An integrated four-examination diagnosis and treatment terminal and system with a built-in traditional Chinese medicine health large model and connected to an internet hospital

Abstract

This invention relates to the technical fields of intelligent diagnostic equipment for traditional Chinese medicine (TCM), medical IoT, and equipment and systems for internet-based TCM hospitals. Specifically, it relates to an integrated terminal and system for the four diagnostic methods (inspection, auscultation, inquiry, and palpation), incorporating a built-in TCM health model, wireless networking capabilities, and access to an internet-based TCM hospital platform for online registration, AI-based diagnosis, physician review, and electronic prescription processing. It aims to address the problems of single-function diagnostic equipment, data disconnect, and lack of a closed-loop diagnosis and treatment system. The terminal includes an integrated shell and integrated multimodal sign acquisition, edge computing core, interactive touch display, and secure communication gateway module. The acquisition module consists of inspection, auscultation, inquiry, and palpation units, synchronously acquiring tongue appearance, voice, chief complaint, and pulse wave features under a unified clock. The edge computing core incorporates a built-in TCM health model, deeply fusing multimodal features through a cross-modal attention mechanism.

Description

Technical Field

[0001] This invention belongs to the field of smart healthcare and TCM auxiliary diagnosis and treatment equipment, specifically involving an integrated four-diagnosis and treatment terminal and system that incorporates a large TCM health model and connects to an internet-based TCM hospital. Background Technology

[0002] With the deep integration of information technology and traditional medicine, the digital, objective, and intelligent transformation of TCM diagnosis and treatment has become a key driving force for improving the level of primary healthcare services, promoting the downward flow of high-quality medical resources, and achieving the goal of "prevention of disease" in health management. Within the existing technological framework, intelligent TCM diagnostic equipment has gradually moved from laboratory research to clinical applications and home health monitoring. Its product forms widely cover digital pulse diagnosis instruments, tongue and facial image acquisition systems, electronic stethoscopes, and various auxiliary diagnostic software. The application of these devices has largely overcome the diagnostic inconsistencies caused by differences in individual physicians' subjective experience in traditional TCM diagnosis and treatment, laying a solid data foundation for the quantitative expression and standardized research of TCM signs.

[0003] Specifically, existing digital data acquisition terminals often focus on improving the precision of a single diagnostic dimension, aiming to simulate the perception process of traditional Chinese medicine's "observation, auscultation, inquiry, and palpation" through advancements in physical sensing technology. For example, high-precision pressure sensor arrays are used to simulate the pressure sensation of a physician's fingers to extract the location, rate, shape, and momentum characteristics of the pulse; or standard light source environments and high-resolution imaging units are used to capture changes in the color and shape of the tongue coating and complexion. These technological advancements have made the digital recording of TCM signs possible and, to some extent, achieved preliminary standardization of the diagnostic and treatment process. However, with the rapid development of medical IoT technology and the increasingly stringent requirements for diagnostic depth, precision, and decision continuity in complex clinical application scenarios, these existing technological solutions, based on modular design and relatively single and independent functions, are gradually revealing deep-seated logical limitations and technical contradictions in practical applications.

[0004] At its root, the core of TCM clinical decision-making lies in the holistic view of "integrated diagnosis and treatment based on syndrome differentiation" and "treatment based on syndrome differentiation." Under the current technological system, due to the fragmentation and discreteness of data acquisition terminals, diagnostic elements of observation, auscultation, inquiry, and palpation are often acquired at different times and on different physical devices. This inevitably leads to severe spatiotemporal disconnects and logical gaps between multimodal symptom data. Lacking a unified physical carrier and synchronous triggering mechanism, it is difficult to achieve true deep coupling and cross-validation of feature data at the algorithmic level, thus greatly weakening the rigor of the diagnostic logic. Furthermore, traditional auxiliary diagnostic systems mostly rely on pre-set expert rule bases or shallow machine learning models. Their reasoning mechanisms are often limited to simple conditional mapping, lacking a deep semantic understanding and knowledge generalization ability regarding ancient TCM classics, massive amounts of medical records from renowned doctors, and complex pathological evolution processes. This significantly limits the system's diagnostic accuracy and clinical reference value when facing variant syndromes and complex concurrent syndromes.

[0005] Meanwhile, the structural imbalance between existing diagnostic equipment and subsequent medical execution pathways has become a core bottleneck restricting the development of TCM internet healthcare. Most current diagnostic terminals only perform "data collection" or "preliminary analysis," and their interaction with internet TCM hospital platforms often remains at the level of simple data uploading, failing to achieve deep integration of business logic. In practical applications, from automatic data collection at hardware terminals to online registration at internet hospitals, to real-time review by remote physicians and the compliant flow of electronic prescriptions, significant collaborative friction and information silos exist between various stages. This disconnect between "diagnosis" and "treatment" not only increases the operational difficulty for patients but also causes remote physicians to worry about the authenticity of data due to a lack of visual understanding of the data collection process, thus posing challenges in risk control and prescription compliance. Especially under the wave of large-scale pre-trained model (Large Model) technology, how to efficiently deploy large-scale TCM health models with powerful cognitive and reasoning capabilities on integrated terminals and form a seamless closed-loop drive with the medical resources of internet hospitals has become a theoretical bottleneck that urgently needs to be overcome in current technological evolution.

[0006] In summary, how to achieve highly integrated synchronous collection of the four diagnostic elements within a single physical terminal, and how to rely on the built-in deep learning model to realize the logical evolution from massive multimodal data to high-quality diagnostic conclusions, while simultaneously establishing a full-link interactive logic between the hardware terminal and the Internet medical ecosystem, and constructing an integrated diagnosis and treatment system from perception and cognition to closed-loop execution, has become a key challenge and an urgent technical problem for those skilled in the art. Summary of the Invention

[0007] This invention provides an integrated diagnostic and treatment terminal and system with a built-in TCM health model and access to an internet-based TCM hospital, in order to solve the technical problems of existing TCM diagnostic equipment, such as limited functionality, spatiotemporal disconnect of multimodal symptom data, lack of deep semantic understanding of diagnostic logic, and missing closed-loop diagnosis and treatment links.

[0008] To achieve the above objectives, this invention provides an integrated four-diagnosis and treatment terminal with a built-in TCM health model and access to an internet-based TCM hospital. The terminal includes an integrated shell and integrated on the shell a multimodal sign acquisition module, an edge computing core module, an interactive touch display module, and a secure communication gateway module. The multimodal sign acquisition module consists of an inspection image acquisition unit, an auscultation acoustic acquisition unit, a consultation interaction unit, and a palpation pulse acquisition unit, used to synchronously acquire the examinee's tongue surface features, voice features, chief complaints, and pulse wave features under a unified clock reference. The edge computing core module has a built-in TCM health model, which generates diagnostic conclusions and suggested prescriptions by deeply fusing and semantically aligning the feature data output by the multimodal sign acquisition module. The secure communication gateway module accesses the internet-based TCM hospital platform through an encrypted tunnel, enabling the uploading of diagnostic data, the receipt of remote physician review results, and the issuance of electronic prescriptions.

[0009] Furthermore, the visual image acquisition unit in the multimodal sign acquisition module includes a standard light source array and a high-resolution image sensor located at the top of the integrated housing cavity. The standard light source array uses LED light sources with a color rendering index (Ra) of not less than 95 and a constant color temperature of 5600K±200K, and achieves uniform light field distribution through a diffuser plate. The high-resolution image sensor uses a 1 / 1.8-inch CMOS photosensitive element with an effective pixel count of not less than 8 million. Combined with an optical lens with distortion correction function, it ensures that the color reproduction error ΔE between the tongue image and the facial image in RGB space and CIE Lab space is less than 3.0. When the visual image acquisition unit is working, it automatically identifies the key facial points and tongue contours of the examinee through image preprocessing algorithms, and realizes automatic cropping and normalization of the region of interest (ROI).

[0010] Furthermore, the auscultation acoustic acquisition unit includes a MEMS microphone array and an audio processing circuit disposed on the front of the integrated housing; the MEMS microphone array has a signal-to-noise ratio of not less than 65dB and a frequency response range of 20Hz to 20kHz, and suppresses ambient background noise through beamforming technology; the audio processing circuit performs 16bit / 44.1kHz pulse code modulation (PCM) sampling on the acquired speech signal, and extracts Mel-frequency cepstral coefficients (MFCC) and linear predictive cepstral coefficients (LPCC) as speech feature vectors, which are input into the TCM health big model for quantitative analysis of speech intensity, pitch, and respiratory frequency.

[0011] Furthermore, the pulse diagnosis acquisition unit includes a pulse wave sensor array, a constant pressure feedback drive mechanism, and an adaptive inflatable cuff. The pulse wave sensor array consists of 3×3 sets of MEMS pressure sensors, installed at the cun, guan, and chi positions on the radial styloid process of the subject. The constant pressure feedback drive mechanism controls the pressure of the inflatable cuff through a high-precision proportional valve to achieve automated switching between superficial, middle, and deep pulse taking, with a pressure control accuracy of ±1 mmHg. The sampling frequency of the pulse wave sensor array is set to 1000 Hz, which can capture the position, number, shape, and momentum characteristics of the pulse wave and convert them into digital signals for output to the edge computing core module.

[0012] Furthermore, the edge computing core module includes a high-performance computing chip, a memory, and a built-in TCM health model. The TCM health model is built on the Transformer architecture, with no less than 7 billion parameters, and has undergone pre-training and supervised fine-tuning (SFT) using a large number of ancient TCM books, modern TCM textbooks, medical records of famous doctors, and structured clinical data. During the inference process, the TCM health model adopts multimodal alignment technology and uses a cross-modal attention mechanism to weight and fuse the feature vectors of the four categories of observation, auscultation, inquiry, and palpation.

[0013] Furthermore, the edge computing core module also integrates a model quantization and compression unit, which uses INT8 or FP16 quantization algorithms to deploy the TCM health model on the terminal side. Under the premise of ensuring that the accuracy of diagnosis decreases by less than 1%, it achieves local millisecond-level inference response, ensuring that the examinee's privacy data can be completed without uploading to the cloud.

[0014] Furthermore, the interactive touch display module consists of a medical-grade capacitive touchscreen and an embedded graphical user interface (GUI), and the consultation interaction unit is implemented based on the GUI. The consultation interaction unit is configured to dynamically generate an adaptive consultation scale based on the logic of the "Ten Questions Song" in traditional Chinese medicine, combined with the examinee's past medical history and the results of observation and auscultation collected in real time. The adaptive consultation scale adopts branching logic and uses NLP (Natural Language Processing) technology to perform real-time semantic parsing of the examinee's voice responses, converting them into standard TCM symptom codes.

[0015] Furthermore, the secure communication gateway module supports multiple access methods such as 5G / Wi-Fi 6 / Ethernet, and has a built-in hardware security module (HSM) based on the SM2 / SM4 national cryptographic algorithm; the secure communication gateway module is configured to establish an encrypted link with the Internet Traditional Chinese Medicine Hospital platform through a Virtual Private Network (VPN); the Internet Traditional Chinese Medicine Hospital platform includes a user management subsystem, an online appointment subsystem, a physician reception subsystem, an AI-assisted diagnosis subsystem, an electronic prescription management subsystem, and a pharmaceutical service subsystem.

[0016] Furthermore, the present invention also provides a diagnostic and treatment system based on the above-mentioned integrated four-diagnosis and treatment terminal, the workflow of which is as follows: As a preferred embodiment of the present invention, the system performs the following steps: Step 1: Identity Authentication and Synchronization. The examinee identifies themselves through the interactive touch display module (facial recognition or medical insurance card reading), and the terminal obtains the examinee's previous electronic medical records and historical diagnosis records from the Internet Traditional Chinese Medicine Hospital platform through the secure communication gateway module.

[0017] Step Two: Synchronized Multimodal Data Acquisition. Under the guidance of the terminal, the subject completes facial and tongue image photography, voice recording, pulse image acquisition, and interactive consultation sequentially or simultaneously. The multimodal sign acquisition module ensures that all data acquisition is completed under a unified absolute timestamp to guarantee the real-time nature and relevance of the signs.

[0018] Step 3: Edge-side large-scale model differentiation. The edge computing core module calls up the built-in TCM health large-scale model to perform real-time reasoning on the collected structured and unstructured data, generating preliminary conclusions including the Eight Principles of Differentiation, Zang-Fu Differentiation, and Qi-Blood-Body Fluid Differentiation, and recommends candidate prescriptions (including prescription name, drug name, dosage, and decoction method) according to preset TCM medication guidelines.

[0019] Step 4: Connecting to the Internet-based Traditional Chinese Medicine Hospital and Physician Review. The terminal uploads the preliminary diagnosis, recommended prescription, and collected original multimodal symptom data packets to the Internet-based Traditional Chinese Medicine Hospital platform; remote practicing physicians retrieve the above data in real time through the physician reception subsystem to visually review and correct the AI ​​diagnosis results; interactive commands during the physician review process are fed back to the terminal in real time through the secure communication gateway module.

[0020] Step 5: Electronic Prescription Circulation and Closed-Loop Execution. After being signed and confirmed by the physician, the electronic prescription is circulated through the pharmaceutical service subsystem to the designated smart pharmacy or third-party logistics system. Simultaneously, the prescription information is updated to the examinee's mobile app and terminal display screen, and the TCM health model automatically generates a personalized health management plan (involving dietary restrictions, acupressure, and lifestyle guidance).

[0021] In a preferred embodiment of the present invention, the internal control logic of the terminal adopts a distributed real-time operating system (RTOS) to ensure that the synchronization error of multi-channel data acquisition is less than 10ms. The driving logic of the pulse wave sensor array includes a motion artifact compensation algorithm, which monitors the vibration of the subject's wrist in real time through a triaxial accelerometer integrated on the sensor substrate, and uses an adaptive filter to cancel motion interference components from the original pressure signal to ensure the purity of pulse extraction.

[0022] As a preferred embodiment of the present invention, the TCM health model employs knowledge enhancement technology, embedding knowledge graphs from classic TCM texts such as the *Huangdi Neijing* and *Shanghan Zabing Lun* into the model's embedding layer. When dealing with complex concurrent symptoms, the model uses Retrieval Enhanced Generation (RAG) technology to retrieve similar medical cases from a local vector database, providing clear classical text sources and logical support for the generated diagnostic conclusions, significantly improving the interpretability of the diagnostic process.

[0023] This invention provides an integrated four-diagnosis and treatment terminal and system with a built-in TCM health model and access to an internet-based TCM hospital. Compared with existing technologies, it has the following advantages: First, through the integration of an integrated physical structure and a unified clock triggering mechanism, it achieves synchronized and standardized collection of TCM four-diagnosis data (inspection, auscultation, inquiry, and palpation), fundamentally solving the problem of spatiotemporal misalignment of multimodal signs caused by traditional discrete diagnostic equipment, and providing a high-fidelity data foundation for "integrated four-diagnosis." Second, by embedding a TCM health model with billions of parameters at the edge, the terminal possesses deep semantic understanding and complex clinical decision-making reasoning capabilities. The model can not only process quantified sensor data but also understand unstructured patient complaints. Through a cross-modal attention mechanism, it achieves accurate identification of TCM syndromes, and its diagnostic logic rigor and accuracy are significantly better than traditional expert systems or shallow learning models. Third, by constructing a closed-loop collaborative system of "edge computing - internet healthcare - remote physicians", a seamless connection from hardware terminal perception to internet healthcare execution is achieved; remote physicians can make decisions based on multi-dimensional and detailed data provided by the terminal, effectively reducing the professional risks of remote diagnosis and treatment, while opening up the last step of electronic prescription circulation and pharmaceutical services, greatly improving the efficiency and coverage of TCM diagnosis and treatment, and has extremely high engineering practical value.

[0024] The technical features described above constitute the core technical solution of this invention. The logical connections and data interactions between its various components are all based on rigorous physical principles and algorithm models, ensuring the stability and advancement of this invention in actual clinical application environments. Furthermore, the diagnostic and treatment system of this invention can be optimized for different application scenarios. For example, for primary community medical institutions, the weight of the model in the differentiation of common diseases can be increased; for home health monitoring, the functions of chronic disease trend analysis and early warning can be enhanced.

[0025] In summary, this invention constructs a novel digital diagnosis and treatment paradigm for Traditional Chinese Medicine (TCM) by deeply integrating high-precision sensing technology, edge artificial intelligence technology, and internet-based medical collaboration technology. The integrated terminal, as the core node of the entire system, not only undertakes the sensing task of data collection but also the cognitive computing task based on a large model. Through secure encrypted communication, it enables bidirectional flow with high-quality medical resources in the cloud, overcoming the systemic problems of strong subjectivity, difficulty in standardization, and disconnect between diagnosis and treatment in TCM diagnosis and treatment from a fundamental technological perspective.

[0026] As a further refinement of the present invention, the pressure sensor array in the pulse diagnosis acquisition unit has a pressure-sensing area ranging from 1.5 mm² to 2.5 mm² for each pressure-sensing unit, with a linearity better than 0.5% FS. This refined layout can accurately simulate the microscopic pressure changes in "determining the guan and cun chi" in traditional Chinese medicine palpation. When the visual image acquisition unit acquires the tongue image, the system automatically starts a white balance correction program. By comparing with the built-in standard color card, it compensates for the interference of ambient light on the color of the tongue body and tongue coating, ensuring that the calculated CIE Lab value has absolute clinical reference significance.

[0027] At the software logic level, the TCM health model adopts dynamic Prompt heuristic technology. During the consultation process, based on the obtained pulse characteristics (such as a pulse rate exceeding 90 beats per minute), it automatically triggers priority inquiries about symptoms such as "palpitation" and "heatiness". This heuristic consultation logic, which simulates the diagnostic thinking of famous doctors, greatly improves the pertinence and efficiency of the consultation.

[0028] In terms of data security and privacy protection, the system architecture of this invention strictly follows the requirements for medical data security level protection. After the original symptom data of all subjects are extracted locally, only the de-identified feature vectors and identification requests are uploaded to the cloud, ensuring the high security of the subjects' biometric information.

[0029] Through the aforementioned meticulously designed hardware structure, highly intelligent built-in large model algorithm, and comprehensive Internet service access mechanism, this invention achieves a leapfrog evolution of the TCM four diagnostic and treatment terminal from "tool attribute" to "intelligent entity attribute," demonstrating significant creativity and promising industrial application prospects. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the overall structure of the integrated four-diagnosis and treatment terminal in this invention; Figure 2 This is a partial structural diagram of the multimodal symptom acquisition module in this invention; Figure 3 This is a logical block diagram of the core edge computing module in this invention; Figure 4 This is a schematic diagram of the overall architecture of the diagnostic and treatment system in this invention; Figure 5 This is a schematic diagram of the workflow of the diagnostic and treatment system in this invention.

[0031] The attached diagram is labeled as follows: 1. Integrated housing; 2. Multimodal sign acquisition module; 21. Visual inspection image acquisition unit; 211. Standard light source array; 212. High-resolution image sensor; 22. Auscultation and olfaction acoustic acquisition unit; 221. MEMS microphone array; 222. Audio processing circuit; 23. Interview unit; 24. Palpation pulse acquisition unit; 241. Pulse wave sensor array; 242. Constant pressure feedback drive mechanism; 243. Adaptive inflatable cuff; 244. Triaxial accelerometer; 3. Edge computing core module; 31. High-performance computing chip; 32. Memory; 33. Traditional Chinese medicine health model; 34. Model quantization and compression unit; 4. Interactive touch display module; 41. Medical-grade capacitive touch screen; 42. Embedded graphical user interface; 5. Secure communication gateway module; 51. Hardware security module; 6. Internet TCM hospital platform; 61. User management subsystem; 62. Online Appointment Subsystem; 63. Physician Reception Subsystem; 64. AI-Assisted Diagnosis Subsystem; 65. Electronic Prescription Management Subsystem; 66. Pharmaceutical Services Subsystem; Detailed Implementation

[0032] This invention provides an integrated diagnostic and treatment terminal and system with a built-in traditional Chinese medicine health model and access to an internet-based traditional Chinese medicine hospital, incorporating four diagnostic methods. To make the objectives, technical solutions, and advantages of this invention clearer, the following detailed description, in conjunction with the accompanying drawings and embodiments, further illustrates the invention. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0033] This invention provides an integrated four-diagnosis terminal with a built-in TCM health model and access to an internet-based TCM hospital. Its core physical carrier is an integrated shell 1, which is made of high-strength medical-grade ABS engineering plastic through a single injection molding process. The interior is divided into a sensing area, a computing area, and an interaction area by reinforcing ribs. Within the sensing area, a highly synchronized multimodal sign acquisition module 2 is integrated. This multimodal sign acquisition module 2 consists of a visual diagnosis image acquisition unit 21, an auscultation and olfaction acoustic acquisition unit 22, a questioning and consultation interaction unit 23, and a palpation pulse acquisition unit 24. To address the issue of sign discrepancies caused by data acquisition at different times in traditional TCM devices, this invention introduces a high-precision crystal oscillator clock synchronization mechanism at the system's bottom layer. Through hard interrupt triggering by a distributed real-time operating system (RTOS), it ensures that all sensing units synchronously acquire the examinee's tongue surface features, voice features, chief complaints, and pulse wave features under a unified clock reference.

[0034] Specifically, the visual image acquisition unit 21 in the multimodal sign acquisition module 2 is installed inside a light shield at the top of the integrated housing 1 to eliminate interference from ambient stray light. This unit is equipped with a standard light source array 211, which consists of 24 SMD LEDs with high color rendering index (Ra≥95) and a constant color temperature of 5600K±200K, perfectly simulating the natural sunlight spectrum. A high-transmittance microprism diffuser is installed in front of the light source, ensuring a light field distribution uniformity of over 90% on the subject's face. The high-resolution image sensor 212 uses a 1 / 1.8-inch large-area CMOS sensor with 8 million effective pixels. Combined with a low-distortion optical lens with multi-layer anti-reflection coating, it can achieve clear imaging within a depth of field range of 10cm to 50cm. During the acquisition process, the visual image acquisition unit 21 executes a white balance correction algorithm through its built-in ISP (Image Signal Processor) and compensates for spectral shift in real time according to the built-in Lab color chart. This ensures that the color reproduction error ΔE between the acquired tongue image and facial image in RGB space and CIE Lab space is less than 3.0, thereby providing high-fidelity underlying data for subsequent large-scale model quantitative analysis of tongue color and coating thickness.

[0035] Furthermore, the auscultation acoustic acquisition unit 22 is positioned on the front of the integrated housing 1 near the subject's lips. This unit integrates a MEMS microphone array 221 and a high-performance audio processing circuit 222. The MEMS microphone array 221 adopts a ring-shaped four-microphone layout, possessing a signal-to-noise ratio of no less than 65dB and a wide frequency response range from 20Hz to 20kHz. Through beamforming technology, the system can automatically lock the subject's vocal orientation and suppress background noise from the surrounding environment. The audio processing circuit 222 performs 16-bit / 44.1kHz high-fidelity pulse code modulation (PCM) sampling on the acquired analog speech signal, and then enters the feature extraction stage, calculating and extracting Mel-frequency cepstral coefficients (MFCC) and linear predictive cepstral coefficients (LPCC) as speech feature vectors. These feature vectors can accurately characterize the formant positions, speech intensity, and breathing frequency of the speech, and are input into the TCM health big data model 33 in the edge computing core module 3 for speech diagnosis analysis.

[0036] In this invention, the pulse diagnosis acquisition unit 24 embodies extremely high engineering precision. This unit includes a pulse wave sensor array 241, a constant pressure feedback drive mechanism 242, and an adaptive inflatable cuff 243. The pulse wave sensor array 241 consists of 3×3 sets of ultra-thin MEMS pressure sensors, with the pressure-receiving area of ​​each pressure-sensing unit precisely set to 2.0 mm², exhibiting linearity better than 0.5% FS, enabling it to sensitively capture weak vascular wall pulsations. The constant pressure feedback drive mechanism 242 employs a closed-loop control system composed of a miniature high-precision proportional valve and a pneumatic pump. It adjusts the pressure inside the inflatable cuff 243 through a PID control algorithm, achieving a pressure control accuracy of ±1 mmHg, thereby automating the switching between the "superficial, middle, and deep" pulse diagnosis methods in traditional Chinese medicine. To eliminate interference from the subject's subtle limb movements, a triaxial accelerometer 244 is integrated on the sensor substrate, utilizing an adaptive filtering algorithm to monitor and compensate for motion artifacts in real time. The sampling frequency of the pulse signal is set to 1000Hz, which is sufficient to capture the slope of the rising limb of the pulse wave, the position of the descending isthmus, and the morphological characteristics of the diatonic wave. These digital signals are transmitted to the edge computing core module 3 in real time.

[0037] The edge computing core module 3 is the brain of the entire terminal, comprising a high-performance computing chip 31, a large-capacity storage device 32, and a built-in TCM health model 33. The high-performance computing chip 31 possesses AI inference computing power of no less than 20 TOPS, capable of supporting the localized real-time operation of complex neural networks. The TCM health model 33 is the core algorithm support of this invention, built on the Transformer architecture and possessing over 7 billion parameters. During its development, the model underwent pre-training on tens of thousands of ancient TCM books, including the *Huangdi Neijing* and *Shanghan Zabing Lun*, modern TCM textbooks, and over one million medical records from renowned doctors. To enable the model to simultaneously process image, audio, numerical, and textual data, this invention employs an innovative multimodal alignment technology, using a cross-modal attention mechanism to weightedly fuse the feature vectors from the four diagnostic methods of observation, auscultation, inquiry, and palpation.

[0038] Furthermore, to address the resource consumption issue of large models running on embedded devices, the edge computing core module 3 also integrates a model quantization and compression unit 34. This unit employs the INT8 quantization algorithm, compressing model weights from FP32 to 8-bit integers through cluster center search and scaling factor calibration. While ensuring that the accuracy attenuation is less than 1%, it achieves millisecond-level inference response locally, ensuring that the examinee's private data (such as facial features and voice) can be preliminarily identified without uploading to the cloud, greatly improving data security.

[0039] The interactive touch display module 4 consists of a 15.6-inch medical-grade capacitive touchscreen 41 and an embedded graphical user interface (GUI) 42. The consultation interaction unit 23 relies on the GUI 42 to realize human-computer interaction. This unit is configured with a logical architecture based on the "Ten Questions Song" of traditional Chinese medicine, and dynamically generates an adaptive consultation scale by combining the examinee's past medical history with the preliminary analysis results of observation and pulse in real time. For example, if the palpation unit detects that the examinee's pulse rate is 105 beats / min (pulse counting), the GUI interface will first pop up questions related to heat syndromes such as "whether there is fever, irritability, or thirst". The adaptive consultation scale adopts complex branching logic and uses a built-in lightweight NLP engine to perform real-time semantic parsing of the examinee's voice answers, converting them into standardized TCM symptom codes, and storing them in the feature matrix of the edge computing core module.

[0040] In terms of data transmission and remote collaboration, the secure communication gateway module 5 provides robust protection. This module supports multiple access methods, including 5G mobile networks, Wi-Fi 6, and Gigabit Ethernet, and incorporates a hardware security module (HSM) 51 based on the SM2 / SM4 national cryptographic algorithms. The secure communication gateway module 5 securely connects the terminal to the Internet-based Traditional Chinese Medicine Hospital platform 6 by establishing an IPsec or SSL VPN encrypted tunnel. The Internet-based Traditional Chinese Medicine Hospital platform 6 is a multi-dimensional service system, including a user management subsystem 61, an online appointment subsystem 62, a physician consultation subsystem 63, an AI-assisted diagnosis subsystem 64, an electronic prescription management subsystem 65, and a pharmaceutical service subsystem 66. Through this architecture, the terminal becomes not only a data collection tool but also a bridge connecting offline patients with high-quality online medical resources.

[0041] This invention also provides a diagnostic and treatment system based on the aforementioned integrated four-diagnosis and treatment terminal, whose workflow design fully embodies the closed-loop logic of clinical diagnosis and treatment. First, step one: identity authentication and synchronization. The examinee performs facial recognition or reads their medical insurance card information via a medical-grade capacitive touchscreen 41. The terminal retrieves the examinee's previous electronic medical records, historical allergies, and diagnostic records from the Internet Traditional Chinese Medicine Hospital platform 6 through the secure communication gateway module 5, providing background reference for this diagnosis. Then, step two: synchronized multimodal acquisition. Guided by the terminal's voice prompts, the examinee places their wrist on the pulse acquisition unit 24, simultaneously facing the visual image acquisition unit 21, and answers questions under the guidance of the consultation interaction unit 23. The multimodal sign acquisition module 2 ensures that all sensors complete data acquisition under a unified absolute timestamp.

[0042] The next core step is step three: edge-side big model diagnosis. The edge computing core module 3 calls the built-in TCM health big model 33 to perform cross-modal fusion reasoning on multi-dimensional data, generating detailed conclusions including the eight principles of diagnosis (yin and yang, exterior and interior, cold and heat, deficiency and excess), organ diagnosis, and qi, blood and body fluid diagnosis, and recommends candidate prescriptions based on the Chinese Pharmacopoeia and TCM medication guidelines. Step four: Internet TCM hospital connection and physician review. The terminal uploads the encrypted diagnosis conclusions and original sign packages (such as the original tongue image and pulse waveform) to the platform. Remote practicing physicians conduct visual review through the physician reception subsystem 63. Finally, step five: electronic prescription circulation and closed-loop execution. After the physician's signature confirmation, the prescription is transferred to the smart pharmacy for automatic dispensing and delivery. At the same time, the TCM health big model 33 automatically generates a personalized health management plan based on the prescription content, including dietary restrictions and acupressure guidance, and feeds it back to the examinee's mobile device.

[0043] As a key engineering detail of this invention, the TCM health model 33 employs knowledge enhancement technology. The research team extracted triples (entity-relationship-entity) from classic TCM texts, constructed a medical knowledge graph with tens of millions of nodes, and embedded it into the model's embedding layer. When dealing with complex symptoms such as "alternating chills and fever," the model not only relies on probabilistic prediction but also uses Retrieval Enhanced Generation (RAG) technology to retrieve passages about Shaoyang syndrome from the *Shanghan Lun* (Treatise on Cold Damage) from a local vector database. This provides clear classical textual evidence for the generated diagnostic conclusions, greatly enhancing the interpretability and authority of the diagnostic process.

[0044] To further demonstrate the superiority of this invention over existing technologies, the research and development team conducted a series of rigorous clinical controlled experiments. The experiments selected 300 volunteers with typical TCM symptoms and divided them into an experimental group (using the integrated terminal of this invention) and a control group (using traditional split-type diagnostic equipment and a common expert system).

[0045] Example 1: Performance Evaluation of the Integrated Four-Diagnosis Terminal of the Present Invention In the experiment, the hardware parameters and diagnostic accuracy of the terminal of the present invention were measured. The average color reproduction error of the visual diagnosis unit was ΔE=2.1, which is far better than the industry-standard ΔE<5.0. The repeatability accuracy of the palpation unit under the switching of pressure in the superficial, middle, and deep positions reached 0.98. The average inference time of the large-scale TCM health model on the local edge side was 120ms, and the power consumption was controlled within 15W.

[0046] Comparative Example 1: The control group, comprised of traditional split-type TCM diagnostic equipment, used a split-type tongue image acquisition device, electronic pulse diagnosis device, and rule-based consultation software. Due to inconsistent data acquisition time points (average interval of 15 minutes) and a lack of unified clock alignment between devices, logical gaps occurred when analyzing complex situations such as "inconsistent pulse and symptoms." Its diagnostic logic primarily relied on simple weighted scoring, lacking a deep understanding of semantics.

[0047] The following is a table of quantitative comparison data obtained from the experiment:

[0048] The data in the table above clearly demonstrates that this invention achieves a qualitative leap in diagnostic accuracy, system response speed, and clinical reliability through deep alignment of multimodal features and the powerful inference capabilities of large edge models. Particularly in terms of data synchronization, this invention reduces errors from minutes to milliseconds, which is crucial for the diagnostic accuracy of traditional Chinese medicine's "pulse and symptom combined analysis."

[0049] Delving further into the underlying algorithm of this invention, the TCM health model 33 employs supervised fine-tuning (SFT) technology during training, and a specialized loss function is designed specifically for the "analogical reasoning" approach unique to TCM. When processing the fusion of tongue image features and pulse wave sequences, the model uses a cross-modal projection layer to map the one-dimensional pulse wave sequence features into a two-dimensional image feature space, seeking the implicit correlation between pulse frequency and facial complexion. This non-linear feature mining capability is not possessed by traditional linear models.

[0050] Furthermore, the internal structure design of the integrated housing 1 also fully considers heat dissipation and electromagnetic compatibility (EMC). The surface of the high-performance computing chip 31 is covered with a copper heat sink with a high thermal conductivity, and heat is dissipated from the honeycomb-shaped heat dissipation holes at the rear of the housing by a micro centrifugal fan, ensuring that the chip temperature fluctuation is less than 5°C when the system operates continuously for more than 8 hours, thereby guaranteeing frequency stability during large model inference. The circuit board layout adopts a multi-layer shielding design, effectively isolating the electromagnetic interference of the 5G communication module to weak pulse pressure signals.

[0051] At the software architecture level, the diagnostic system of this invention adopts a microservice design. The AI-assisted diagnostic subsystem 64 in the Internet Traditional Chinese Medicine Hospital platform 6 can perform secondary reinforcement learning based on the anonymized data uploaded by the terminal, continuously iterating the model parameters. Every cycle, the updated lightweight model weights are pushed to the terminal side through the secure communication gateway module 5, completing the silent upgrade of the local large model. This "cloud-based training, edge inference" model ensures both the system's continuous evolution capability and the terminal's response speed and privacy protection.

[0052] For the pharmaceutical service subsystem 66, this invention designs a full-process traceability mechanism. The electronic prescription not only includes the ingredients and dosage, but also the processing requirements for each ingredient. After receiving the instruction, the smart pharmacy automatically matches the origin traceability code and feeds back the key parameters of the decoction process (such as heat, water volume, and time) to the examinee's mobile APP via the Internet, forming a complete industrial closed loop from TCM perception, intelligent diagnosis, expert review to pharmaceutical execution.

[0053] In summary, this invention, through its precisely integrated hardware structure, deeply fused multimodal perception algorithms, and secure and reliable internet-based medical collaboration mechanism, not only overcomes the shortcomings of traditional Chinese medicine diagnosis, such as strong subjectivity and low standardization, but also provides expert-level diagnostic support for primary healthcare institutions and family health monitoring scenarios through the deployment of large-scale edge models. Its high degree of integration in engineering implementation and rigorous data processing ensure that this invention possesses significant advancements and broad industrial application prospects in the field of digital TCM. Every optimization of technical details, from the selection of sensor pressure-sensing area to the configuration of the number of Transformer model layers, serves the core goal of improving the objectivity and clinical applicability of diagnostic conclusions. Through this comprehensive technological innovation, this invention truly realizes the digital rebirth and systematic integration of the four diagnostic methods of TCM.

Claims

1. An integrated diagnostic and treatment terminal with a built-in traditional Chinese medicine health model and access to an internet-based traditional Chinese medicine hospital, characterized in that: The terminal includes: an integrated shell (1), the interior of which is divided into a sensing area, a computing area, and an interaction area by a reinforcing rib structure; a multimodal sign acquisition module (2) integrated in the sensing area, the multimodal sign acquisition module (2) consisting of a visual diagnosis image acquisition unit (21), an auscultation and olfaction acoustic acquisition unit (22), a consultation and interaction unit (23), and a palpation pulse acquisition unit (24), used to synchronously acquire the tongue surface features, voice features, chief complaint symptoms, and pulse wave features of the examinee under the unified clock reference of the distributed real-time operating system; and an edge computing core module (3) integrated in the computing area, which has a built-in TCM health model (33). The edge computing core module (3) is used to perform deep fusion and semantic alignment of the heterogeneous feature data output by the multimodal symptom acquisition module (2) to generate diagnostic conclusions and suggested prescriptions; the interactive touch display module (4) integrated in the interactive area includes a medical-grade capacitive touch screen (41) and an embedded graphical user interface (42) to present diagnostic interaction instructions and diagnostic results; and a secure communication gateway module (5) configured to access the Internet Traditional Chinese Medicine Hospital platform (6) through a hardware encrypted tunnel to realize the desensitized uploading of diagnostic data, the reception of remote review instructions from physicians, and the issuance of electronic prescriptions.

2. The integrated four-diagnosis and treatment terminal according to claim 1, characterized in that, The visual image acquisition unit (21) is located inside the light shield at the top of the inner cavity of the integrated housing (1). It includes: a standard light source array (211), which consists of multiple LED light sources with a color rendering index Ra of not less than 95 and a color temperature of 5600K±200K. A microprism diffuser is provided in front of the standard light source array (211) to achieve uniform distribution of the light field on the subject's face; a high-resolution image sensor (212), which uses a 1 / 1.8-inch CMOS photosensitive element with an effective pixel of not less than 8 million. Combined with an optical lens with distortion correction function, it ensures that the color reproduction error ΔE of the acquired tongue image and facial image in RGB space and CIE Lab space is less than 3.

0. When the visual image acquisition unit (21) is working, it automatically identifies the key points of the subject's face and the outline of the tongue through an image preprocessing algorithm, and performs real-time white balance correction according to the built-in standard color card to complete the automatic cropping and normalization of the region of interest (ROI).

3. The integrated four-diagnosis and treatment terminal according to claim 1, characterized in that, The auscultation acoustic acquisition unit (22) is located on the upper front of the integrated housing (1), and includes: a MEMS microphone array (221) with a signal-to-noise ratio of not less than 65dB and a frequency response range of 20Hz to 20kHz. The MEMS microphone array (221) suppresses ambient background noise and locks the subject's voice direction through beamforming technology; an audio processing circuit (222) is used to sample the speech signal acquired by the MEMS microphone array (221) using 16bit / 44.1kHz pulse code modulation, and extract the Mel-frequency cepstral coefficient (MFCC) and linear predictive cepstral coefficient (LPCC) as speech feature vectors. The speech feature vectors are input to the TCM health model (33) for quantitative analysis of speech intensity, pitch, and respiratory frequency.

4. The integrated four-diagnosis and treatment terminal according to claim 1, characterized in that, The pulse diagnosis acquisition unit (24) includes a pulse wave sensor array (241), consisting of 3×3 sets of MEMS pressure sensors, with each pressure-sensing unit having a pressure-receiving area ranging from 1.5 mm² to 2.5 mm² and a linearity better than 0.5%. FS, installed on the wrist support position of the integrated housing (1), is used to correspond to the cun, guan, and chi positions at the radial styloid process of the examinee; constant pressure feedback drive mechanism (242) controls the internal pressure of the adaptive inflation cuff (243) through a high-precision proportional valve to realize the automatic switching of the pulse taking pressure of floating, middle and deep pulse taking, and its pressure control accuracy reaches ±1mmHg; triaxial accelerometer (244), integrated on the substrate of the pulse wave sensor array (241), is used to monitor the vibration displacement of the examinee's wrist in real time; wherein, the sampling frequency of the palpation pulse acquisition unit (24) is set to 1000Hz, and the output data of the triaxial accelerometer (244) is used to adaptively filter the original pressure signal through the motion artifact compensation algorithm to extract the position, number, shape and momentum characteristics of the pulse wave.

5. The integrated four-diagnosis and treatment terminal according to claim 1, characterized in that, The edge computing core module (3) includes a high-performance computing chip (31), a memory (32), and a built-in TCM health big model (33). The TCM health big model (33) is built on the Transformer architecture, with no less than 7 billion parameters, and has been pre-trained and supervised fine-tuned using ancient TCM books, modern TCM textbooks, famous doctors' medical records, and structured clinical data. The TCM health big model (33) uses knowledge enhancement technology to embed the knowledge graph of classic TCM books into the embedding layer of the model. When dealing with complex concurrent symptoms, it uses retrieval enhancement generation technology to retrieve similar medical records from the local vector database, providing the ancient book source basis for the generated diagnostic conclusions.

6. The integrated four-diagnosis and treatment terminal according to claim 1, characterized in that, The consultation interaction unit (23) is configured to dynamically generate an adaptive consultation scale based on the logic of the "Ten Questions Song" of traditional Chinese medicine, combined with the examinee's past medical history and the real-time collected observation results or pulse characteristics. The adaptive consultation scale adopts branch jump logic and uses NLP natural language processing technology to perform real-time semantic parsing of the examinee's voice answers and convert them into standard TCM symptom codes. During the consultation, if the pulse position or pulse number detected by the palpation pulse acquisition unit (24) triggers a preset threshold, the consultation interaction unit (23) automatically jumps to the corresponding priority question subsequence using dynamic Prompt heuristic technology.

7. The integrated four-diagnosis and treatment terminal according to claim 1, characterized in that, The edge computing core module (3) also integrates a model quantization compression unit (34). The model quantization compression unit (34) uses the INT8 or FP16 quantization algorithm to deploy the TCM health big model (33) on the terminal side to achieve local millisecond-level inference response. The secure communication gateway module (5) has a built-in hardware security module (51) based on the SM2 / SM4 national cryptographic algorithm and is configured to establish an encrypted link with the Internet TCM hospital platform (6) through a virtual private network (VPN). The secure communication gateway module (5) performs feature extraction and desensitization processing on the uploaded original symptom data and only transmits the desensitized feature vector and diagnosis request to the cloud.

8. A diagnostic and treatment system based on the integrated four-diagnosis and treatment terminal according to any one of claims 1 to 7, characterized in that, The system includes the integrated four-diagnosis and treatment terminal and an Internet TCM hospital platform (6) connected to the terminal via a network. The Internet TCM hospital platform (6) includes: a user management subsystem (61) for maintaining the examinee's previous electronic medical records and historical diagnosis records; a physician reception subsystem (63) for remote practicing physicians to retrieve the multimodal symptom data packets uploaded by the terminal in real time and to perform visual verification and correction of the AI ​​diagnosis results; an AI-assisted diagnosis subsystem (64) for performing secondary reinforcement learning and iteration of model weights based on the desensitized data uploaded by the terminal; an electronic prescription management subsystem (65) for storing and circulating electronic prescriptions confirmed by physician signature; and a pharmaceutical service subsystem (66) for circulating electronic prescriptions to smart pharmacies and providing examinees with information on the traceability of medicinal materials and instructions for decoction.

9. The diagnostic and treatment system according to claim 8, characterized in that, The system's workflow is designed as follows: Step 1, Identity Authentication and Synchronization: The terminal performs identity recognition through the interactive touch display module (4) and obtains the examinee's previous electronic medical records from the Internet Traditional Chinese Medicine Hospital platform (6) through the secure communication gateway module (5); Step 2, Synchronized Multimodal Acquisition: Under the guidance of the terminal, the examinee synchronously completes facial and tongue image photography, voice recording, pulse image acquisition, and interactive consultation. The multimodal sign acquisition module (2) ensures that all sensor data are collected under a unified timestamp; Step 3, Edge-side Large Model Differentiation: The edge computing... The core module (3) calls the built-in TCM health big model (33) to perform cross-modal fusion reasoning and generate preliminary diagnosis conclusions and candidate prescriptions; Step 4, physician review: The terminal uploads the diagnosis conclusions and original symptom data to the Internet TCM hospital platform (6), and the remote practicing physician reviews and corrects them through the physician reception subsystem (63); Step 5, closed-loop execution: The electronic prescription confirmed by the physician is transferred to the pharmaceutical service subsystem (66), and at the same time, the TCM health big model (33) automatically generates a personalized health management plan based on the prescription content and feeds it back to the examinee.

Images