A traditional chinese medicine pulse condition analysis method and system based on image enhancement

Abstract

The application discloses a traditional Chinese medicine pulse condition diagnosis and treatment analysis method and system based on image enhancement, relates to the technical field of image recognition, and comprises the following steps: synchronously collecting static pressure data and dynamic pulse wave data of cun, guan and chi positions, converting pulse response intensity, collection time and pulse-taking pressure into gray scale distribution, transverse time sequence and vertical pressure stratification respectively, and forming a cun-guan-chi pulse condition gray scale diagram; performing gray scale background flattening processing on the cun-guan-chi pulse condition gray scale diagram, extracting main wave edges, double pulse wave details and descending middle isthmus textures through differential multi-scale morphological top-hat enhancement, and synthesizing a cun-guan-chi enhanced pulse condition diagram; and corresponding registration of the cun-guan-chi enhanced pulse condition diagram and the static pressure data, determination of main wave skeletons, double pulse wave skeletons, descending middle isthmus texture skeletons and corresponding pulse-taking pressures. The application highlights main wave edges, double pulse wave details and descending middle isthmus textures through differential multi-scale morphological top-hat enhancement, and reduces background interference caused by pressure progression and contact state.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a method and system for traditional Chinese medicine pulse diagnosis and analysis based on image enhancement. Background Technology

[0002] Digital research on pulse diagnosis in Traditional Chinese Medicine (TCM) typically relies on pressure sensors or multi-channel pulse diagnosis equipment to collect dynamic pulse wave data at the cun, guan, and chi positions. Pulse location, pulse strength, pulse rate, tension, and fluency are quantified using parameters such as main wave amplitude, pulse cycle, waveform slope, dicrotic wave characteristics, and pulse pressure. With the development of image recognition technology, some methods have begun to convert pulse wave sequences into grayscale images or feature maps, and then use digital image processing techniques such as edge enhancement, texture analysis, and skeleton extraction to describe pulse morphology, gradually expanding pulse analysis from single waveform parameter calculation to image feature recognition.

[0003] While existing technologies can extract parameters or display images of pulse waves, they typically rely on the peaks, troughs, periods, and amplitudes under a single pulse-taking pressure as the primary criteria for judgment. They lack the ability to express the continuity of pulse structure at the cun, guan, and chi levels under different pressure layers. Even when image processing methods are used, they often remain at the level of enhanced display or ordinary texture extraction, failing to uniformly encode the main wave edge, diphtheria wave details, descending isthmus texture, and the topological relationship between pressure layers. This results in weak pulse features being easily masked by background grayscale shifts, making it difficult to form stable and interpretable structured recognition results for diagnostic features such as pulse position, pulse strength, tension, and fluency. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a TCM pulse diagnosis and analysis method based on image enhancement to solve the problems of weak texture in pulses with multiple pressure layers, which are difficult to enhance and pulse structure, which are difficult to identify.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for TCM pulse diagnosis and analysis based on image enhancement, comprising,

[0008] Static pressure data and dynamic pulse wave data at the Cun, Guan, and Chi positions are collected simultaneously. The pulse response intensity, collection time, and pulse taking pressure are converted into grayscale distribution, horizontal time sequence, and vertical pressure layering, respectively, to form a grayscale image of the Cun, Guan, and Chi pulse.

[0009] The grayscale image of the Cun, Guan, and Chi pulses was flattened, and the edge of the main wave, details of the diphthop wave, and texture of the descending septum were extracted by differential multi-scale morphological top cap enhancement to synthesize the enhanced Cun, Guan, and Chi pulse image.

[0010] The enhanced pulse image of Cun, Guan, and Chi is registered with the static pressure data. The main wave skeleton, diphtheria wave skeleton, descending isthmus texture skeleton and corresponding pulse taking pressure are determined by phase correlation registration, morphological skeleton extraction and connected domain labeling to form the enhanced pulse taking image of Cun, Guan, and Chi.

[0011] The polar coordinates of the pulse skeleton in the enhanced image of Cun Guan Chi pulse taking are expanded, gray-level co-occurrence features and directional gradient features are extracted, and a pulse topology map and graph Laplacian spectrum sequence are constructed through pressure layer pulse topology coding. The graph Laplacian spectrum sequence is associated with the pulse taking pressure, gray-level co-occurrence features and directional gradient features to form the Cun Guan Chi pulse map feature group.

[0012] By comparing the differences in pulse pressure, pulse pattern topology, and Laplace spectral sequence in the pulse image feature group of Cun, Guan, and Chi, the pulse position, pulse strength, pulse rate, tension, and fluency are identified, and the results of TCM pulse diagnosis analysis are generated.

[0013] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, wherein: the formation of the grayscale image of the cun, guan, and chi pulses specifically includes:

[0014] Static pressure data, dynamic pulse wave data, and location identifiers at the cun, guan, and chi positions are acquired simultaneously and organized into a multi-location pulse acquisition sequence according to the acquisition time.

[0015] The pulse acquisition sequence from multiple locations is divided into Cun-Guan-Chi pulse sequences according to the location identifier, and the pulse response intensity in the Cun-Guan-Chi pulse sequences is converted into grayscale values ​​to obtain the Cun-Guan-Chi grayscale response sequence.

[0016] The acquisition time in the Cun-Guan-Chi grayscale response sequence is arranged as a horizontal time sequence, and the pulse taking pressure is arranged as a vertical pressure layer to generate a Cun-Guan-Chi pulse grayscale image.

[0017] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, the grayscale background smoothing process specifically includes:

[0018] The background grayscale field generated by the pressure progression is separated from the grayscale image of the Cun, Guan, and Chi pulses to obtain the pressure-layered background image;

[0019] The grayscale difference between the pressure layer background image and the Cun-Guan-Chi pulse grayscale image is corrected, and the location identification, horizontal time sequence and vertical pressure layer are preserved to generate a flat pulse grayscale image.

[0020] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, the step of extracting the main wave edge, diphthoplasty details, and descending isthmus texture through differential multi-scale morphological cap enhancement specifically includes:

[0021] Small-scale structuring elements are used to perform white cap transformation on the flat pulse grayscale image to extract diphtheria wave details and obtain a bright detail image;

[0022] The bright detail map is morphologically enhanced using mesoscale structural elements to extract the rising edge and falling edge of the main wave, thus obtaining the main wave edge map;

[0023] A black cap transform was performed on the flat pulse grayscale image using large-scale structuring elements to extract the dark texture of the mid-straight canyon and obtain the dark texture image.

[0024] The main wave edge image, bright detail image, and dark texture image are differentially synthesized to generate an enhanced pulse image of the cun, guan, and chi positions.

[0025] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, the step of registering the enhanced pulse images of the cun, guan, and chi positions with static pressure data specifically includes:

[0026] Enhanced image blocks of adjacent pressure layers at the same location are extracted from the enhanced pulse image of Cun, Guan, and Chi, resulting in a sequence of image blocks between pressure layers.

[0027] Phase correlation calculation is performed on the pressure interlayer image block sequence, and the vein position in the pressure interlayer image block sequence is corrected based on the obtained lateral displacement to generate a pressure interlayer registration pulse map.

[0028] By correlating the pressure interlayer registration pulse map with the static pressure data, a pressure registration enhanced pulse map is obtained.

[0029] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, wherein determining the main wave skeleton, dicrotic wave skeleton, descending isthmus texture skeleton, and corresponding pulse-taking pressure specifically includes:

[0030] The main wave region, diabetic wave region, and descending isthmus region are extracted from the pressure registration enhanced pulse map to obtain the pulse pattern region map;

[0031] Morphological skeleton extraction was performed on the vein pattern region map to generate the main wave skeleton, diphtheria wave skeleton and descending isthmus texture skeleton, and connected component marking was performed to obtain the skeleton connected marking map.

[0032] The continuous state of the skeleton within the same pressure layer and the skeleton continuity relationship between adjacent pressure layers are identified by skeleton connectivity marker maps to determine the pulse image region.

[0033] The pulse image area is mapped to the static pressure data to determine the pulse pressure at the cun, guan, and chi positions, and to generate enhanced pulse images at the cun, guan, and chi positions.

[0034] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, the extraction of gray-level co-occurrence features and directional gradient features specifically includes:

[0035] The main wave skeleton, diabetic wave skeleton, and descending isthmus texture skeleton in the enhanced pulse diagnosis diagrams of the cun, guan, and chi positions are determined as the center line of the pulse texture skeleton.

[0036] The main wave texture area, diphtheria wave texture area and descending isthmus texture area are extracted along the center line of the vein texture skeleton to form the vein texture region;

[0037] The vein texture area is unfolded in polar coordinates according to the skeleton direction to obtain the vein unfolding diagram;

[0038] The gray-level co-occurrence relationship in the vein pattern unfolding is statistically analyzed and the directional change is extracted to generate gray-level co-occurrence features and directional gradient features;

[0039] Using the positional correspondence of the center line of the vein skeleton as an index, gray-level co-occurrence features and directional gradient features are mapped to the same vein coordinates to obtain vein image features.

[0040] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, wherein: the formation of the cun-guan-chi pulse feature group specifically includes:

[0041] Identify the endpoints of the main wave skeleton, the ridges of the main wave peak, the inflection points of the diphtheria wave, and the ridges of the descending valley from the features of the pulse pattern image to form a set of pulse pattern nodes;

[0042] The skeleton connection segments within the same cycle are defined as the periodic vein connection set, and the vein nodes corresponding to the positions in adjacent pressure layers are defined as the interlayer vein connection set.

[0043] A vein topology map is constructed based on horizontal temporal sequence, vertical pressure stratification, periodic vein connection set, and interlayer vein connection set.

[0044] Using the set of vein nodes as matrix indices, the set of periodic vein connections and the set of interlayer vein connections are transformed into a vein connection matrix, and the vein connection matrix is ​​decomposed into a graph Laplace decomposition to obtain a graph Laplace spectrum sequence.

[0045] By associating the pulse topology map, the Laplacian spectrum sequence, the pulse pressure, and the pulse image features, a pulse feature group of cun, guan, and chi is formed.

[0046] As a preferred embodiment of the image enhancement-based TCM pulse diagnosis and analysis method of the present invention, the generation of TCM pulse diagnosis and analysis results specifically includes:

[0047] The pulse characteristics of the Cun, Guan, and Chi pulse diagrams are matched according to the Cun, Guan, and Chi regions to obtain the correspondence between the three pulse diagrams.

[0048] By comparing the vein sampling pressure, vein pattern topology map and graph Laplacian spectrum sequence in the correspondence of the three vein maps, the differences in vein sampling pressure, vein pattern topology and pressure layer continuity are obtained.

[0049] Based on the correspondence of the three pulse diagrams, pulse image features are identified by pulse pressure difference features, pulse pattern topology difference features, pressure layer continuity difference features, gray-scale co-occurrence features, and directional gradient features, resulting in pulse position, pulse force, pulse rate, tension, and fluency.

[0050] The pulse position, pulse strength, pulse rate, tension, and fluency are matched with the corresponding viscera and meridians of the cun, guan, and chi positions to generate TCM pulse diagnosis analysis results.

[0051] Secondly, the present invention provides a TCM pulse diagnosis and analysis system based on image enhancement, including a pulse image construction module, which simultaneously collects static pressure data and dynamic pulse wave data at the cun, guan, and chi positions, and converts the pulse response intensity, collection time, and pulse taking pressure into grayscale distribution, horizontal time sequence, and vertical pressure layering, respectively, to form a cun-guan-chi pulse grayscale image.

[0052] The image enhancement module performs grayscale background flattening on the grayscale image of the Cun, Guan, and Chi pulses, and extracts the main wave edge, diphtheria wave details, and descending isthmus texture through differential multi-scale morphological top cap enhancement to synthesize the enhanced Cun, Guan, and Chi pulse image.

[0053] The pulse map determination module registers the enhanced pulse map of Cun, Guan and Chi with the static pressure data. Through phase correlation registration, morphological skeleton extraction and connected domain labeling, the main wave skeleton, diabetic wave skeleton, descending isthmus texture skeleton and corresponding pulse pressure are determined to form the enhanced pulse map of Cun, Guan and Chi.

[0054] The feature encoding module expands the polar coordinates of the pulse skeleton in the enhanced image of Cun-Guan-Chi pulse taking, extracts gray-level co-occurrence features and directional gradient features, and constructs a pulse topology map and graph Laplacian spectrum sequence through pressure layer pulse topology encoding. The graph Laplacian spectrum sequence is associated with the pulse taking pressure, gray-level co-occurrence features and directional gradient features to form a Cun-Guan-Chi pulse map feature group.

[0055] The diagnostic analysis module compares the differences in pulse pressure, pulse pattern topology, and Laplace spectral sequence in the characteristic groups of the Cun, Guan, and Chi pulse diagrams to identify pulse position, pulse strength, pulse rate, tension, and fluency, and generates TCM pulse diagnosis analysis results.

[0056] The beneficial effects of this invention are as follows: By converting static pressure data and dynamic pulse wave data at the Cun, Guan, and Chi positions into a grayscale image of the Cun, Guan, and Chi pulse with lateral temporal sequence, longitudinal pressure layering, and grayscale distribution, the pulse data can be incorporated into image enhancement and image recognition processes. Differential multi-scale morphological cap enhancement is employed to highlight the main wave edge, dicrotic wave details, and descending isthmus texture, respectively. Compared to ordinary grayscale enhancement or simple filtering, this method can simultaneously preserve bright details, dark textures, and the main wave edge structure in the pulse image, reducing background interference caused by pressure progression and contact states. Furthermore, this invention no longer determines the pulse location solely based on the maximum amplitude. Instead, this invention determines the continuity of pulse patterns between pressure layers through phase correlation registration, morphological skeleton extraction, and connected component labeling, giving the enhanced pulse image a more stable image structure foundation. More importantly, this invention extracts gray-level co-occurrence features and directional gradient features after unfolding the pulse pattern skeleton in polar coordinates, and constructs a pulse pattern topology map and graph Laplacian spectrum sequence through pulse pattern topology coding between pressure layers. This elevates the main wave, diphtheria wave, and descending isthmus from ordinary image textures to comparable spectral structure features, which is beneficial to improving the stability, interpretability, and correspondence with the TCM syndrome differentiation of pulse position, pulse strength, pulse rate, tension, and fluency recognition. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart of a traditional Chinese medicine pulse diagnosis and analysis method based on image enhancement.

[0059] Figure 2 Flowchart for constructing and enhancing grayscale images of the Cun, Guan, and Chi pulses.

[0060] Figure 3 A flowchart for determining pressure interlayer registration and pulse enhancement mapping is provided.

[0061] Figure 4 This is a flowchart of vein topology coding and dialectical analysis. Detailed Implementation

[0062] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0063] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0064] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0065] Reference Figure 1 This is one embodiment of the present invention, which provides a method for TCM pulse diagnosis analysis based on image enhancement, including the following steps:

[0066] S1. Reference Figure 2 The static pressure data and dynamic pulse wave data at the cun, guan, and chi positions are collected simultaneously. The pulse response intensity, collection time, and pulse pressure are converted into grayscale distribution, horizontal time sequence, and vertical pressure layering, respectively, to form a grayscale image of the cun, guan, and chi pulse.

[0067] S1.1. Static pressure data, dynamic pulse wave data, and location identification are simultaneously acquired at the cun, guan, and chi positions on the subject's wrist. The static pressure data, dynamic pulse wave data, and location identification acquired at the same time are then organized into a multi-site pulse acquisition sequence according to the acquisition time. Among them, static pressure data is used to represent the pulse taking pressure, dynamic pulse wave data is used to represent the pulse response intensity, and location identification is used to distinguish the cun, guan, and chi positions. Each set of data in the multi-site pulse acquisition sequence includes the acquisition time, location identification, pulse taking pressure, and pulse response intensity.

[0068] The pulse acquisition sequence for multiple locations is divided into Cun-Guan-Chi pulse sequences according to the location identifier, so that the Cun-Guan-Chi pulse sequences retain the acquisition time, pulse pressure and pulse response intensity corresponding to the Cun, Guan and Chi positions respectively.

[0069] Among them, the cun, guan, and chi position data at the same acquisition time are kept synchronously corresponding, and the different pulse pressure data at the same location are arranged in the order of pressure progression.

[0070] The pulse response intensity in the Cun-Guan-Chi pulse sequence is converted to grayscale to obtain the Cun-Guan-Chi grayscale response sequence. The grayscale conversion uses the linear grayscale normalization method commonly used in digital image processing to map the pulse response intensity within the same location and pressure layer to grayscale values, enabling dynamic pulse wave data to be converted into a grayscale distribution recognizable by image enhancement processing. The current pulse response intensity is converted to grayscale using the maximum and minimum values ​​of the pulse response intensity within the same location and pressure layer, expressed as:

[0071] ;

[0072] in, Indicates the location of the part, indicating the position of the inch, the gate, and the ruler; Indicates the longitudinal pressure stratification number; Indicates the collection time sequence number; The part is marked as The longitudinal pressure stratification number is The collection time sequence number is The intensity of the pulse response at that time; The part is marked as The longitudinal pressure stratification number is The maximum pulse response intensity within; The part is marked as The longitudinal pressure stratification number is The minimum pulse response intensity within; This represents the converted grayscale value, and its range corresponds to the grayscale range of the digital image. This represents the rounding operation, used to convert the normalized grayscale calculation result into an integer grayscale value.

[0073] when At that time, there is no pulse response change that can be used for image enhancement within the vertical pressure layer, and the gray value is taken as the intermediate gray value of the same vertical pressure layer to maintain the continuity of the gray response sequence.

[0074] The acquisition times in the Cun-Guan-Chi grayscale response sequence are arranged as a horizontal time sequence, and the pulse-taking pressure is arranged as a vertical pressure layer. Grayscale values ​​are then filled into the corresponding horizontal time sequence positions and vertical pressure layer positions to generate a Cun-Guan-Chi pulse grayscale image. The correspondence between image coordinates and the Cun-Guan-Chi grayscale response sequence is represented by a two-dimensional coordinate mapping relationship in digital images as follows:

[0075] ;

[0076] in, The part is marked as Horizontal temporal position in the pulse grayscale image Longitudinal pressure stratification location The grayscale value at that location; Indicates the collection time sequence number The corresponding horizontal time sequence position; Indicates the longitudinal pressure stratification number The corresponding longitudinal pressure stratification location.

[0077] S2. The grayscale image of the Cun, Guan, and Chi pulses is flattened, and the edges of the main wave, details of the diphthoplasty wave, and texture of the descending mid-spine are extracted by differential multi-scale morphological top cap enhancement to synthesize the enhanced Cun, Guan, and Chi pulse image.

[0078] S2.1. Read the grayscale images of the pulse corresponding to the Cun, Guan, and Chi positions according to the location identifiers, and keep the horizontal temporal sequence, vertical pressure layering, and grayscale distribution of the Cun, Guan, and Chi pulse grayscale images unchanged; among them, the horizontal temporal sequence is used to represent the sequential relationship of the acquisition time, the vertical pressure layering is used to represent the progressive relationship of the pulse acquisition pressure, and the grayscale distribution is used to represent the imaged result of the pulse response intensity, so as to obtain the original grayscale images of Cun, Guan, and Chi that can be image-enhanced.

[0079] The background grayscale field generated by the pressure progression is separated from the original grayscale image of the Cun, Guan, and Chi pulses to obtain a pressure-layered background image. The pressure-layered background image is used to represent the overall grayscale changes caused by wrist contact, sensor contact surface, and pulse taking pressure progression. The pressure-layered background image is not used to represent the main wave edge, diphtheria wave details, and descending isthmus texture. The pressure-layered background image is extracted using grayscale morphological opening operation, which is commonly used in digital image processing. Grayscale morphological opening operation is derived from erosion and dilation operations in mathematical morphology. It first weakens local fine bright textures and then restores the background outline, which is suitable for separating the slowly changing background grayscale field from the Cun, Guan, and Chi pulse grayscale image.

[0080] The original grayscale image of the Cun Guan Chi is subjected to grayscale morphological opening operation using background structuring elements. The expression is:

[0081] ;

[0082] in, The part is marked as Pressure stratification background map in horizontal time position Longitudinal pressure stratification location The background grayscale value at that location; This indicates the horizontal temporal position of the grayscale image of the Cun, Guan, and Chi pulses. Longitudinal pressure stratification location The grayscale value at that location; This represents the background structuring element used to extract the background grayscale field. The scale of the background structuring element is larger than the local texture width of the diphthral wave details and the main wave edge in the grayscale image of the pulse at the cun, guan, and chi positions. This represents the grayscale morphological opening operation.

[0083] S2.2. Perform grayscale difference correction on the pressure layer background image and the original grayscale image of Cun Guan Chi, and retain the location identification, horizontal time sequence and vertical pressure layer to generate a flat pulse grayscale image; the grayscale difference correction is calculated by the grayscale difference value of the same horizontal time sequence position and the same vertical pressure layer position, so that the background grayscale field generated by the pressure progression is suppressed, and the main wave edge, diphtheria wave details and descending isthmus texture are retained in the flat pulse grayscale image as objects for subsequent image enhancement.

[0084] When the grayscale difference is less than zero, the grayscale value at the corresponding position in the flat pulse grayscale image is set to zero to ensure that the flat pulse grayscale image still conforms to the grayscale range of the digital image.

[0085] Small-scale structuring elements are used to perform white-cap transformation on the grayscale image of a flat pulse to extract diphthrombus details and obtain a bright detail image. The white-cap transformation is derived from the opening operation difference processing in mathematical morphology and is used to highlight bright details smaller than the small-scale structuring elements in the grayscale image of a flat pulse. Diphthrombus details are usually manifested as local bright textures after the main wave, so small-scale structuring elements are used to extract diphthrombus details.

[0086] The expression for calculating the highlight detail map by the difference between the grayscale image of the smooth pulse and the grayscale image of the smooth pulse is as follows:

[0087] ;

[0088] in, The part is marked as The bright detail image in the horizontal time sequence position Longitudinal pressure stratification location The grayscale value at that location; This indicates the horizontal temporal position of the flat pulse grayscale image. Longitudinal pressure stratification location The grayscale value at that location; This represents a small-scale structural element, the scale of which matches the local width of the diphtheria wave detail in the flat pulse grayscale image. This represents the grayscale morphological opening operation; the bright detail image serves as the image source for extracting the main wave edge.

[0089] S2.3. Morphological enhancement of the bright detail map is performed using mesoscale structuring elements to extract the rising and falling edges of the main wave, resulting in a main wave edge map. The main wave edge map represents the gray-level abrupt change regions of the main wave in the lateral temporal direction. The scale of the mesoscale structuring elements matches the texture width of the rising and falling edges of the main wave in the bright detail map. The main wave edges are extracted using gray-level morphological gradients, which are derived from dilation and erosion operations in mathematical morphology. The difference between the dilation and erosion results is used to highlight the image edges.

[0090] The expression for calculating the grayscale morphological gradient of the bright detail image using mesoscale structuring elements is as follows:

[0091] ;

[0092] in, The part is marked as The main wave edge map in the lateral time sequence position Longitudinal pressure stratification location The grayscale value at that location; Indicates the horizontal time sequence position of the highlight detail image. Longitudinal pressure stratification location The grayscale value at that location; Represents mesoscale structural elements; This represents the grayscale morphological dilation operation; This represents the grayscale morphological erosion operation.

[0093] S2.4. A black hat transformation is performed on the flat pulse grayscale image using a large-scale structuring element to extract the dark texture of the mid-valley and obtain a dark texture image. The black hat transformation is derived from the difference processing of the closing operation in mathematical morphology and is used to highlight dark textures smaller than the large-scale structuring element. The mid-valley in the pulse grayscale image is represented by a local low grayscale area after the main wave. The dark texture of the mid-valley is extracted by the large-scale structuring element.

[0094] The expression for calculating the dark texture map by the difference between the grayscale morphological closing operation result of the flattened pulse grayscale image and the flattened pulse grayscale image is as follows:

[0095] ;

[0096] in, The part is marked as Dark texture map in lateral temporal position Longitudinal pressure stratification location The grayscale value at that location; This indicates the horizontal temporal position of the flat pulse grayscale image. Longitudinal pressure stratification location The grayscale value at that location; This represents a large-scale structural element, the scale of which matches the local width of the descending canyon dark texture in the flat pulse grayscale image. This represents the grayscale morphological closing operation.

[0097] The main wave edge map, bright detail map, and dark texture map are differentially synthesized to generate the Cun-Guan-Chi enhanced pulse map. The differential synthesis preserves the main wave rising edge and main wave falling edge in the main wave edge map, superimposes the diphtheria wave details in the bright detail map, and enhances the descending isthmus texture in the dark texture map, so that the Cun-Guan-Chi enhanced pulse map simultaneously contains the main wave edge, diphtheria wave details, and descending isthmus texture.

[0098] The grayscale values ​​of the main wave edge map, bright detail map, and dark texture map at the same horizontal time position and the same vertical pressure stratification position are differentially synthesized, and the expression is:

[0099] ;

[0100] in, The part is marked as The enhanced pulse image of Cun, Guan, and Chi positions in the horizontal time sequence Longitudinal pressure stratification location The grayscale value at that location.

[0101] When the differential synthesis result exceeds the digital image grayscale range (0-255), that is, when the differential synthesis result is less than 0, the corresponding grayscale value is limited to 0; when the differential synthesis result is greater than 255, the corresponding grayscale value is limited to 255, so that the grayscale value of the Cun-Guan-Chi enhanced pulse image is kept within the eight-bit grayscale image range of 0 to 255, so as to meet the image enhancement input requirements for subsequent pressure layer inter-pulse correspondence and pulse image area determination.

[0102] S3. Reference Figure 3 The enhanced pulse image of Cun, Guan, and Chi was registered with the static pressure data. The main wave skeleton, diphtheria wave skeleton, descending isthmus texture skeleton, and corresponding pulse taking pressure were determined by phase correlation registration, morphological skeleton extraction, and connected domain labeling to form the enhanced pulse taking image of Cun, Guan, and Chi.

[0103] S3.1. Extract the enhanced pulse images of Cun, Guan and Chi corresponding to the Cun, Guan and Chi parts according to the location identification, and extract the enhanced image blocks of adjacent pressure layers in each part according to the longitudinal pressure layering order to obtain the pressure layer image block sequence; each group of enhanced image blocks in the pressure layer image block sequence retains the horizontal time sequence position, the longitudinal pressure layering position and the gray value.

[0104] Among them, the horizontal temporal position is used to represent the unfolding position of the vein in the acquisition time direction, the vertical pressure layer position is used to represent the progressive relationship between adjacent pulse acquisition pressure layers, and the gray value is used to represent the main wave edge, diphtheria wave details and descending isthmus texture after image enhancement.

[0105] Phase correlation calculations are performed on the inter-pressure layer image block sequence. Based on the obtained lateral displacement, the vein positions in the inter-pressure layer image block sequence are corrected to generate an inter-pressure layer registration pulse map. The phase correlation calculation adopts the existing method based on the Fourier shift theorem in digital image registration. The cross power spectrum is calculated through the Fourier transform results of adjacent pressure layer enhanced image blocks. Then, the cross power spectrum is subjected to inverse Fourier transform to obtain the correlation response map. The peak position in the correlation response map corresponds to the displacement between adjacent pressure layer enhanced image blocks. Since the vein changes between adjacent pressure layers in the same location are mainly manifested as lateral temporal shifts, the lateral displacement corresponding to the peak value is extracted from the correlation response map, and the lateral displacement is used to correct the lateral position of the next pressure layer enhanced image block.

[0106] The expression for calculating the correlation response map between adjacent pressure-layered enhanced image patches using phase correlation is as follows:

[0107] ;

[0108] in, The part is marked as The The first longitudinal pressure stratification and the first The correlation response diagram between the longitudinal pressure strata at location The response value at the location; Indicates the inverse Fourier transform; The part is marked as The Fourier transform results of a longitudinal pressure-layered image block; The part is marked as The Fourier transform conjugate of a vertical pressure-layered image patch; Represents frequency domain coordinates; This indicates the modulo operation.

[0109] The coordinates corresponding to the peak position in the correlation response graph are: ,in, Indicates the first The first longitudinal pressure stratification and the first Lateral displacement between longitudinal pressure layers This indicates the longitudinal displacement.

[0110] During the registration of adjacent pressure stratification at the same location, take As the pulse position correction amount, and based on the lateral displacement, the position of the next pressure-layered enhanced image block is corrected, the expression is:

[0111] ;

[0112] in, Indicates the first after lateral displacement correction A vertical pressure-layered image patch at a horizontal temporal location Longitudinal pressure stratification location The grayscale value at that location; Indicates the first before correction A vertical pressure-layered image patch at a horizontal temporal location Longitudinal pressure stratification location The grayscale value at that location; This indicates the corresponding lateral displacement.

[0113] When multiple peaks appear in the correlation response map, the lateral displacement corresponding to the peak with the largest response value is taken; when the lateral displacement exceeds the lateral temporal range of the enhanced image block, the excess part is limited within the boundary of the enhanced image block to ensure the continuity of the registration results between pressure layers.

[0114] The horizontal time series range is determined by the number of acquisition time numbers within the same location and the same longitudinal pressure layer. For example, dynamic pulse wave data for 10 seconds is acquired at a sampling frequency of 100 Hz within each longitudinal pressure layer, forming 1000 acquisition time numbers. The horizontal time series range is from the first acquisition time number to the 1000th acquisition time number. Enhanced image blocks in the image block sequence between pressure layers are all registered within this horizontal time series range.

[0115] S3.2. Corresponding the pressure-coordinated pulse image to the static pressure data, a pressure-coordinated enhanced pulse image is obtained. Each pulse pressure value in the static pressure data corresponds one-to-one with the longitudinal pressure stratification position in S1. Each enhanced image block in the pressure-coordinated pulse image retains the corresponding longitudinal pressure stratification position. Therefore, the pressure-coordinated pulse image is correlated with the static pressure data according to the longitudinal pressure stratification position, so that each registered enhanced image block corresponds to a specific pulse pressure value, resulting in a pressure-coordinated enhanced pulse image that simultaneously contains image enhancement results and pulse pressure information.

[0116] The pulse pattern region map is obtained by extracting the main wave region, dicrotic wave region, and descending isthmus region from the pressure registration enhanced pulse map. The main wave region is used to represent the main wave structure with the most obvious and continuous gray-scale changes in the pulse cycle. The dicrotic wave region is used to represent the local fluctuation structure that reappears after the main wave. The descending isthmus region is used to represent the local low gray-scale transition structure between the main wave and the dicrotic wave. Within each longitudinal pressure layer, the main wave region, descending isthmus region, and dicrotic wave region are identified sequentially according to the lateral temporal direction. The location identifier, lateral temporal position, and longitudinal pressure layer position are preserved to obtain the pulse pattern region map used for morphological skeleton extraction.

[0117] S3.3. Morphological skeleton extraction is performed on the vein pattern region map to generate the main wave skeleton, diphtheria wave skeleton, and descending isthmus texture skeleton, and connected component labeling is applied to obtain a skeleton connectivity label map. Morphological skeleton extraction adopts the morphological thinning method in digital image processing. While maintaining the topological structure of the vein pattern region map, the main wave region, diphtheria wave region, and descending isthmus region are gradually thinned to a single-pixel width of the center line. Specifically, the main wave region is thinned to obtain the main wave skeleton, the diphtheria wave region is thinned to obtain the diphtheria wave skeleton, and the descending isthmus region is thinned to obtain the descending isthmus texture skeleton. After obtaining the main wave skeleton, diphtheria wave skeleton, and descending isthmus texture skeleton, connected component labeling is used to assign connectivity numbers to the skeleton pixels in the same longitudinal pressure layer, so that pixels belonging to the same continuous skeleton have the same connected component label, resulting in a skeleton connectivity label map.

[0118] The continuity of the skeleton within the same pressure stratum and the continuity relationship between adjacent pressure stratums are identified by skeleton connectivity markers to determine the pulse image region. Skeleton continuity indicates whether the main wave skeleton, diphtheria wave skeleton, and descending isthmus texture skeleton remain continuous within the same longitudinal pressure stratum. Skeleton continuity indicates whether the corresponding skeletons between adjacent longitudinal pressure stratums maintain stable continuity in the transverse temporal direction. When the main wave skeleton, diphtheria wave skeleton, and descending isthmus texture skeleton are continuous within the same longitudinal pressure stratum, and the main wave skeleton, diphtheria wave skeleton, and descending isthmus texture skeleton between adjacent longitudinal pressure stratums maintain positional continuity after registration between pressure strata, the corresponding enhanced image region is determined as the pulse image region. When multiple pulse image regions satisfying the conditions exist in the same location, the enhanced image region with the longest continuous length of the main wave skeleton, the most stable position of the diphtheria wave skeleton, and the clearest continuity relationship of the descending isthmus texture skeleton is selected as the pulse image region to ensure that the image enhancement result in the pulse image region can stably reflect the pulse morphology of that location.

[0119] S3.4. Perform pressure layer mapping between the pulse image area and the static pressure data to determine the pulse pressure at the cun, guan, and chi positions, and generate enhanced pulse images at the cun, guan, and chi positions. The pressure layer mapping is performed one-to-one with the pulse pressure value in S1 according to the longitudinal pressure layer position of the pulse image area. When the pulse image area is located at the longitudinal pressure layer position corresponding to the cun position, the cun pulse pressure is obtained and the cun pulse enhancement image is generated. When the pulse image area is located at the longitudinal pressure layer position corresponding to the guan position, the guan pulse pressure is obtained and the guan pulse enhancement image is generated. When the pulse image area is located at the longitudinal pressure layer position corresponding to the chi position, the chi pulse pressure is obtained and the chi pulse enhancement image is generated. The enhanced pulse images at the cun, guan, and chi positions retain the corresponding main wave skeleton, dicrotic wave skeleton, and descending isthmus texture skeleton, respectively.

[0120] S4. Reference Figure 4 The polar coordinates of the pulse skeleton in the enhanced image of Cun-Guan-Chi pulse taking are expanded, gray-level co-occurrence features and directional gradient features are extracted, and a pulse topology map and a graph Laplacian spectrum sequence are constructed through pressure layer pulse topology coding. The graph Laplacian spectrum sequence is associated with the pulse taking pressure, gray-level co-occurrence features and directional gradient features to form a Cun-Guan-Chi pulse map feature group.

[0121] S4.1. Extract the enhanced pulse images from the Cun, Guan, and Chi regions according to the location identifiers. Extract the main wave skeleton, diphtheria wave skeleton, and descending isthmus texture skeleton from the Cun, Guan, and Chi regions respectively. Determine the main wave skeleton, diphtheria wave skeleton, and descending isthmus texture skeleton as the center line of the pulse skeleton. The center line of the pulse skeleton is used to represent the central direction of the main wave structure, diphtheria wave structure, and descending isthmus structure in the image enhancement result.

[0122] Along the center line of the vein skeleton, extract the main wave texture area corresponding to the main wave skeleton, the diphtheria wave texture area corresponding to the diphtheria wave skeleton, and the descending isthmus texture area corresponding to the descending isthmus texture skeleton to both sides. So that the main wave texture area, the diphtheria wave texture area, and the descending isthmus texture area retain the lateral temporal position, the longitudinal pressure layer position, and the gray value in the cun-pin pulse enhancement image, guan-pin pulse enhancement image, and chi-pin pulse enhancement image, respectively, to obtain the vein texture region used for image enhancement feature extraction.

[0123] The vein texture region is expanded in polar coordinates according to the skeleton direction of the vein skeleton centerline to obtain the vein expansion image. The polar coordinate expansion adopts the rectangular coordinate to polar coordinate transformation method commonly used in digital image processing, and the expansion direction is established in combination with the local direction of the vein skeleton centerline, so that the main wave texture region, diphtheria wave texture region and descending isthmus texture region are expanded into a continuous texture image along the direction of the vein skeleton centerline.

[0124] The expression for polar coordinate expansion of the vein texture region using the sampling point positions on the center line of the vein skeleton and the local orientation at the sampling points is as follows:

[0125] ;

[0126] in, The part is marked as The vein pattern unfolding diagram in the first At the sampling point of the center line of the vein skeleton, the extreme diameter is Polar angle is The grayscale value of the location; The part is marked as The vein-textured area; and The part is marked as The The sampling points of the center line of the vein skeleton are located at the horizontal temporal position and the vertical pressure layer position in the original image. The part is marked as The Local orientation angle at the sampling point of the center line of the vein skeleton; This represents the radial distance from the sampling point to the center line of the vein skeleton; This indicates the angle of expansion around the center line of the vein skeleton.

[0127] The local orientation angle is calculated from the position difference between adjacent sampling points on the center line of the vein skeleton, so that the polar coordinate unfolding result is consistent with the actual orientation of the center line of the vein skeleton, and a vein unfolding map is obtained for gray-level co-occurrence relationship statistics and orientation change extraction.

[0128] S4.2. Perform gray-level co-occurrence relationship statistics and direction change extraction on the vein pattern unfolded image to generate gray-level co-occurrence features and directional gradient features. The gray-level co-occurrence relationship statistics employ the gray-level co-occurrence matrix method commonly used in digital image processing. This method statistically analyzes the occurrence of gray-level combinations of adjacent pixels in the vein pattern unfolded image, representing the texture density, gray-level variations, and local uniformity of the main wave texture area, diphtheria wave texture area, and descending isthmus texture area. The direction change extraction uses the directional gradient histogram method commonly used in digital image processing. This method segmentally analyzes the direction of pixel gray-level changes in the vein pattern unfolded image, representing the main wave rising direction, the main wave falling direction, and the diphtheria wave turning direction. The statistical directions in the gray-level co-occurrence matrix can be the horizontal temporal direction, the vertical pressure layering direction, and the diagonal direction. The statistical directions in the directional gradient histogram are recorded in segments according to the gradient direction, so that the gray-level co-occurrence features and directional gradient features can jointly characterize the texture differences and orientation differences in the image enhancement results, and obtain the gray-level co-occurrence features and directional gradient features corresponding to the cun-bu pulse enhancement image, guan-bu pulse enhancement image, and chi-bu pulse enhancement image, respectively.

[0129] Using the positional correspondence of the vein skeleton centerline as an index, gray-level co-occurrence features and directional gradient features are mapped to the same vein coordinates to obtain vein image features. The positional correspondence of the vein skeleton centerline is used to maintain consistent correspondence between gray-level co-occurrence features and directional gradient features at the same horizontal temporal position, the same vertical pressure layer position, and the same vein category. The vein categories include the main wave texture area, the diphtheria wave texture area, and the descending isthmus texture area. After completing the one-to-one correspondence between gray-level co-occurrence features and directional gradient features through the positional correspondence of the vein skeleton centerline, the main wave texture area, the diphtheria wave texture area, and the descending isthmus texture area form unified vein image features under the same vein coordinates. These vein image features simultaneously retain the texture and orientation information from the image enhancement result.

[0130] S4.3. Identify the endpoints of the main wave skeleton, the ridges of the main wave peak, the inflection points of the diphthora wave, and the ridges of the descending valley from the features of the pulse pattern image to form a pulse pattern node set. The endpoints of the main wave skeleton are used to indicate the start and end positions of the main wave skeleton in the lateral temporal direction; the ridges of the main wave peak are used to indicate the positions where the gray-level bulges of the main wave skeleton are most obvious; the inflection points of the diphthora wave are used to indicate the positions where the directional changes of the diphthora wave skeleton are most obvious; and the ridges of the descending valley are used to indicate the positions where the gray-level troughs of the descending valley texture skeleton are most obvious. After identification, record the spatial positions of the endpoints of the main wave skeleton, the ridges of the main wave peak, the inflection points of the diphthora wave, and the ridges of the descending valley according to the lateral temporal position and the longitudinal pressure stratification position to obtain the pulse pattern node set.

[0131] The skeletal connection segments within the same cycle are defined as the periodic vein connection set, and the vein nodes corresponding to the positions in adjacent pressure layers are defined as the interlayer vein connection set. A vein topology map is constructed according to the transverse temporal sequence, longitudinal pressure layer, periodic vein connection set, and interlayer vein connection set. The periodic vein connection set represents the connection relationship between the endpoints of the main wave skeleton, the ridge point of the main wave peak, the diphthoplast inflection point, and the ridge point of the descending valley within the same pulse cycle. The interlayer vein connection set represents the continuity relationship of corresponding vein nodes between adjacent longitudinal pressure layers. The transverse temporal sequence describes the sequential order of vein nodes within the same pulse cycle, and the longitudinal pressure layer describes the interlayer distribution order of vein nodes under different pulse-taking pressures. The periodic vein connection set and the interlayer vein connection set together define the connection mode between vein nodes, resulting in vein topology maps corresponding to the cun, guan, and chi regions, respectively.

[0132] S4.4. Using the set of vein nodes as the matrix index, the set of periodic vein connections and the set of interlayer vein connections are transformed into a vein connection matrix. The vein connection matrix is ​​then decomposed into a graph Laplace matrix to obtain a graph Laplace spectrum sequence. The graph Laplace decomposition adopts the graph Laplace matrix decomposition method in spectral theory. First, the vein connection matrix is ​​constructed based on the set of vein nodes and the connection relationship. Then, the degree matrix and the graph Laplace matrix are obtained based on the vein connection matrix. Finally, the graph Laplace spectrum sequence is obtained through the eigenvalue decomposition of the graph Laplace matrix.

[0133] The expressions for the Graph Laplace matrix and the Graph Laplace spectral sequence are:

[0134] ;

[0135] ;

[0136] in, The part is marked as The graph Laplace matrix; The part is marked as The degree matrix, where the diagonal elements represent the number of connections between each vein node and other vein nodes in the vein node set. The part is marked as The vein connection matrix, where the matrix elements indicate whether there is a periodic vein connection relationship or an interlayer vein connection relationship between two corresponding vein nodes. The part is marked as The 1 eigenvector; The part is marked as The Each feature value.

[0137] The eigenvalues ​​are arranged in ascending order to form a graphical Laplacian spectrum sequence, which is used to characterize the overall connectivity structure and interlayer continuity structure of the vein topology.

[0138] The pulse pattern topology map, Laplace spectral sequence, pulse pressure, and pulse pattern image features are correlated to form pulse pattern feature groups for the Cun, Guan, and Chi regions. This correlation is based on the same location, the same longitudinal pressure layer, and the same transverse temporal position. The Cun pulse pattern feature group includes Cun pulse pressure, Cun pulse pattern image features, Cun pulse pattern topology map, and Cun Laplace spectral sequence; the Guan pulse pattern feature group includes Guan pulse pressure, Guan pulse pattern image features, Guan pulse pattern topology map, and Guan Laplace spectral sequence; and the Chi pulse pattern feature group includes Chi pulse pressure, Chi pulse pattern image features, Chi pulse pattern topology map, and Chi Laplace spectral sequence. Together, these three groups form the Cun-Guan-Chi pulse pattern feature group.

[0139] S5. By comparing the differences in pulse pressure, pulse pattern topology, and Laplace spectral sequence in the pulse image feature group of Cun, Guan, and Chi, the pulse position, pulse strength, pulse rate, tension, and fluency are identified, and the results of TCM pulse diagnosis analysis are generated.

[0140] S5.1. The pulse feature groups of Cun, Guan and Chi are matched according to the Cun, Guan and Chi parts to obtain the correspondence of the three pulse maps; the correspondence of the three pulse maps is used to maintain the image enhancement feature correspondence of Cun, Guan and Chi under the same horizontal time sequence, the same vertical pressure layer and the same pulse pattern category.

[0141] By comparing the pulse-taking pressure in the corresponding relationship of the three pulse diagrams, the pulse-taking pressure difference characteristics are obtained. The pulse-taking pressure difference characteristics are used to represent the relative changes in pulse-taking pressure at the cun, guan, and chi positions. The pulse-taking pressure difference characteristics are determined by the difference between the pulse-taking pressure at the cun, guan, and chi positions.

[0142] The calculation of pulse pressure difference uses the absolute difference operation in mathematics. The absolute difference operation is used to represent the distance relationship between two pulse pressures, and the expression is:

[0143] ;

[0144] in, Indicates location with part Characteristics of differences in pulse-taking pressure between individuals; Indicates location The pressure applied during pulse taking; Indicates location The pressure applied during pulse taking; the location and parts Take any two parts from the cun, guan, and chi sections respectively; This represents absolute value operations.

[0145] For example, if the pulse pressure at the cun position is 60 mmHg, the pulse pressure at the guan position is 80 mmHg, and the pulse pressure at the chi position is 100 mmHg, then the characteristic difference in pulse pressure between the cun and guan positions is 20 mmHg, the characteristic difference in pulse pressure between the guan and chi positions is 20 mmHg, and the characteristic difference in pulse pressure between the cun and chi positions is 40 mmHg.

[0146] S5.2. Compare the vein topology diagrams in the correspondence of the three vein maps to obtain the vein topology difference features. The vein topology difference features are used to represent the differences in the connection structure of the main wave skeleton endpoints, main wave peak ridges, diphtheria inflection points, and descending valley ridges between the Cun, Guan, and Chi sections. In the comparison process, firstly, establish the correspondence of similar vein nodes between the Cun, Guan, and Chi sections according to the vein node set, then compare the connection status between similar vein nodes according to the periodic vein connection set and the interlayer vein connection set, and finally count the vein node connection relationships with differences in the vein topology diagram to obtain the vein topology difference features. The vein topology difference features continue to retain the location identifier, lateral temporal position, and longitudinal pressure stratification position.

[0147] By comparing the graph Laplace spectral sequences in the correspondence of the three pulse maps, the characteristics of the continuity difference between pressure layers are obtained; the characteristics of the continuity difference between pressure layers are used to represent the overall changes in the pulse connection structure of the cun, guan and chi parts during the pressure progression.

[0148] The difference between the Graph Laplace spectral sequences is calculated using the Euclidean distance in mathematics. The Euclidean distance is used to represent the overall difference between two spectral sequences, and its expression is:

[0149] ;

[0150] in, Indicates location with part The interlayer pressure variation characteristics between them; Indicates location The first of the Laplace spectral sequences One eigenvalue; Indicates location The first of the Laplace spectral sequences One eigenvalue; Indicates the number of feature values ​​involved in the comparison; location and parts Take any two parts from the cun, guan, and chi sections respectively.

[0151] S5.3. Based on the correspondence of the three pulse diagrams, pulse image features are identified by pulse pressure difference features, pulse pattern topology difference features, pressure layer continuity difference features, gray-scale co-occurrence features, and directional gradient features, and pulse position, pulse force, pulse rate, tension, and fluency are obtained.

[0152] Specifically, the pulse pressure difference characteristics are correlated with the longitudinal pressure stratification location. The superficial, intermediate, and deep pulse-taking directions are determined according to the low-pressure, intermediate-pressure, and high-pressure stratifications, respectively, to obtain the pulse location. The gray-level variations and local uniformity in the gray-level co-occurrence characteristics are correlated with the main wave texture area, dicrotic wave texture area, and descending isthmus texture area to determine pulse strength and fluency. The upward and downward directions of the main wave and the turning direction of the dicrotic wave in the directional gradient characteristics are correlated with the direction changes of the pulse pattern skeleton centerline to determine tension. The pulse rate is determined by corresponding the repetition interval between adjacent main wave peaks and ridges in the transverse time series to the pulse pattern period repetition relationship. The pulse pattern topological difference features and pressure layer continuity difference features are used to check the connection stability of the main wave skeleton, diphtheria wave skeleton and descending isthmus texture skeleton in different pressure layers to obtain the pulse pattern structure stability. Through the correspondence between pulse position, pulse strength, pulse rate, tension, fluency and pulse pattern structure stability, the image enhancement results of the cun, guan and chi parts are converted into pulse position, pulse strength, pulse rate, tension and fluency.

[0153] The pulse position, pulse strength, pulse rate, tension, and fluency are matched with the corresponding viscera and meridians of the cun, guan, and chi positions to generate TCM pulse diagnosis analysis results. The viscera and meridians corresponding to the cun, guan, and chi positions are used to limit the diagnosis of pulse characteristics in different locations. The pulse position, pulse strength, pulse rate, tension, and fluency are used to represent the pulse manifestation corresponding to the pulse pattern structure after image enhancement. After matching, TCM pulse diagnosis analysis results containing pulse name analysis results, visceral deficiency and excess tendencies, and meridian blockage tendencies are generated.

[0154] This embodiment also provides a TCM pulse diagnosis and analysis system based on image enhancement, including: a pulse image construction module, which synchronously collects static pressure data and dynamic pulse wave data at the cun, guan, and chi positions, and converts the pulse response intensity, collection time, and pulse taking pressure into grayscale distribution, horizontal time sequence, and vertical pressure layering, respectively, to form a cun-guan-chi pulse grayscale image.

[0155] The image enhancement module performs grayscale background flattening on the grayscale image of the Cun, Guan, and Chi pulses, and extracts the main wave edge, diphtheria wave details, and descending isthmus texture through differential multi-scale morphological top cap enhancement to synthesize the enhanced Cun, Guan, and Chi pulse image.

[0156] The pulse map determination module registers the enhanced pulse map of Cun, Guan and Chi with the static pressure data. Through phase correlation registration, morphological skeleton extraction and connected domain labeling, the main wave skeleton, diabetic wave skeleton, descending isthmus texture skeleton and corresponding pulse pressure are determined to form the enhanced pulse map of Cun, Guan and Chi.

[0157] The feature encoding module expands the polar coordinates of the pulse skeleton in the enhanced image of Cun-Guan-Chi pulse taking, extracts gray-level co-occurrence features and directional gradient features, and constructs a pulse topology map and graph Laplacian spectrum sequence through pressure layer pulse topology encoding. The graph Laplacian spectrum sequence is associated with the pulse taking pressure, gray-level co-occurrence features and directional gradient features to form a Cun-Guan-Chi pulse map feature group.

[0158] The diagnostic analysis module compares the differences in pulse pressure, pulse pattern topology, and Laplace spectral sequence in the characteristic groups of the Cun, Guan, and Chi pulse diagrams to identify pulse position, pulse strength, pulse rate, tension, and fluency, and generates TCM pulse diagnosis analysis results.

[0159] In summary, this invention transforms static pressure data and dynamic pulse wave data at the Cun, Guan, and Chi positions into a grayscale image of the Cun, Guan, and Chi pulse with lateral temporal sequence, longitudinal pressure layering, and grayscale distribution, enabling the pulse data to enter the image enhancement and image recognition process. Differential multi-scale morphological cap enhancement is employed to highlight the main wave edge, diphtheria wave details, and descending isthmus texture, respectively. Compared to ordinary grayscale enhancement or simple filtering, this simultaneously preserves bright details, dark textures, and the main wave edge structure in the pulse image, reducing background interference caused by pressure progression and contact states. Furthermore, this invention no longer determines the pulse location solely based on the maximum amplitude, but... This invention determines the continuity of pulse patterns between pressure layers through phase correlation registration, morphological skeleton extraction, and connected component labeling, giving the enhanced pulse image a more stable image structure foundation. More importantly, after unfolding the pulse pattern skeleton in polar coordinates, this invention extracts gray-level co-occurrence features and directional gradient features, and constructs a pulse pattern topology map and graph Laplacian spectrum sequence through pulse pattern topology coding between pressure layers. This elevates the main wave, diphtheria wave, and descending isthmus from ordinary image textures to comparable spectral structure features, which is beneficial to improving the stability, interpretability, and correspondence with the TCM syndrome differentiation of pulse position, pulse strength, pulse rate, tension, and fluency recognition.

[0160] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for TCM pulse diagnosis based on image enhancement, characterized in that: include, Static pressure data and dynamic pulse wave data at the Cun, Guan, and Chi positions are collected simultaneously. The pulse response intensity, collection time, and pulse taking pressure are converted into grayscale distribution, horizontal time sequence, and vertical pressure layering, respectively, to form a grayscale image of the Cun, Guan, and Chi pulse. The grayscale image of the Cun, Guan, and Chi pulses was flattened, and the edge of the main wave, details of the diphthop wave, and texture of the descending septum were extracted by differential multi-scale morphological top cap enhancement to synthesize the enhanced Cun, Guan, and Chi pulse image. The enhanced pulse image of Cun, Guan, and Chi is registered with the static pressure data. The main wave skeleton, diphtheria wave skeleton, descending isthmus texture skeleton and corresponding pulse taking pressure are determined by phase correlation registration, morphological skeleton extraction and connected domain labeling to form the enhanced pulse taking image of Cun, Guan, and Chi. The polar coordinates of the pulse skeleton in the enhanced image of Cun Guan Chi pulse taking are expanded, gray-level co-occurrence features and directional gradient features are extracted, and a pulse topology map and graph Laplacian spectrum sequence are constructed through pressure layer pulse topology coding. The graph Laplacian spectrum sequence is associated with the pulse taking pressure, gray-level co-occurrence features and directional gradient features to form the Cun Guan Chi pulse map feature group. By comparing the differences in pulse pressure, pulse pattern topology, and Laplace spectral sequence in the pulse image feature group of Cun, Guan, and Chi, the pulse position, pulse strength, pulse rate, tension, and fluency are identified, and the results of TCM pulse diagnosis analysis are generated.

2. The method for TCM pulse diagnosis based on image enhancement as described in claim 1, characterized in that: The formation of the grayscale image of the cun, guan, and chi pulses specifically includes: Static pressure data, dynamic pulse wave data, and location identifiers at the cun, guan, and chi positions are acquired simultaneously and organized into a multi-location pulse acquisition sequence according to the acquisition time. The pulse acquisition sequence from multiple locations is divided into Cun-Guan-Chi pulse sequences according to the location identifier, and the pulse response intensity in the Cun-Guan-Chi pulse sequences is converted into grayscale values ​​to obtain the Cun-Guan-Chi grayscale response sequence. The acquisition time in the Cun-Guan-Chi grayscale response sequence is arranged as a horizontal time sequence, and the pulse taking pressure is arranged as a vertical pressure layer to generate a Cun-Guan-Chi pulse grayscale image.

3. The method for TCM pulse diagnosis based on image enhancement as described in claim 2, characterized in that: The grayscale background smoothing process specifically includes: By separating the background grayscale field generated by the pressure progression from the grayscale image of the Cun, Guan, and Chi pulses, a pressure-layered background image is obtained. The grayscale difference between the pressure layer background image and the Cun-Guan-Chi pulse grayscale image is corrected, and the location identification, horizontal time sequence and vertical pressure layer are preserved to generate a flat pulse grayscale image.

4. The image enhancement-based TCM pulse diagnosis and analysis method as described in claim 3, characterized in that: The extraction of the main wave edge, diphtheria wave details, and descending isthmus texture through differential multi-scale morphological cap enhancement specifically includes: Small-scale structuring elements are used to perform white cap transformation on the flat pulse grayscale image to extract diphtheria wave details and obtain a bright detail image; The bright detail map is morphologically enhanced using mesoscale structural elements to extract the rising edge and falling edge of the main wave, thus obtaining the main wave edge map; A black cap transform was performed on the flat pulse grayscale image using large-scale structuring elements to extract the dark texture of the mid-straight canyon and obtain the dark texture image. The main wave edge image, bright detail image, and dark texture image are differentially synthesized to generate an enhanced pulse image of the cun, guan, and chi positions.

5. The image enhancement-based TCM pulse diagnosis and analysis method as described in claim 4, characterized in that: The registration of the enhanced pulse image at the cun, guan, and chi positions with the static pressure data specifically includes: Enhanced image blocks of adjacent pressure layers at the same location are extracted from the enhanced pulse image of Cun, Guan, and Chi, resulting in a sequence of image blocks between pressure layers. Phase correlation calculation is performed on the pressure interlayer image block sequence, and the vein position in the pressure interlayer image block sequence is corrected based on the obtained lateral displacement to generate a pressure interlayer registration pulse map. By correlating the pressure interlayer registration pulse map with the static pressure data, a pressure registration enhanced pulse map is obtained.

6. The method for TCM pulse diagnosis analysis based on image enhancement as described in claim 5, characterized in that: The determination of the main wave skeleton, diphtheria wave skeleton, descending isthmus texture skeleton, and corresponding pulse taking pressure specifically includes: The main wave region, diabetic wave region, and descending isthmus region are extracted from the pressure registration enhanced pulse map to obtain the pulse pattern region map; Morphological skeleton extraction was performed on the vein pattern region map to generate the main wave skeleton, diphtheria wave skeleton and descending isthmus texture skeleton, and connected component marking was performed to obtain the skeleton connected marking map. The continuous state of the skeleton within the same pressure layer and the skeleton continuity relationship between adjacent pressure layers are identified by skeleton connectivity marker maps to determine the pulse image region. The pulse image area is mapped to the static pressure data to determine the pulse pressure at the cun, guan, and chi positions, and to generate enhanced pulse images at the cun, guan, and chi positions.

7. The method for TCM pulse diagnosis based on image enhancement as described in claim 6, characterized in that: The extraction of gray-level co-occurrence features and directional gradient features specifically includes: The main wave skeleton, diabetic wave skeleton, and descending isthmus texture skeleton in the enhanced pulse diagnosis diagrams of the cun, guan, and chi positions are determined as the center line of the pulse texture skeleton. Extract the main wave texture area, diphtheria wave texture area and descending isthmus texture area along the center line of the vein skeleton to form the vein texture region; The vein texture area is unfolded in polar coordinates according to the skeleton direction to obtain the vein unfolding diagram; The gray-level co-occurrence relationship in the vein pattern unfolding is statistically analyzed and the directional change is extracted to generate gray-level co-occurrence features and directional gradient features; Using the positional correspondence of the center line of the vein skeleton as an index, gray-level co-occurrence features and directional gradient features are mapped to the same vein coordinates to obtain vein image features.

8. The method for TCM pulse diagnosis based on image enhancement as described in claim 7, characterized in that: The formation of the characteristic group of the cun-guan-chi pulse diagram specifically includes: Identify the endpoints of the main wave skeleton, the ridges of the main wave peak, the inflection points of the diphtheria wave, and the ridges of the descending valley from the features of the pulse pattern image to form a set of pulse pattern nodes; The skeleton connection segments within the same cycle are defined as the periodic vein connection set, and the vein nodes corresponding to the positions in adjacent pressure layers are defined as the interlayer vein connection set. A vein topology map is constructed based on horizontal temporal sequence, vertical pressure stratification, periodic vein connection set, and interlayer vein connection set. Using the set of vein nodes as matrix indices, the set of periodic vein connections and the set of interlayer vein connections are transformed into a vein connection matrix, and the vein connection matrix is ​​decomposed into a graph Laplace decomposition to obtain a graph Laplace spectrum sequence. By associating the pulse topology map, the Laplacian spectrum sequence, the pulse pressure, and the pulse image features, a pulse feature group of cun, guan, and chi is formed.

9. The method for TCM pulse diagnosis based on image enhancement as described in claim 8, characterized in that: The generated TCM pulse diagnosis analysis results specifically include: The pulse characteristics of the Cun, Guan, and Chi pulse diagrams are matched according to the Cun, Guan, and Chi regions to obtain the correspondence between the three pulse diagrams. By comparing the vein sampling pressure, vein pattern topology map and graph Laplacian spectrum sequence in the correspondence of the three vein maps, the differences in vein sampling pressure, vein pattern topology and pressure layer continuity are obtained. Based on the correspondence of the three pulse diagrams, pulse image features are identified by pulse pressure difference features, pulse pattern topology difference features, pressure layer continuity difference features, gray-scale co-occurrence features, and directional gradient features, resulting in pulse position, pulse force, pulse rate, tension, and fluency. The pulse position, pulse strength, pulse rate, tension, and fluency are matched with the corresponding viscera and meridians of the cun, guan, and chi positions to generate TCM pulse diagnosis analysis results.

10. A TCM pulse diagnosis and analysis system based on image enhancement, based on the TCM pulse diagnosis and analysis method based on image enhancement as described in any one of claims 1 to 9, characterized in that: include, The pulse image construction module simultaneously collects static pressure data and dynamic pulse wave data at the cun, guan, and chi positions, and converts the pulse response intensity, collection time, and pulse taking pressure into grayscale distribution, horizontal time sequence, and vertical pressure layering, respectively, to form a grayscale image of the cun, guan, and chi pulse. The image enhancement module performs grayscale background flattening on the grayscale image of the Cun, Guan, and Chi pulses, and extracts the main wave edge, diphtheria wave details, and descending isthmus texture through differential multi-scale morphological top cap enhancement to synthesize the enhanced Cun, Guan, and Chi pulse image. The pulse map determination module registers the enhanced pulse map of Cun, Guan and Chi with the static pressure data. Through phase correlation registration, morphological skeleton extraction and connected domain labeling, the main wave skeleton, diabetic wave skeleton, descending isthmus texture skeleton and corresponding pulse pressure are determined to form the enhanced pulse map of Cun, Guan and Chi. The feature encoding module expands the polar coordinates of the pulse skeleton in the enhanced image of Cun-Guan-Chi pulse taking, extracts gray-level co-occurrence features and directional gradient features, and constructs a pulse topology map and graph Laplacian spectrum sequence through pressure layer pulse topology encoding. The graph Laplacian spectrum sequence is associated with the pulse taking pressure, gray-level co-occurrence features and directional gradient features to form a Cun-Guan-Chi pulse map feature group. The diagnostic analysis module compares the differences in pulse pressure, pulse pattern topology, and Laplace spectral sequence in the characteristic groups of the Cun, Guan, and Chi pulse diagrams to identify pulse position, pulse strength, pulse rate, tension, and fluency, and generates TCM pulse diagnosis analysis results.

Images