An augmented reality-based intelligent electrocardiogram lead positioning system and method

Abstract

This invention belongs to the field of biomedical engineering, and specifically discloses an intelligent electrocardiogram (ECG) lead positioning system and method based on augmented reality. It aims to address the clinical pain points of traditional ECG chest lead (V1-V6) positioning, which relies on experience, has large errors (±1-2cm), and is time-consuming (3.2±0.8 minutes). The system includes: (1) an AR display device (integrating a high frame rate camera and environmental sensors); (2) a modified lead patch (with an optically recognizable reflective array / QR code on the surface, physically isolated from the electrodes); (3) a processing unit (running a multi-scale feature fusion CNN model to identify anatomical landmarks such as the sternal angle and intercostal spaces in real time); (4) a dynamic projection engine and a body position adaptive calibration module to correct the projection position; and (5) a data communication module (encrypting data transmission to the ECG machine and generating a quality control report). The method achieves accurate lead positioning through AR scanning, dynamic projection guiding patch application, real-time offset detection, and feedback adjustment.

Description

Technical Field

[0001] This invention belongs to the field of biomedical engineering, and specifically discloses an intelligent electrocardiogram lead positioning system and method based on augmented reality. Background Technology

[0002] 1. Clinical challenges in ECG lead localization Electrocardiography (ECG), as a core diagnostic tool for cardiovascular diseases, relies heavily on the precise localization of the chest leads (V1-V6) patches for signal quality. Traditional localization methods require medical staff to manually palpate anatomical landmarks (such as the sternal angle, intercostal spaces, and midclavicular line), which has three inherent drawbacks: High error rate: Obese patients and patients with chest deformities have blurred body landmarks, and the error rate of novice operators is >30% (clinical research data). Furthermore, changes in body position or respiratory movements can cause the patch to shift by ±1-2cm, which directly affects the accuracy of ST segment analysis. Inefficient: A single V1-V6 positioning takes 3-4 minutes (delaying the golden period for rescue in emergency situations). Lack of standardization: Reliance on operator experience, lack of real-time visualization tools, and a qualification rate of less than 68% in primary hospitals.

[0003] Existing auxiliary solutions (such as anatomical atlases and positioning molds) only simplify the disinfection process and do not solve the core positioning problem in dynamic environments.

[0004] 2. Limitations of AR Medical Technology Applications Augmented reality (AR) technology has made progress in areas such as surgical navigation and anatomy teaching, but its application in electrocardiogram lead localization still faces key bottlenecks: Static image dependence: Existing AR medical systems (such as National Taiwan University Light Field AR and Meilin H2H system) are mostly based on CT / MRI to reconstruct static models, lacking dynamic compensation mechanisms such as breathing and body movement, and cannot adapt to real-time positioning requirements; High hardware costs: It relies on dedicated AR glasses such as HoloLens (unit price > $3500), and the refresh rate < 30Hz causes projection ghosting; Algorithm disconnect: The multi-lead ECG algorithm focuses on signal analysis but is not combined with lead spatial positioning. Furthermore, existing AR marking technologies (such as QR code patches) are easily affected by skin folds, resulting in a detection error >0.5cm.

[0005] Studies have shown (IEEE TMI 2023) that traditional AR projections drift by 4.7 mm during respiratory movements, far exceeding the tolerance of ECG patches (<2 mm).

[0006] 3. Technological Gaps and Innovation Opportunities There are two major gaps in the current field: Lack of dynamic compensation mechanism: There is currently no technology that can compensate for soft tissue deformation caused by breathing / position in real time; Insufficient collaboration between medical-grade AR hardware: The lead patch is not deeply coupled with the AR system, and adding optical markers to the existing patch will interfere with the conductivity of the electrodes (impedance increase >15Ω). Summary of the Invention

[0007] To address the aforementioned problems, this invention provides an intelligent electrocardiogram lead localization system and method based on augmented reality. This invention achieves a technological breakthrough for the first time through an anatomical dynamic recognition algorithm, an improved lead patch structure, and a cross-platform AR engine, realizing "millimeter-level dynamic localization, low-cost hardware adaptation, and a closed-loop medical data system," thus filling the aforementioned gap. To achieve the above objectives, the technical solution adopted by the present invention is as follows: An augmented reality-based intelligent electrocardiogram lead localization system, characterized in that it includes: (a) AR display device: with built-in camera and environmental sensors; (b) Six independent modified lead patches V1-V6: Each patch surface is provided with an optically identifiable mark, which is a reflective dot or a QR code array, and the marked area is physically isolated from the ECG electrodes; (c) Processing unit: runs anatomical landmark recognition algorithm and dynamic projection engine; (d) Projection unit: Real-time overlay of positioning markers of leads V1-V6 onto the AR display device; (e) Data interoperability module: Communicates with the electrocardiograph and generates a lead connection quality report; The anatomical landmark recognition algorithm in processing unit (c) employs a multi-scale feature fusion convolutional neural network (CNN), including: The first branch network extracts global features of the sternal angle and clavicle; The second branch network focuses on the local texture features of the interrib gaps; The feature fusion layer dynamically weights and outputs the coordinates of anatomical landmarks, with a positioning error ≤0.3cm; The dynamic projection engine in processing unit (c) includes: (c1) Respiratory movement compensation submodule, which tracks the movement trajectory of the chest cavity in real time through a camera and adjusts the projection position; (c2) Body position adaptive calibration submodule, which identifies supine / sitting posture based on device gyroscope data and dynamically corrects the projection angle.

[0008] The above-mentioned technical solution is the first AR lead-based positioning closed-loop system: through the design of physical isolation between optical markers and electrodes (patch), it solves the problem of interference with electrode conductivity in traditional AR QR codes (impedance increase <3Ω). Furthermore, it features dynamic accuracy assurance: multi-scale CNN (global + local feature fusion) controls the anatomical landmark recognition error to ≤0.3cm, overcoming the clinical pain point of blurred body surface landmarks in obese patients; the above scheme also has a real-time feedback mechanism: respiratory compensation (c1) and body position calibration (c2) sub-modules, eliminating the drift (>4.7mm) of traditional static AR models under respiratory / body movement.

[0009] Furthermore, in the aforementioned system, the optically identifiable markings of the improved lead patch are a ring-shaped reflective array. Medical conductive adhesive is embedded at the center of the array as an ECG electrode. A ring-shaped insulating groove is provided between the reflective array and the conductive adhesive. The density of the reflective array is positively correlated with the importance of the lead position, with V3 and V4 having a higher density than V1 and V6. The insulating groove design completely isolates the reflective material from interference with the electrodes (traditional patches have an impedance increase of >15Ω), while the graded density of the reflective array (high density in V3 / V4) improves the positioning accuracy of key leads. The ring-shaped layout avoids the defect of single-point reflection being easily blocked (e.g., by skin folds), ensuring robust identification under complex body surface conditions.

[0010] Furthermore, in the aforementioned system, the annular insulating groove of the improved lead patch is filled with a flexible reflective material. This material has a reflectivity of ≥90% in the infrared spectrum, and the surface in contact with the skin is covered with a medical pressure-sensitive adhesive layer. The high infrared reflectivity, combined with the infrared camera of the AR device, maintains a recognition rate of >99% in dark environments (<5 Lux), such as operating rooms. The medical pressure-sensitive adhesive improves the comfort of patch adhesion and reduces allergic reactions (compared to traditional conductive adhesives that directly contact the skin).

[0011] Furthermore, the aforementioned system also includes an error feedback mechanism: When the camera detects that the actual position of the patch deviates from the projected mark by more than 0.5cm, a multimodal alarm is triggered. The multimodal alarm consists of screen bright flashing, voice prompts, and vibration warnings. The system automatically records deviation data and uploads it to the cloud to optimize the model. In the above solution, multi-sensory alarms forcefully intervene to correct errors, avoiding the accumulation of errors by novice medical staff (the error rate of traditional methods is >30%); the cloud collects deviation data to continuously train the CNN model, dynamically improving positioning accuracy (such as optimization for special body types).

[0012] Furthermore, the aforementioned system also incorporates a medical data security architecture: End-to-end encrypted transmission of patch position data is employed; Real-time calculations are performed on the local device, and the raw image data does not leave the device. The dynamic compliance verification module complies with HIPAA and ISO 13485 standards. Local destruction of original images meets GDPR / HIPAA requirements, preventing the risk of leakage of patient chest images; encryption ensures low-latency communication (<200ms) by transmitting only coordinate data (not images).

[0013] Furthermore, the above system supports cross-platform adaptation to low computing power requirements: By using layered rendering technology, an AR projection frame rate of ≥60FPS can be maintained on low-end devices; The anatomical recognition algorithm employs a lightweight MobileNetV3 model, achieving a latency of ≤200ms on devices with 4GB of memory. In some embodiments, it runs smoothly on mid-to-low-end devices (tablets costing around 1000 RMB), breaking down the cost barrier of dedicated hardware such as HoloLens. 25,000).

[0014] This invention also discloses an augmented reality-based intelligent electrocardiogram lead localization method, which utilizes the aforementioned system and includes the following steps: Step S1: Scan the patient's chest with an AR device and use a CNN model to identify the sternal angle, clavicle, and intercostal spaces in real time; Step S2: Calculate the standard positions of V1-V6 based on the coordinates of the anatomical landmarks and generate AR projection markers; Step S3: Guide the user to paste the lead patch according to the projection position in the order of V1→V2→V3→V4→V5→V6, and detect the patch offset in real time through optical marks; AR projection guides step by step (V1→V6 order), solving the problem of traditional palpation relying on experience.

[0015] Step S4: If the offset exceeds the threshold, dynamically adjust the projection position and provide a correction instruction; in some embodiments, real-time offset detection (threshold 0.5cm) + correction instruction feedback can improve the clinical accuracy to 98%.

[0016] Step S5: After verifying all patch positions, synchronize the data to the electrocardiograph and output a connection quality score.

[0017] Furthermore, in the above method, the dynamic adjustment of the projection mark in step S2 includes: Based on ARKit spatial anchoring technology, the projection coordinates are bound to the real-time position of the patient's body surface; Introducing a soft tissue deformation compensation factor, the calculation formula is: ΔP=k BMI sin(2πt / T), where ΔP is the position correction, k is the elasticity coefficient (k is clinically validated to be 0.3-0.5, which is positively correlated with chest wall elasticity), BMI is the patient's body fat index, and T is the respiratory cycle. Introducing BMI and respiratory cycle (T) to quantify soft tissue deformation (e.g., a 1.2mm outward shift during inspiration) solves the respiratory drift problem of traditional AR static projection; ARKit anchoring technology ensures that the projection moves with the patient's body surface (e.g., minor positional adjustments), eliminating the need for repeated scanning.

[0018] Furthermore, in the above method, the offset detection in step S3 uses: QR code recognition technology analyzes the orientation angle of the patch; Deformation analysis technology of reflective dot array is used to determine whether the patch is wrinkled or detached; deformation analysis (wrinkle / detachment) combined with position detection reduces signal distortion (in some embodiments, the target achievement rate for children / burn patients is >93%).

[0019] When attaching the lead patch in step S3: The system uses a camera to track the reflective markings of the pasted patches in real time; If the current numbered patch is not detected, the projection display of subsequent patches is paused until the deviation correction is completed. The current lead is forced to complete correction before the next lead is projected, avoiding the mis-order pasting issues of traditional processes.

[0020] This invention also discloses a storage chip, characterized in that the chip stores an executable program, which drives the system to implement the positioning method. Preferably, the chip integrates core algorithms such as respiratory compensation and body position calibration, enabling third-party devices (such as electrocardiographs) to quickly access the system.

[0021] Compared with the prior art, the present invention has the following beneficial effects: This invention achieves breakthroughs in precise positioning, efficiency improvement, and clinical applicability by integrating augmented reality (AR) spatial computing, multimodal anatomical landmark recognition algorithms, and medical-grade hardware collaboration technology. Based on a deep learning model (multi-scale feature fusion convolutional neural network), it identifies anatomical landmarks such as the sternal angle and intercostal spaces in real time, reducing the error of traditional palpation from ±1-2cm to ±0.3mm, making it particularly suitable for obese patients or those with chest deformities. Combined with ARKit dynamic anchoring and reflective marker collaborative calibration technology, it supports supine / lateral decubitus respiratory motion compensation, reducing lead patch placement time from the traditional 5-8 minutes to ≤2 minutes, and increasing clinical accuracy from 68% to 98%. Simultaneously, the lead patch integrates an ECG acquisition circuit, ensuring signal quality and simultaneously generating automated reports. Furthermore, it features a cross-platform AR engine adaptation solution (compatible with various iOS / Android models, 60FPS rendering) and an end-to-end encryption architecture (compliant with HIPAA / ISO). The 13485 standard significantly reduces equipment upgrade costs and data leakage risks, providing standardized and low-cost ECG lead positioning solutions for primary healthcare, and promoting the precise and widespread application of intelligent medical devices. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of an intelligent electrocardiogram lead positioning system based on augmented reality disclosed in this invention; Figure 2This is a schematic diagram of an intelligent electrocardiogram lead localization method based on augmented reality disclosed in this invention. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Glossary of terms in this invention: I. Core System Components AR (Augmented Reality): Using computer technology to overlay virtual information onto a real scene, this system uses it to project ECG lead positioning markers in real time.

[0025] Lead patches (V1-V6): Electrode patches attached to the chest for acquiring electrocardiogram (ECG) signals. V1 to V6 represent the six standard chest lead positions.

[0026] Optically identifiable markings: Reflective dots or QR code arrays on the patch surface for AR camera to track location (physically isolated from electrodes).

[0027] Annular insulating groove: An annular groove on the lead patch that isolates the reflective material from the conductive adhesive to prevent signal interference.

[0028] Processing unit: Hardware modules (such as embedded processors) that run anatomical recognition algorithms and dynamic projection engines.

[0029] II. Algorithms and Techniques CNN (Convolutional Neural Network): A deep learning model used to identify anatomical landmarks such as the sternal angle and intercostal spaces.

[0030] Multi-scale feature fusion: The dual-branch structure of CNN extracts global (sternum / clavicle) and local (intercostal) features simultaneously.

[0031] Dynamic projection engine: A software module that adjusts the AR projection position in real time, including breathing compensation and body position calibration sub-modules.

[0032] Respiratory movement compensation: The camera tracks chest wall movement and corrects the projection offset according to the formula ΔP=k·BMI·sin(2πt / T) (ΔP: correction amount; k: elasticity coefficient; BMI: body fat index; T: respiratory cycle).

[0033] Body position adaptive calibration: Based on gyroscope data, the patient's supine / sitting posture is identified, and the projection angle is dynamically adjusted.

[0034] Dynamic weighting of feature fusion layer: A network layer in CNN that integrates features from different branches, outputting anatomical coordinates with an error ≤0.3cm.

[0035] III. Medical Hardware and Security Medical conductive adhesive: The conductive material at the center of the lead patch, with an impedance ≤5Ω, ensuring the quality of ECG signals.

[0036] End-to-end encryption: Data is encrypted throughout the transmission process from the AR device to the electrocardiograph (compliant with HIPAA standards).

[0037] HIPAA (Health Insurance Portability and Accountability Act): U.S. healthcare data security regulations that require the protection of patient privacy.

[0038] ISO 13485: International standard for quality management systems for medical devices.

[0039] Impedance: The resistance value between the patch and the skin. The standard requirement is <5kΩ (this system is optimized to <5Ω).

[0040] IV. Clinical Parameters and Indicators Louis's angle: The prominence at the junction of the manubrium and body of the sternum, a key landmark for V2 lead localization.

[0041] Intercostal space: The soft tissue space between two ribs, used to determine the height of leads V3-V6.

[0042] BMI (Body Fat Index): Weight (kg) / Height (m) 2 This is used to quantify the impact of soft tissue deformation on positioning.

[0043] Signal-to-noise ratio (SNR): The ratio of the intensity of useful components to noise in an electrocardiogram signal, measured in dB (≥34.7 dB in this system).

[0044] ST segment analysis: A key waveform in ECG diagnosis of myocardial ischemia, which depends on the accuracy of lead position.

[0045] V. Extended Technical Terminology Lightweight MobileNetV3: A streamlined CNN model suitable for mobile devices with a latency of ≤200ms.

[0046] Layered rendering technology: Reduces the precision of AR models on low-computing-power devices to maintain a frame rate of ≥60FPS.

[0047] ARKit Spatial Anchoring: A feature of Apple's AR development framework that binds virtual coordinates to the surface of a real object.

[0048] Deformation analysis technology: Determines whether a patch is wrinkled or detached by deforming the reflective dot array.

[0049] Negative pressure bag: A miniature adsorption device in the burn patient adaptation program, generating ≤0.1 N / cm². 2 Safe suction power.

[0050] VI. Organizations and Standards AHA (American Heart Association): The authoritative body that sets the standards for electrocardiogram lead positioning.

[0051] IEC 60601-2-25: International standard for safety of medical electrocardiogram equipment.

[0052] GDPR (General Data Protection Regulation): The EU's patient privacy protection regulations, which require the local destruction of original images.

[0053] Example 1 like Figure 1 As shown, an augmented reality-based intelligent electrocardiogram lead localization system includes: (a) AR display device: with built-in camera and environmental sensors; (b) Six independent modified lead patches V1-V6: Each patch surface is provided with an optically identifiable mark, which is a reflective dot or a QR code array, and the marked area is physically isolated from the ECG electrodes; (c) Processing unit: runs anatomical landmark recognition algorithm and dynamic projection engine; (d) Projection unit: Real-time overlay of positioning markers of leads V1-V6 onto the AR display device; (e) Data interoperability module: Communicates with the electrocardiograph and generates a lead connection quality report; The anatomical landmark recognition algorithm in processing unit (c) employs a multi-scale feature fusion convolutional neural network (CNN), including: The first branch network extracts global features of the sternal angle and clavicle; The second branch network focuses on the local texture features of the interrib gaps; The feature fusion layer dynamically weights and outputs the coordinates of anatomical landmarks, with a positioning error ≤0.3cm; The dynamic projection engine in processing unit (c) includes: (c1) Respiratory movement compensation submodule, which tracks the movement trajectory of the chest cavity in real time through a camera and adjusts the projection position; (c2) Body position adaptive calibration submodule, which identifies supine / sitting posture based on device gyroscope data and dynamically corrects the projection angle.

[0054] Preferably, the optically identifiable markings of the improved lead patch are a ring-shaped reflective array, with medical conductive adhesive embedded in the center of the array as an ECG electrode. A ring-shaped insulating groove is provided between the reflective array and the conductive adhesive, and the arrangement density of the reflective array is positively correlated with the importance of the lead position, with the density of V3 and V4 being greater than that of V1 and V6.

[0055] Preferably, the annular insulating groove of the improved lead patch is filled with a flexible reflective material, which has a reflectivity of ≥90% in the infrared spectrum band and is covered with a medical pressure-sensitive adhesive layer on the surface in contact with the skin.

[0056] As a preferred option, an error feedback mechanism is also included: When the camera detects that the actual position of the patch deviates from the projected mark by more than 0.5cm, a multimodal alarm is triggered. The multimodal alarm consists of screen bright flashing, voice prompts, and vibration warnings. The system automatically records deviation data and uploads it to the cloud to optimize the model.

[0057] As a preferred option, a medical data security architecture is also provided: End-to-end encrypted transmission of patch position data is employed; Real-time calculations are performed on the local device, and the raw image data does not leave the device. The dynamic compliance verification module complies with HIPAA and ISO 13485 standards.

[0058] Preferably, the above system supports cross-platform low-computing-power adaptation: By using layered rendering technology, an AR projection frame rate of ≥60FPS can be maintained on low-end devices; The anatomical recognition algorithm uses a lightweight MobileNetV3 model, with a latency of ≤200ms on a 4GB memory device.

[0059] like Figure 2 The illustrated method is an intelligent electrocardiogram lead localization method based on augmented reality. This method requires the use of the aforementioned system and includes the following steps: Step S1: Scan the patient's chest with an AR device and use a CNN model to identify the sternal angle, clavicle, and intercostal spaces in real time; Step S2: Calculate the standard positions of V1-V6 based on the coordinates of the anatomical landmarks and generate AR projection markers; Step S3: Guide the user to paste the lead patch according to the projection position in the order of V1→V2→V3→V4→V5→V6, and detect the patch offset in real time through optical marks; Step S4: If the offset exceeds the threshold, dynamically adjust the projection position and provide a correction instruction. Step S5: After verifying all patch positions, synchronize the data to the electrocardiograph and output a connection quality score.

[0060] Preferably, the dynamic adjustment of the projection marks in step S2 includes: Based on ARKit spatial anchoring technology, the projection coordinates are bound to the real-time position of the patient's body surface; Introducing a soft tissue deformation compensation factor, the calculation formula is: ΔP=k BMI sin(2πt / T), where ΔP is the position correction, k is the elasticity coefficient, BMI is the patient's body fat index, and T is the respiratory cycle.

[0061] Preferably, the offset detection in step S3 uses: QR code recognition technology analyzes the orientation angle of the patch; Deformation analysis technology of reflective dot arrays is used to determine whether the patch is wrinkled or detached; When attaching the lead patch in step S3: The system uses a camera to track the reflective markings of the pasted patches in real time; If the current numbered patch is not detected, the projection display of subsequent patches will be paused until the deviation correction is completed.

[0062] Example 2 System Architecture and Core Components This embodiment provides an augmented reality-based intelligent electrocardiogram lead localization system, including the following core components: 1. AR display devices: It uses lightweight AR glasses or mobile terminals (such as iPads) with a high refresh rate (≥60Hz) and a built-in 1080P camera, infrared sensor, gyroscope and accelerometer.

[0063] Preferred solution: Use customized AR glasses that support infrared spectral imaging to enhance the ability to recognize reflective markers.

[0064] Another preferred option: It uses the Apple ARKit 6.0 or Google ARCore 9.0 engine and is compatible with both iOS and Android devices (such as the iPad Pro M2 chip or Samsung Galaxy Tab S9).

[0065] Infrared enhancement module: Customized 850nm infrared camera (1280×720 resolution@60FPS) to penetrate the epidermis and identify deep anatomical landmarks.

[0066] Motion sensor: Built-in 6-axis IMU (gyroscope + accelerometer), sampling rate ≥200Hz, real-time capture of body position changes (supine / sitting angle deviation ≤0.5°).

[0067] 2. Improved lead patch (V1-V6): The patch uses a layered design: Surface layer: Circular reflective array, made of flexible material with high reflectivity (≥90% in infrared band), array arrangement density is graded according to the importance of leads (V3, V4 density > V1, V6). Middle layer: Annular insulating groove, filled with medical-grade silicone to isolate the reflective layer and electrodes; Bottom layer: Medical conductive gel electrode, with silver nanowires ensuring impedance ≤5Ω.

[0068] Preferred: Three-layer composite design: Reflective layer: The ring array uses a microprism reflective film (3M). Scotchlite Infrared reflectivity ≥92%, array density grading: V3 / V4: 120 points / cm 2 (High-precision area) V1 / V6: 80 points / cm 2 (Standard Area) Insulating layer: Annular silicone groove (width 0.5mm), which isolates the reflective layer from the electrode, and the impedance increase is <3Ω (actual measurement data).

[0069] Electrode layer: Silver nanowire conductive adhesive (impedance 2.5±0.3Ω), which is embedded with the insulating layer through laser micropore technology.

[0070] QR code assistance solution: A miniature QR code (1×1mm in size) is etched on the edge of the patch to store the lead ID and calibration parameters.

[0071] 3. Processing Unit: Anatomical landmark recognition algorithm: A two-branch CNN architecture is adopted: the first branch (ResNet-18) extracts global features of the sternal angle / clavicle; the second branch (MobileNetV3) focuses on local textures of the intercostal spaces. The feature fusion layer dynamically weights the output coordinates, with a positioning error of ≤0.3cm (measured data).

[0072] Dynamic projection engine: The respiratory compensation submodule tracks the movement trajectory of the chest cavity through a camera and corrects the projection position in real time using the formula ΔP = k·BMI·sin(2πt / T) (ΔP is the correction amount, k is the elasticity coefficient, and T is the respiratory cycle). Posture calibration submodule: Based on gyroscope data, it identifies supine / sitting positions and dynamically adjusts the projection angle.

[0073] 4. Error feedback mechanism: If the camera detects a deviation of more than 0.5cm between the actual position of the patch and the AR marker, a level 3 alarm will be triggered: The AR screen flashes a bright red warning frame. The voice prompt reads "Lead VX position offset"; AR device vibration warning.

[0074] The preferred alarm mechanism is as follows Table 1 Alarm Mechanism 5. Data security architecture: Location data is transmitted after being encrypted with AES-256. The original images are destroyed immediately after processing on the device (in compliance with GDPR Medical Annex). Example 3 1. Positioning Method and Flow The positioning method of the system described in Example 1 or 2 includes the following steps: Step S1: Anatomical landmark scanning By scanning the patient's chest with an AR device, a CNN model identifies the sternal angle, midclavicular line, and 4th intercostal space (key anchor points) in real time.

[0075] Step S2: Dynamic projection generation The V1-V6 coordinates are calculated based on international standards (AHA guidelines), and the projection is bound to the patient's body surface using ARKit spatial anchoring technology; The breathing compensation module adjusts the marker position based on real-time chest wall movement.

[0076] Step S3: Lead patch application and monitoring Guide the pasting process in the order of V1→V6, and track the reflective array of the patch using an infrared camera; Real-time detection technology: QR code recognition: Analyzing the orientation angle of the patch; Reflective array deformation analysis: to determine wrinkles / detachment; If the current numbered patch is not detected, pause subsequent projections until the correction is complete.

[0077] Step S4: Offset Correction When the offset is greater than 0.5cm, the projection position is dynamically adjusted and a voice prompt is given: "Please adjust the VX patch".

[0078] Step S5: Data Synchronization and Quality Control After verifying the location of all patches, the coordinates are transmitted to the electrocardiograph via encrypted Bluetooth. The electrocardiograph provides feedback lead impedance data and generates a connection quality score (excellent / good / poor).

[0079] As a preferred option, the following example is provided: Step S1: Patient scanning and anatomical calibration The patient lies supine with the chest area exposed, and the AR device is used to scan the body surface at a distance of 30-50cm.

[0080] Key steps: The infrared camera automatically adjusts the exposure to 100-150 lux (to avoid overexposure due to reflections). The CNN model outputs the coordinates of 6 anatomical landmarks (sternal angle, midpoint between left and right clavicles, and 4th / 5th intercostal space) within 0.5 seconds. Construct a 3D point cloud model of the thoracic cavity (accuracy ±1mm). Step S2: Dynamic projection generation The rules for calculating V1-V6 coordinates are shown in Table 2 below. Table 2 Rules for Calculating Coordinates V1-V6 Demonstration of respiratory compensation: When inspiration (thoracic expansion) is detected, the projection marker automatically shifts outward by ΔP (approximately 1.2 mm for patients with BMI=28). Step S3: Patch application and real-time monitoring Reflective array deformation detection algorithm: def detect_deformation(reflect_array): # Calculate the variance of the distance between reflective points dist_variance = np.var(calculate_distances(reflect_array)) if dist_variance > 0.15: # Threshold determined through clinical testing return "WRINKLED" # Wrinkled state elif count_missing_points(reflect_array) > 10%: return "DETACHED" # Detached status else: return "NORMAL" Sequential control logic: The system only activates the projection of the next lead after the error of the current lead is less than 0.5cm. Step S4: Data Synchronization and Quality Control Report Connection quality scoring formula Score = 0.4xR pos +0.3xR imp +0.3xSNR R pos Location matching degree (0-1) R imp Impedance compliance rate (1 for impedance < 5kΩ) SNR: Normalized signal-to-noise ratio of electrocardiogram.

[0081] 2. Clinical validation data Test subjects: 52 patients (including 22 with BMI>30 and 5 with pectus carinatum). The results are shown in Table 3. Table 3 Comparison of Results Special case: In a patient with a BMI of 35, the traditional method mistakenly placed the V3 affix on the costal arch. This system automatically corrected the position by recognizing the texture of the intercostal spaces.

[0082] Example 4 Cross-platform adaptation and cost control solutions 1. Lightweight Engine Design Layered rendering technology: High-end equipment: Full-precision CNN model (ResNet18) + real-time point cloud rendering (60FPS) Low-end devices (such as Android tablets with 4GB of RAM): The anatomical recognition model was quantized to INT8 (size reduced from 18MB to 4.3MB). The projection uses a simplified grid (face count from 50k to 5k). The frame rate is stable at 45-50 FPS. 2. Lead patch mass production solution Table 4 Cost Comparison Table: Compatibility testing: It can be used plug-and-play with mainstream ECG machines such as GE MAC5500 and Philips PageWriter TC70.

[0083] Conclusions and Innovations: 1. Cost controllability: Despite a 133% increase in the cost per surface mount device (SMD), the annual consumable cost only increased. 540 per thousand people The incremental costs were offset by improved operational efficiency (the time for a single positioning was reduced from 5 minutes to 2 minutes, saving medical staff manpower costs).

[0084] 2. Technological barriers: The processing of reflective arrays and insulating layers are core manufacturing processes (which competitors cannot avoid). The cost structure proves the irreplaceability of the technology (traditional patches cannot achieve AR collaboration).

[0085] 3. Achieving Medical Compliance Dynamic compliance verification module workflow: Detect device location upon startup (use IP positioning when GPS is disabled). If located in the EU, GDPR encryption mode will be automatically enabled (key length increased to 384 bits). The operation record is signed locally and then uploaded to the HIPAA certified cloud server.

[0086] Example 5 This embodiment provides a memory chip, which adopts an eMMC 5.1 embedded package (size 11.5×13mm), and its internal division is as follows: Secure storage area: OTP memory for embedding lightweight MobileNetV3 model and breathing compensation algorithm (for ΔP formula: ΔP=k) BMI sin(2πt / T)) and body position calibration logic; Data interaction area: Receives chest movement data collected by the camera via the SPI bus and outputs the projection coordinates to the ARKit engine in real time; Encryption engine: The chip's physical ID is bound to the program key, and data self-destruction is triggered when the chip is illegally disassembled.

[0087] Workflow: After the device is started, the chip BootLoader loads the anatomical recognition algorithm into the AR device memory, and generates AR projection markers by dynamically calling the breathing compensation submodule (calculating ΔP).

[0088] Example 6 Special patient adaptation program In-depth analysis of technical issues Burn patients cannot have lead patches directly applied to their skin wounds. Children (especially those under 6 years old) have a chest size that is only 1 / 3 that of adults and their intercostal spaces are not ossified, resulting in traditional AR positioning errors as high as ±2.5cm. Existing solutions rely on empirical estimations, leading to ECG signal distortion rates >40%.

[0089] Innovative technical solutions 1. Non-contact positioning system Biodegradable stent design: Material: Medical-grade polylactic acid (PLA) 3D printed, 0.3mm thick, light transmittance ≥85% 2. Smart scaling algorithm for children Parametric modeling: Structure: Hexagonal honeycomb grid support (1.5mm aperture), surface sputtered with nano-zinc oxide reflective layer (92% infrared reflectivity) Adsorption mechanism: Miniature negative pressure cells (3mm in diameter) are arranged at the edge, and a micro-pump generates ≤0.1N / cm² adsorption.2 Adhesion force (wound safety threshold) Dynamic compensation enhancement: The respiratory compensation coefficient k is increased to 0.5 (formula: ΔP=0.5·BMI·sin(2πt / T)). The AR projection markers have been switched to a circular dashed frame (to avoid obstructing wound observation). Parametric modeling % Enter your age (in years) and weight (in kg) if Age < 12 Thorax_Scale = 0.82 + 0.03*Age; % Thorax scaling factor (derived from a database of pediatric thoracic CT scans, sample size ≥ 2000 cases) Rib_Sensitivity = 1.4 - 0.05*Age; % Rib intervertebral space detection sensitivity end CNN model optimization: The weights of the local branches (MobileNetV3) are increased by 40%, and the convolution kernel is changed from 5×5 to 3×3. The training dataset has been expanded to include over 2000 pediatric chest CT reconstruction models. Clinical validation data are shown in Table 5. Table 5 Clinical Validation Testing standards: The gold standard is the position of the CT 3D reconstruction leads, and the ECG signal-to-noise ratio is ≥30dB to be considered qualified.

[0090] Conclusions and Innovations: 1. Breakthrough precision: The error for children is 1.2mm (traditional AR ≥ 2.5mm), which is due to the age scaling algorithm (Thorax_Scale = 0.82 + 0.03*Age). A 93% pass rate for burn patients fills a gap in the industry (where wounds cannot be covered with patches → negative pressure stent adsorption).

[0091] 2. Clinical value: Signal qualification rate >93% meets AHA ECG acquisition standards (SNR≥30dB); The universality of the compensation coefficient k=0.5 (BMI range 18-35) was verified.

[0092] Example 7 Optimization of robustness in extreme scenarios Technical challenges In the bumpy environment of an ambulance, the equipment vibration frequency reaches 5-10Hz, and the low light in the operating room (<5 Lux) causes the recognition rate of reflective marks to plummet to 60%.

[0093] Innovative technical solutions 1) Multi-source motion compensation engine Jitter compensation algorithm: The device's angular velocity ω is collected in real time using a gyroscope, and the displacement amplitude A is calculated using the camera's optical flow method. 2) Additional correction amount for projected position: ΔD = 0.2·A·sin(2πωt) (The coefficient 0.2 was optimized after 200 bump tests) 3) Multispectral fusion recognition: Infrared illumination (850nm LED) + laser speckle positioning (Class 1 safety level) 4) Image fusion formula: I_fused = 0.7*IR + 0.3*Laser / / Weights are switched to 0.5:0.5 in low light. The formula parameters are defined in Table 6. Table 6. Formula Parameter Definitions Conclusions and Innovations: 1) Creativity of weighting coefficients: The 0.7:0.3 ratio has been rigorously verified in experiments, resulting in a 16dB improvement in signal-to-noise ratio under low light conditions; The dynamic correlation between weights and illumination creates an algorithmic barrier.

[0094] 2) Hardware Collaborative Innovation: The hardware cost of the dual-mode laser speckle + infrared imaging system only increases by ¥50 per device. Solve the problem of failure in low-light scenes in operating rooms / ambulances.

[0095] 2. Emergency Downgrade Mode Dynamic resource scheduling, see Table 7: Table 7 Dynamic Resource Scheduling Scheme Extreme environment tests are shown in Table 8. Table 8 Extreme Environment Testing Test equipment: Electromagnetic motion simulation table (frequency 5-15Hz), dark box illumination 0.1-5 Lux Conclusions and Innovations: 1. Robust breakthrough: The error in bumpy environments is 2.1mm (traditional AR > 5mm), which is derived from the shake compensation formula ΔD = 0.2·A·sin(2πωt); A 99% recognition rate validates the effectiveness of multispectral fusion (I_fused).

[0096] 2. Intelligent degradation mechanism: Dynamic scheduling increases device battery life by 3 times (a necessity in emergency scenarios); The simplified CNN model still maintains a frame rate of 40fps (competitors generally have a frame rate of <20fps).

[0097] Based on the above embodiments, it can be concluded that this invention, through a five-dimensional closed-loop technology chain of "anatomical dynamic recognition - AR virtual-real fusion - medical hardware collaboration - cross-platform optimization - data security", advances ECG lead localization from subjective palpation to a spatial computing paradigm, improving operational efficiency by more than 70% and expanding clinical coverage to more than 95% (including special populations such as children and burn victims), providing a precise and inclusive solution for primary healthcare.

[0098] The above are merely a few preferred embodiments of the present invention, described in a relatively specific and detailed manner, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A method for intelligent electrocardiogram lead localization based on augmented reality, characterized in that, The method utilizes an augmented reality-based intelligent electrocardiogram lead localization system. The system includes: (a) AR display device: with built-in camera and environmental sensors; (b) Six independent modified lead patches V1-V6: Each patch surface is provided with an optically identifiable mark, which is a reflective dot or a QR code array, and the marked area is physically isolated from the ECG electrodes; (c) Processing unit: runs anatomical landmark recognition algorithm and dynamic projection engine; (d) Projection unit: Real-time overlay of positioning markers of leads V1-V6 onto the AR display device; (e) Data interoperability module: Communicates with the electrocardiograph and generates a lead connection quality report; The anatomical landmark recognition algorithm in processing unit (c) employs a multi-scale feature fusion convolutional neural network (CNN), including: The first branch network extracts global features of the sternal angle and clavicle; The second branch network focuses on the local texture features of the interrib gaps; The feature fusion layer dynamically weights and outputs the coordinates of anatomical landmarks, with a positioning error ≤0.3cm; The dynamic projection engine in processing unit (c) includes: (c1) Respiratory movement compensation submodule: Tracks the movement trajectory of the chest cavity in real time through a camera and adjusts the projection position; (c2) Body position adaptive calibration submodule: Based on the gyroscope data in the AR display device, it identifies supine / sitting postures and dynamically corrects the projection angle; The optically identifiable markings of the improved lead patch are a ring-shaped reflective array. Medical conductive adhesive is embedded in the center of the array as an ECG electrode. A ring-shaped insulating groove is provided between the reflective array and the conductive adhesive. The arrangement density of the reflective array is positively correlated with the importance of the lead position, with V3 and V4 having a density greater than V1 and V6. The improved lead patch has a ring-shaped insulating groove filled with a flexible reflective material. This material has a reflectivity of ≥90% in the infrared spectrum band and is covered with a medical pressure-sensitive adhesive layer on the surface in contact with the skin. The system also includes an error feedback mechanism: When the camera detects that the actual position of the patch deviates from the projected mark by more than 0.5cm, a multimodal alarm is triggered. The multimodal alarm consists of screen bright flashing, voice prompts, and vibration warnings. The system automatically records deviation data and uploads it to the cloud to optimize the model; The system also features a medical data security architecture: End-to-end encrypted transmission of patch position data is employed; Real-time calculations are performed on the local device, and the raw image data does not leave the device. The dynamic compliance verification module complies with HIPAA and ISO 13485 standards. The system supports cross-platform low-computing-power adaptation: By using layered rendering technology, an AR projection frame rate of ≥60FPS can be maintained on low-end devices; The anatomical recognition algorithm uses a lightweight MobileNetV3 model, with a latency of ≤200ms on a device with 4GB of memory. Based on the above system, the augmented reality-based intelligent electrocardiogram lead localization method includes the following steps: Step S1: Scan the patient's chest with an AR device and use a CNN model to identify the sternal angle, clavicle, and intercostal spaces in real time; Step S2: Calculate the standard positions of V1-V6 based on the coordinates of the anatomical landmarks and generate AR projection markers; Step S3: Guide the user to paste the lead patch according to the projection position in the order of V1→V2→V3→V4→V5→V6, and detect the patch offset in real time through optical marks; Step S4: If the offset exceeds the threshold, dynamically adjust the projection position and provide a correction instruction. Step S5: After verifying all patch positions, synchronize the data to the electrocardiograph and output a connection quality score; The dynamic adjustment of the projection marks in step S2 includes: Based on ARKit spatial anchoring technology, the projection coordinates are bound to the real-time position of the patient's body surface; Introducing a soft tissue deformation compensation factor, the calculation formula is: ΔP=k BMI sin(2πt / T), where ΔP is the position correction, k is the elasticity coefficient, BMI is the patient's body fat index, and T is the respiratory cycle.

2. The method according to claim 1, characterized in that, The offset detection in step S3 uses: QR code recognition technology analyzes the orientation angle of the patch; Deformation analysis technology of reflective dot arrays is used to determine whether the patch is wrinkled or detached; When attaching the lead patch in step S3: The system uses a camera to track the reflective markings of the pasted patches in real time; If the current numbered patch is not detected, the projection display of subsequent patches will be paused until the deviation correction is completed.

3. A memory chip, characterized in that, The chip stores an executable program, which is used to implement the positioning method as described in claim 1.

Images