A method for electrocardiogram positioning and comparison based on photogrammetry

Abstract

The present invention discloses an electrocardiogram positioning and comparison method based on photogrammetry, and relates to the technical field of photogrammetry. The present invention comprises the following steps: obtaining an electrocardiogram that has been monitored, and preprocessing the electrocardiogram; extracting Hog features from the preprocessed electrocardiogram; obtaining a large number of abnormal electrocardiograms, and extracting distortion features from the abnormal electrocardiograms; calculating the amplitude and cycle length of the waveform by analyzing the peak and trough of the electrocardiogram waveform; matching the Hog features of the electrocardiogram to be analyzed with the distortion features in the abnormal electrocardiogram; locating the abnormal points in the electrocardiogram, and calculating the probability density of the abnormal points to assist doctors in diagnosis. The present invention matches the Hog features of the electrocardiogram to be analyzed with the distortion features in the abnormal electrocardiogram, locates the abnormal points in the electrocardiogram to be analyzed, and calculates the probability density of the abnormal points, thereby improving the efficiency of electrocardiogram processing and assisting doctors in diagnosis.

Description

Technical Field

[0001] The invention belongs to the technical field of photogrammetry, and in particular relates to an electrocardiogram positioning and comparison method based on photogrammetry. Background Art

[0002] Photogrammetry is a technique that uses photography to make precise measurements. It is widely used in fields such as geographic information systems, architecture, engineering, manufacturing, and medicine. It relies on taking images of objects and analyzing them to obtain accurate three-dimensional data about the objects. In the medical field, especially in the analysis and processing of electrocardiograms, photogrammetry can help more accurately identify and analyze graphic representations of heart activity.

[0003] Among them, the ECG image positioning and comparison method based on photogrammetry is an ECG analysis method combined with photogrammetry technology. Its purpose is to improve the accuracy and readability of ECG data using image processing technology. Through precise image capture and analysis, specific features and waveforms on the ECG, such as P waves, QRS complex waves, and T waves, can be more accurately located, which is crucial to ensuring the accuracy of ECG interpretation, especially when diagnosing heart disease, evaluating heart function, or conducting heart studies.

[0004] Traditional ECG methods have limitations in accuracy and efficiency. They usually rely on relatively basic image processing technology, which makes it difficult to capture subtle changes and details in ECG waveforms, leading to the neglect of certain important ECG features and affecting the accuracy of diagnosis. Lacking advanced image correction technology, ECG images are susceptible to distortion and noise, which reduces the readability of the waveform and the accuracy of analysis. Traditional methods are inefficient when processing large amounts of data, leading to delays in the diagnosis and treatment process, especially when processing complex or ambiguous ECG images, which poses a challenge to doctors' judgment and may sometimes even lead to diagnostic errors or delays. Summary of the invention

[0005] The purpose of the present invention is to provide an electrocardiogram positioning and comparison method based on photogrammetry, which locates abnormal points in the electrocardiogram to be analyzed by matching the Hog features of the electrocardiogram to be analyzed with the distortion features in the abnormal electrocardiogram, and calculates the probability density of the abnormal points, thereby solving the problems of low efficiency of existing electrocardiogram processing methods when processing large amounts of data and easy diagnosis errors or delays when processing complex or fuzzy electrocardiogram images.

[0006] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0007] The present invention is an electrocardiogram positioning and comparison method based on photogrammetry, comprising the following steps:

[0008] Step S1: obtaining the monitored electrocardiogram and preprocessing the electrocardiogram;

[0009] Step S2: extracting Hog features from the preprocessed electrocardiogram;

[0010] Step S3: obtaining a large number of abnormal electrocardiograms and extracting distortion features from the abnormal electrocardiograms;

[0011] Step S4: Calculate the amplitude and cycle length of the waveform by analyzing the peak and trough of the ECG waveform;

[0012] Step S5: matching the Hog feature of the electrocardiogram to be analyzed with the distortion feature in the abnormal electrocardiogram;

[0013] Step S6: locating the abnormal point in the electrocardiogram and calculating the probability density of the abnormal point;

[0014] Step S7: Match the corresponding abnormal symptoms according to the probability density of the abnormal points on the electrocardiogram to assist doctors in diagnosis.

[0015] As a preferred technical solution, in step S1, the electrocardiogram preprocessing operation includes:

[0016] Step S11, reducing the image: using the resize method in OpenCV to reduce the resolution of the electrocardiogram, the image resolution of 1600*1200 is reduced to 800*600 through the resize method in OpenCV;

[0017] Step S12, cropping: After obtaining the outline of the electrocardiogram, find its circumscribed rectangle, crop it into a new sample image according to the four coordinate points of the circumscribed rectangle, and remove irrelevant pixels in the image;

[0018] Step S13, background removal and edge removal: performing bitwise AND and bitwise OR operations on the pixels of the image to remove the background;

[0019] Step S14, rescaling: scale the cropped image to a uniform size. Since the cropped images have different sizes, a uniform size facilitates the extraction of features of the same dimension. In general, the samples are uniformly adjusted to a resolution of 224*224.

[0020] As a preferred technical solution, in step S2, the specific steps of Hog feature extraction are as follows:

[0021] Step S21: grayscale the preprocessed electrocardiogram image;

[0022] Step S22: normalizing the grayscaled image;

[0023] Step S23: Calculate the gradient of each image and divide the image into 8*8 pixel grids;

[0024] Step S24: Count the gradient histogram of each grid;

[0025] Step S25: Connect the feature descriptors of each grid in series to obtain the Hog feature descriptor of the area;

[0026] Step S26: connect all Hog feature descriptors in the entire image in series to obtain a feature vector of the entire image.

[0027] After feature extraction, feature classification is achieved through machine learning methods. This paper proposes to use the support vector machine (SVM) method to achieve classification. SVM has good robustness and strong generalization ability for unknown data. This experiment uses the SVM classifier in the machine learning code library sklearn library to classify defects, and uses the grid search method to select the best hyperparameter c and gamma combination. Use the support vector machine classification algorithm to classify the extracted features, map the data through the kernel function, capture the optimal segmentation hyperplane, distinguish between data points that meet and do not meet the standard ECG signal characteristics, classify and evaluate the performance of the ECG machine, and generate feature compliance classification results;

[0028] SVM is a powerful machine learning algorithm that maps data to a high-dimensional space by selecting a suitable kernel function, and then finds the optimal segmentation hyperplane in the space to distinguish data points of different categories. In this step, SVM is used to distinguish data points that meet and do not meet the characteristics of standard ECG signals, and classify and evaluate the performance of the ECG machine. This classification result provides an intuitive representation of the performance of the ECG machine, which can clearly distinguish the pros and cons of the ECG machine in reproducing standard ECG signals.

[0029] Based on the feature conformity classification results, the decision tree algorithm is used to comprehensively evaluate the performance of the electrocardiograph. By constructing a tree structure model, each branch represents a decision path, and each node represents a feature judgment condition, the influence of the feature is evaluated, and whether the electrocardiograph accurately restores the features of the standard signal is judged, and the performance evaluation index of the electrocardiograph is generated. For example, according to the feature conformity classification results, the decision tree algorithm can judge whether the amplitude, period, shape and other features of the waveform meet the preset standard ECG signal features. The output of the decision tree model is the comprehensive evaluation index of the performance of the electrocardiograph. These indicators comprehensively consider the various performances of the electrocardiograph in restoring the standard ECG signal, including accuracy, consistency and reliability. The generated electrocardiograph performance evaluation index provides a comprehensive reference basis for the calibration and performance evaluation of the electrocardiograph, which helps to determine the performance level and improvement direction of the electrocardiograph.

[0030] As a preferred technical solution, in step S4, any peak point on the ECG waveform is obtained. And any valley point is ; The Euclidean distance between the peak and the trough is calculated as:

[0031] ;

[0032] Then any peak point is And any valley point is The Manhattan distance between them is:

[0033] .

[0034] As a preferred technical solution, in step S5, the process of matching the Hog feature of the electrocardiogram to be analyzed with the distortion feature in the abnormal electrocardiogram is as follows:

[0035] Step S51: selecting the best matching sub-block from the distortion features in the abnormal electrocardiogram Hog feature to be analyzed;

[0036] Step S52: extract the distortion features as the best matching points, which are respectively denoted as A and B, and use brute force matching to find the initial matching pair;

[0037] Step S53: first filtering, using optical flow tracking to find matching alignment and feature loss, and calculate the optical flow displacement of the feature;

[0038] Step S54: Second filtering, using the RANSAC algorithm to filter out redundant matching pairs.

[0039] As a preferred technical solution, in step S54, the homography matrix between the analyzed electrocardiogram and the abnormal electrocardiogram is calculated by obtaining the first filtered matching pairs in step S53. , use affine transformation to find the image The projected pixel coordinates of the image to be matched ,but and The reprojection error is , the specific calculation formula is:

[0040] ;

[0041] The reprojection error is equal to the preset threshold For comparison, if , it indicates that the image matching point is a point inside the abnormal ECG, otherwise it is a point outside the abnormal ECG.

[0042] As a preferred technical solution, in step S6, the probability density function of the abnormal point is:

[0043] ;

[0044] In the formula, is the error estimate, is the probability weighted value, is the Gaussian distribution variance, and W is the uniform distribution window width.

[0045] As a preferred technical solution, there are multiple abnormal points, and the cost function is:

[0046] C =− ∑ i = 1 n log [ a ( 1 2 πσ 2 )exp( − e i 2 σ 2 ) + ( 1 − a ) 1 W ] .

[0047] As a preferred technical solution, in step S7, the decision tree algorithm comprehensively evaluates the performance of the electrocardiograph by constructing a tree structure model, in which each branch represents a decision path, each node represents a feature judgment condition, the influence of the feature is evaluated, and it is determined whether the electrocardiograph accurately restores the features of the standard signal, thereby generating an electrocardiograph performance evaluation index.

[0048] The present invention has the following beneficial effects:

[0049] The present invention matches the Hog features of the electrocardiogram to be analyzed with the distortion features in the abnormal electrocardiogram, locates the abnormal points in the electrocardiogram to be analyzed, calculates the probability density of the abnormal points, improves the electrocardiogram processing efficiency, and assists doctors in diagnosis.

[0050] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0052] Figure 1 A flow chart of an electrocardiogram positioning and comparison method based on photogrammetry of the present invention;

[0053] Figure 2 This is the flow chart of electrocardiogram preprocessing;

[0054] Figure 3 This is the Hog feature extraction flow chart. DETAILED DESCRIPTION

[0055] 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.

[0056] In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0057] In order to make the purpose, technical solution and advantages of this application clearer, the following Figure 1 The implementation methods of the present application are described in further detail.

[0058] Before introducing the embodiments of the present application, the video measurement technology is first described.

[0059] Photogrammetry is the process of taking pictures with an optical camera and processing them to obtain the shape, size, position, characteristics and their relationships of the objects being photographed.

[0060] The main task of photogrammetry is to measure and produce topographic maps of various scales, establish topographic databases, and provide spatial basic data for various geographic information systems, land information systems, and various projects. At the same time, it serves non-topographic fields such as industry, architecture, biology, medicine, archaeology, etc.

[0061] Traditional photogrammetry is a science and technology that uses optical cameras to take pictures and study and determine the shape, size, position and relationship of the objects photographed. It includes: obtaining images of the objects photographed, studying the processing methods of single or multiple pictures, including theories, equipment and techniques, and methods and equipment for outputting the measured results in graphical or digital form. Its main tasks are to measure and produce topographic maps of various scales, establish topographic databases, and provide basic surveying and mapping data for geographic information systems and various engineering applications.

[0062] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-3 It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] See also Figure 1 As shown, the present invention is an electrocardiogram positioning and comparison method based on photogrammetry, comprising the following steps:

[0064] Step S1: obtaining the monitored electrocardiogram and preprocessing the electrocardiogram;

[0065] Step S2: extracting Hog features from the preprocessed electrocardiogram;

[0066] Step S3: obtaining a large number of abnormal electrocardiograms and extracting distortion features from the abnormal electrocardiograms;

[0067] Step S4: Calculate the amplitude and cycle length of the waveform by analyzing the peak and trough of the ECG waveform;

[0068] Step S5: matching the Hog feature of the electrocardiogram to be analyzed with the distortion feature in the abnormal electrocardiogram;

[0069] Step S6: locating the abnormal point in the electrocardiogram and calculating the probability density of the abnormal point;

[0070] Step S7: Match the corresponding abnormal symptoms according to the probability density of the abnormal points on the electrocardiogram to assist doctors in diagnosis.

[0071] See also Figure 2 As shown, in step S1, the preprocessing operation of the electrocardiogram includes:

[0072] Step S11, reducing the image: using the resize method in OpenCV to reduce the resolution of the electrocardiogram, the image resolution of 1600*1200 is reduced to 800*600 through the resize method in OpenCV;

[0073] Step S12, cropping: After obtaining the outline of the electrocardiogram, find its circumscribed rectangle, crop it into a new sample image according to the four coordinate points of the circumscribed rectangle, and remove irrelevant pixels in the image;

[0074] Step S13, background removal and edge removal: performing bitwise AND and bitwise OR operations on the pixels of the image to remove the background;

[0075] Step S14, rescaling: scale the cropped image to a uniform size. Since the cropped images have different sizes, a uniform size facilitates the extraction of features of the same dimension. In general, the samples are uniformly adjusted to a resolution of 224*224.

[0076] See also Figure 3 As shown, in step S2, the specific steps of Hog feature extraction are as follows:

[0077] Step S21: grayscale the preprocessed electrocardiogram image;

[0078] Step S22: normalizing the grayscaled image;

[0079] Step S23: Calculate the gradient of each image and divide the image into 8*8 pixel grids;

[0080] Step S24: Count the gradient histogram of each grid;

[0081] Step S25: Connect the feature descriptors of each grid in series to obtain the Hog feature descriptor of the area;

[0082] Step S26: connect all Hog feature descriptors in the entire image in series to obtain a feature vector of the entire image.

[0083] After feature extraction, feature classification is achieved through machine learning methods. This paper proposes to use the support vector machine (SVM) method to achieve classification. SVM has good robustness and strong generalization ability for unknown data. This experiment uses the SVM classifier in the machine learning code library sklearn library to classify defects, and uses the grid search method to select the best hyperparameter c and gamma combination. Use the support vector machine classification algorithm to classify the extracted features, map the data through the kernel function, capture the optimal segmentation hyperplane, distinguish between data points that meet and do not meet the standard ECG signal characteristics, classify and evaluate the performance of the ECG machine, and generate feature compliance classification results;

[0084] SVM is a powerful machine learning algorithm that maps data to a high-dimensional space by selecting a suitable kernel function, and then finds the optimal segmentation hyperplane in the space to distinguish data points of different categories. In this step, SVM is used to distinguish data points that meet and do not meet the characteristics of standard ECG signals, and classify and evaluate the performance of the ECG machine. This classification result provides an intuitive representation of the performance of the ECG machine, which can clearly distinguish the pros and cons of the ECG machine in reproducing standard ECG signals.

[0085] Based on the feature conformity classification results, the decision tree algorithm is used to comprehensively evaluate the performance of the electrocardiograph. By constructing a tree structure model, each branch represents a decision path, and each node represents a feature judgment condition, the influence of the feature is evaluated, and whether the electrocardiograph accurately restores the features of the standard signal is judged, and the performance evaluation index of the electrocardiograph is generated. For example, according to the feature conformity classification results, the decision tree algorithm can judge whether the amplitude, period, shape and other features of the waveform meet the preset standard ECG signal features. The output of the decision tree model is the comprehensive evaluation index of the performance of the electrocardiograph. These indicators comprehensively consider the various performances of the electrocardiograph in restoring the standard ECG signal, including accuracy, consistency and reliability. The generated electrocardiograph performance evaluation index provides a comprehensive reference basis for the calibration and performance evaluation of the electrocardiograph, which helps to determine the performance level and improvement direction of the electrocardiograph.

[0086] In step S4, any peak point on the ECG waveform is obtained as And any valley point is ; The Euclidean distance between the peak and the trough is calculated as:

[0087] ;

[0088] Then any peak point is And any valley point is The Manhattan distance between them is:

[0089] .

[0090] In step S5, the process of matching the Hog feature of the electrocardiogram to be analyzed with the distortion feature in the abnormal electrocardiogram is as follows:

[0091] Step S51: selecting the best matching sub-block from the distortion features in the abnormal electrocardiogram Hog feature to be analyzed;

[0092] Step S52: extract the distortion features as the best matching points, which are respectively denoted as A and B, and use brute force matching to find the initial matching pair;

[0093] Step S53: first filtering, using optical flow tracking to find matching alignment and feature loss, and calculate the optical flow displacement of the feature;

[0094] Step S54: Second filtering, using the RANSAC algorithm to filter out redundant matching pairs.

[0095] In step S54, the homography matrix between the analyzed electrocardiogram and the abnormal electrocardiogram is calculated by using the matching pairs obtained in the first filtering in step S53. , use affine transformation to find the image The projected pixel coordinates of the image to be matched ,but and The reprojection error is , the specific calculation formula is:

[0096] ;

[0097] Reprojection error and preset threshold For comparison, if , it indicates that the image matching point is a point inside the abnormal ECG, otherwise it is a point outside the abnormal ECG.

[0098] In step S6, the probability density function of the abnormal point is:

[0099] ;

[0100] In the formula, is the error estimate, is the probability weighted value, is the Gaussian distribution variance, and W is the uniform distribution window width.

[0101] There are multiple abnormal points, and their cost function is:

[0102] C =− ∑ i = 1 n log [ a ( 1 2 πσ 2 )exp( − e i 2 σ 2 ) + ( 1 − a ) 1 W ] .

[0103] In step S7, the decision tree algorithm comprehensively evaluates the performance of the electrocardiograph by constructing a tree structure model, in which each branch represents a decision path and each node represents a feature judgment condition. The influence of the feature is evaluated to determine whether the electrocardiograph accurately restores the features of the standard signal and to generate an electrocardiograph performance evaluation index.

[0104] It is worth noting that in the above system embodiment, the various units included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0105] In addition, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiments can be completed by instructing related hardware through a program, and the corresponding program can be stored in a computer-readable storage medium.

[0106] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An electrocardiogram positioning and comparison method based on photogrammetry, characterized in that: The steps include: Step S1: obtaining the monitored electrocardiogram and preprocessing the electrocardiogram; Step S2: extracting Hog features from the preprocessed electrocardiogram; Step S3: obtaining a large number of abnormal electrocardiograms and extracting distortion features from the abnormal electrocardiograms; Step S4: Calculate the amplitude and cycle length of the waveform by analyzing the peak and trough of the ECG waveform; Step S5: matching the Hog feature of the electrocardiogram to be analyzed with the distortion feature in the abnormal electrocardiogram; Step S6: locating the abnormal point in the electrocardiogram and calculating the probability density of the abnormal point; Step S7: matching corresponding abnormal symptoms according to the probability density of abnormal points on the electrocardiogram to assist doctors in diagnosis; In step S4, any peak point on the ECG waveform is obtained. And any valley point is ; The Euclidean distance between the peak and the trough is calculated as: ； Then any peak point is And any valley point is The Manhattan distance between them is: ； In step S5, the process of matching the Hog feature of the electrocardiogram to be analyzed with the distortion feature in the abnormal electrocardiogram is as follows: Step S51: selecting the best matching sub-block from the distortion features in the abnormal electrocardiogram Hog feature to be analyzed; Step S52: extract the distortion features as the best matching points, which are respectively denoted as A and B, and use brute force matching to find the initial matching pair; Step S53: first filtering, using optical flow tracking to find matching alignment and feature loss, and calculate the optical flow displacement of the feature; Step S54: Second filtering, using the RANSAC algorithm to filter out redundant matching pairs; Among them, in step S54, the homography matrix between the analyzed electrocardiogram and the abnormal electrocardiogram is calculated by obtaining the first filtered matching pairs in step S53 , use affine transformation to find the image The projected pixel coordinates of the image to be matched ,but and The reprojection error is , the specific calculation formula is: ； The reprojection error is equal to the preset threshold For comparison, if , it indicates that the image matching point is a point inside the abnormal ECG, otherwise it is a point outside the abnormal ECG.

2. The electrocardiogram positioning and comparison method based on photogrammetry according to claim 1, characterized in that: In step S1, the preprocessing operation of the electrocardiogram includes: Step S11, reducing the image: using the resize method in OpenCV to reduce the resolution of the electrocardiogram; Step S12, cropping: after obtaining the outline of the electrocardiogram, find its circumscribed rectangle, and crop it into a new sample image according to the four coordinate points of the circumscribed rectangle; Step S13, background removal and edge removal: performing bitwise AND and bitwise OR operations on the pixels of the image to remove the background; Step S14, rescaling: scaling the cropped image to a uniform size.

3. The electrocardiogram positioning and comparison method based on photogrammetry according to claim 1, characterized in that: In step S2, the specific steps of Hog feature extraction are as follows: Step S21: grayscale the preprocessed electrocardiogram image; Step S22: normalizing the grayscaled image; Step S23: Calculate the gradient of each image and divide the image into 8*8 pixel grids; Step S24: Count the gradient histogram of each grid; Step S25: Connect the feature descriptors of each grid in series to obtain the Hog feature descriptor of the area; Step S26: connect all Hog feature descriptors in the entire image in series to obtain a feature vector of the entire image.

4. The electrocardiogram positioning and comparison method based on photogrammetry according to claim 1, characterized in that: In step S6, the probability density function of the abnormal point is: ； In the formula, is the error estimate, is the probability weighted value, is the Gaussian distribution variance, and W is the uniform distribution window width.

5. The electrocardiogram positioning and comparison method based on photogrammetry according to claim 4, characterized in that: There are multiple abnormal points, and their cost function is: 。 6. The electrocardiogram positioning and comparison method based on photogrammetry according to claim 1, characterized in that: In step S7, the decision tree algorithm comprehensively evaluates the performance of the electrocardiograph by constructing a tree structure model, in which each branch represents a decision path, each node represents a feature judgment condition, the influence of the feature is evaluated, whether the electrocardiograph accurately restores the features of the standard signal, and the performance evaluation index of the electrocardiograph is generated.

Images