Automatic positioning algorithm for cun-guan-chi three parts of radial artery

Abstract

The invention discloses a radial artery cun-guan-chi three-part automatic positioning algorithm, and relates to the technical field of image extraction. The invention provides a radial artery cun-guan-chi three-part automatic positioning algorithm, which comprises the steps of building a cun-mouth blood perfusion image acquisition platform, and obtaining video stream image data; a region of interest is calibrated, feature points are selected, and a Lucas-Kanade optical flow method is used for motion detection; correcting the video stream image data based on the wrist motion state obtained by the motion detection; based on a maximum combination ratio algorithm, extracting the signal intensity of the corrected video stream image data pixels, and obtaining a pulse wave in each pixel; based on the pulse waves, the position of the radial artery is determined through the ratio of the power of the signal intensity in the pulse rate area range to the noise of a filter passband; and obtaining the positions of the cunkou pulse, the guan pulse and the chi pulse by using a machine vision algorithm. Therefore, by the adoption of the method, the positioning accuracy of the Cun pulse, the Guan pulse and the Chi pulse can be improved, and pulse diagnosis visualization is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of medical image extraction, and in particular to an automatic positioning algorithm for the three parts of the radial artery, Cun, Guan and Chi. Background Art

[0002] Chinese medicine pulse diagnosis emphasizes "differentiation and treatment", that is, comprehensive analysis of the cause and pathogenesis of the disease according to the different manifestations of the disease. Due to the differences in the human body of the patient and the different experience of traditional Chinese medicine practitioners, there is a lack of unified and accurate standards for the recognition and identification of pulse conditions, making pulse diagnosis highly subjective; at the same time, the mastery of pulse diagnosis theory requires a sufficient knowledge reserve system and long-term learning and practical experience. Therefore, the obscure pulse evaluation and diagnostic standards pose a great challenge to the promotion and inheritance of pulse diagnosis.

[0003] Therefore, there is an urgent need to develop an innovative Cunkou intelligent acupoint-finding system that can automatically identify and locate the positions of the Cunkou pulse, Guan pulse and Chi pulse, simulate the techniques of finding the pulse in traditional Chinese medicine, and simulate and analyze the pulse-picking process based on the finite element analysis of pulse diagnosis, so as to provide a reference for the innovation of modern Chinese medicine pulse diagnosis technology and optimization of methods. Summary of the invention

[0004] The purpose of the present invention is to provide an automatic positioning algorithm for the three parts of the radial artery, which can visualize the Cunkou pulse, Guan pulse and Chi pulse, and automatically identify and locate them.

[0005] To achieve the above object, the present invention provides an automatic positioning algorithm for the three parts of the radial artery, including the following steps: S1. Build a Cunkou blood perfusion image acquisition platform to obtain video stream image data; S2, calibrate the region of interest, select feature points, use the Lucas-Kanade optical flow method to perform motion detection on the video stream image data, and obtain the wrist motion state; S3, based on the obtained wrist movement state, correcting the video stream image data to obtain corrected video stream image data; S4, based on the maximum merging ratio algorithm, using the corrected video stream image data, extracting the signal intensity of the pixel, obtaining the time domain characteristics of the pulse wave in each pixel, and then obtaining the pulse rate; S5. Based on the frequency domain characteristics of the pulse wave, the power of the signal intensity within the pulse rate region and the noise ratio of the filter passband are obtained, thereby determining the position of the radial artery; S6. Based on the position of the radial artery, the positions of the Cunkou pulse, Guan pulse and Chi pulse are obtained using a machine vision algorithm.

[0006] Preferably, calibrating the region of interest includes binarizing the first frame image in the acquired video stream image data, detecting and acquiring the pulse pillow contour, calculating the minimum circumscribed rectangle of the pulse pillow contour, and calibrating it as the region of interest.

[0007] Preferably, the correction processing of the video stream image data includes correcting the video stream image data using the first frame image as a reference frame.

[0008] Preferably, correcting the video stream image data includes setting a threshold value of the wrist movement deformation, and using the average Euclidean distance of the feature points to determine the correction method, which is specifically as follows: When the average Euclidean distance of the feature points is greater than a threshold value of the wrist motion deformation, the current frame of the video stream image data is updated as a reference frame; When the average Euclidean distance of the feature points is less than the threshold of the wrist motion deformation and the area of ​​the circumscribed rectangle of the feature points changes, the current frame of the video stream image data is updated as the reference frame; When the average Euclidean distance of the feature points is less than the threshold of the wrist motion deformation and the area of ​​the circumscribed rectangle of the feature points does not change, the current frame of the video stream image data is corrected using affine transformation.

[0009] Preferably, the average Euclidean distance of the feature points is as follows: In the formula, is the average Euclidean distance of feature points, is the number of feature points, The current frame The image coordinates of feature points in the reference frame coordinate system, The reference frame The image coordinates of the feature points.

[0010] Preferably, in step S4, the pulse rate is calculated as follows: In the formula, for Moment The pulse wave estimation of pixels, is the number of pixels, For the The optimal weight coefficient of pixels, for Moment The signal strength of each pixel, is the indicator function, , They are In the time window The maximum amplitude and minimum amplitude within is the signal strength threshold.

[0011] Preferably, in step S5, the power of the signal intensity within the pulse rate region and the noise ratio of the filter passband are obtained as follows: Where PR is the pulse rate, is a constant, for The ratio of the power in the pulse rate region to the noise in the filter passband, for exist The power spectral density between , are the lower cutoff frequency and upper cutoff frequency of the bandpass filter respectively.

[0012] Preferably, in step S6, obtaining the positions of the Cunkou pulse, Guan pulse and Chi pulse comprises the following steps: S61, using findContours function to detect and determine key corner points; S62, combining the key corner point and the position of the radial artery to obtain the image coordinates of the Cunkou pulse; S63, establishing a mapping relationship between the image coordinates and the spatial rectangular coordinate system to obtain the rectangular coordinates of the Cunkou pulse; S64. According to the TCM theory of "same body inch", the rectangular coordinates of the guan pulse and the chi pulse are determined based on the rectangular coordinates of the cun pulse.

[0013] Therefore, the present invention adopts the above-mentioned radial artery Cun, Guan and Chi three-part automatic positioning algorithm, which has the following technical effects: (1) Through the time domain characteristics of the pulse wave, the power of the signal intensity in the pulse rate area and the noise ratio of the filter passband are obtained, which can realize the visualization of the radial artery; (2) Use the findContours function to perform edge detection and obtain key corner points. The positions of the Cunkou, Guan, and Chi pulses can be obtained to achieve automatic positioning, providing a reference for medical pulse diagnosis.

[0014] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a video stream image processing flow chart in an embodiment of an automatic positioning algorithm for radial artery Cun, Guan and Chi; Figure 2 A pulse wave obtained in an embodiment of a radial artery Cun, Guan and Chi three-part automatic positioning algorithm, wherein: Figure 2 (a) is the time domain waveform. Figure 2(b) is the frequency domain waveform; Figure 3 The present invention is a schematic diagram of radial artery position extraction in an embodiment of an automatic positioning algorithm for radial artery Cun, Guan and Chi; Figure 4 A rectangular coordinate system established in an embodiment of an automatic positioning algorithm for the radial artery, cun, guan and chi parts; Figure 5 The present invention is a schematic diagram of the positions of Cun, Guan and Chi in an embodiment of an automatic positioning algorithm for the Cun, Guan and Chi parts of the radial artery. DETAILED DESCRIPTION

[0016] The present invention can be explained in more detail by the following examples. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following examples.

[0017] like Figure 1 As shown, the present invention provides an automatic positioning algorithm for the radial artery Cun, Guan and Chi parts, comprising the following steps: S1. Build a Cunkou blood perfusion image acquisition platform, place the subject's wrist on the fixed pulse pillow, close to the pulse pillow, with the arm and the heart roughly at the same level, and keep it in a resting state, record the image acquisition for 10 seconds, and obtain the video stream image data.

[0018] Among them, the Cunkou blood perfusion image acquisition platform includes a camera, several light-emitting diodes (LEDs) installed around the camera lens, a polarizer and a visual bracket.

[0019] S2. Calibrate the region of interest, select feature points at the edge of the wrist in the region of interest, use the first frame image as the reference frame, use the Lucas-Kanade optical flow method to perform motion detection on the video stream image data, and obtain the wrist movement state (translation, scaling or deformation).

[0020] Among them, the region of interest is marked as follows: The first frame of the video stream image data is binarized to find the contour of the pulse pillow, and the minimum circumscribed rectangle of the contour is calculated and marked as the region of interest (ROI).

[0021] Among them, since the pulse pillow position runs through the Cunkou position, the pulse pillow area is calibrated as ROI to reduce the calculation amount of image processing.

[0022] S3. Based on the obtained wrist movement state, perform alignment correction processing on the video stream image data to obtain corrected video stream image data.

[0023] When processing deformed wrist images, a more complex and sophisticated method is used to determine the correspondence of feature points to ensure the accuracy of the region of interest (ROI) and the precision of alignment. However, this inevitably introduces alignment errors, which distorts the PPG signal waveform.

[0024] Therefore, the alignment correction method adopted in this embodiment is as follows: (1) When the wrist motion deformation is not obvious or the deformation is small compared with the reference frame, the current frame is corrected and aligned using the reference frame, and the PPG signal of the area of ​​interest is extracted; (2) When the degree of wrist motion deformation is larger than that of the reference frame, the frame is marked as invalid and the reference frame is updated to the current frame to ensure the accuracy of PPG signal extraction in subsequent video stream images.

[0025] In this embodiment, the average Euclidean distance of the feature points is used to measure the wrist movement deformation, and the affine transformation is used for correction and alignment, as follows: In the formula, is the average Euclidean distance of feature points, is the number of feature points, The current frame The image coordinates of feature points in the reference frame coordinate system, The reference frame The image coordinates of the feature points.

[0026] In order to filter the waveform distortion caused by movement in the PPG signal, a threshold a is set, a=2. For the problem of wrist movement deformation, if d>a, update the reference frame; if d<a, when the area of ​​the rectangle circumscribed by the reference point changes, update the reference frame; if d<a, when the area of ​​the rectangle circumscribed by the reference point does not change, average the displacement of the feature points, and use affine transformation to align the ROI image.

[0027] S4, based on the maximum ratio merging algorithm (MRC algorithm), the signal intensity of the processed video stream image pixels is used to obtain the time domain characteristics of the pulse wave in each pixel, such as Figure 2 As shown in (a), Fourier transform is used to convert it into frequency domain features, such as Figure 2 As shown in (b), by analyzing the extracted waveform, it can be seen that the initially extracted signal is interfered by noise of different frequencies. Therefore, in this embodiment, a bandpass filter of [0.7Hz, 6Hz] is used for processing to filter out the out-of-band components and out-of-band noise reflected by the skin surface, and the filtered signal is obtained. .

[0028] because, It contains the same expected signal received from each small area, accompanied by strength differences, and has noise influences of different intensities. Therefore, this embodiment obtains the combined pulse wave estimation value by weighted averaging the signals in all small areas, as follows: In the formula, for Moment The pulse wave estimation of pixels, is the number of pixels; For the The optimal weight coefficient of pixels; for Moment The signal strength of each pixel; is the indicator function; , They are In the time window Maximum amplitude and minimum amplitude within; is the threshold value.

[0029] S5, the pulse rate is equal to the fundamental frequency of the PPG signal oscillation, and the spectral power of the PPG signal is concentrated in a small frequency band around the pulse rate, and the spectral power of the noise is distributed in the passband of the filter [0.5Hz, 5Hz]. Based on this, this embodiment determines the position of the radial artery by the ratio of the power of the signal intensity within the pulse rate region to the noise of the filter passband, as follows: In the formula, for The ratio of the power in the pulse rate region to the noise in the filter passband, for exist The power spectral density between is the pulse rate; It is a small area around the pulse rate of the human wrist. By setting the upper and lower limits of the pulse rate, the changes in the pulse rate can be better analyzed. , are the lower cutoff frequency and upper cutoff frequency of the bandpass filter respectively.

[0030] like Figure 3 As shown, The highest value is represented by a red pixel. The low value is represented by the blue pixel. Since the pulse frequency at the Cunkou position is high, the position of the red pixel is the Cunkou position of the radial artery.

[0031] S6. Based on the position of the radial artery, the positions of the Cunkou pulse, Guan pulse and Chi pulse are obtained using a machine vision algorithm.

[0032] S61, performing histogram equalization processing on the region of interest of the reference frame to improve the contrast of the region of interest, and using the findContours function to detect and obtain the polygonal contour of the wrist, screening all corner points, selecting the corner point with the largest y coordinate, and determining it as the key corner point, such as Figure 5 shown.

[0033] S62. Draw a line parallel to the x-axis through the key corner point, intersecting it with the obtained position of the radial artery, to obtain the image coordinates of the Cunkou pulse.

[0034] S63. Use convex hull detection to obtain the image coordinates of the four vertices of the region of interest and use the image coordinates of the four vertices to obtain the conversion parameters between the image coordinates and the spatial rectangular coordinate system. Use the conversion parameters to convert the image coordinates of the Cunkou pulse into actual rectangular coordinates.

[0035] like Figure 4 As shown, a rectangular coordinate system is established in the horizontal plane with the Cunkou pulse as the origin and the direction of the Cun, Guan, and Chi line as the X-axis, wherein the Y-axis is perpendicular to the X-axis, and the direction toward the thumb is the positive direction.

[0036] Among them, the image coordinates of the four vertices of the region of interest are (56, 45), (56, 395), (406, 395) and (406, 395) respectively; since the actual length and width of the occipital artery are both 70 mm, the actual size corresponding to each pixel is 0.2 mm × 0.2 mm.

[0037] S64. According to the TCM theory of "same body inch", the inch pulse is one inch and nine points long. X The axial range is 0.1 / 11~0.7 / 11 of the forearm length. X The axial range is 0.7 / 11~1.3 / 11 of the forearm length. X The range of the axial direction is 1.3 / 11~2 / 11 of the forearm length. The rectangular coordinate system of the Guan pulse and the Chi pulse is obtained. Based on the rectangular coordinates of the Cunkou pulse, the rectangular coordinates of the Guan pulse and the Chi pulse are determined.

[0038] Therefore, the present invention adopts the above-mentioned radial artery Cun, Guan and Chi three-part automatic positioning algorithm, which can improve the positioning accuracy of Cun pulse, Guan pulse and Chi pulse, realize pulse diagnosis visualization, and provide a reference for the objectification and modernization of traditional Chinese medicine pulse diagnosis.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. An automatic positioning algorithm for the radial artery Cun, Guan and Chi parts, characterized in that: The following steps are involved: S1. Build a Cunkou blood perfusion image acquisition platform to obtain video stream image data; S2, calibrate the region of interest, select feature points, use the Lucas-Kanade optical flow method to perform motion detection on the video stream image data, and obtain the wrist motion state; S3, based on the obtained wrist movement state, correcting the video stream image data to obtain corrected video stream image data; S4, based on the maximum merging ratio algorithm, using the corrected video stream image data, extracting the signal intensity of the pixel, obtaining the time domain characteristics of the pulse wave in each pixel, and then obtaining the pulse rate; S5. Based on the frequency domain characteristics of the pulse wave, the power of the signal intensity within the pulse rate region and the noise ratio of the filter passband are obtained, thereby determining the position of the radial artery; S6. Based on the position of the radial artery, the positions of the Cunkou pulse, Guan pulse and Chi pulse are obtained using a machine vision algorithm.

2. The automatic positioning algorithm of radial artery Cun, Guan and Chi according to claim 1, characterized in that: In step S2, the region of interest is marked, including binarizing the first frame of the acquired video stream image data, detecting and acquiring the contour of the pulse pillow, calculating the minimum circumscribed rectangle of the contour of the pulse pillow, and marking it as the region of interest.

3. The automatic positioning algorithm of radial artery Cun, Guan and Chi according to claim 1, characterized in that: In step S3, the correction processing of the video stream image data includes correcting the video stream image data using the first frame image as a reference frame.

4. The automatic positioning algorithm of radial artery Cun, Guan and Chi according to claim 3, characterized in that: Correction of the video stream image data includes setting a threshold of wrist motion deformation and using the average Euclidean distance of feature points to determine the correction method, as follows: When the average Euclidean distance of the feature points is greater than a threshold value of the wrist motion deformation, the current frame of the video stream image data is updated as a reference frame; When the average Euclidean distance of the feature points is less than the threshold of the wrist motion deformation and the area of ​​the circumscribed rectangle of the feature points changes, the current frame of the video stream image data is updated as the reference frame; When the average Euclidean distance of the feature points is less than the threshold of the wrist motion deformation and the area of ​​the circumscribed rectangle of the feature points does not change, the current frame of the video stream image data is corrected using affine transformation.

5. The automatic positioning algorithm of radial artery Cun, Guan and Chi according to claim 4, characterized in that: The average Euclidean distance of feature points is as follows: In the formula, is the average Euclidean distance of feature points, is the number of feature points, The current frame The image coordinates of feature points in the reference frame coordinate system, The reference frame The image coordinates of feature points.

6. The automatic positioning algorithm of radial artery Cun, Guan and Chi according to claim 1, characterized in that: In step S4, the pulse rate is calculated as follows: In the formula, for Moment The pulse wave estimation of pixels, is the number of pixels, For the The optimal weight coefficient of pixels, for Moment The signal strength of each pixel, is the indicator function, , They are In the time window The maximum amplitude and minimum amplitude within is the signal strength threshold.

7. The automatic positioning algorithm of radial artery Cun, Guan and Chi according to claim 1, characterized in that: In step S5, the power of the signal intensity within the pulse rate region and the noise ratio of the filter passband are obtained, as follows: Where PR is the pulse rate, is a constant, for The ratio of the power in the pulse rate region to the noise in the filter passband, for exist The power spectral density between , are the lower cutoff frequency and upper cutoff frequency of the bandpass filter respectively.

8. The automatic positioning algorithm of radial artery Cun, Guan and Chi according to claim 1, characterized in that: In step S6, the positions of the Cunkou pulse, Guan pulse and Chi pulse are obtained, including the following steps: S61, using findContours function to detect and determine key corner points; S62, combining the key corner point and the position of the radial artery to obtain the image coordinates of the Cunkou pulse; S63, establishing a mapping relationship between the image coordinates and the spatial rectangular coordinate system to obtain the rectangular coordinates of the Cunkou pulse; S64. According to the TCM theory of "same body inch", the rectangular coordinates of the guan pulse and the chi pulse are determined based on the rectangular coordinates of the cun pulse.

Images