Individualized electrocardiogram operation process based on iconography examination and AI large model

Abstract

The invention relates to the field of electrocardiogram treatment, and discloses an individualized electrocardiogram operation process relying on iconography examination and an AI large model, which comprises the following steps: S1, chest CT image acquisition: a CT scanner is used for acquiring chest image data of a patient, the CT scanner can generate high-resolution multilayer CT images, and the data format comprises but is not limited to a DICOM format; s2, heart projection position analysis: carrying out image preprocessing, heart segmentation, heart center point calculation and heart edge positioning on the acquired CT image data; and S3, lead position optimization: generating a lead electrode position based on a heart projection position by using an AI large model. According to the method, the actual position and form of the heart of each patient in the thoracic cavity can be accurately determined through CT image acquisition and heart projection position analysis, and the optimized lead position is generated by utilizing the AI large model based on the individualized projection positions, so that the lead electrode can more accurately capture the electrical activity of the heart.

Description

Technical Field

[0001] The present invention relates to the field of electrocardiogram (ECG) treatment, and in particular to an individualized ECG operation process based on imaging examination and an AI large model. Background Art

[0002] The electrocardiogram (ECG) is an important clinical diagnostic tool used to record changes in the heart's electrical activity and has been the most widely used method for ECG recording. However, the traditional 12-lead system has certain limitations in positioning the lead electrodes, relying mainly on fixed standard positions that do not fully take into account the anatomical differences between individuals. Therefore, although the standard 12-lead system can provide useful information in most cases, in some specific cases, especially in patients with abnormal heart position and size, the traditional lead position may not accurately reflect the true electrical activity of the heart, thus affecting the accuracy of diagnosis.

[0003] The heart is a complex three-dimensional structure, and its position and shape in the chest cavity vary from person to person. For example, the heart may shift position due to congenital malformations, acquired pathologies (such as myocardial hypertrophy, cardiac enlargement), or other physiological factors (such as body shape changes). In this case, placing lead electrodes according to a fixed standard position may cause distortion of the ECG waveform or even miss important electrical activity information. Summary of the invention

[0004] In order to make up for the above shortcomings, the present invention provides an individualized ECG operation process based on imaging examination and AI large model, aiming to improve the problem mentioned in the prior art that "the standard 12-lead system can provide useful information in most cases, but in some specific cases, especially in patients with abnormal heart position and size, the traditional lead position may not accurately reflect the true electrical activity of the heart."

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: an individualized electrocardiogram operation process based on imaging examination and AI large model, comprising the following steps:

[0006] S1: Chest CT Image Acquisition

[0007] Using a CT scanner to collect chest image data of the patient, the CT scanner can generate high-resolution multi-slice CT images, and the data format includes but is not limited to DICOM format;

[0008] S2: Analysis of cardiac projection position

[0009] Perform image preprocessing, heart segmentation, heart center point calculation and heart edge positioning on the acquired CT image data;

[0010] S3: Lead position optimization

[0011] Generate lead electrode positions based on the heart projection position using an AI big model, and optimize the lead positions through simulation verification and fine-tuning, wherein the AI ​​big model includes but is not limited to a convolutional neural network (CNN), a generative adversarial network (GAN), and a recurrent neural network (RNN);

[0012] S4: Individualized ECG Recording

[0013] Place electrodes according to the optimized lead positions and record the ECG waveform;

[0014] S5: Data Integration and Analysis

[0015] Integrate the recorded ECG data with the CT image data, and use the AI ​​model for comprehensive analysis.

[0016] As a further description of the above technical solution:

[0017] The scanning layer thickness of the scanner used for CT image acquisition in S1 is 1 mm, the scanning layer thickness is 0.5 mm, the scanning range is from the sternum to the diaphragm, the scanning speed is 0.5 seconds / layer, and the radiation dose is sufficient to ensure the image quality. The scanning mode can be spiral scanning or non-spiral scanning. The appropriate scanning mode can also be selected according to the specific situation of the patient.

[0018] As a further description of the above technical solution:

[0019] In S2, the image preprocessing technology includes but is not limited to median filtering, Gaussian filtering, adaptive histogram equalization and wavelet transform, and the technology can significantly improve the quality of image data and reduce the impact of artifacts and noise.

[0020] As a further description of the above technical solution:

[0021] In S2, the heart segmentation model adopts an optimized deep learning model, including but not limited to U-Net, MaskR-CNN and ResNet, and the model can accurately extract the three-dimensional projection position of the heart in the CT image.

[0022] As a further description of the above technical solution:

[0023] In S2, the heart center point calculation method adopts the centroid algorithm or the geometric center algorithm, which can accurately calculate the center position of the heart in the chest cavity and provide a reliable reference for the optimization of the lead position.

[0024] As a further description of the above technical solution:

[0025] In S3, the AI ​​large model uses a convolutional neural network (CNN) or a generative adversarial network (GAN) to generate optimized lead positions. The model can generate the most suitable lead positions based on the three-dimensional projection position of the heart, thereby improving the accuracy and personalization level of the electrocardiogram.

[0026] As a further description of the above technical solution:

[0027] In S3, simulation verification is performed through a cardiac anatomical model and an electrophysiological model to ensure the rationality of the lead position and the accuracy of capturing the cardiac electrical activity. The cardiac anatomical model includes but is not limited to a three-dimensional model of the heart's front-back, left-right, and top-bottom, and the electrophysiological model includes but is not limited to a simulation model of the heart's electrical activity.

[0028] As a further description of the above technical solution:

[0029] In the S4, the electrocardiogram device has the characteristics of high fidelity and low noise, and can accurately record the electrocardiogram waveform of the optimized lead position. The electrocardiogram device includes but is not limited to a standard 12-lead electrocardiograph and a multi-lead electrocardiograph. The electrode placement method includes but is not limited to manual placement, robot-assisted placement, and automatic placement. The electrode placement has high accuracy, can ensure good contact with the skin, and reduce contact resistance and signal interference.

[0030] As a further description of the above technical solution:

[0031] In S5, an AI model is used to perform comprehensive analysis to generate an individualized ECG report. The comprehensive analysis includes but is not limited to ECG waveform feature extraction, abnormal waveform detection, reconstruction of a three-dimensional heart model, and simulation of cardiac electrical activity. The ECG waveform feature extraction includes but is not limited to extraction of QRS complex, ST segment, and T wave features. The abnormal waveform detection includes but is not limited to detection of abnormal waveforms such as ST segment elevation, T wave inversion, and QRS complex widening.

[0032] As a further description of the above technical solution:

[0033] The heart three-dimensional model reconstruction can generate a detailed three-dimensional model of the heart based on CT image data, and the heart electrical activity simulation can simulate the electrocardiogram waveform of the optimized lead position, thereby improving the accuracy and reliability of diagnosis.

[0034] The present invention has the following beneficial effects:

[0035] 1. In the present invention, through CT image acquisition and heart projection position analysis, the actual position and shape of each patient's heart in the chest cavity can be accurately determined. Based on these individualized projection positions, the AI ​​large model is used to generate optimized lead positions to ensure that the lead electrodes can more accurately capture the electrical activity of the heart.

[0036] 2. In the present invention, a detailed model of the heart is generated by three-dimensional reconstruction technology, and the positions of the heart edges between different ribs are marked on the model to ensure that the lead electrodes can be accurately placed in key areas of the heart. This optimized lead position can capture more electrocardiogram waveform features, reduce the omission of important electrical activity information, and improve the reliability of diagnosis.

[0037] 3. In the present invention, the individualized electrocardiogram recording method not only improves the accuracy of diagnosis, but also provides important reference information for doctors to formulate personalized treatment plans, thereby enhancing the application effect of personalized medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 The figure is a complete flowchart of an individualized electrocardiogram operation process based on imaging examination and AI large model in the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Embodiment 1

[0041] Reference Figure 1 The present invention provides an individualized electrocardiogram operation process based on imaging examination and AI large model, comprising the following steps:

[0042] S1: Chest CT Image Acquisition

[0043] A CT scanner is used to collect chest imaging data of the patient. The CT scanner can generate high-resolution multi-slice CT images. The data format includes but is not limited to the DICOM format. Specifically, the scanning layer thickness of the scanner for CT image acquisition is 1 mm, the scanning layer thickness is 0.5 mm, the scanning range is from the sternum to the diaphragm, the scanning speed is 0.5 seconds / layer, and the radiation dose only needs to ensure the image quality. The scanning mode is selected as spiral scanning or non-spiral scanning. The appropriate scanning mode can also be selected according to the specific situation of the patient;

[0044] S2: Analysis of cardiac projection position

[0045] The collected CT image data is subjected to image preprocessing, heart segmentation, heart center point calculation and heart edge positioning. Specifically, the image preprocessing technology includes but is not limited to median filtering, Gaussian filtering, adaptive histogram equalization and wavelet transform. Specifically, the median filtering reference formula is:

[0046] g(x,y)=median{f(x+i,y+j)|(i,j)∈W}

[0047] Among them, g is the processed image, f is the original image, and W is a window centered at (x, y); Gaussian filtering can refer to the formula:

[0048]

[0049] Among them, G(i,j) is a two-dimensional Gaussian distribution function, and k is the radius of the Gaussian kernel;

[0050] The adaptive histogram equalization method can refer to the formula:

[0051]

[0052] Among them, H is the cumulative distribution function of the local histogram, L is the number of gray levels, and N is the total number of pixels in the local area;

[0053] The wavelet transform can refer to the formula:

[0054]

[0055] Among them, W f (a,b) is the result of wavelet transform, f(t) is the original signal, ψ is the mother wavelet function, a is the scale parameter, and b is the translation parameter.

[0056] The technology can significantly improve the quality of image data and reduce the impact of artifacts and noise. In addition, the heart segmentation model uses an optimized deep learning model, including but not limited to U-Net, MaskR-CNN and ResNet. Among them, U-Net can refer to the following formula:

[0057]

[0058] Among them, y i is the true label, p i is the predicted probability and N is the total number of pixels.

[0059] MaskR-CNN can refer to the following formula:

[0060]

[0061] in, is the classification loss, is the bounding box regression loss, is the segmentation mask loss.

[0062] The model can accurately extract the three-dimensional projection position of the heart in the CT image. The heart center point calculation method uses the centroid algorithm or the geometric center algorithm. The algorithm can accurately calculate the center position of the heart in the chest cavity and provide a reliable reference for the optimization of the lead position. The heart center point calculation is based on the segmentation result, and the centroid algorithm or the geometric center algorithm is used to calculate the center point position of the heart. The centroid algorithm calculates the centroid position of the heart voxel and is often used to calculate the center position of a certain area in the image. Suppose we have a binary image (that is, the segmented heart area), where a pixel value of 1 indicates that the pixel belongs to the heart area, and a pixel value of 0 indicates that it does not belong to the heart area. The following formula can be used as a reference:

[0063]

[0064] Where N is the total number of pixels within the heart region.

[0065] The geometric center algorithm calculates the geometric center position of the heart voxel. The geometric center algorithm is used to calculate the center point of a three-dimensional object. You can refer to the following formula:

[0066]

[0067] Where N is the total number of points within the heart region.

[0068] The heart edge positioning generates a three-dimensional model of the heart through three-dimensional reconstruction technology, and marks the heart edge positions of different intercostal spaces on the model. The three-dimensional reconstruction technology includes but is not limited to surface reconstruction, volume reconstruction and multi-plane reconstruction;

[0069] S3: Lead position optimization

[0070] Generate lead electrode positions based on the projection position of the heart using an AI big model, and optimize the lead positions through simulation verification and fine-tuning. The AI ​​big model includes but is not limited to convolutional neural networks (CNN), generative adversarial networks (GAN), and recurrent neural networks (RNN). Specifically, the AI ​​big model uses convolutional neural networks (CNN) or generative adversarial networks (GAN) to generate optimized lead positions. The model can generate the most appropriate lead positions based on the three-dimensional projection position of the heart, thereby improving the accuracy and personalization of the electrocardiogram. In addition, simulation verification is performed using a cardiac anatomical model and an electrophysiological model to ensure the rationality of the lead position and the accuracy of capturing cardiac electrical activity. The cardiac anatomical model includes but is not limited to a three-dimensional model of the heart in front and back, left and right, and up and down. The electrophysiological model includes but is not limited to a simulation model of cardiac electrical activity. The cardiac anatomical model is used in a virtual environment to simulate the distribution of lead electrodes on the surface of the heart.

[0071] S4: Individualized ECG Recording

[0072] Electrodes are placed according to the optimized lead positions, and electrocardiogram waveforms are recorded. Specifically, the electrocardiogram equipment has the characteristics of high fidelity and low noise, and can accurately record the electrocardiogram waveforms of the optimized lead positions. The electrocardiogram equipment includes but is not limited to a standard 12-lead electrocardiograph and a multi-lead electrocardiograph. Electrode placement methods include but are not limited to manual placement, robot-assisted placement, and automatic placement. The electrode placement has high accuracy, can ensure good contact with the skin, and reduce contact resistance and signal interference;

[0073] S5: Data Integration and Analysis

[0074] The recorded ECG data is integrated with the CT image data, and an AI model is used for comprehensive analysis. Specifically, the AI ​​model is used for comprehensive analysis to generate an individualized ECG report. The comprehensive analysis includes but is not limited to ECG waveform feature extraction, abnormal waveform detection, heart three-dimensional model reconstruction, and cardiac electrical activity simulation. ECG waveform feature extraction includes but is not limited to the extraction of QRS complex, ST segment, and T wave features. Abnormal waveform detection includes but is not limited to the detection of ST segment elevation, T wave inversion, and QRS complex widening abnormal waveforms. Heart three-dimensional model reconstruction can generate a detailed three-dimensional model of the heart based on CT image data. Heart electrical activity simulation can simulate the ECG waveform of the optimized lead position to improve the accuracy and reliability of diagnosis.

[0075] Embodiment 2:

[0076] Diagnosis of ST-segment elevation myocardial infarction

[0077] First, understand the patient's basic situation. Patient's gender: male; age: 52 years old; clinical symptoms:

[0078] Chest pain, palpitations; past medical history: hypertension, hyperlipidemia.

[0079] Implementation process:

[0080] 1. Before undergoing a chest CT scan, the patient should take off his or her top and metal objects and remain calm and relaxed.

[0081] 2. Use a standard CT scanner, set the scanning layer thickness to 1 mm, the scanning layer spacing to 0.5 mm, the scanning range from the sternum to the diaphragm, the scanning speed to 0.5 seconds per layer, and the radiation dose to be as low as possible.

[0082] 3. Collect multi-slice CT image data and store them in DICOM format files.

[0083] 4. Use adaptive histogram equalization technology to denoise and enhance the acquired CT images to improve image quality.

[0084] 5. Use the optimized U-Net model to automatically segment the heart in the CT image and generate a three-dimensional model of the heart.

[0085] 6. Based on the segmentation results, the center point of the heart is calculated using the centroid algorithm, and a three-dimensional model of the heart is generated through three-dimensional reconstruction technology, and the heart edge positions of different intercostals are marked on the model.

[0086] 7. Use convolutional neural network to generate lead electrode positions based on the heart projection position.

[0087] 8. The distribution of lead electrodes on the heart surface is simulated by the cardiac anatomical model, and the electrophysiological model is used to simulate the generated ECG waveform to ensure that the optimized lead position can accurately capture the electrical activity of the heart.

[0088] 9. Based on the verification results, fine-tune the lead position to ensure that the final lead position is physiologically and anatomically reasonable and can most accurately reflect the electrical activity of the heart.

[0089] 10. The doctor places the electrodes on the patient's chest according to the optimized lead position. When placing the electrodes, ensure that they are accurately aligned with the optimized position and maintain good contact with the skin.

[0090] 11. Use high-fidelity, low-noise ECG equipment to record ECG waveforms.

[0091] 12. Integrate the recorded ECG data with the CT image data, use the AI ​​model for comprehensive analysis, and generate an individualized ECG report.

[0092] The above steps yielded the following results: In the generated personalized ECG report, ST segment elevation is clearly visible in leads V1 to V4, consistent with the chest pain symptoms. Compared with the traditional lead positions, the optimized lead positions are closer to the actual heart position of the patient, reducing the possibility of waveform distortion. This personalized ECG record not only improves the accuracy of diagnosis, but also provides important reference information for subsequent treatment plans.

[0093] Embodiment three:

[0094] Monitoring of myocardial hypertrophy

[0095] First, understand the patient's basic condition: patient gender: female; age: 65 years old; clinical symptoms: dyspnea, fatigue; past medical history: hypertension, diabetes.

[0096] Implementation process: The process different from the second embodiment is that in step 4, the adaptive histogram equalization technology is replaced by Gaussian filtering technology; the U-Net model in step 5 is replaced by the Mask R-CNN model; the centroid algorithm in step 6 is replaced by the geometric center algorithm; the other operation processes are the same.

[0097] The above steps yielded the following results: In the generated individualized ECG report, the QRS complex was significantly widened in leads V5 and V6, suggesting possible myocardial hypertrophy. In addition, the T wave was also inverted in these leads, further supporting the diagnosis of myocardial hypertrophy. Compared with the traditional lead position, the optimized lead position was more consistent with the patient's cardiac anatomy. Combined with the CT imaging data and ECG waveform, the doctor confirmed that the patient had left ventricular hypertrophy.

[0098] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An individualized electrocardiogram operation process based on imaging examination and AI large model, characterized by: The following steps are involved: S1: Chest CT Image Acquisition Using a CT scanner to collect chest image data of the patient, the CT scanner can generate high-resolution multi-slice CT images, and the data format includes but is not limited to DICOM format; S2: Analysis of cardiac projection position Perform image preprocessing, heart segmentation, heart center point calculation and heart edge positioning on the acquired CT image data; S3: Lead position optimization Generate lead electrode positions based on the heart projection position using an AI big model, and optimize the lead positions through simulation verification and fine-tuning, wherein the AI ​​big model includes but is not limited to a convolutional neural network (CNN), a generative adversarial network (GAN), and a recurrent neural network (RNN); S4: Individualized ECG Recording Place electrodes according to the optimized lead positions and record the ECG waveform; S5: Data Integration and Analysis Integrate the recorded ECG data with the CT image data, and use the AI ​​model for comprehensive analysis.

2. According to claim 1, a personalized electrocardiogram operation process based on imaging examination and AI large model is characterized by: The scanning layer thickness of the scanner used for CT image acquisition in S1 is 1 mm, the scanning layer thickness is 0.5 mm, the scanning range is from the sternum to the diaphragm, the scanning speed is 0.5 seconds / layer, and the radiation dose is sufficient to ensure the image quality. The scanning mode can be spiral scanning or non-spiral scanning. The appropriate scanning mode can also be selected according to the specific situation of the patient.

3. According to claim 1, a personalized electrocardiogram operation process based on imaging examination and AI large model is characterized by: In S2, the image preprocessing technology includes but is not limited to median filtering, Gaussian filtering, adaptive histogram equalization and wavelet transform, and the technology can significantly improve the quality of image data and reduce the impact of artifacts and noise.

4. According to claim 3, a personalized electrocardiogram operation process based on imaging examination and AI large model is characterized in that: In S2, the heart segmentation model adopts an optimized deep learning model, including but not limited to U-Net, MaskR-CNN and ResNet, and the model can accurately extract the three-dimensional projection position of the heart in the CT image.

5. According to claim 4, a personalized electrocardiogram operation process based on imaging examination and AI large model is characterized in that: In S2, the heart center point calculation method adopts the centroid algorithm or the geometric center algorithm, which can accurately calculate the center position of the heart in the chest cavity and provide a reliable reference for the optimization of the lead position.

6. According to claim 1, a personalized electrocardiogram operation process based on imaging examination and AI large model is characterized by: In S3, the AI ​​large model uses a convolutional neural network (CNN) or a generative adversarial network (GAN) to generate optimized lead positions. The model can generate the most suitable lead positions based on the three-dimensional projection position of the heart, thereby improving the accuracy and personalization level of the electrocardiogram.

7. According to claim 6, a personalized electrocardiogram operation process based on imaging examination and AI large model is characterized by: In S3, simulation verification is performed through a cardiac anatomical model and an electrophysiological model to ensure the rationality of the lead position and the accuracy of capturing the cardiac electrical activity. The cardiac anatomical model includes but is not limited to a three-dimensional model of the heart's front-back, left-right, and top-bottom, and the electrophysiological model includes but is not limited to a simulation model of the heart's electrical activity.

8. According to claim 1, a personalized electrocardiogram operation process based on imaging examination and AI large model is characterized by: In the S4, the electrocardiogram device has the characteristics of high fidelity and low noise, and can accurately record the electrocardiogram waveform of the optimized lead position. The electrocardiogram device includes but is not limited to a standard 12-lead electrocardiograph and a multi-lead electrocardiograph. The electrode placement method includes but is not limited to manual placement, robot-assisted placement, and automatic placement. The electrode placement has high accuracy, can ensure good contact with the skin, and reduce contact resistance and signal interference.

9. The individualized electrocardiogram operation process based on imaging examination and AI large model according to claim 1 is characterized in that: In S5, an AI model is used to perform comprehensive analysis to generate an individualized ECG report. The comprehensive analysis includes but is not limited to ECG waveform feature extraction, abnormal waveform detection, reconstruction of a three-dimensional heart model, and simulation of cardiac electrical activity. The ECG waveform feature extraction includes but is not limited to extraction of QRS complex, ST segment, and T wave features. The abnormal waveform detection includes but is not limited to detection of abnormal waveforms such as ST segment elevation, T wave inversion, and QRS complex widening.

10. The individualized electrocardiogram operation process based on imaging examination and AI large model according to claim 9 is characterized in that: The heart three-dimensional model reconstruction can generate a detailed three-dimensional model of the heart based on CT image data, and the heart electrical activity simulation can simulate the electrocardiogram waveform of the optimized lead position, thereby improving the accuracy and reliability of diagnosis.

Images