invention discloses a method, computing device, and storage medium for generating treatment plans based on transdermal drug delivery. The method includes: using a large language model to perform semantic understanding and field information extraction processing on patient questionnaire data, outputting symptom feature description information; generating TCM diagnostic information based on the symptom feature description information and individual patient information, combined with a pre-constructed medical knowledge base; determining the acupoints and number to be applied, the type and dosage of the drug, and the operating parameters of the application plate using the TCM diagnostic information; and converting the acupoints and number to be applied, the drug type and dosage, and the operating parameters of the application plate into structured data to generate a treatment plan. This invention solves the problems of existing transdermal drug delivery technologies, such as reliance on physician experience, subjective acupoint selection, lack of quantitative parameters, and lack of objective indicators for efficacy evaluation, achieving end-to-end intelligent and standardized generation from patient information input to an executable application treatment plan.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of medical technology, specifically to a method, computing device, and storage medium for generating treatment plans based on transdermal drug delivery. Background Technology

[0002] Transdermal drug delivery is a non-invasive method of drug administration that delivers medication into the body through the skin. A transdermal drug delivery device is the key equipment for implementing this technology, comprising a patch that adheres to the skin. The patch typically has a central cavity and an outer ring cavity. The outer ring cavity uses negative pressure to hold the patch to the skin surface, while the central cavity contains the drug and provides the necessary temperature and pressure to promote transdermal absorption.

[0003] Although transdermal drug delivery devices are relatively mature in terms of structure, their clinical application still heavily relies on the personal experience of physicians, leading to significant bottlenecks in promotion and standardization. Specifically: acupoint selection depends on the physician's diagnostic experience, resulting in high subjectivity and low standardization; key parameters such as negative pressure, temperature, duration, and drug dosage lack quantitative decision-making tools, making it difficult to achieve individualized and precise treatment; expert physicians proficient in this technology are scarce, hindering its widespread adoption; and efficacy evaluation relies on subjective feelings, lacking objective quantitative indicators.

[0004] Therefore, there is an urgent need in this field for a method that can reduce reliance on experience and achieve end-to-end intelligent generation of patient information into an executable solution for applying the treatment plate. Summary of the Invention

[0005] In view of the above problems, the present invention provides a method, computing device and storage medium for generating diagnosis and treatment plans based on transdermal drug delivery, so as to at least partially solve the above technical problems.

[0006] According to a first aspect of the present invention, a method for generating a treatment plan based on transdermal drug delivery is provided, executed in a computing device. The transdermal drug delivery method includes: loading a drug into a patch of a transdermal drug delivery device, and allowing the drug to be absorbed transdermally through the adhesion of the patch to the skin surface. The method includes: using a large language model to perform semantic understanding and field information extraction processing on patient questionnaire data, and outputting symptom feature description information; generating traditional Chinese medicine (TCM) diagnostic information based on the symptom feature description information and patient individual information, combined with a pre-constructed medical knowledge base; determining the acupoints and number to be applied, the drug type and dosage, and the patch operating parameters through the TCM diagnostic information; and converting the acupoints and number to be applied, the drug type and dosage, and the patch operating parameters into structured data to generate a treatment plan.

[0007] Optionally, the acupoints and their quantity to be applied are determined using the TCM diagnostic information, including: obtaining the meridian tropism rules corresponding to the syndrome and treatment principles in the TCM diagnostic information from the TCM knowledge graph; traversing the meridian knowledge base based on the meridian tropism rules to select acupoints that meet the meridian tropism conditions to generate a candidate acupoint set; and using a large language model to perform matching and selection in the candidate acupoint set according to the syndrome and treatment principles to determine the acupoints and their quantity to be applied, wherein the acupoints to be applied include main acupoints and auxiliary acupoints.

[0008] Optionally, the drug type is determined using the TCM diagnostic information, including: matching a first drug set from a pre-constructed syndrome-drug indication function mapping table based on the syndrome in the TCM diagnostic information; filtering a second drug set matching the selected acupoints and meridians from a pre-constructed acupoint-drug meridian association table based on the main acupoints and auxiliary acupoints; taking the intersection of the first and second drug sets as a candidate drug set; enumerating and traversing all combinations of all drugs in the candidate drug set and scoring them based on the TCM diagnostic information, acupoint information, patient constitution type, and age, and selecting the drug combination with the highest score as the initial drug combination; inputting the initial drug combination into the TCM compatibility contraindication rule library for verification, and if the initial drug combination contains any contraindication pairing, removing the combination and sequentially selecting the next highest-scoring combination until the contraindication verification is passed, thereby determining the final drug type.

[0009] Optionally, the dosage is determined using the TCM diagnostic information, including: determining a skin permeability correction coefficient based on skin characteristic data in the patient's individual information, which is positively correlated with skin thickness and stratum corneum density; determining an inflammation / aging degree correction coefficient based on inflammation and / or aging degree data in the patient's individual information, which is positively correlated with inflammation-related and / or aging-related indicator values ​​and increases in stages or continuously as the indicator values ​​increase; determining an individual difference correction coefficient based on the TCM constitution type and age in the patient's individual information, which is the product of the constitution correction coefficient and the age correction coefficient, wherein the constitution correction coefficient is determined according to the preset coefficient value corresponding to each constitution type, and the age correction coefficient is determined according to the preset coefficient value corresponding to the preset age range; and using the product of the baseline transdermal dose, skin permeability correction coefficient, inflammation / aging degree correction coefficient, and individual difference correction coefficient as the transdermal drug delivery dose, which is the product of the oral equivalent dose and the transdermal absorption rate correction coefficient.

[0010] Optionally, the operating parameters include at least negative pressure value, drug administration temperature, and treatment duration. The operating parameters of the application plate are determined based on the TCM diagnostic information, including: determining an initial negative pressure value and an initial drug administration temperature based on the tissue characteristics of the body surface area corresponding to the determined acupoint; if the initial negative pressure value meets a preset negative pressure value constraint, then the initial negative pressure value is used as the negative pressure value; otherwise, the negative pressure value constraint boundary value adjacent to the initial negative pressure value is used as the negative pressure value; updating the drug administration temperature based on the attribute information of the determined drug type; determining the initial treatment duration based on the drug administration temperature; if the cumulative heat dose value under the combination of the drug administration temperature and the initial treatment duration exceeds the cumulative heat dose safety threshold, then adjusting the treatment duration until the cumulative heat dose value does not exceed the cumulative heat dose safety threshold, and using the adjusted value as the treatment duration. The cumulative heat dose safety threshold is determined based on the tissue characteristics of the body surface area corresponding to the acupoint.

[0011] Optionally, the description of the disease characteristics includes at least one or more of the following: the site of inflammation, the type of inflammation, the degree of inflammation, accompanying symptoms, and age-related lesions and functional states.

[0012] Optionally, the examination data includes consultation information and examination reports, and utilizes a large language model to perform semantic understanding and field information extraction processing on the patient's examination data to output disease feature description information, including: using the large language model to perform semantic understanding on the consultation information, identifying and extracting symptom description fields; using the large language model to perform semantic understanding on the examination report, identifying and extracting test indicator numerical fields; and performing structured encoding on the extracted symptom description fields and test indicator numerical fields to generate disease feature description information.

[0013] Optionally, based on the symptom description information and patient individual information, combined with a pre-built medical knowledge base, TCM diagnostic information is generated, including: obtaining the constitution type and four diagnostic methods information from the patient individual information, the four diagnostic methods information including at least one or more of tongue appearance, pulse appearance, complexion, and voice; using the symptom description information, constitution type, and four diagnostic methods information as input conditions, matching a set of candidate syndrome types corresponding to the symptom location information and accompanying symptom information from a pre-built TCM knowledge graph, the TCM knowledge graph being constructed with symptom location, symptom, syndrome type, pathogenesis, and treatment principle as entities, and the relationships between entities as edges; filtering out candidate syndrome types incompatible with the constitution type and / or the four diagnostic methods information from the candidate syndrome type set to obtain TCM diagnostic information, the TCM diagnostic information including at least the syndrome type name, the pathogenesis analysis corresponding to the syndrome type, and the treatment principle and method corresponding to the pathogenesis analysis.

[0014] Optionally, the tissue characteristics of the body surface area include skin thickness, elasticity, and blood supply characteristics; the disease characteristic description information includes lesion depth, degree of inflammation; and, based on the tissue characteristics of the body surface area corresponding to the determined acupoint to be applied, determining the initial negative pressure value and the initial drug administration temperature includes: determining the initial negative pressure value based on the lesion depth, skin thickness, and elasticity; and determining the initial drug administration temperature based on the lesion depth, degree of inflammation, blood supply characteristics, and the attribute information of the determined drug type.

[0015] Optionally, the TCM diagnostic information includes inflammatory syndromes and aging syndromes. The inflammatory syndromes include at least one of the following: phlegm obstructing the lungs, damp-heat accumulation, and qi stagnation and blood stasis. The aging syndromes include at least one of the following: kidney essence deficiency, qi and blood deficiency, imbalance of the Chong and Ren meridians, and liver and kidney yin deficiency.

[0016] Optionally, it also includes: collecting the patient's objective physiological indicators during the treatment process based on the generated treatment plan; comparing the objective physiological indicators with the physiological indicators before treatment using a large language model; and adjusting the treatment plan based on the comparison results.

[0017] According to another aspect of the present invention, a computing device is provided, comprising: at least one processor; and a memory storing program instructions, wherein the program instructions are configured to be executed by the at least one processor, the program instructions including instructions for performing the methods described above.

[0018] According to another aspect of the present invention, a readable storage medium storing program instructions is provided, which, when read and executed by a computing device, causes the computing device to perform the method described above.

[0019] The technical solution provided by this invention has the following advantages compared with the prior art:

[0020] Existing transdermal drug delivery technologies rely on physician experience, are subjective in acupoint selection, lack quantitative parameters, and lack objective indicators for efficacy evaluation. Although some solutions have attempted to use artificial intelligence to assist in TCM diagnosis or acupoint recommendation, none of them have solved the problem of seamless integration from patient information to executable plans for application treatment plates, nor have they achieved the automatic generation of specific drug dosage and physical safety constraints for transdermal drug delivery scenarios.

[0021] This invention uses a large language model to perform semantic understanding and field information extraction on consultation information and examination reports, outputting structured descriptions of disease features. Based on this, it generates TCM diagnostic information by combining it with a pre-built medical knowledge base. Then, it uses this diagnostic information to determine the acupoints to be applied, the type of medicine and the transdermal dosage, as well as the operating parameters of the application plate. Finally, all the information is integrated into a structured treatment plan that can be directly sent to the transdermal drug delivery device for execution. This achieves fully automated generation of the entire process from patient information input to the executable plan of the application treatment plate, significantly reducing the reliance of this technology on the experience of highly skilled physicians.

[0022] Based on this, the present invention adopts a two-stage acupoint selection mechanism of syndrome type and treatment principle → meridian tropism rule → candidate acupoint set → large language model matching and screening, so that the acupoint selection process has traceable and verifiable intermediate evidence, thereby improving the transparency and clinical acceptability of acupoint selection.

[0023] By using a two-way matching of syndrome type and drug, combination enumeration scoring, and verification of the rules of incompatibilities of traditional Chinese medicine such as the Eighteen Incompatibilities and Nineteen Antagonisms, an objective and safe screening of drug combinations has been achieved.

[0024] By establishing a dedicated conversion model for oral equivalent dose to transdermal dose (with the baseline transdermal dose adjusted for skin permeability, inflammation / aging degree, and individual differences), the technical challenge of the long-standing lack of quantitative evidence for transdermal doses has been solved.

[0025] By embedding three physical safety constraints—negative pressure safety zone cutoff, CEM43 thermal dose cumulative verification, and drug-temperature negative feedback suppression—in the process of determining the operating parameters of the dressing tray, equipment-related safety accidents are eliminated from the root of the algorithm.

[0026] Furthermore, by collecting objective physiological indicators before and after treatment and comparing the efficacy with a large language model, dynamic closed-loop optimization of the treatment plan was achieved. The entire plan, from data collection, diagnostic reasoning, and plan generation to equipment execution and efficacy feedback, is seamlessly integrated, effectively improving the standardization, treatment precision, and clinical reproducibility of transdermal drug delivery technology.

[0027] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0028] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0029] Figure 1 A block diagram of the physical components (i.e., hardware) of a computing device is shown;

[0030] Figure 2 A flowchart of a treatment plan generation method 200 based on transdermal drug delivery according to an embodiment of the present invention is shown;

[0031] Figure 3 A schematic diagram of a treatment plan generation device 300 based on transdermal drug delivery according to an embodiment of the present invention is shown. Detailed Implementation

[0032] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0033] Transdermal drug delivery is a non-invasive method of drug administration that delivers medication into the body through the skin to achieve local or systemic treatment. A transdermal drug delivery device is the key equipment for this technology, containing a patch (also called a controlled-release patch or treatment patch) that adheres to the skin surface. During treatment, the patch is placed on specific areas or acupoints on the body. The pressure generated by its internal structure or other physical actions promotes drug penetration and absorption through the skin, thereby exerting a therapeutic effect.

[0034] Transdermal drug delivery devices are typically designed with a central cavity and an outer ring cavity that are isolated from each other. During treatment, the outer ring cavity is connected to a negative pressure source, and the resulting negative pressure suction force slightly lifts the underlying skin tissue, thus allowing the entire patch to adhere stably to the skin surface. This suction ensures the seal between the central cavity and the skin contact area, preventing the drug contained within the central cavity from leaking into the external environment, thereby improving drug utilization and treatment safety. Furthermore, positive pressure and temperature control can be combined to create a high-temperature drug delivery environment with positive pressure sealing on the skin surface. Under the action of positive and negative pressure, an expanding and orderly continuous drug channel is formed, allowing the active ingredients to penetrate the stratum corneum and reach deep into the tissue, thereby forming a high-concentration drug reservoir around the target tissue.

[0035] The structure of the transdermal drug delivery device (the central cavity and outer ring cavity of the dressing plate, negative pressure adsorption sealing, positive and negative pressure and temperature control to promote penetration) has been disclosed in many existing technologies and is the technical basis of this invention.

[0036] Although transdermal drug delivery devices are relatively mature in terms of structure and physical permeation enhancement, the current clinical application of this technology still heavily relies on the personal experience of senior physicians, resulting in significant bottlenecks in its promotion and standardization, specifically in the following aspects.

[0037] First, the selection of treatment sites (especially acupoints in Traditional Chinese Medicine) relies heavily on experience. Acupoint selection requires a complex theory of syndrome differentiation and treatment, demanding that the practitioner possess a solid foundation in Traditional Chinese Medicine theory and rich clinical experience. Different physicians have varying understandings of the same disease, leading to highly subjective and low-standardized acupoint selection schemes, making it difficult to consistently reproduce treatment effects.

[0038] Second, key treatment parameters lack quantitative decision support. Parameters such as negative pressure, treatment duration, drug dosage and ratio need to be dynamically adjusted according to the patient's individual condition (physical condition, age, skin characteristics, disease stage), but currently there is a lack of quantitative decision tools. Parameter setting mainly relies on the physician's personal experience and repeated trials, making it difficult to achieve precise and optimized treatment for each individual.

[0039] Third, the scarcity of high-quality medical resources restricts the dissemination of technology to lower levels. The training period for expert physicians who are proficient in this technology is long and costly, making it difficult to effectively implement the technology in primary healthcare institutions or resource-scarce areas, thus limiting its reach.

[0040] Fourth, the evaluation of treatment efficacy lacks objective quantitative indicators. Treatment efficacy evaluation relies heavily on patients' subjective feelings and physicians' clinical observations, lacking objective and quantitative physiological indicators for real-time monitoring and efficacy verification. This hinders timely adjustments to treatment plans and objective comparisons of treatment outcomes.

[0041] To address the problems existing in the prior art, this invention proposes a solution. This application proposes a method for generating treatment plans based on transdermal drug delivery. This method, executed in a computing device, achieves end-to-end intelligent and standardized generation of an executable treatment plan from patient information input to the application treatment plate using a large language model. First, the method utilizes a large language model to perform semantic understanding and field information extraction on the patient's questionnaire data to obtain structured disease feature descriptions. Then, based on these disease feature descriptions and individual patient information, combined with a pre-built medical knowledge base, traditional Chinese medicine diagnostic information is inferred. Next, based on this diagnostic information, the acupoints to be applied, the type and dosage of the medication, and the operating parameters of the application plate are determined. Finally, all information is integrated into structured data output, forming a customized treatment plan that can be directly sent to the transdermal drug delivery device.

[0042] The treatment plan generation method based on transdermal drug delivery can be executed on a computing device. This computing device can be a server, such as an application server or web server; it can also be a desktop computer, laptop, processor chip, tablet computer, etc., but is not limited to these. Figure 1 A block diagram of the physical components (i.e., hardware) of a computing device is shown. In a basic configuration, the computing device includes at least one processing unit and system memory. According to one aspect, depending on the configuration and type of the computing device, the system memory includes, but is not limited to, volatile memory (e.g., random access memory), non-volatile memory (e.g., read-only memory), flash memory, or any combination of such memory.

[0043] According to one aspect, system memory includes the operating system. System memory also includes program modules. According to one aspect, the operating system, for example, is suitable for controlling the operation of a computing device. Furthermore, examples are practiced in conjunction with graphics libraries, other operating systems, or any other applications, and are not limited to any particular application or system. Figure 1 The basic configuration is illustrated by the components within the dotted lines. According to one aspect, the computing device has additional features or functions. For example, according to one aspect, the computing device includes additional data storage devices (removable and / or non-removable), such as disks, optical discs, or magnetic tapes. This additional storage... Figure 1 The image shows removable storage devices and non-removable storage devices.

[0044] As stated above, according to one aspect, program modules are stored in system memory. According to one aspect, the program modules can be implemented as one or more computer program products, and this application does not limit the type of computer program product. In some embodiments according to this application, computer programs / instructions related to the transdermal drug delivery-based treatment plan generation method are encapsulated as a computer program product, which, when executed by a processor (i.e., a processing unit), implements the transdermal drug delivery-based treatment plan generation method according to this application.

[0045] According to one aspect, examples can be practiced on circuits including discrete electronic components, packaged or integrated electronic chips containing logic gates, circuits utilizing microprocessors, or on a single chip containing electronic components or a microprocessor. For example, it can be practiced via wherein... Figure 1Each or many of the components shown can be implemented as an example by integrating a System-on-a-Chip (SOC) on a single integrated circuit. According to one aspect, such an SOC device may include one or more processing units, graphics units, communication units, system virtualization units, and various application functions, all integrated (or “burned in”) onto a chip substrate as a single integrated circuit. When operating via the SOC, the functions described herein can be operated via dedicated logic integrated on a single integrated circuit (chip) with other components of the computing device. Other techniques capable of performing logical operations can also be used to practice embodiments of the invention; these other techniques include, but are not limited to, mechanical, optical, fluid, and quantum technologies. Additionally, embodiments of the invention can be practiced within a general-purpose computer or in any other circuit or system.

[0046] According to one aspect, the computing device may also have one or more input devices, such as a keyboard, mouse, pen, voice input device, touch input device, etc. It may also include output devices, such as a monitor, speaker, printer, etc. The foregoing devices are examples and other devices may also be used. The computing device may include one or more communication connections that allow communication with other computing devices, such as transdermal drug delivery devices or cloud platform servers. Examples of suitable communication connections include, but are not limited to: RF transmitters, receivers, and / or transceiver circuitry; Universal Serial Bus (USB); parallel and / or serial ports.

[0047] As used herein, the term computer-readable medium includes computer storage medium. Computer storage medium can include volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information (e.g., computer-readable instructions, data structures, or program modules). System memory, removable storage devices, and non-removable storage devices are all examples of computer storage media (i.e., memory storage). Computer storage media can include random access memory (RAM), read-only memory (ROM), electrically erasable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital universal disc (DVD) or other optical storage, magnetic tape, magnetic tape, disk storage or other magnetic storage devices, or any other article of manufacture that can be used to store information and is accessible by a computing device. According to one aspect, any such computer storage medium can be part of a computing device. Computer storage media does not include carrier waves or other transmitted data signals.

[0048] According to one aspect, a communication medium is implemented by computer-readable instructions, data structures, program modules, or other data in a modulated data signal (e.g., a carrier wave or other transmission mechanism), and includes any information transmission medium. According to one aspect, the term modulated data signal describes a signal having one or more sets of features or altered in a manner that encodes information in the signal. By way of example and not limitation, a communication medium includes wired media such as wired networks or direct wired connections, and wireless media such as acoustic, radio frequency (RF), infrared, and other wireless media.

[0049] In an embodiment of the invention, a computing device is configured to execute a transdermal drug delivery-based treatment plan generation method according to the invention. The computing device includes one or more processors and one or more readable storage media storing program instructions, which, when executed by the one or more processors, cause the computing device to execute the transdermal drug delivery-based treatment plan generation method of the present invention.

[0050] Figure 2 A flowchart of a treatment plan generation method 200 based on transdermal drug delivery according to an embodiment of the present invention is shown. Method 200 is adapted to be executed in a computing device (e.g., the aforementioned computing device).

[0051] like Figure 2 As shown, the purpose of method 200 is to realize a fully automated and intelligent generation method that inputs multi-source diagnosis and treatment information from patients to outputs executable treatment plans from the patch treatment plate.

[0052] Method 200 includes:

[0053] Step S202: Using a large language model, perform semantic understanding and field information extraction processing on the patient's question data, and output disease feature description information;

[0054] Step S204: Based on the description of the disease characteristics and the patient's individual information, combined with the pre-built medical knowledge base, generate TCM diagnosis information;

[0055] Step S206: Based on the TCM diagnosis information, determine the acupoints and number to be applied, the type and dosage of the medicine, and the operating parameters of the application plate;

[0056] Step S208: Convert the acupoints and quantity to be applied, the drug formulation and dosage, and the operating parameters of the application plate into structured data to generate a treatment plan.

[0057] This disclosed method for generating treatment plans based on transdermal drug delivery utilizes a large language model to perform semantic understanding and field information extraction from consultation information and examination reports, outputting structured descriptions of disease characteristics to ensure a solid data foundation for subsequent syndrome differentiation. Through explicit lesion location → meridian tropism → acupoint selection, combined with constraints from a traditional Chinese medicine knowledge graph, the acupoint selection process has traceable and verifiable intermediate evidence. By establishing a dose conversion model for transdermal drug delivery, the method achieves scientific quantification from equivalent oral doses to transdermal drug delivery doses. By embedding physical safety constraints in the treatment parameter determination process, the method eliminates equipment-related safety accidents at the algorithmic level. Finally, by directly sending structured data to the transdermal drug delivery device for execution, the method achieves fully automated generation from patient information input to an executable treatment plan applied to the treatment plate.

[0058] Compared to existing technologies that rely on physicians' personal experience for diagnosis, acupoint selection, drug formulation, and parameter setting, this disclosure fully utilizes the semantic understanding and reasoning capabilities of large language models and the structured knowledge expression capabilities of knowledge graphs. It achieves full-chain automation from patient information collection to structured treatment plan output, significantly reducing the dependence of transdermal drug delivery technology on high-level physician experience and improving the reproducibility and accuracy of treatment.

[0059] The steps of the treatment plan generation method based on transdermal drug delivery disclosed herein will now be described in detail using an exemplary process.

[0060] Step S202: Using a large language model, perform semantic understanding and field information extraction processing on the patient's question data, and output disease feature description information.

[0061] In this step, we first collect multi-source / multi-type medical data from the patient, including basic patient information, consultation information, and examination reports.

[0062] Patient basic information includes age, gender, constitution type (e.g., the nine constitution types in Traditional Chinese Medicine), past medical history, allergy history, and skin characteristics. Consultation information includes chief complaint, symptom description (location, nature, severity, triggers, duration), and information from the four diagnostic methods of Traditional Chinese Medicine (tongue appearance, pulse, complexion, and voice).

[0063] The examination report includes imaging reports, laboratory test results (especially inflammation-related indicators such as C-reactive protein, interleukin-6, tumor necrosis factor-α, erythrocyte sedimentation rate, etc.; as well as aging-related indicators such as sex hormone levels, bone mineral density, oxidative stress markers, etc.), and pathology reports.

[0064] Data collection is completed through client applications. Specifically, patients fill in structured forms or describe their chief complaints and symptoms in natural language through mobile devices or web pages, and import test reports through interfaces to the hospital information system (HIS) or laboratory information management system (LIS).

[0065] For natural language text, entity recognition and standardization are performed using a large language model, mapping free descriptions such as cough, sputum, shortness of breath, and wheezing into structured symptom fields; inflammatory indicators (C-reactive protein, interleukin-6, tumor necrosis factor-α, erythrocyte sedimentation rate, etc.) and aging indicators (sex hormones, bone mineral density, oxidative stress markers, etc.) in the examination report are extracted and quantified.

[0066] In one specific implementation, patients submit the following consultation information via a mobile application:

[0067] I've had a persistent cough for the past three months, especially when I wake up in the morning. The phlegm is white and quite thick. I feel short of breath and have a tightness in my chest when I walk quickly.

[0068] By performing semantic understanding on the natural language description using a large language model, the following structured symptom fields were extracted: symptom name "cough", nature "recurrent", location "lungs", trigger "worsened in the morning", accompanying symptoms "coughing up sputum (white and viscous), shortness of breath, chest tightness", and duration "three months".

[0069] Meanwhile, the patient's examination reports were imported from the hospital information system via an interface: pulmonary function test showed "obstructive ventilatory dysfunction", chest CT report showed "bilateral interstitial changes", laboratory test showed C-reactive protein 25 mg / L (normal reference value <10 mg / L) and interleukin-6 18 pg / mL (normal reference value <7 pg / mL).

[0070] In another specific implementation, a 45-year-old female patient submitted the following consultation information:

[0071] My menstrual flow has been decreasing, and for the past six months I have been experiencing frequent bouts of hot flashes and sweating. I also have trouble sleeping at night, lower back pain, weak legs, and dry skin.

[0072] At this point, the extracted structured symptom fields are: symptom name "scanty menstruation", accompanying symptoms "hot flashes, night sweats, insomnia, lower back and knee pain, dry skin", and duration of illness "six months". Simultaneously, imported test reports show: estradiol (E2) level is 15 pg / mL (normal reference range: 30-120 pg / mL during the follicular phase), follicle-stimulating hormone (FSH) is 45 mIU / mL (normal reference range: 4-15 mIU / mL during the follicular phase), and bone mineral density T-score is -2.0 (indicating osteopenia).

[0073] The data acquisition module encodes unstructured natural language consultation information and semi-structured examination reports into structured input that can be processed by the large language model. Specifically, in a preferred embodiment, all the collected information is assembled into a structured prompt word template, which includes three parts: "patient basic information", "chief complaint and symptoms (structured fields)" and "examination report (quantitative indicators)", as the input context for the large language model.

[0074] Then, a large language model is used to perform semantic understanding and field information extraction on the structured input. The large language model can be a commercial model such as GPT-4, Claude 3, Wenxin Yiyan, or Tongyi Qianwen, or an open-source model fine-tuned for the medical field (such as ChatGLM-6B or Baichuan-13B). In a preferred embodiment, a large language model enhanced with knowledge from the field of Traditional Chinese Medicine is used, and prompt word engineering guides the model to output structured data in a fixed format.

[0075] By sending prompt words in the following format to the large language model:

[0076] You are a TCM transdermal drug delivery expert. Based on the following patient information, please extract the characteristic descriptions of the symptoms and output them in JSON format. Output fields must include: location of inflammation, type of inflammation (acute / subacute / chronic), degree of inflammation (mild / moderate / severe), accompanying symptoms, location of aging lesions, and functional status due to aging.

[0077] The large language model outputs structured descriptions of disease features based on the input patient information. For the aforementioned example of chronic respiratory inflammation, the model outputs:

[0078] Location of inflammation: Lungs; Type of inflammation: Chronic; Severity of inflammation: Moderate to severe; Accompanying symptoms: Cough, expectoration of sputum (white and viscous), shortness of breath, chest tightness; Location of aging lesions: None; Functional status due to aging: Normal.

[0079] For the aforementioned example of ovarian insufficiency, the model output is:

[0080] Location of inflammation: None; Type of inflammation: None; Degree of inflammation: None; Accompanying symptoms: Scanty menstruation, hot flashes, night sweats, insomnia, lower back and knee pain, dry skin; Location of aging: Ovaries (Chong and Ren meridians); Functional state of aging: Decreased ovarian function.

[0081] The symptom description information includes at least the location of inflammation, type of inflammation, degree of inflammation, accompanying symptoms, and one or more of the following: age-related lesion locations and functional states. It is worth emphasizing that the explicit output of symptom description information in this step differs from the existing end-to-end black-box mapping of symptoms to acupoints. This invention explicitly outputs lesion location as an intermediate result, providing traceable and verifiable intermediate evidence for subsequent syndrome differentiation and acupoint selection. Physicians can verify the accuracy of the lesion location output by the model, and if errors are found, they can be corrected promptly, thereby improving the reliability and clinical acceptability of the entire system.

[0082] Step S204: Based on the description of disease characteristics and individual patient information, combined with a pre-built medical knowledge base, generate TCM diagnosis information.

[0083] In this step, based on the symptom feature description information output in step S102, and combined with the constitution type and four diagnostic information in the patient's individual information, relevant knowledge is retrieved from the pre-constructed TCM knowledge graph through retrieval enhancement generation (RAG) technology, and TCM diagnostic information is output after comprehensive reasoning by the large language model.

[0084] First, the patient's individual information is used to obtain their constitution type and diagnostic information (including the four diagnostic methods). The constitution type includes the nine traditional Chinese medicine constitution classifications (balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-dampness constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution, and special constitution). The four diagnostic methods include at least one or more of the following: tongue appearance, pulse appearance, complexion, and voice. This information has already been filled out by the patient or entered by the physician during the data collection phase in step S102.

[0085] In a preferred embodiment, patient information is obtained as follows: When a patient first uses the system, they complete a Traditional Chinese Medicine (TCM) constitution identification questionnaire. The system automatically determines the patient's constitution type based on the questionnaire results. For example, a patient with excessive phlegm, obesity, oily face, and a thick, greasy tongue coating may be identified as having a phlegm-dampness constitution; a patient who is thin, has a dry mouth and throat, and hot palms and soles may be identified as having a yin deficiency constitution. The four diagnostic methods (palp. 1, 2, 3, 4) are collected and entered into the system by the attending physician during the consultation, such as a pale red tongue, a white, greasy tongue coating, and a slippery, rapid pulse.

[0086] Then, using the description of disease characteristics (including disease location information and accompanying symptom information), constitution type and four diagnostic information as input conditions, a set of candidate syndrome types is matched from the pre-constructed TCM knowledge graph.

[0087] The Traditional Chinese Medicine (TCM) knowledge graph is one of the core foundational data of this invention. This knowledge graph is constructed using disease location, symptoms, syndrome types, pathogenesis, and treatment principles as entities, and the relationships between these entities as edges. Specifically, the basic unit of the knowledge graph is a triple, namely <entity 1, relation, entity 2>.

[0088] For example: <Phlegm-dampness obstructing the lungs syndrome, main symptoms: cough and phlegm>, <Phlegm-dampness obstructing the lungs syndrome, location of the disease: lung>, <Feishu acupoint, meridian tropism: Bladder meridian of Foot Taiyang>, <Kidney essence deficiency syndrome, treatment principle: tonify the kidney and replenish essence>. The knowledge graph is constructed based on authoritative textbooks such as "Traditional Chinese Medicine Diagnostics", "Traditional Chinese Materia Medica", and "Acupuncture and Moxibustion" as well as the knowledge of clinical experts, and is completed through a combination of manual annotation and semi-automatic extraction.

[0089] In a preferred embodiment, the knowledge graph contains more than 5,000 entities and more than 30,000 relationships, covering the TCM syndrome differentiation and treatment knowledge of common internal medicine chronic diseases.

[0090] During the retrieval phase, the system employs Retrieval Enhancement Generation (RAG) technology: lesion location (e.g., lung) and accompanying symptoms (e.g., cough, sputum, shortness of breath) from the disease feature description information are used as query vectors. Candidate syndrome types related to these lesion locations and symptoms are then retrieved from the vector index of the knowledge graph. The vector index is implemented by encoding entities and relationships in the knowledge graph into dense vectors and building a FAISS index, supporting millisecond-level semantic retrieval.

[0091] For the aforementioned example of chronic respiratory inflammation, the system used the location of the disease (lungs) and symptoms (cough, sputum, shortness of breath, chest tightness) as search criteria to retrieve the following candidate syndromes from the knowledge graph: phlegm-dampness obstructing the lungs, phlegm-heat accumulating in the lungs, lung qi deficiency, and lung yin deficiency. Further, considering the patient's constitution (assuming the patient has a phlegm-dampness constitution) and the four diagnostic methods (white, greasy tongue coating, slippery and rapid pulse), syndromes incompatible with the constitution and diagnostic information were excluded. The phlegm-dampness constitution highly matched the phlegm-dampness obstructing the lungs syndrome; the white, greasy tongue coating indicated excessive phlegm-dampness; and the slippery and rapid pulse indicated phlegm-heat or phlegm-dampness. After comprehensive judgment, phlegm-heat accumulating in the lungs syndrome (inconsistent with a white, greasy tongue coating) and lung yin deficiency syndrome (inconsistent with a phlegm-dampness constitution) were eliminated, while phlegm-dampness obstructing the lungs syndrome and lung qi deficiency syndrome were retained as candidates. Subsequently, the large language model performed a final screening based on the completeness of the pathogenesis analysis, outputting the following TCM diagnostic information:

[0092] Syndrome name: Phlegm-dampness obstructing the lungs; Pathogenesis analysis: Phlegm-dampness obstructing the lungs, causing the lungs to lose their ability to disperse and descend; Treatment principle and method: Resolve phlegm and turbidity, disperse lung qi and relieve asthma.

[0093] For the aforementioned example of ovarian hypofunction, the system used the following search criteria: location of the disease: ovary (Chong and Ren meridians); symptoms: scanty menstruation, hot flashes, night sweats, and lower back and knee weakness. The system retrieved the following candidate syndrome types: Kidney Essence Deficiency Syndrome, Kidney Yin Deficiency Syndrome, Chong and Ren Meridian Imbalance Syndrome, and Qi and Blood Deficiency Syndrome. Combining the patient's constitution type (assumed to be Yin deficiency) and the four diagnostic methods (red tongue with little coating, thready and rapid pulse), syndrome types incompatible with the constitution and diagnostic information were filtered out. Kidney Essence Deficiency Syndrome and Kidney Yin Deficiency Syndrome highly matched the Yin deficiency constitution; a red tongue with little coating indicated Yin deficiency with internal heat; and a thready and rapid pulse indicated Yin and Blood deficiency with concurrent deficiency heat. After comprehensive judgment, the final output was:

[0094] Syndrome name: Kidney essence deficiency syndrome; Pathogenesis analysis: Insufficient kidney essence, malnourishment of Chong and Ren meridians; Treatment principle and method: Tonify the kidney and replenish essence, regulate Chong and Ren meridians.

[0095] In this step, the generation of TCM diagnostic information relies on the organic combination of the reasoning ability of the pre-constructed TCM knowledge graph and the large language model: the knowledge graph provides structured TCM domain knowledge (the association between disease location, symptoms, syndrome type, pathogenesis, and treatment principles), ensuring that the reasoning results do not deviate from the TCM theoretical framework; the large language model provides semantic understanding and flexible reasoning ability for unstructured input (the description of the four diagnostic methods in the patient's individual information). The two complement each other through retrieval-enhanced generation (RAG) technology.

[0096] Step S206: Based on the TCM diagnosis information, determine the acupoints and number to be applied, the type and dosage of the medicine, and the operating parameters of the application plate.

[0097] This step comprises three parallel subtasks: acupoint identification, medication determination, and operational parameter determination. These are described in detail below.

[0098] 1. Identify acupoints.

[0099] The goal of acupoint selection is to choose the most suitable combination of acupoints for the current patient based on the TCM diagnostic information (syndrome type, pathogenesis, and treatment principle) obtained in step S204.

[0100] In a preferred embodiment, the acupoints are determined using a two-stage method: first, a set of candidate acupoints is generated based on rules, and then a large language model is used for refined screening.

[0101] The first stage involves acquiring the meridian affiliation rules corresponding to the syndrome types and treatment principles from the TCM diagnostic information, based on the TCM knowledge graph. TCM theory holds that each meridian has its specific pathway and therapeutic area, and acupoint selection must follow the principle of "where the meridian passes, where the meridian's function extends." For example, lung diseases typically involve selecting the back-shu points of the Lung Meridian of Hand-Taiyin and the Bladder Meridian of Foot-Taiyang; kidney-related diseases typically involve selecting acupoints of the Kidney Meridian of Foot-Shaoyin and the Ren Meridian.

[0102] The knowledge graph stores the mapping relationship between syndrome types and meridian tropism. For example:

[0103] Phlegm-dampness obstructing the lungs syndrome → Hand Taiyin Lung Meridian, Foot Taiyang Bladder Meridian

[0104] Kidney essence deficiency syndrome → Foot Shaoyin Kidney Meridian, Ren Meridian, Foot Taiyin Spleen Meridian

[0105] Based on the aforementioned meridian tropism rules, the system traverses the meridian and acupoint knowledge base, filters out acupoints that meet the meridian tropism criteria, and generates a candidate acupoint set. The meridian and acupoint knowledge base contains detailed information on 361 acupoints of the fourteen meridians of the human body, including acupoint name, meridian tropism, location, and indications. For example, for phlegm-dampness obstructing the lungs syndrome, the system filters out 11 acupoints from the Lung Meridian of Hand-Taiyin, including Zhongfu, Yunmen, Tianfu, Xiabai, Chize, Kongzui, Lieque, Jingqu, Taiyuan, Yuji, and Shaoshang, as well as back-shu points from the Bladder Meridian of Foot-Taiyang, such as Fengmen, Feishu, Jueyinshu, Xinshu, Dushu, and Geshu, which together constitute the candidate acupoint set.

[0106] In the second stage, the large language model performs a combination screening of acupoints from the candidate acupoint set based on syndrome differentiation and treatment principles. In this embodiment, the following prompt words can be sent to the large language model:

[0107] You are a Traditional Chinese Medicine acupuncture expert. The patient's syndrome is phlegm-dampness obstructing the lungs, and the treatment principle is to resolve phlegm, clear turbidity, and relieve asthma. Please select the main acupoint (targeting the affected organ) and auxiliary acupoints (modified based on syndrome differentiation) from the following candidate acupoints, following the principle of monarch, minister, assistant, and guide acupoint pairing, and explain the rationale for your selection. Candidate acupoints: [Zhongfu, Yunmen, Chize, Lieque, Taiyuan, Feishu, Fengmen, Tanzhong, Tiantu, Fenglong, Zusanli].

[0108] Large language model output:

[0109] Main acupoints: Feishu (Bladder Meridian of Foot Taiyang, Back-Shu point of the Lung, directly regulates lung function), Tanzhong (Ren Meridian, meeting point of Qi, relieves chest tightness and regulates Qi);

[0110] Acupoints: Tian Tu (Ren Meridian, for clearing the lungs and relieving cough), Feng Long (Stomach Meridian of Foot Yangming, a key acupoint for resolving phlegm), Zu San Li (Stomach Meridian of Foot Yangming, for strengthening the spleen and resolving phlegm);

[0111] Reasons for selecting acupoints: Lung Shu (BL13) and Tan Zhong (CV17) are the main acupoints, which can promote lung function and regulate Qi; Tian Tu (CV22) stops cough and resolves phlegm; Feng Long (ST40) is a key acupoint for resolving phlegm; Zu San Li (ST36) strengthens the spleen to prevent the source of phlegm production.

[0112] Simultaneously, the model's output acupoint names are associated with the surface location information in the acupoint knowledge base, outputting a list of acupoints to be applied, including acupoint names, meridian affiliations, and surface location coordinates (or standard anatomical location descriptions). For the example above, the output is:

[0113] Feishu (primary acupoint), Bladder Meridian of Foot Taiyang, 1.5 cun lateral to the spinous process of the 3rd thoracic vertebra; Tanzhong (primary acupoint), Ren Meridian, anterior midline, level with the 4th intercostal space; Tiantu (secondary acupoint), Ren Meridian, midline of the suprasternal notch; Fenglong (secondary acupoint), Stomach Meridian of Foot Yangming, 8 cun above the tip of the lateral malleolus, 1 cun lateral to Tiaokou; Zusanli (secondary acupoint), Stomach Meridian of Foot Yangming, 3 cun below Dubi, one finger-width lateral to the anterior crest of the tibia.

[0114] The main acupoints are selected for the affected organ (lung), and the auxiliary acupoints are selected according to the syndrome differentiation (phlegm-resolving, spleen-strengthening). There are 5 acupoints to be treated, and 5 corresponding treatment plates (one treatment plate for each acupoint).

[0115] For the example of ovarian hypofunction, the acupoint scheme output by the large language model is:

[0116] Main acupoints: Shenshu (Bladder Meridian of Foot Taiyang, Back-Shu point of the Kidney, tonifies the Kidney and benefits essence), Guanyuan (Ren Meridian, location of Dantian, warms the Kidney and strengthens Yang).

[0117] Acupuncture points: Sanyinjiao (Spleen Meridian of Foot-Taiyin, regulates Chong and Ren meridians), Zusanli (Stomach Meridian of Foot-Yangming, tonifies Qi and blood), Taixi (Kidney Meridian of Foot-Shaoyin, nourishes kidney Yin).

[0118] 2. Determination of drug type.

[0119] The goal of determining the drug type is to select the most suitable drug combination for the current patient from the drug library based on TCM diagnosis information and acupoint selection results.

[0120] In a preferred embodiment, the drug type is determined by a combination of multi-step screening and optimization scoring.

[0121] The first step is to match the first set of medicines from a pre-constructed syndrome-medicine indication mapping table based on the syndrome types in the TCM diagnostic information. This mapping table records the recommended medicines (single herbs or compound prescriptions) corresponding to each syndrome type. For example:

[0122] Phlegm-dampness obstructing the lungs syndrome → Pinellia ternata, Citrus reticulata peel, Poria cocos, Glycyrrhiza uralensis, Perilla frutescens seed, Sinapis alba seed, Raphanus sativus seed, etc.

[0123] Kidney essence deficiency syndrome → Rehmannia glutinosa, Cornus officinalis, Dioscorea opposita, Lycium barbarum, Cuscuta chinensis, Eucommia ulmoides, Dipsacus asper, etc.

[0124] For the example of phlegm obstructing the lungs, the first prescription is: [Pinellia ternata, Citrus reticulata peel, Poria cocos, Glycyrrhiza uralensis, Perilla frutescens seed, Sinapis alba seed, Raphanus sativus seed, Prunus armeniaca seed, Fritillaria thunbergii bulb, Trichosanthes kirilowii fruit].

[0125] The second step involves selecting a second set of medications that match the selected acupoints' meridian affinities from a pre-constructed acupoint-medicine meridian association table, based on the primary and secondary acupoints. This association table records recommended medications (medicines that enter that meridian) for each acupoint's meridian affinity. For example:

[0126] Hand Taiyin Lung Meridian → Ephedra, Apricot Kernel, Platycodon, Peucedanum, Aster, Coltsfoot Flower, Stemona, etc.

[0127] Foot Taiyang Bladder Meridian → Notopterygium root, Angelica pubescens root, Saposhnikovia root, Schizonepeta root, Ephedra root, etc.

[0128] Ren Meridian → Ginseng, Astragalus, Angelica sinensis, Rehmannia glutinosa, White peony root, etc.

[0129] Foot Shaoyin Kidney Meridian → Rehmannia glutinosa, Cornus officinalis, Lycium barbarum, Cuscuta chinensis, Eucommia ulmoides, etc.

[0130] For the aforementioned acupoint scheme (Feishu-Bladder Meridian, Tanzhong-Ren Meridian, Tiantu-Ren Meridian), the second set of medicines is: [Qianghuo, Duhuo, Fangfeng, Jingjie, Mahuang, Renshen, Huangqi, Danggui, Shudihuang, Baishao].

[0131] The third step is to take the intersection of the first and second drug sets as the candidate drug set. In the example above, the intersection is an empty set (the two drug sets have no common elements). If the intersection is too small (e.g., less than 3 drugs), the screening scope of the second drug set is expanded by extending the acupoint meridian affiliation to include both the acupoint meridian affiliation and adjacent meridians, thereby increasing the number of candidate drugs.

[0132] In some implementations, the generation of candidate drug sets is not limited to taking the intersection, but rather uses a weighted union method: the drug weight of the first drug set is 0.6, the drug weight of the second drug set is 0.4, and the drugs whose sum of weights is greater than 0.3 are selected as candidate drug sets. This avoids the problem of insufficient candidate drugs due to the intersection being too small.

[0133] The fourth step involves enumerating and scoring all combinations of all drugs in the candidate drug set, using TCM diagnostic information, acupoint information, patient constitution type, and age as constraints. Specifically, first, a power set (i.e., all possible combinations) of all drugs in the candidate drug set is generated. After excluding the empty set, a comprehensive score is calculated for each combination. The formula for calculating the comprehensive score is:

[0134] Overall Score = w1 × Syndrome Type Matching Degree + w2 × Acupoint Matching Degree + w3 × Constitution Matching Degree + w4 × Age Matching Degree

[0135] Among them, the syndrome matching degree measures the degree of suitability of each drug in the drug combination to the current syndrome (based on the confidence level in the syndrome-drug indication function mapping table); the acupoint matching degree measures the degree of meridian matching between each drug in the drug combination and the selected acupoints; the constitution matching degree measures the suitability of the drug combination to the patient's constitution type (e.g., patients with phlegm-dampness constitution should not use overly nourishing drugs); and the age matching degree measures the suitability of the drug combination to the patient's age group (e.g., elderly patients should use mild drugs and avoid strong drugs). w1, w2, w3, and w4 are preset weighting coefficients, which are set to 0.4, 0.3, 0.2, and 0.1 respectively in a preferred embodiment.

[0136] Iterate through all drug combinations and select the drug combination with the highest score as the initial drug combination. For example, for the phlegm-dampness obstructing the lungs syndrome example, suppose the candidate drug set is [Pinellia ternata, Citrus reticulata peel, Poria cocos, Glycyrrhiza uralensis, Perilla frutescens seed, Apricot kernel], the system enumerates all non-empty combinations (63 in total) and calculates the comprehensive score of each combination. Suppose the combination with the highest score is [Pinellia ternata, Citrus reticulata peel, Poria cocos, Perilla frutescens seed, Apricot kernel].

[0137] The fifth step is to input the initial drug combination into the Chinese herbal medicine incompatibilities rule database for verification. The Chinese herbal medicine incompatibilities rule database must contain at least one or more sets of incompatibilities pairing rules from the Eighteen Incompatibilities and Nineteen Antagonisms.

[0138] The Eighteen Incompatibilities include: licorice is incompatible with Euphorbia kansui, Euphorbia pekinensis, Sargassum, and Daphne genkwa; aconite is incompatible with Fritillaria cirrhosa, Trichosanthes kirilowii, Pinellia ternata, Ampelopsis japonica, and Bletilla striata; Veratrum nigrum is incompatible with ginseng, Adenophora stricta, Salvia miltiorrhiza, Scrophularia ningpoensis, Asarum heterotropoides, and Paeonia lactiflora.

[0139] The nineteen incompatibilities include: sulfur is incompatible with sodium sulfate, mercury is incompatible with arsenic, wolfsbane is incompatible with litharge, croton is incompatible with morning glory, clove is incompatible with turmeric, aconite and aconitum carmichaelii are incompatible with rhinoceros horn, potassium nitrate is incompatible with sparganium rhizome, cinnamon is incompatible with litharge, and ginseng is incompatible with trogopterus dung.

[0140] If the initial drug combination contains any incompatible pairing, that combination is discarded, and the next highest-scoring combination is selected sequentially until the incompatibility check is passed, thus determining the final drug type. For example, if the highest-scoring combination contains both Pinellia ternata and Aconitum carmichaelii (one of the eighteen incompatible pairs), the system will reject that combination and select the next highest-scoring combination, until a combination without any incompatible pairings is found.

[0141] 3. Determine the dosage of the medicine.

[0142] The goal of drug dosage determination is to calculate the specific dosage suitable for transdermal administration for each identified drug type.

[0143] In a preferred embodiment, the dosage of the agent is determined by the following method:

[0144] First, the skin permeability correction coefficient is determined based on the skin characteristic data in the patient's individual information;

[0145] Then, an inflammation / aging correction factor is determined based on the inflammation level data and / or aging level data in the patient's individual information;

[0146] Finally, individual difference correction coefficients were determined based on the patient's TCM constitution type and age in the individual information; the transdermal drug delivery dose was calculated as the product of the baseline transdermal dose, the skin permeability correction coefficient, the inflammation / aging degree correction coefficient, and the individual difference correction coefficient.

[0147] The skin permeability correction factor is determined based on the following: it is positively correlated with skin thickness and stratum corneum density. The thicker the skin or the denser the stratum corneum, the greater the resistance to transdermal drug absorption, and the higher the required dosage.

[0148] In a preferred embodiment, skin characteristic data is obtained by having the patient complete a skin characteristic questionnaire, including:

[0149] What is your skin type? (A. Dry B. Oily C. Normal D. Combination)

[0150] Is your skin prone to allergies? (Yes / No)

[0151] Do you have a history of skin diseases? (Yes / No, please specify if yes).

[0152] Skin permeability is comprehensively assessed based on questionnaire results, combined with factors such as age and gender.

[0153] In some implementations, a transdermal water loss (TEWL) meter is used to measure the patient's skin barrier function. A higher TEWL value indicates better skin permeability and a smaller correction factor. The skin permeability correction factor typically ranges from 0.5 to 2.0.

[0154] The inflammation / aging degree correction coefficient is determined based on the following: the inflammation / aging degree correction coefficient is positively correlated with the values ​​of inflammation-related indicators and / or aging-related indicators, and increases in stages or continuously as the indicator values ​​increase. The higher the degree of inflammation or the more severe the aging, the higher the demand for local blood supply and metabolism, and the greater the required drug dosage.

[0155] In some implementations, inflammation-related indicators (such as C-reactive protein and interleukin-6) and aging-related indicators (such as sex hormone levels and oxidative stress markers) are graded and mapped: a coefficient of 1.0 corresponds to an indicator value falling within the normal reference range, a coefficient of 1.2 to 1.5 corresponds to a value falling within the mildly abnormal range, and a coefficient of 1.5 to 2.0 corresponds to a value falling within the moderately to severely abnormal range. When multiple indicators are abnormal simultaneously, the maximum value or weighted average of the coefficients is taken.

[0156] The individual difference correction factor is determined based on the product of the constitution correction factor and the age correction factor. The constitution correction factor is determined according to the preset coefficient values ​​corresponding to each constitution type. For example, the correction factor for a balanced constitution is 1.0; for Qi deficiency, it is 1.1; for Yang deficiency, it is 1.15; for Yin deficiency, it is 0.9 (Yin deficiency patients tend to have drier skin, which may result in slightly lower absorption); for phlegm-dampness, it is 1.2 (phlegm-dampness patients have oilier skin, which may affect drug absorption); for damp-heat, it is 1.1; for blood stasis, it is 1.05; for Qi stagnation, it is 1.0; and for special constitution (allergic constitution), it is 0.85 (the dosage needs to be reduced to decrease the risk of allergies). The age correction factor is determined according to the preset coefficient values ​​corresponding to the preset age range. For example, 18-40 years old is 1.0; 41-60 years old is 1.1; 61-75 years old is 1.2; and over 75 years old is 1.3. As people age, their skin barrier function weakens, but their ability to metabolize drugs also declines, resulting in a slight increase in the overall coefficient.

[0157] The baseline transdermal dose is the product of the oral equivalent dose and the transdermal absorption correction factor. The oral equivalent dose refers to the standard dose of the drug for the same syndrome type administered orally in traditional Chinese medicine. For example, the oral equivalent dose of a certain expectorant Chinese medicine is 6g / dose. The transdermal absorption correction factor reflects the difference in bioavailability between transdermal and oral administration of the drug. Its value is determined by in vitro transdermal experiments or the ratio of transdermal absorption rate to oral bioavailability reported in the literature. For example, if the oral bioavailability of a drug is 100% and the transdermal bioavailability is 23%, then the transdermal absorption correction factor is 100% / 23%≈4.35, and the baseline transdermal dose is 6g×4.35≈26.1g (that is, approximately 26.1g of drug needs to be absorbed through the skin to achieve the same therapeutic effect as 6g of oral administration). The specific value of the transdermal absorption rate correction coefficient varies depending on the type of drug. In a preferred embodiment, the computing device has a built-in lookup table of transdermal absorption rate correction coefficients for common Chinese medicines. For drugs not included in the table, the user is prompted to enter the value or is advised to conduct an in vitro transdermal test.

[0158] Final transdermal dose = baseline transdermal dose × skin permeability correction factor × inflammation / aging degree correction factor × individual difference correction factor.

[0159] For example, consider a patient with phlegm-dampness obstructing the lungs. The baseline transdermal dose of the herb Pinellia ternata is 20 mg. This patient has neutral skin characteristics (skin permeability correction factor 1.0), C-reactive protein of 25 mg / L (mildly abnormal, inflammation / aging correction factor 1.3), a constitution type of phlegm-dampness (constitution correction factor 1.2), and an age of 52 years (age correction factor 1.1, individual difference correction factor = 1.2 × 1.1 = 1.32). Therefore, the transdermal dose of Pinellia ternata = 20 mg × 1.0 × 1.3 × 1.32 ≈ 34.3 mg.

[0160] The computing device calculates the transdermal dosage for each selected drug and outputs the drug regimen, including: drug type, single dose, ratio, and preparation parameters of the drug-impregnated film / pad (such as drug concentration, drug volume, etc.).

[0161] 4. Determine the operating parameters of the dressing tray.

[0162] The goal of determining the operating parameters of the dressing plate is to generate the various physical parameters that the transdermal drug delivery device needs to be set in actual treatment. The operating parameters include at least the negative pressure value, drug delivery temperature, and treatment duration.

[0163] In a preferred embodiment, the determination of operating parameters follows the process of initialization → constraint verification → output.

[0164] First, based on the tissue characteristics of the body surface area corresponding to the identified acupoints to be treated, the initial negative pressure value and the initial drug administration temperature are determined.

[0165] In this step, the tissue characteristics of the body surface area include skin thickness, elasticity, and blood supply. Disease characteristic description information includes lesion depth and degree of inflammation.

[0166] The initial negative pressure value is determined based on the depth of the lesion, skin thickness, and elasticity. When the lesion is deep (such as a deep lung lesion) or the body surface tissue is dense (such as thick back muscles), a larger negative pressure is required to form an effective drug delivery channel; when the lesion is shallow or the body surface tissue is thin (such as the suprasternal notch or abdomen), a smaller negative pressure is required to avoid discomfort or damage.

[0167] In a preferred embodiment, the initial negative pressure value is determined according to the following rules: 15 kPa for lesion depth ≤ 1 cm (e.g., epidermis, subcutaneous tissue); 25 kPa for lesion depth 1-3 cm (e.g., muscle layer); and 35 kPa for lesion depth > 3 cm (e.g., deep thoracic or abdominal cavity). This value is further modified based on the tissue characteristics of the acupoint area: the negative pressure increases by 2 kPa for every 1 mm increase in skin thickness; and decreases by 5 kPa for poor skin elasticity (e.g., loose skin in elderly patients).

[0168] The initial drug administration temperature is determined based on the depth of the lesion, the degree of inflammation, the blood supply characteristics, and the properties of the identified drug type. For deeper lesions or more severe inflammation, the temperature should be appropriately increased to enhance drug penetration; for areas with abundant blood supply, temperature should be increased cautiously to avoid excessive vasodilation.

[0169] In a preferred embodiment, the initial administration temperature is determined according to the following rules: the base temperature is set at 38°C; the temperature is increased by 1°C for mild inflammation and by 2°C for moderate to severe inflammation; the temperature is increased by 0.5°C for every 1cm increase in lesion depth; and the temperature is decreased by 1°C for areas with rich blood supply (such as the face and neck). The adjustment of temperature based on drug properties will be completed in subsequent steps.

[0170] Next, the initial negative pressure value is constrained and verified. A safe negative pressure range is determined based on the skin sensitivity level in the skin characteristic data. The lower limit of the safe negative pressure range is the minimum negative pressure (usually 10 kPa) to prevent the dressing from detaching, and the upper limit is the maximum negative pressure (usually 40 kPa) to prevent skin damage. If the initial negative pressure value meets the preset negative pressure value constraint (i.e., it is within the safe negative pressure range), then the initial negative pressure value is used as the negative pressure value; otherwise, the negative pressure value boundary value adjacent to the initial negative pressure value is used as the negative pressure value. For example, if the initial negative pressure value is 42 kPa, exceeding the upper limit of 40 kPa, then 40 kPa is used as the negative pressure value; if the initial negative pressure value is 8 kPa, below the lower limit of 10 kPa, then 10 kPa is used as the negative pressure value.

[0171] Next, the administration temperature is updated based on the attribute information of the determined drug type.

[0172] The attribute information of a drug type includes its medicinal properties (cold, hot, warm, cool) and whether it has a microvascular dilation effect. When a drug has a microvascular dilation effect (such as drugs containing blood-activating and stasis-removing ingredients such as chuanxiong, danshen, and safflower), a negative adjustment is applied to the initial administration temperature (lowering the administration temperature) to prevent adverse reactions caused by excessive vasodilation.

[0173] In a preferred embodiment, the negative adjustment is as follows: decrease by 1°C when containing 1-2 blood-activating herbs; decrease by 2°C when containing 3 or more blood-activating herbs. For medicines that do not contain blood-activating ingredients, the administration temperature remains unchanged or is slightly adjusted according to the properties of the medicine (+0.5°C for warm medicines, -0.5°C for cold medicines).

[0174] Finally, the treatment duration is determined based on the administration temperature.

[0175] There is a negative correlation between drug administration temperature and treatment duration: the higher the drug administration temperature, the shorter the treatment duration.

[0176] In a preferred embodiment, the basic treatment duration is set to 20 minutes (corresponding to 38°C). For every 1°C increase in drug administration temperature, the treatment duration is shortened by 2 minutes; for every 1°C decrease in drug administration temperature, the treatment duration is extended by 2 minutes.

[0177] Subsequently, the computing device determines the cumulative heat dose safety threshold based on the skin's heat tolerance level in the skin characteristic data. Cumulative heat dose is an important indicator for measuring the risk of thermal injury, and its calculation is based on the CEM43°C heat dose model. The CEM43 cumulative formula is: CEM43 = Σtᵢ·R^(43−Tᵢ), where R is taken as 0.5 when the local tissue temperature T>43°C and 0.25 when T≤43°C. A commonly used skin safety threshold is approximately 240 minutes of equivalent value; exceeding this threshold will lead to irreversible necrosis of the skin's basal layer.

[0178] The calculation device obtains the cumulative caloric dose value under the combination of drug administration temperature and initial treatment duration. If this value does not exceed the cumulative caloric dose safety threshold, the initial treatment duration is used as the treatment duration; if it exceeds the threshold, the treatment duration is adjusted until the cumulative caloric dose value does not exceed the cumulative caloric dose safety threshold. The adjustment method is to gradually shorten the treatment duration (by 1 minute each time) and recalculate the cumulative caloric dose until the safety requirements are met.

[0179] For example, suppose the administration temperature is 40℃ and the initial treatment duration is 20 minutes. According to the CEM43 model: T = 40℃ (≤43℃), R = 0.25, CEM43 = 20 × 0.25^(43−40) = 20 × 0.25^3 = 20 × 0.015625 = 0.3125 minutes of equivalent value, which is far below the 240-minute safety threshold. Therefore, this temperature-duration combination is safe. If the administration temperature is 44℃ and the initial treatment duration is 20 minutes, then CEM43 = 20 × 0.5^(43−44) = 20 × 0.5^(-1) = 20 × 2 = 40 minutes of equivalent value (for the portion T>43℃, segmented calculation is required; this is a simplified example), and it is still safe. However, if the administration temperature is 46℃ and the treatment duration is 60 minutes, then CEM43 = 60 × 0.5^(43−46) = 60 × 8 = 480 minutes of equivalent value, which exceeds the safety threshold, and the treatment duration will be automatically shortened.

[0180] The final output operating parameters include: negative pressure value (e.g., 25 kPa), drug delivery temperature (e.g., 39.5℃), treatment duration (e.g., 18 minutes), treatment frequency (e.g., once a day), treatment course arrangement (e.g., 10 treatments as one course), as well as the number of application plates (i.e., the number of acupoints to be applied, e.g., 5) and treatment intensity level (e.g., medium intensity).

[0181] Step S208: Convert the acupoints and quantity to be applied, the drug formulation and dosage, and the operating parameters of the application plate into structured data to generate a treatment plan.

[0182] In this step, all the information determined in step S206 is integrated into a structured treatment plan.

[0183] Specifically, the system generates data with the following structure:

[0184] Patient Information: Patient ID is P20240821001, name is Zhang Moumou, age is 52, gender is male.

[0185] Traditional Chinese Medicine diagnostic information: The syndrome is phlegm-dampness obstructing the lungs, the pathogenesis is phlegm-dampness accumulating in the lungs and the lungs failing to disperse and descend, and the treatment principle is to resolve phlegm and turbidity, and to disperse lung heat and relieve asthma.

[0186] Acupoint scheme: The main acupoints are Feishu (Bladder Meridian of Foot Taiyang, 1.5 cun lateral to the spinous process of the 3rd thoracic vertebra) and Tanzhong (Ren Meridian, anterior midline, level with the 4th intercostal space); the auxiliary acupoints are Tiantu (Ren Meridian, in the middle of the suprasternal notch), Fenglong (Stomach Meridian of Foot Yangming, 8 cun above the tip of the lateral malleolus, 1 cun lateral to Tiaokou acupoint) and Zusanli (Stomach Meridian of Foot Yangming, 3 cun below Dubi acupoint, one finger-width lateral to the anterior crest of the tibia); the number of application plates is 5.

[0187] Drug regimen: Pinellia ternata 34.3mg (2 parts), Citrus reticulata peel 28.5mg (1.5 parts), Poria cocos 42.0mg (2.5 parts), Perilla frutescens seed 22.0mg (1 part), Prunus armeniaca seed 18.5mg (1 part); Preparation parameters for soaking film: drug concentration 5%, drug volume 2mL / pan, soaking time 10 minutes.

[0188] Operating parameters: negative pressure 25 kPa, drug delivery temperature 39.5℃, treatment duration 18 minutes, treatment frequency once a day, treatment course of 10 treatments, number of application plates 5, treatment intensity level moderate.

[0189] Generation time: 2024-08-21T14:30:00Z.

[0190] This structured data can be directly sent to the transdermal drug delivery device and its supporting cloud platform to guide the transdermal drug delivery device in performing acupoint application, drug loading, and setting operating parameters.

[0191] Specifically, the transdermal drug delivery device receives this structured data through a communication interface and automatically completes the following operations:

[0192] (1) Based on the acupoint location information, the robotic arm or manual assistance will apply the plaster to each acupoint in sequence;

[0193] (2) According to the drug regimen, the corresponding type and dosage of drug solution is automatically drawn and injected into the central drug delivery chamber of each dressing tray;

[0194] (3) The pressure value of the negative pressure pump, the temperature value of the heating module, and the treatment timer are automatically set according to the operating parameters. The entire process requires no manual intervention, realizing complete automation from plan generation to physical execution.

[0195] After treatment, the efficacy feedback and plan adjustment steps can be optionally implemented: during the treatment process, objective physiological indicators (such as the results of inflammation index re-examination, skin condition, and patient symptom scores) are collected, and the indicators before and after treatment are compared using a large language model to evaluate the efficacy. Based on the efficacy, the acupoints, dosages, and parameters in the subsequent treatment plan are dynamically adjusted to form a closed loop of generation → execution → evaluation → adjustment, thereby achieving objective comparison of efficacy and optimization of the plan.

[0196] In a preferred embodiment, the specific process for efficacy feedback is as follows: After completing a course of treatment (e.g., 10 sessions), the patient is prompted to recheck relevant indicators (such as inflammatory markers like C-reactive protein and interleukin-6, or aging indicators like sex hormone levels), and symptom scores (such as cough frequency, shortness of breath, and frequency of hot flashes) are collected through a client application. The post-treatment indicators are compared with baseline data before treatment, and a large language model automatically generates an efficacy evaluation report. If the efficacy does not meet expectations (e.g., inflammatory markers decrease by less than 30%), the system automatically analyzes possible causes (whether the acupoint selection is accurate, whether the dosage is sufficient, or whether the parameters are optimized), and adjusts the treatment plan for the next course accordingly (e.g., adding acupoints, increasing the dosage, adjusting the temperature, etc.).

[0197] To more clearly illustrate the technical solution of the present invention and its application in actual clinical practice, two specific application examples are provided below.

[0198] Example 1: Generation of a treatment plan for chronic respiratory inflammation (COPD).

[0199] Patient Wang, male, 65 years old, with a 20-year smoking history, was diagnosed with chronic obstructive pulmonary disease (COPD) 5 years ago. Chief complaint: recurrent cough with sputum, worsening in the morning, white and viscous sputum, shortness of breath after walking even slightly, chest tightness, worsening in the past month. The above consultation information was collected through a client application, and the patient's examination reports were imported via an interface: pulmonary function test showed FEV1 / FVC = 58% (suggesting moderate obstructive ventilatory dysfunction), chest CT showed increased lung markings and signs of emphysema in both lungs, and laboratory tests showed C-reactive protein 28 mg / L and interleukin-6 22 pg / mL.

[0200] Using a large language model, semantic understanding and field information extraction were performed on the above data, and the following disease feature description information was output: the site of inflammation is the lung, the type of inflammation is chronic, the degree of inflammation is moderate, the accompanying symptoms are cough, sputum (white and viscous), wheezing, and chest tightness, and the lesion location of aging is none.

[0201] Further combining the patient's individual information (constitution type: phlegm-dampness constitution; four diagnostic information: white and greasy tongue coating, slippery and rapid pulse), the candidate syndrome type is matched from the TCM knowledge graph through retrieval enhancement generation technology, and the TCM diagnosis information is output: the syndrome type is phlegm-dampness obstructing the lungs, the pathogenesis is phlegm-dampness obstructing the lungs and the lungs failing to disperse and descend, and the treatment principle is to resolve phlegm and turbidity, and to disperse the lungs and relieve asthma.

[0202] Based on the syndrome differentiation and treatment principles, a set of candidate acupoints is generated through the meridian tropism rules. After matching and screening by the big language model, the acupoint scheme is determined: the main acupoints are Feishu (BL13) and Tanzhong (CV17); the auxiliary acupoints are Tiantu (CV22), Fenglong (ST40), and Zusanli (ST36); the number of application plates is 5.

[0203] Based on the syndrome differentiation, a first set of drugs is matched from the syndrome differentiation-drug indication function mapping table, and a second set of drugs is matched based on the acupoint meridian tropism. The intersection of these sets yields a candidate drug set. After enumeration, scoring, and incompatibility checks, the drug types are determined as follows: Pinellia ternata, Citrus reticulata peel, Poria cocos, Perilla frutescens seed, and Prunus armeniaca seed. The transdermal dosage of each drug is calculated based on the patient's skin characteristics (neutral skin, moderate skin thickness), inflammation level (C-reactive protein 28 mg / L, mildly abnormal, correction factor 1.3), constitution type (phlegm-dampness constitution, constitution correction factor 1.2), and age (65 years old, age correction factor 1.2). Taking Pinellia ternata as an example: the baseline transdermal dose is 20 mg, with a skin permeability correction factor of 1.0, an inflammation / aging degree correction factor of 1.3, and an individual difference correction factor of 1.2 × 1.2 = 1.44. The transdermal dose is 20 × 1.0 × 1.3 × 1.44 = 37.4 mg.

[0204] Based on the location of the lesion (lungs, approximately 3cm deep) and the tissue characteristics of the acupoint area (thick back muscles, moderate skin thickness), the initial negative pressure value was determined to be 25kPa. After safety constraint verification (the safe negative pressure range is 10-40kPa), the output negative pressure value was 25kPa. Based on the depth of the lesion, the degree of inflammation, and the blood supply characteristics of the acupoint, the initial drug administration temperature was determined to be 39℃. Based on the properties of the drug (containing warming herbs such as Pinellia ternata and Citrus reticulata peel, with no significant blood-activating components), the drug administration temperature was maintained at 39℃. Based on the drug administration temperature of 39℃, the initial treatment duration was determined to be 18 minutes. After verification by CEM43 thermal dosing, safety was confirmed, and the output treatment duration was 18 minutes.

[0205] The final structured treatment plan is generated and directly sent to the transdermal drug delivery device for execution. Patients receive treatment once daily for 18 minutes each time, with 10 treatments constituting one course.

[0206] Example 2: Generation of a diagnosis and treatment plan for chronic diseases such as ovarian hypofunction (premature ovarian failure).

[0207] Patient Li, female, 42 years old, diagnosed with premature ovarian failure 2 years ago. Chief complaint: Increasingly light menstrual flow; frequent bouts of hot flashes and sweating over the past six months; poor sleep at night; lower back pain and leg weakness; and dry skin. Medical records and examination reports were collected: Estradiol (E2) 18 pg / mL, Follicle-stimulating hormone (FSH) 42 mIU / mL, Anti-Müllerian hormone (AMH) 0.3 ng / mL, and bone mineral density T-score -1.8.

[0208] The large language model is used to output the description of the symptoms: the site of inflammation is none, the type of inflammation is none, the degree of inflammation is none, the accompanying symptoms are scanty menstruation, hot flashes, night sweats, insomnia, soreness of the lower back and knees, and dry skin, the aging site is the ovaries (Chong and Ren meridians), and the aging functional state is ovarian dysfunction.

[0209] Based on the patient's individual information (constitution type: Yin deficiency; four diagnostic methods: red tongue with little coating, thready and rapid pulse), the TCM diagnosis information is output: the syndrome is kidney essence deficiency syndrome, the pathogenesis is insufficient kidney essence and malnourishment of the Chong and Ren meridians, and the treatment principle is to tonify the kidney and replenish essence, and regulate the Chong and Ren meridians.

[0210] The acupoints were determined as follows: the main acupoints were Shenshu (BL23) and Guanyuan (CV4); the auxiliary acupoints were Sanyinjiao (SP6), Zusanli (ST36), and Taixi (KI3); and the number of application plates was 5.

[0211] The following medications were identified: Rehmannia glutinosa (processed), Cornus officinalis, Dioscorea opposita, Lycium barbarum, and Cuscuta chinensis. Based on the patient's skin characteristics (dry skin, thin skin, skin permeability correction factor 1.2), degree of aging (significantly decreased sex hormone levels, moderate to severe abnormality, correction factor 1.6), constitution type (Yin deficiency constitution, constitution correction factor 0.9), and age (42 years old, age correction factor 1.1), the transdermal dosage of each medication was calculated. For example, with Rehmannia glutinosa (processed): the baseline transdermal dose is 30 mg, and the transdermal dosage = 30 × 1.2 × 1.6 × 0.9 × 1.1 ≈ 57.0 mg.

[0212] Determine the operating parameters: Based on the lesion location (ovary, depth approximately 5cm) and the tissue characteristics of the acupoint area (thin skin in the waist and abdomen with moderate elasticity), determine the initial negative pressure value as 35kPa. After safety constraint verification, output the negative pressure value as 35kPa. Based on the lesion depth, inflammation level (no inflammation), and acupoint blood supply characteristics (rich blood supply in the waist and abdomen), determine the initial drug administration temperature as 38℃. Based on the drug properties (containing nourishing herbs such as Rehmannia glutinosa and Cornus officinalis, with no significant blood-activating components), maintain the drug administration temperature at 38℃. Based on the drug administration temperature of 38℃, determine the initial treatment duration as 20 minutes. After CEM43 thermal dose verification to confirm safety, output the treatment duration as 20 minutes.

[0213] Finally, a structured treatment plan is generated and issued for implementation. Patients receive treatment once daily for 22 minutes each time, with 10 treatments constituting one course of treatment.

[0214] In practical clinical applications, expert-annotated data in the field of transdermal drug delivery is extremely scarce, making it difficult to directly train a reliable large language model parameter generation capability. To address this issue, in a preferred embodiment of this invention, a text-conditionally driven physical constraint variational autoencoder and a latent flow matching algorithm are used to generate dressing tray operating parameters.

[0215] This preferred embodiment includes the following steps:

[0216] The first step is to model the operating parameters to be output as a heterogeneous parameter space that is a mixture of continuous and discrete parameters.

[0217] Continuous parameters (negative pressure, administration temperature, treatment duration) are modeled using a truncated Gaussian distribution to avoid out-of-boundary probability leakage caused by the infinite support set of the standard Gaussian distribution.

[0218] Discrete parameters (number of dressings, treatment intensity level) are sampled using discrete reparameterization technology. Specifically, for count parameters such as the number of dressings, they are encoded into one-hot vectors and then sampled using Gumbel-Softmax; for ordinal parameters such as treatment intensity level, cumulative probability encoding is used to maintain the level order logic.

[0219] The second step is to construct a manifold in the latent variable space and then perform continuous-time probabilistic flow modeling through latent flow matching.

[0220] First, a variational autoencoder (VAE) is trained to encode high-dimensional patient features (disease description information, individual information, diagnosis information, etc.) into low-dimensional latent variables, and then the latent variables are decoded into running parameters.

[0221] Then, in the latent space, the optimal transmission path from the baseline noise distribution to the empirical distribution of the patient's target parameters is learned through flow matching. Flow matching is achieved by regressing a time-dependent vector field that progressively transmits the prior distribution (standard Gaussian) to the target distribution (empirical distribution of patient parameters). This method can generate massive amounts of medically logically safe and diverse synthetic treatment parameter data with only a small number of real expert-annotated samples (e.g., 50-100 cases).

[0222] The third step is to serialize the generated numerical parameters into natural language training samples using a structured text template with a fixed format, and then perform supervised fine-tuning (SFT) on the basic large language model.

[0223] The training samples are formatted as follows: Patient information: [symptom description, individual information, diagnosis information] → Operating parameters: [negative pressure value, drug administration temperature, treatment duration]. In this way, the large language model learns the ability to accurately infer and map from unstructured patient information to heterogeneous treatment parameters.

[0224] The core advantage of this preferred implementation is that, under conditions of scarce real clinical data, through synthetic data amplification and model fine-tuning, the large language model can still generate safe and personalized treatment parameters, thereby reducing reliance on the experience of highly skilled physicians while ensuring the safety and reproducibility of the output parameters.

[0225] According to the treatment plan generation method based on transdermal drug delivery provided by the present invention, semantic understanding and field information extraction of consultation information and examination reports are performed through a large language model to output structured disease feature description information. Based on this, TCM diagnosis information is generated in conjunction with a pre-built medical knowledge base. Then, the acupoints to be applied, drug type and transdermal dosage, and application plate operating parameters are determined through linkage of the diagnosis information. Finally, all information is integrated into a structured treatment plan that can be directly sent to the transdermal drug delivery device for execution. This realizes the fully automated generation of the entire process from patient information input to the executable plan of the application treatment plate, which significantly reduces the dependence of this technology on the experience of highly skilled physicians.

[0226] Based on this, the present invention adopts a two-stage acupoint selection mechanism of syndrome type and treatment principle → meridian tropism rule → candidate acupoint set → large language model matching and screening, so that the acupoint selection process has traceable and verifiable intermediate evidence, thereby improving the transparency and clinical acceptability of acupoint selection.

[0227] By using a two-way matching of syndrome type and drug, combination enumeration scoring, and verification of the rules of incompatibilities of traditional Chinese medicine such as the Eighteen Incompatibilities and Nineteen Antagonisms, an objective and safe screening of drug combinations has been achieved.

[0228] By establishing a dedicated conversion model for oral equivalent dose to transdermal dose (with the baseline transdermal dose adjusted for skin permeability, inflammation / aging degree, and individual differences), the technical challenge of the long-standing lack of quantitative evidence for transdermal doses has been solved.

[0229] By embedding three physical safety constraints—negative pressure safety zone cutoff, CEM43 thermal dose cumulative verification, and drug-temperature negative feedback suppression—in the process of determining the operating parameters of the dressing tray, equipment-related safety accidents are eliminated from the root of the algorithm.

[0230] Furthermore, by collecting objective physiological indicators before and after treatment and comparing the efficacy with a large language model, dynamic closed-loop optimization of the treatment plan was achieved. The entire plan, from data collection, diagnostic reasoning, and plan generation to equipment execution and efficacy feedback, is seamlessly integrated, effectively improving the standardization, treatment precision, and clinical reproducibility of transdermal drug delivery technology.

[0231] like Figure 3 As shown, another embodiment of this disclosure provides a treatment plan generation device 300 based on transdermal drug delivery, which resides in a computing device. The device 300 includes a processing module 302, a generation module 304, a determination module 306, and a conversion module 308 that are coupled to each other.

[0232] The processing module 302 is configured to use a large language model to perform semantic understanding and field information extraction on the patient's question data and output disease feature description information.

[0233] The generation module 304 is configured to generate TCM diagnostic information based on the symptom description information and the patient's individual information, combined with a pre-built medical knowledge base.

[0234] The module 306 is configured to determine the acupoints and number of patches to be applied, the type and dosage of the medicine, and the operating parameters of the patching plate based on the TCM diagnosis information.

[0235] The conversion module 308 is configured to convert the acupoints and quantity to be applied, the drug type and dosage, and the operating parameters of the application plate into structured data to generate a treatment plan.

[0236] It should be noted that the working principle and process of device 300 are similar to those of the aforementioned method 200. For relevant details, please refer to the description of method 100. These details will not be repeated here.

[0237] The various techniques described herein can be implemented in combination with hardware or software, or a combination thereof. Thus, the methods and apparatus of the present invention, or certain aspects or portions thereof, can take the form of program code (i.e., instructions) embedded in a tangible medium, such as a removable hard disk, USB flash drive, floppy disk, CD-ROM, or any other machine-readable storage medium, wherein when the program is loaded into and executed by a machine such as a computer, the machine becomes an apparatus for practicing the present invention.

[0238] When the program code is executed on a programmable computer, the computing device generally includes a processor, a processor-readable storage medium (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. The memory is configured to store program code; the processor is configured to execute the method of the present invention according to instructions in the program code stored in the memory.

[0239] By way of example, and not limitation, readable media include readable storage media and communication media. Readable storage media stores information such as computer-readable instructions, data structures, program modules, or other data. Communication media generally embodies computer-readable instructions, data structures, program modules, or other data in the form of modulated data signals such as carrier waves or other transmission mechanisms, and includes any information delivery medium. Any combination of the above is also included within the scope of readable media.

[0240] In the specification provided herein, the algorithms and displays are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used with the examples of this invention. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing preferred embodiments of the invention.

[0241] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0242] Those skilled in the art will understand that modules, units, or components of the devices disclosed in the examples herein can be arranged in the devices described in this embodiment, or alternatively, can be located in one or more devices different from the devices in this example. The modules in the foregoing examples can be combined into a single module or, in addition, can be divided into multiple sub-modules.

[0243] Furthermore, some of the embodiments described herein are methods or combinations of method elements that can be implemented by a processor of a computer system or by other means of performing the functions. Therefore, a processor having the necessary instructions for implementing the methods or method elements forms means for implementing the methods or method elements. Furthermore, the elements described herein in the apparatus embodiments are examples of means for implementing the functions performed by elements for the purposes of carrying out the invention.

[0244] As used herein, unless otherwise specified, the use of ordinal numbers such as “first,” “second,” “third,” etc., to describe ordinary objects merely indicates different instances of similar objects and is not intended to imply that the objects being described must have a given order in time, space, ordering, or any other manner.

[0245] Although the invention has been described with reference to a limited number of embodiments, those skilled in the art will understand from the foregoing description that other embodiments are conceivable within the scope of the invention described herein. Furthermore, it should be noted that the language used in this specification has been chosen primarily for readability and instructional purposes, and not for the purpose of interpreting or limiting the subject matter of the invention. Therefore, many modifications and alterations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims.

Claims

1. A method for generating a treatment plan based on transdermal drug delivery, executed in a computing device, wherein the transdermal drug delivery method includes: The method involves loading a drug into a patch tray of a transdermal drug delivery device, allowing the drug to be absorbed transdermally through the adhesion of the patch tray to the skin surface. Using a large language model, semantic understanding and field information extraction are performed on the patient's question data to output disease feature description information; Based on the description of the disease characteristics and the individual patient information, combined with a pre-built medical knowledge base, TCM diagnostic information is generated. Based on the TCM diagnostic information, determine the acupoints and number to be applied, the type and dosage of the medicine, and the operating parameters of the application plate; The acupoints and their quantity to be applied, the drug formulation and dosage, and the operating parameters of the application plate are converted into structured data to generate a treatment plan.

2. The method as described in claim 1, wherein, Based on the TCM diagnostic information, the acupoints and number of points to be treated are determined, including: Based on the syndrome and treatment principles in the TCM diagnostic information, obtain the meridian tropism rules corresponding to the syndrome and treatment principles in the TCM knowledge graph; Based on the meridian tropism rules, the acupoint knowledge base is traversed to filter out acupoints that meet the meridian tropism conditions to generate a candidate acupoint set; The large language model performs a matching and screening of the candidate acupoints in the set of acupoints according to the syndrome type and treatment principle to determine the acupoints to be applied and the number of acupoints, which include the main acupoints and the auxiliary acupoints.

3. The method as described in claim 2, wherein, Based on the TCM diagnostic information, the type of medication is determined, including: Based on the syndrome type in the TCM diagnosis information, the first drug set is matched from the pre-constructed syndrome type-drug indication function mapping table; Based on the main acupoints and auxiliary acupoints, a second set of medicines that matches the selected acupoints and meridians is selected from the pre-constructed acupoint-medicine meridian association table; The intersection of the first drug set and the second drug set is taken as the candidate drug set; Using the TCM diagnosis information, acupoint information, patient constitution type, and age as constraints, all combinations of all drugs in the candidate drug set are enumerated and traversed for scoring, and the drug combination with the highest score is selected as the initial drug combination. The initial drug combination is input into the Chinese medicine compatibility rule library for verification. If the initial drug combination contains any incompatible pairing, the combination is removed and the next highest-scoring combination is selected in turn until the incompatibility verification is passed, so as to determine the final drug type.

4. The method of claim 3, wherein, Based on the TCM diagnostic information, the dosage of the medicine is determined, including: The skin permeability correction coefficient is determined based on skin characteristic data in the patient's individual information. The skin permeability correction coefficient is positively correlated with skin thickness and stratum corneum density. The inflammation / aging degree correction coefficient is determined based on the inflammation degree data and / or aging degree data in the patient's individual information. The inflammation / aging degree correction coefficient is positively correlated with the values ​​of inflammation-related indicators and / or aging-related indicators, and increases in stages or continuously as the indicator values ​​increase. Based on the patient's individual information, including their TCM constitution type and age, an individual difference correction coefficient is determined. This individual difference correction coefficient is the product of the constitution correction coefficient and the age correction coefficient. The constitution correction coefficient is determined based on the preset coefficient value corresponding to each constitution type, and the age correction coefficient is determined based on the preset coefficient value corresponding to the preset age range in which the age falls. The transdermal dose is the product of the baseline transdermal dose, the skin permeability correction factor, the inflammation / aging degree correction factor, and the individual difference correction factor. The baseline transdermal dose is the product of the oral equivalent dose and the transdermal absorption rate correction factor.

5. The method of claim 3, wherein, The operating parameters include at least the negative pressure value, drug delivery temperature, and treatment duration, as well as the operating parameters of the application plate determined based on the TCM diagnostic information, including: Based on the tissue characteristics of the body surface area corresponding to the identified acupoints to be treated, the initial negative pressure value and the initial drug administration temperature are determined. If the initial negative pressure value meets the preset negative pressure value constraint, then the initial negative pressure value is used as the negative pressure value; otherwise, the negative pressure value constraint boundary value adjacent to the initial negative pressure value is used as the negative pressure value. Based on the attribute information of the determined drug type, the drug administration temperature is updated to obtain the drug administration temperature; The initial treatment duration is determined based on the drug administration temperature. If the cumulative thermal dose value under the combination of the drug administration temperature and the initial treatment duration exceeds the cumulative thermal dose safety threshold, the treatment duration is adjusted until the cumulative thermal dose value does not exceed the cumulative thermal dose safety threshold, and the adjusted value is used as the treatment duration. The cumulative thermal dose safety threshold is determined based on the tissue characteristics of the body surface area corresponding to the acupoint to be applied.

6. The method of claim 1, wherein, The description of the disease characteristics includes at least one or more of the following: the location of inflammation, the type of inflammation, the degree of inflammation, accompanying symptoms, and age-related lesions and functional states.

7. The method of claim 1, wherein, The data obtained from the medical records includes consultation information and examination reports, as well as... Using a large language model, semantic understanding and field information extraction are performed on patient questionnaire data to output disease feature description information, including: The large language model is used to perform semantic understanding on the consultation information, and to identify and extract symptom description fields. The inspection report is semantically understood using the large language model to identify and extract numerical fields of inspection indicators. The extracted symptom description fields and test indicator numerical fields are structured and encoded to generate the disease feature description information.

8. The method of claim 1, wherein, Based on the symptom description information and individual patient information, combined with a pre-built medical knowledge base, TCM diagnostic information is generated, including: Obtain the patient's constitution type and four diagnostic information from the individual patient information, wherein the four diagnostic information includes at least one or more of the following: tongue appearance, pulse appearance, complexion, and voice. Using the disease feature description information, constitution type and four diagnostic information as input conditions, a set of candidate syndromes corresponding to the disease location information and accompanying symptom information is matched from a pre-constructed TCM knowledge graph. The TCM knowledge graph is constructed with disease location, symptoms, syndromes, pathogenesis and treatment principles as entities and the relationship between entities as edges. Candidate syndromes that are incompatible with the constitution type and / or the four diagnostic methods are screened out from the candidate syndrome set to obtain TCM diagnostic information. The TCM diagnostic information includes at least the syndrome name, the pathogenesis analysis corresponding to the syndrome, and the treatment principles and methods corresponding to the pathogenesis analysis.

9. The method of claim 5, wherein, The tissue characteristics of the body surface area include skin thickness, elasticity, and blood supply characteristics; the disease feature description information includes lesion depth, degree of inflammation, and... Based on the tissue characteristics of the body surface area corresponding to the identified acupoints to be treated, the initial negative pressure value and initial drug administration temperature are determined, including: The initial negative pressure value is determined based on the depth of the lesion, skin thickness, and elasticity. The initial administration temperature is determined based on the lesion depth, inflammation level, blood supply characteristics, and the attribute information of the determined drug type.

10. The method of claim 1, wherein, The TCM diagnostic information includes inflammatory syndromes and aging syndromes. The inflammatory syndromes include at least one of the following: phlegm-dampness obstructing the lungs syndrome, damp-heat accumulation syndrome, and qi stagnation and blood stasis syndrome. The aging syndromes include at least one of the following: kidney essence deficiency syndrome, qi and blood deficiency syndrome, imbalance of the Chong and Ren meridians syndrome, and liver and kidney yin deficiency syndrome.

11. The method of claim 1, wherein, Also includes: During the treatment process based on the generated treatment plan, objective physiological indicators of the patient are collected; The objective physiological indicators are compared with the physiological indicators before treatment using a large language model, and the treatment plan is adjusted based on the comparison results.

12. A computing device, comprising: At least one processor; and A memory storing program instructions, wherein the program instructions are configured to be executed by the at least one processor, the program instructions including instructions for performing the method as described in any one of claims 1-11.

13. A readable storage medium storing program instructions that, when read and executed by a computing device, cause the computing device to perform the method as described in any one of claims 1-11.