Thyroid nodule home intervention system and method based on ar navigation and drl

Abstract

The application discloses a thyroid nodule home intervention system and method based on AR navigation and DRL, relates to the technical field of thyroid nodule home management and intervention, and discloses a thyroid nodule home intervention system based on AR navigation and DRL, which comprises a client module, a cloud service module and the like. The client module is arranged on a user mobile terminal and is used for guiding the user to perform self-operation, collect data and receive feedback in a home environment. The cloud service module is connected with the client module through internet communication and is used for remote data processing, model operation and personalized intervention scheme management. Through the cooperation of multiple modules, an innovative system for thyroid nodule home management is constructed, that is, the AR navigation unit of the client is used to fit the facial feature points and the statistical shape model of the neck, so that the maximum error of acupoint positioning is 3mm, and the problem that acupoints cannot be accurately found in traditional home operation is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of home management and intervention of thyroid nodules, in particular to a thyroid nodule home intervention system and method based on AR navigation and DRL. BACKGROUND

[0002] Thyroid nodule is one of the most common endocrine diseases. In recent years, with the popularization of high-resolution ultrasound technology and the enhancement of health examination awareness, the detection rate of thyroid nodules has shown a significant upward trend. At present, the standard management of thyroid nodules at home and abroad mainly follows the "Guidelines for Diagnosis and Treatment of Thyroid Nodules and Differentiated Thyroid Carcinoma". For benign nodules, the clinical standard practice is to perform thyroid ultrasound review every 6-12 months, and by comparing the changes in size, shape, boundary and other characteristics of the nodules, the nature and development trend of the nodules are evaluated. This traditional management mode of waiting and observation has obvious defects: first, during the long follow-up period, the patient is in a completely passive state, lacking any active intervention means, which is easy to cause anxiety and panic; second, frequent hospital review brings time cost and economic burden to the patient; finally, ultrasound examination can only provide static morphological information and cannot reflect the dynamic changes of the nodules in the development process.

[0003] In the prior art field, although there are some health management mobile application software that attempts to provide thyroid health related services, the functions are mostly single, mainly focusing on the following aspects: health knowledge popularization and information push, medication reminder and review schedule management, simple symptom recording and tracking. These applications cannot solve the core problem of thyroid nodule management, i.e. lack of evidence-based medical support, quantifiable home intervention means.

[0004] A few studies attempt to apply traditional therapies such as Chinese massage to the adjuvant therapy of thyroid nodules, but there are obvious limitations in actual operation: on the one hand, patients have difficulty in accurately finding neck-related acupoints such as Renying, Tiantu, Futu and foot reflex zones, resulting in large randomness of operation; on the other hand, the intensity, frequency and duration of massage cannot be quantified and standardized, and the effect is difficult to guarantee and evaluate. More importantly, these operations lack objective data recording, and doctors cannot obtain any process data when inquiring, and can only rely on the patient's subjective description and ultrasound image results for judgment.

[0005] In terms of technology, augmented reality technology has been applied in the fields of medical training and anatomy education, but its application in the field of chronic disease home management is still blank. Similarly, the application of artificial intelligence technology in the generation of personalized intervention programs is still in the preliminary exploration stage, and there is no report of combining AR technology and intelligent hardware for thyroid nodule management.

[0006] Therefore, the AR navigation and DRL based thyroid nodule home intervention system and method are proposed to solve the problems in the prior art. SUMMARY

[0007] The AR navigation and DRL based thyroid nodule home intervention system and method are provided to solve the problem that the AR technology in the market is not applied to the chronic disease home management field.

[0008] To achieve the above object, the application provides the following technical scheme.

[0009] The AR navigation and DRL based thyroid nodule home intervention system comprises:

[0010] The client module is arranged on the user mobile terminal and is used for guiding the user to perform self-operation, collecting data and receiving feedback in a home environment.

[0011] The cloud service module is in communication connection with the client module through the Internet and is used for remote data processing, model operation and individualized intervention scheme management; supports the HL7FHIR standard open interface and is used for integration with a hospital information system and an ultrasonic equipment system; the model is updated incrementally every day for 24 hours, and the client supports edge reasoning and cache update content;

[0012] The doctor end module is in communication connection with the cloud service module and is used for providing a remote monitoring and management interface for medical staff.

[0013] The client module comprises:

[0014] The AR navigation unit is specifically configured to call a depth perception component of the user mobile terminal, fit based on real-time acquired facial key points and a preset neck statistical shape model, calculate and render and display a virtual mark of a target acupoint on the user body surface in real time, and guide the self-operation of the user.

[0015] The data collection and interaction unit is specifically configured to establish a communication connection with an external intelligent pressure sensing device, receive and display pressure time series data from the device in real time when the user operates, and provide a health care task execution interface.

[0016] The AI question and answer unit is specifically configured to support voice input interaction based on a thyroid field medical knowledge graph and a fine-tuning large model, automatically transfer human customer service for sensitive questions, and provide real-time health consultation.

[0017] The daily task and clock-in unit is specifically configured to generate an individualized massage task list, record the task completion state and support clock-in, issue points based on the completion rate, and use the points to exchange ultrasonic examination coupons.

[0018] The cloud service module comprises:

[0019] The DRL recommendation engine unit, specifically, constructs a state space with user attributes, nodule characteristics, symptom scores, and historical operation compliance data, constructs an action space with executable massage acupoint combinations, operation force, and operation frequency, constructs a reward function with predicted nodule volume changes and quality of life improvements, and generates a personalized thyroid health intervention scheme for user home execution through a trained strategy network.

[0020] Specifically, the state space includes user seven-day massage completion rate; in the reward function, the quality of life improvement weight λ is dynamically adjusted according to the user TI-RADS classification.

[0021] A thyroid nodule home intervention method based on AR navigation and DRL, comprising the following steps:

[0022] Step 1: remotely receiving user registration information and baseline medical data, analyzing and storing user baseline archives;

[0023] Step 2: based on the initial state of the user, generating an initial personalized thyroid health intervention scheme suitable for home execution through a DRL recommendation model, and issuing the scheme to the user client;

[0024] Step 3: in the user's home environment, guiding the user to accurately position the acupoints specified in the scheme through the AR navigation function of the client;

[0025] Step 4: in the process of user self-operation, collecting real-time biomechanical data through external equipment, and processing and extracting features of the data to generate real-time guidance feedback;

[0026] Step 5: record the behavior data of each self-operation of the user, and upload to the cloud server;

[0027] Step 6: the cloud server periodically aggregates user data, updates user state, and optimizes the DRL recommendation model using incremental learning to generate an optimized home intervention scheme for the next period;

[0028] Step 7: at the preset follow-up node, remotely remind the user to review, and generate a therapeutic effect analysis report according to the review results obtained remotely and the baseline data.

[0029] Compared with the prior art, the beneficial effects of the present application are:

[0030] The application constructs an innovative system of home management of thyroid nodules through multi-module cooperation, that is, through the AR navigation unit of the client, the facial feature points are fitted with the neck statistical shape model, the error of acupoint positioning is less than or equal to 3mm, and the problem of inaccurate acupoint positioning in traditional home operation is solved. And the intelligent pressure sensor and the data processing unit realize the quantification of the pressing force, frequency and other operation quantities through filtering, feature extraction and two-classification model, so that the effective pressing judgment accuracy is greatly improved, and the experience operation is converted into a standardized process.

[0031] The application constructs a complete hospital diagnosis, home intervention and remote supervision closed-loop management system, so that patients can still receive standardized and effective health management after leaving the hospital, changes the traditional mode of regular observation of thyroid nodules, and fills the gap in clinical management.

[0032] The application converts the traditional experience operation into a standardized and quantifiable digital therapy, the deep reinforcement learning algorithm used can dynamically adjust the intervention scheme according to the real-time data of each user, the system performs model incremental update every 24 hours, and the recommended scheme is always adapted to the latest state of the user.

[0033] The above summary is only for the purpose of the description and is not intended to limit in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the application will be readily apparent to those skilled in the art by reference to the drawings and the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 A flowchart of the method for home intervention of thyroid nodules based on AR navigation and DRL of the application;

[0035] Figure 2 A module block diagram of the home intervention system for thyroid nodules based on AR navigation and DRL of the application. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0037] Embodiment one

[0038] As shown in the accompanying drawings, Figure 2 A home intervention system for thyroid nodules based on AR navigation and DRL, comprising:

[0039] 1. A client module installed on a user's mobile terminal, used to guide the user to perform self-operation, collect data and receive feedback in a home environment;

[0040] In the user's mobile terminal, the user uploads a thyroid ultrasound report through the App, the system analyzes the report through the OCR and NLP engines, extracts the nodule location, size, and TI-RADS classification, automatically calculates the nodule volume, and the user completes the QoL-THY life quality scale to generate an initial symptom score.

[0041] For facial and neck data collection, start the TrueDepth camera, capture RGB-D images, call the MediaPipeFaceMesh model, extract 468 facial 3D key points, use the nose tip (Landmark1) and chin center (Landmark152) as stable points for coordinate system alignment and rigid transformation; load the pre-trained neck statistical shape model to generate a personalized 3D neck grid.

[0042] 2. A cloud service module connected to the client module through an Internet communication connection, used for remote data processing, model operation and personalized intervention plan management; supports HL7 FHIR standard open interface for integration with hospital information systems and ultrasound equipment systems; performs model incremental updates every day for 24 hours, and the client supports edge reasoning and cache update content;

[0043] 3. A doctor's end module connected to the cloud service module, used to provide a remote monitoring and management interface for medical staff;

[0044] 4. An external intelligent pressure sensing device, which collects pressure time series data F_raw(t) at a sampling rate of 50Hz and transmits it to the client through Bluetooth 5.2; uses a 10Hz second-order Butterworth low-pass filter to smooth F_raw(t) and filter out high-frequency noise to obtain the filtered signal F_filtered(t).

[0045] The sliding window size is 3s, containing 150 data points, and the root mean square value and peak factor are calculated;

[0046] Input the feature vector [RMS, CF] into the pre-trained random forest model, the training set contains 12,478 labeled samples, of which 8,923 are valid pressing samples;

[0047] The model contains 500 decision trees, each tree has a depth of at least 15 layers, and outputs the "effective pressing" prediction probability P.

[0048] When P≥0.85 and the condition is met for 5 consecutive windows, it is determined to be an effective operation. The model AUC=0.92, with an accuracy of 89.7%.

[0049] To train the DRL, first, the PPO-Clip algorithm is adopted, and the training period is 5000 epochs, and each epoch contains 1000 interaction steps;

[0050] The policy network is a 3-layer Transformer encoder with a hidden layer dimension of 256 and 8 attention heads per layer. The value network shares the first two layers of weights with the policy network, and the third layer outputs the state value estimate separately. The experience replay pool capacity is 1 million, and the network parameters are updated with a batch size of 256 samples each time. The clipping parameter is 0.2, and the initial learning rate is 5e-5, which decays with training.

[0051] The client module includes:

[0052] 1.1, AR navigation unit, through the MediaPipe library to extract multiple feature points of the face, based on the stable feature points of the nose tip and chin to align the coordinate system and rigid transformation, load the predefined statistical shape model SSM of the neck, and fit to generate a personalized three-dimensional grid model of the neck.

[0053] According to the pre-stored mapping relationship of acupoint anatomical position, the three-dimensional space coordinates of the target acupoints including the Heart-Contral Channel, the Heavenly Rush, the Discharge Spring, the Fushu, the Sanyinjiao and the Sanyinjiao are calculated on the three-dimensional grid model.

[0054] The positioning error of the three-dimensional space coordinates is maximally 3mm, and the voice broadcast acupoint avoidance area is supported.

[0055] Specifically, the depth perception component of the user's mobile terminal is called, the real-time acquired facial key points are fitted based on the pre-set statistical shape model of the neck, the virtual identification of the target acupoint is calculated and rendered in real time on the user's body surface, and the user's self-operation is guided; through the MediaPipe library to extract multiple feature points of the face, based on the stable feature points of the nose tip and chin to align the coordinate system and rigid transformation, load the predefined statistical shape model SSM of the neck, and fit to generate a personalized three-dimensional grid model of the neck.

[0056] In the acupoint positioning algorithm, the input is the RGB-D frame and the facial key points, and the process is: MediaPipeFaceMesh, coordinate system alignment, SSM fitting, and output of 6 major acupoint 3D coordinates.

[0057] The error is controlled to be maximally 3mm, and the voice prompt avoids the dangerous area; real-time rendering: using ARKit / ARCore to superimpose virtual identification on the user's body surface.

[0058] 1.2, data acquisition and interaction unit, low-pass filter processing is performed on the received raw pressure data stream; the time domain characteristics including root mean square RMS and peak factor are extracted from the filtered data; the time domain characteristics are input into the pre-trained binary classification machine learning model to determine whether the current operation is an effective pressing, and the determination result is fed back to the user interface in real time.

[0059] Specifically: a communication connection is established with an external intelligent pressure sensing device, and pressure time series data from the device during user operation is received and displayed in real time; a health care task execution interface is also provided;

[0060] 1.3, AI question and answer unit, specifically: based on the thyroid field medical knowledge graph and fine-tuning large model, voice input interaction is supported, sensitive questions are automatically transferred to human customer service, and real-time health consultation is provided;

[0061] 1.4, daily task and clock-in unit, specifically: a personalized massage task list is generated, the task completion status is recorded and clock-in is supported, points are awarded based on completion rate, and points are used to exchange ultrasound examination coupons.

[0062] The cloud service module includes:

[0063] 2.1, DRL recommendation engine unit, the algorithm used is the proximal policy optimization PPO algorithm, and the policy network and the value network both use the Transformer encoder architecture. The Transformer encoder is a 3-layer structure, the policy network and the value network share weights, and the clipping parameter ε=0.2 when using the PPO-Clip algorithm.

[0064] Specifically: the state space is constructed based on user attributes, nodule characteristics, symptom scores and historical operation compliance data, the action space is constructed based on executable massage acupoint combinations, operation force and operation frequency, the reward function is constructed based on predicted nodule volume changes and quality of life improvement, and the personalized thyroid health intervention scheme for user home execution is generated through the trained policy network.

[0065] Among them, the state space includes the user's seven-day massage completion rate; in the reward function, the quality of life improvement weight λ is dynamically adjusted according to the user's TI-RADS classification.

[0066] Embodiment two

[0067] 1. When the user initiates AR navigation, the phone camera captures a frame of 1080p RGB image and its corresponding depth map. The frame image is input into a lightweight MediaPipeFaceMesh model, which outputs normalized 468 3D landmark coordinates in 15ms. With the nose tip (landmark 1) and chin center (landmark 152) as stable reference points, a 4x4 transformation matrix is calculated by singular value decomposition algorithm to convert the feature points from model coordinate system to camera world coordinate system. The system pre-stores a statistical shape model based on 368 cases of neck CT data, and uses the aligned mandibular contour feature points (landmarks 334-353) as input to fit the user's personalized neck 3D grid through principal component analysis model. In the SSM model, the positions of each acupoint have been predefined as specific vertices on the grid. The system directly obtains the 3D world coordinates [x, y, z] = [0.123, -0.045, 0.387] of these vertices. Finally, using the SceneKit rendering engine of ARKit, a 3D glowing sphere virtual marker with a diameter of 8mm is drawn at this coordinate position, achieving accurate positioning with an error of <2.5mm.

[0068] Specifically, the system's test on 368 volunteers showed that the average positioning error of the Renying and Tiandu points was 2.1mm. For users with a neck fat thickness greater than 2cm, the SSM model can additionally compensate for subcutaneous tissue deformation, and the error can be controlled within 2.5mm. The voice prompt module will detect the carotid artery position in real time and trigger the broadcast of "move up 0.5cm to avoid the carotid artery" when the user is close to it.

[0069] 2. The raw pressure signal F_raw(t) from the external intelligent pressure sensing device is first filtered by a 2nd order Butterworth low-pass filter with a cutoff frequency of 10Hz to obtain the smoothed signal F_filtered(t). Within a 3-second wide sliding window (150 data points), the root mean square value (RMS=3.8N) and the crest factor (CrestFactor=2.1) of F_filtered(t) are calculated in real time. The feature vector [3.8, 2.1] is input into a pre-trained random forest classifier, which is trained on 12,478 labeled samples containing 500 decision trees. The classifier outputs a prediction probability P(effective)=0.93. Since the threshold value P>0.85 and 5 consecutive windows are judged to be "effective pressing", the system gives continuous green waveform feedback on the UI. When a window is judged to be invalid due to a high peak factor CrestFactor=4.8>3.0 (P(effective)=0.21), the system immediately triggers the linear vibration motor in the device to produce a 100ms vibration pulse, achieving millisecond-level tactile alarm.

[0070] Specifically, the policy network of the DRL model adopts a 3-layer Transformer encoder with a hidden layer dimension of 256 and 8 attention heads per layer; the value network shares the first 2 layers of weights with the policy network, and the 3rd layer outputs the value estimate separately. During training, the model is initialized with 1 million virtual interaction data, fine-tuned with 2,147 clinical data, and incrementally updated every 24 hours using the user's operation data. The client locally caches the model parameters for nearly 7 days to reduce latency.

[0071] More specifically, in the input layer of the acupoint positioning algorithm, the RGB-D frame collected by the mobile TrueDepth camera; the 468 facial key points are output synchronously.

[0072] In the processing layer, first, take the nose tip (landmark 1) and chin center (landmark 152) as the reference points, calculate the 3x3 rotation matrix and 3x1 translation vector through singular value decomposition, and construct the conversion matrix from the model coordinate system to the camera world coordinate system, to realize the rigid alignment of the face and neck coordinate system.

[0073] Then, load the SSM based on 368 neck CT data, which contains 100 principal components with an explanation rate of ≥95%, input the aligned mandibular contour feature points (landmarks 334-353), and fit the user's personalized neck 3D mesh model through least squares iterative optimization.

[0074] Finally, according to the pre-stored acupoint-mesh vertex mapping relationship, the world coordinates of the target acupoint are directly extracted from the fitted 3D mesh, and converted to screen 2D coordinates through the ARKit rendering engine, and superimposed with a green virtual circle.

[0075] In the output layer, the 3D spatial coordinates of the 6 major acupoints and the thyroid reflex area of the foot, with a maximum positioning error of 3mm.

[0076] Example Three

[0077] As shown in the accompanying Figure 1 , the user's process for home intervention of nodules is as follows:

[0078] User Zhang, female, 38 years old, physical examination found that there was a TI-RADS 3 nodule in the right lobe of the thyroid gland, with a size of 6.2x4.8x5.0mm. The specific process of using the AR navigation and DRL-based thyroid nodule home intervention system in the home environment is as follows:

[0079] 1. Zhang downloads and installs the APP of the system through the mobile phone application store at home, completes the mobile phone number registration and identity verification. According to the system instructions, he uses the mobile phone to shoot and upload his latest thyroid ultrasound report. The OCR and NLP analysis engine of the system automatically extracts the key information from the report: nodule location: right lobe, size: 6.2x4.8x5.0mm, TI-RADS classification: 3. According to the ellipsoid formula, the initial volume of the nodule is automatically calculated . Then, he completes the thyroid disease special quality of life scale evaluation built in the system, and the score is 52. All data are uploaded to the cloud after encryption to form his exclusive digital health record.

[0080] 2. The cloud deep reinforcement learning recommendation engine receives Zhang's initial state vector s0: {age: 38, BMI: 22.1, TI-RADS: 3, nodule volume: 0.078, QoL score: 52, historical compliance rate: 0}. The engine inputs s0 into the trained PPO policy network with a 3-layer Transformer encoder structure and a hidden layer dimension of 256. After network inference, the action vector a0 is output, which generates his first 14-day personalized health care plan after decoding: ① perform 1 time per day, at night; ② acupoints and operations: bilateral Anmian points, 2.5 minutes each, intensity 3.8N, clockwise rubbing and pressing, Tianfu point 1.5 minutes, intensity 2.8N, gentle tapping, bilateral thyroid reflex areas of the feet, 3.5 minutes each, pushing from bottom to top. The plan is delivered to his mobile phone APP through HTTPS protocol.

[0081] 3. Every night before going to bed, open the APP and enter "today's task" and click "start AR navigation". The rear TrueDepth camera of the mobile phone is started, and the AR navigation unit detects 468 feature points of the face in real time through the MediaPipeFaceMesh model, calculates the coordinates of the neck acupoints, and displays a virtual green circle on the screen to accurately cover the Anmian and Tianfu points on the neck, with a voice prompt "please rub and press in this area". Wear the smart finger pressing device, which is connected to the APP through Bluetooth 5.2. During the operation, the device collects pressure data F(t) at a sampling rate of 50Hz and transmits it to the APP in real time. The APP data processing unit performs 10Hz low-pass filtering on F(t) and calculates the real-time root mean square value. When the pressure is maintained in the range of 3.5-4.2N, the screen displays a green waveform; when the pressure reaches 5.8N at a certain moment, a red spike appears on the waveform, and the device produces a short vibration to remind that the intensity is too large.

[0082] 4. After each task is completed, the complete data of the operation is encrypted and uploaded to the cloud. The cloud engine aggregates all user de-identified data every morning and initiates an incremental learning process to fine-tune and optimize the DRL model using new data. After 14 consecutive days of check-in, 100 points are awarded, and the system automatically pushes a "50-yuan ultrasound examination coupon" redeemable at nearby hospitals. Clicking on the coupon will take you to the hospital's appointment system.

[0083] 5. After using the system for 3 months, the system reminds Zhang through an APP push and a text message to make an appointment for a review. After the review at the local hospital, the new report is uploaded to the system. The system analysis shows that the nodule volume has decreased to 0.062 cm 3 , with a volume change rate AV = -20.5%. The QoL score has improved to 38. The cloud engine receives this positive feedback and stores it as a high-reward-value sample in the experience replay pool. The doctor sees the automatically generated follow-up report on the doctor's end system, which shows that the nodule volume has significantly decreased and the quality of life has significantly improved, and remotely approves the system's automatically generated next-stage optimization plan.

[0084] Specifically, during the operation, the AI question-answering robot replies within 1.2 seconds based on the knowledge graph: "TI-RADS class 3 nodules that increase by less than 20% in the short term and have no malignant features should be observed first." When she asks if she can continue to massage during pregnancy, the system recognizes it as a sensitive question and jumps to an online consultation with a human thyroid specialist.

[0085] Embodiment Four

[0086] In a clinical trial at a certain third-level hospital, 126 TI-RADS class 3 nodule patients used the system for 3 months of home-based intervention management, and the results showed that:

[0087] (1) Nodule change: After intervention, the average nodule volume decreased by (15.8 ± 6.3)%, with 78.6% (99 / 126) of patients experiencing a volume reduction of >10%, 18.3% (23 / 126) of patients experiencing stable volume changes (within ±10%), and 3.2% (4 / 126) of patients experiencing a volume increase of >10%.

[0088] (2) Quality of life improvement: The QoL-THY score improved by an average of (21.7 ± 8.4) points (from a baseline of 52.3 ± 9.1 to 30.6 ± 7.2).

[0089] (3) Compliance and satisfaction: The average operation compliance rate was 85.7%, and the user satisfaction score was 4.6 / 5.0. 92.1% of patients reported a significant reduction in psychological anxiety.

[0090] (4) Clinical efficiency is improved: the efficiency of doctor follow-up is improved by about 45%, and the average follow-up time of each patient is shortened from 15.2 minutes to 8.4 minutes.

[0091] The above examples show that the system of the application can effectively realize the home intervention management of thyroid nodule patients, and shows good effect in promoting nodule shrinkage and improving life quality, and improves the utilization efficiency of medical resources.

[0092] Example Five

[0093] All data communication between the client and the cloud server is encrypted and transmitted using the TLS1.3 protocol, uses the AES-256-GCM algorithm to encrypt the data payload, and enables two-way certificate authentication to prevent man-in-the-middle attacks.

[0094] User static data is stored in the cloud database using the AES-256 algorithm, and the encryption key is managed uniformly by a special key management service, realizing the separate storage of data and keys.

[0095] The data set used for model training is protected by differential privacy technology, adding Laplace noise, and the privacy budget is set to epsilon=0.9, ensuring that no sensitive information of any single user can be inferred from the training data.

[0096] The system implements role-based access control, and all data access operations are recorded in an unalterable audit log, and regular security audits and vulnerability scans are performed.

[0097] The embodiments of the application are given for the purpose of example and description, although the embodiments of the application have been shown and described above, it is understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the application, and the changes, modifications, replacements and modifications of the above-mentioned embodiments within the scope of the application should be included in the protection scope of the application.

Claims

1. An AR navigation and DRL-based thyroid nodule home intervention system, characterized in that, The application relates to a thyroid health care system, which comprises: a client module arranged on a user mobile terminal, used for guiding the user to perform self-operation, collecting data and receiving feedback in a home environment; a cloud service module in communication connection with the client module through the Internet, used for remote data processing, model operation and individualized intervention scheme management; supporting the HL7 FHIR standard open interface, used for integration with a hospital information system and an ultrasonic equipment system; a doctor terminal module in communication connection with the cloud service module, used for providing a remote monitoring and management interface for medical staff; wherein the client module comprises: an AR navigation unit, which is specifically configured to call a depth perception component of the user mobile terminal, fit based on real-time acquired facial key points and a pre-defined neck statistical shape model, calculate and render a virtual mark of a target acupoint on the user body surface in real time, and guide the self-operation of the user; a data collection and interaction unit, which is specifically configured to establish a communication connection with an external intelligent pressure sensing device, receive and display real-time pressure time series data from the device when the user operates, and provide a health care task execution interface; an AI question and answer unit, which is specifically configured to support voice input interaction based on a thyroid field medical knowledge graph and a fine-tuned large model, automatically transfer to a human customer service for sensitive questions, and provide real-time health consultation; a daily task and punch-in unit, which is specifically configured to generate an individualized massage task list, record the task completion status and support punch-in, issue points based on the completion rate, and use the points to exchange ultrasonic examination coupons; the cloud service module comprises: a DRL recommendation engine unit, which is specifically configured to construct a state space with user attributes, nodule characteristics, symptom scores and historical operation compliance data, construct an action space with executable massage acupoint combinations, operation force and operation frequency, construct a reward function with predicted nodule volume changes and life quality improvement, and generate an individualized thyroid health care intervention scheme for the user to execute at home through a trained strategy network; specifically, the state space comprises a seven-day massage completion rate of the user; in the reward function, the life quality improvement weight lambda is dynamically adjusted according to the TI-RADS classification of the user.

2. The AR navigation and DRL-based thyroid nodule home intervention system of claim 1, wherein, The algorithm used by the DRL recommendation engine unit is a proximal policy optimization (PPO) algorithm, and the strategy network and the value network of the PPO algorithm both adopt a Transformer encoder architecture; the Transformer encoder is a 3-layer structure, the strategy network and the value network share weights, and the clipping parameter epsilon of the PPO-Clip algorithm is 0.

2.

3. The AR navigation and DRL-based thyroid nodule home intervention system of claim 1, wherein, The cloud service module further comprises a data security unit, which is specifically configured to encrypt the stored user static data by using an AES-256 algorithm, and add noise to a data set used for model training by using a differential privacy technology with a privacy budget epsilon of 1.

0.

4. The AR navigation and DRL-based thyroid nodule home intervention system of claim 1, wherein, The AR navigation unit extracts multiple feature points of a face through a MediaPipe library, aligns and rigidly transforms coordinate systems based on stable feature points of a nose tip and a chin, loads a pre-defined neck statistical shape model (SSM), and fits to generate an individualized neck three-dimensional grid model. According to the pre-stored mapping relationship between the anatomical positions of the acupoints, three-dimensional space coordinates of the target acupoints of the Renying acupoint, the Tiantu acupoint, the Lianquan acupoint, the Futu acupoint, the Zusanli acupoint and the Sanyinjiao acupoint are calculated on the three-dimensional grid model; The positioning error of the three-dimensional space coordinates is less than or equal to 3mm, and the acupoint avoidance area is broadcasted through voice.

5. The AR navigation and DRL-based thyroid nodule home intervention system of claim 1, wherein, The data acquisition and interaction unit specifically comprises: The received original pressure data stream is subjected to low-pass filtering processing; The time domain characteristics of the root mean square (RMS) and the peak factor are extracted from the filtered data; The time domain characteristics are input into a pre-trained binary classification machine learning model to determine whether the current operation is an effective pressing, and the determination result is fed back to the user interface in real time.

6. The AR navigation and DRL-based thyroid nodule home intervention system of claim 5, wherein, The binary classification machine learning model is a random forest model, the sampling frequency of the received pressure data is 50Hz, and the cutoff frequency of the low-pass filtering is 10Hz.

7. The AR navigation and DRL-based thyroid nodule home intervention system of claim 1, wherein, Further comprising: An external intelligent pressure sensing device internally provided with a pressure sensor and an inertial measurement unit (IMU) and communicating with the client module through a Bluetooth protocol, used for collecting and uploading pressure data during user operation.

Images