A method for intelligent identification of acupuncture point reaction zones using infrared thermal imaging

Abstract

This invention relates to the field of image analysis technology, specifically to an intelligent identification method for acupuncture point reaction zones combined with infrared thermal imaging. The method acquires time-series infrared thermal images of the target area and acupuncture reference points; based on the spatial gradient and temporal temperature variation characteristics of pixels, a diffusion feature matrix is ​​constructed to accurately quantify heat transfer resistance; the diffusion feature matrix is ​​mapped to a heat transfer cost network, the cumulative transfer cost is calculated, and the topological connection relationship of the minimum cost path tree is determined; based on the topological connection relationship, path backtracking statistics are performed to generate a path overlap feature field; utilizing the difference between the convergence characteristics of meridian heat transfer and the radiation characteristics of traumatic heat diffusion, combined with the cumulative transfer cost, the heat transfer skeleton line is accurately extracted; based on the skeleton line, a dynamic active state signal reflecting the patency of the meridians is obtained. This invention effectively removes interference from traumatic thermal spots, achieving accurate identification and quantitative evaluation of meridian sensing trajectories.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, specifically to a method for intelligent recognition of acupuncture point reaction zones combined with infrared thermal imaging. Background Technology

[0002] Acupuncture, as an important component of traditional Chinese medicine, has therapeutic effects closely related to the transmission of sensations along the meridians. In recent years, infrared thermal imaging technology, due to its non-contact, non-destructive, and visual characteristics, has been widely used in clinical research and efficacy evaluation of acupuncture, aiming to objectively display the phenomenon of sensation transmission along the meridians by monitoring changes in body surface temperature.

[0003] Existing infrared thermography analysis methods are mostly based on single temperature threshold segmentation or region growing algorithms to qualitatively observe meridian sensing areas. However, in practical applications, acupuncture can cause local mechanical damage and microvascular dilation around the needle insertion point, forming a traumatic hot spot centered on the needle puncture. This appears as a high-intensity circular high-temperature area in infrared thermography, making the temperature values ​​of the traumatic hot spot and the meridian sensing area extremely similar. Existing infrared thermography analysis methods struggle to effectively remove this strong background interference, easily misinterpreting traumatic heat diffusion as a meridian reaction, or failing to identify weak distal meridian transmission trajectories due to the masking effect of the traumatic hot spot. This makes it difficult to objectively quantify the patency of meridian sensing, limiting the standardization and intelligentization of acupuncture efficacy evaluation. Summary of the Invention

[0004] To address the technical problem that existing infrared thermography analysis methods cannot accurately identify meridian transmission trajectories under strong traumatic thermal spot interference, thus preventing the objective quantification of meridian transmission patency, this invention aims to provide an intelligent identification method for acupuncture point reaction zones combined with infrared thermography. The specific technical solution adopted is as follows: This invention provides a method for intelligent identification of acupuncture point reaction zones using infrared thermal imaging, the method comprising the following steps: Acquire time-series infrared thermal images and needle-puncture reference points of the target area; Based on the spatial temperature gradient features and temporal temperature variation features of each pixel in the infrared thermal image, the diffusion feature value of each pixel is obtained, and then a diffusion feature matrix is ​​constructed. The diffusion feature matrix is ​​mapped to a heat transport cost network to obtain the cumulative transport cost from the needle-punching reference point to each pixel and to determine the topological connection relationship of each pixel in the minimum cost path tree. Based on the topological connection relationship, path backtracking statistics are performed on each pixel to generate a path overlap feature field. The heat transport skeleton line with high transport significance is extracted by combining the cumulative transport cost. Based on the heat transfer skeleton line, the dynamic active status signal of the heat transfer channel is obtained.

[0005] Furthermore, the method for obtaining the diffusion feature value is as follows: For any pixel in the infrared thermal image at the current moment, extract the temperature sequence of that pixel within the preceding sliding time window at the current moment; The slope of the straight line fitted to the element in the temperature sequence is used as the temperature time-domain change analysis value of that pixel. The Sobel operator is used to obtain the temperature gradient of the pixel in the horizontal and vertical directions, and the magnitude of the temperature gradient is used as the temperature spatial gradient value of the pixel. When the temperature time-domain change analysis value is greater than or equal to the preset thermal activity threshold, the ratio of the temperature spatial gradient value to the sum of the temperature time-domain change analysis value and the preset smoothing factor is used as the diffusion feature value of that pixel. When the temperature time-domain change analysis value is less than the preset thermal activity threshold, the preset maximum penalty value is used as the diffusion feature value of the pixel.

[0006] Furthermore, the method for mapping the diffusion feature matrix to a heat transport cost network is as follows: The preset proportional quantile values ​​of all diffusion feature values ​​in the diffusion feature matrix are used as the boundary reference values; The diffusion feature values ​​in the diffusion feature matrix that are greater than the boundary reference value are truncated to the boundary reference value, and the truncated diffusion feature matrix is ​​normalized to the preset numerical range through linear transformation. In the normalized diffusion feature matrix, the product of the mean of the normalized diffusion feature values ​​of two adjacent pixels and their Euclidean distance is used as the weight of the connection edge between the two adjacent pixels, thereby constructing the heat transport cost network.

[0007] Furthermore, the method for obtaining the cumulative transmission cost is as follows: For any pixel in the hot transport cost network, the optimal path from the needle reference point to the pixel is obtained by the shortest path search algorithm. The sum of the weights of the connecting edges traversed by the optimal path is used as the cumulative transport cost from the needle reference point to the pixel.

[0008] Furthermore, the method for obtaining the topological connection relationship is as follows: The direct predecessor pixel that minimizes the cumulative transmission cost of this pixel will be used as the predecessor related pixel of this pixel. The index mapping relationship between the pixel and its predecessor related pixels is used as the topological connection relationship of the pixel in the minimum cost path tree.

[0009] Furthermore, the method for obtaining the path overlap feature field is as follows: The pixels in the hot transport cost network are arranged in descending order of cumulative transport cost to obtain the reference sequence; Initialize a counting matrix of the same size as the infrared thermal image, and set the initial count value of all pixels to 1; Traverse each pixel in the reference sequence from left to right, find the predecessor related pixels of each pixel according to the topological connection relationship, and add the count value of each pixel to the count value of its predecessor related pixels until the traversal is completed. A logarithmic transformation is performed on the final counting matrix to obtain the path overlap of each pixel in the counting matrix, and a path overlap feature field is generated; wherein, the path overlap of pixels within a preset shielding radius from the needle puncture reference point is forcibly set to 0.

[0010] Furthermore, the method for obtaining the heat transfer skeleton wire is as follows: The ratio of the path overlap of each pixel to the sum of the cumulative transmission cost and a preset minimum positive number is used as the unit resistance overlap density of each pixel. The product of the unit resistance overlap density and the theoretical reference cost of each pixel is used as the heat transfer significance value of each pixel, thereby generating a heat transfer significance map; the theoretical reference cost is the product of the Euclidean distance from the corresponding pixel to the needle reference point and the normalized average diffusion characteristic value of the whole map. Adaptive threshold segmentation is performed on the heat transfer saliency map to obtain the foreground connected regions; A morphological thinning algorithm is applied to the foreground connected region to iteratively peel off edge pixels until the foreground connected region is shrunk to a connected curve with a width of one pixel, which serves as a heat transfer skeleton line.

[0011] Furthermore, the method for obtaining the dynamic active state signal is as follows: For any pixel on the heat transfer skeleton line, the length of the trajectory line between the pixel and the needle reference point is taken as the geodesic distance between the pixel and the needle reference point. The pixel on the heat transfer skeleton line corresponding to the maximum geodesic distance is taken as the farthest endpoint of the skeleton. The sum of the heat transfer saliency values ​​of all pixels on the heat transfer skeleton line is taken as the skeleton saliency sum. The product of the sum of skeleton saliency and the length of the heat transport skeleton line is taken as the generalized transport efficiency product; The ratio of the generalized transmission efficiency product to the sum of the cumulative transmission cost at the farthest point of the skeleton and a preset minimum positive number is used as the heat transmission efficiency index. Based on the change in the heat transfer efficiency index at the current moment compared to the heat transfer efficiency index at historical moments, obtain the heat transfer efficiency change analysis value; When the heat transfer efficiency change analysis value is greater than the first preset threshold, the dynamic active state signal of the heat transfer channel is an excited state signal. When the heat transfer efficiency change analysis value is less than or equal to the first preset threshold and greater than the second preset threshold, the dynamic active state signal of the heat transfer channel is a saturated state signal. When the heat transfer efficiency change analysis value is less than or equal to the second preset threshold, the dynamic active state signal of the heat transfer channel is a decaying state signal.

[0012] Furthermore, the method for acquiring the time-series infrared thermal image is as follows: The original infrared thermal image sequence of the target area is acquired at a preset sampling frequency, and the first frame is used as the reference frame; Rigid registration based on feature points is performed on each frame of the original infrared thermal image sequence, except for the reference frame, to align it with the reference frame in spatial coordinates, thereby generating a time-series infrared thermal image.

[0013] Furthermore, the method for obtaining the acupuncture reference point is as follows: When the acquisition time exceeds the preset time threshold, the average infrared thermal image of all infrared thermal images within the preset time window is obtained. The location of the pixel with the highest gray value in the average infrared thermal image is used as the reference point for needle puncture.

[0014] The present invention has the following beneficial effects: This invention first obtains the diffusion feature value of each pixel based on the spatial gradient and temporal variation characteristics of temperature in infrared thermography, accurately reflecting the thermal transport resistance characteristics of each pixel. This allows for the construction of a diffusion feature matrix, which is helpful in initially distinguishing between the low-resistance characteristics of meridian channels and the high-resistance characteristics of traumatic hot spots. To transform the physical feature values ​​into edge weights required for graph theory analysis, the diffusion feature matrix is ​​mapped to a thermal transport cost network. This facilitates the unification of dimensions and the establishment of connectivity between pixels, thereby obtaining the cumulative transport cost from the acupuncture reference point to each pixel and determining the topological connection relationship of each pixel in the minimum cost path tree. This is beneficial for quantifying the physical resistance of thermal signal transmission. The algorithm constructs a backtracking path chain for the entire graph; then, based on the topological connection relationship, it performs path backtracking statistics on each pixel to generate a path overlap feature field. This is beneficial for feature separation by utilizing the difference between the convergence characteristics of meridian heat transmission and the radiation characteristics of traumatic heat diffusion, eliminating artifacts, and extracting heat transmission skeleton lines with high transmission significance by combining cumulative transmission cost. This accurately determines the physical path and morphology of meridian sensation transmission, which is conducive to achieving precise positioning of meridian trajectories. Furthermore, based on the heat transmission skeleton lines, it obtains the dynamic active state signal of the heat transmission channel, accurately reflecting the physical evolution process of the meridian from excitation, saturation to decay. This is beneficial for quantitatively evaluating the patency of meridian sensation transmission and providing an objective basis for evaluating the efficacy of clinical acupuncture. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0016] Figure 1 A schematic flowchart illustrating an intelligent identification method for acupuncture point reaction zones combined with infrared thermal imaging, provided in one embodiment of the present invention. Figure 2 This is a structural diagram of an intelligent recognition system for acupuncture point reaction areas combined with infrared thermal imaging, provided in one embodiment of the present invention. Figure 3 This is a schematic diagram of a computer device provided according to an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method for intelligent identification of acupuncture point reaction areas combined with infrared thermal imaging proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following describes in detail, with reference to the accompanying drawings, a specific scheme of the intelligent identification method for acupuncture point reaction areas combined with infrared thermal imaging provided by the present invention.

[0020] Example 1: This invention proposes an intelligent identification method for acupuncture point reaction zones combined with infrared thermal imaging. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates a schematic flowchart of an intelligent identification method for acupuncture point reaction areas combined with infrared thermal imaging, according to an embodiment of the present invention. The method includes the following steps: Step S1: Obtain time-series infrared thermal images and needle puncture reference points of the target area.

[0021] Specifically, this embodiment uses a single subject for analysis, and all subsequent analyses will be based on this subject. Similarly, this embodiment uses a single acupuncture point for analysis, and all subsequent analyses will refer to that acupuncture point. To reduce interference from ambient noise, this embodiment uses the subject's acupuncture site and its surrounding skin area as the target region, ensuring that thermal features are extracted only from the effective human skin surface. To ensure that the rigid registration algorithm effectively eliminates motion artifacts, complex non-rigid deformations caused by breathing or muscle contraction must be avoided. Therefore, this embodiment requires the subject to remain still and maintain stable breathing during the data acquisition process to ensure that skin deformation in the target region primarily manifests as rigid displacement.

[0022] Since the acupuncture monitoring process typically lasts from several minutes to tens of minutes, the subject's breathing or unconscious micro-movements can cause displacement of the human body contour in the infrared thermal image. Without correction, subsequent time-series analysis of the same coordinate point will exhibit severe deviations (i.e., motion artifacts). Therefore, the system first performs image registration. Specifically, in this embodiment, the system first acquires the original infrared thermal image sequence of the target area using an infrared thermal imager at a preset sampling frequency. The first frame at the start of acquisition is used as the reference frame. Each frame of the original infrared thermal image sequence, excluding the reference frame, is processed using a feature-point-based image registration algorithm: first, corner features are extracted from the reference frame and each other frame; then, feature point matching is performed; the homography matrix between the two frames is calculated; finally, the homography matrix is ​​used to perform geometric transformations (translation and rotation corrections) on the corresponding frames to ensure strict alignment with the reference frame in spatial coordinates, thereby generating a time-series infrared thermal image. In this embodiment, the preset sampling frequency is set to 10Hz. Implementers can set the preset sampling frequency according to actual conditions; this is not limited here. It should be noted that for non-rigid deformation areas such as the abdomen that may experience significant respiratory fluctuations, as an alternative, this step can also employ non-rigid registration algorithms based on optical flow or free-form deformation to further eliminate local deformation interference. This embodiment uses rigid registration as an example for simplification. The time-series infrared thermal images after registration ensure that the same coordinates always correspond to the same physical location on the target area surface. The image registration algorithm, optical flow method, and non-rigid registration algorithm based on free-form deformation are all well-known technologies and will not be elaborated further.

[0023] It is known that the core algorithm of this embodiment relies on the thermal steady-state characteristics during acupuncture needle retention. At the moment of needle insertion and in the initial stage (e.g., the first 1-2 minutes), the temperature at the needle puncture site fluctuates drastically due to acute trauma, which does not conform to the algorithm's assumptions. Therefore, it is necessary to define an applicable time window and lock the calculation benchmark during this stage. This embodiment sets a preset time threshold of 2 minutes to ensure the system avoids the unstable fluctuation stage in the initial needle insertion phase. The implementer can set the size of the preset time threshold according to the acupuncture technique or the subject's physical condition; this is not limited here. When the acquisition time exceeds the preset time threshold, it indicates that the thermal response has entered a relatively stable observation period. To obtain a stable temperature benchmark, the average infrared thermal image of all infrared thermal images within the preset time window is obtained, ensuring the elimination of random noise in single-frame images, which is beneficial for improving the accuracy of benchmark positioning. This embodiment sets the preset time window to 10 seconds to ensure sufficient sample size without excessive computational resource consumption. The implementer can set the size of the preset time window according to the sampling frequency; this is not limited here, but the initial time of the preset time window is the end time corresponding to the preset time threshold.

[0024] Furthermore, to shield against environmental background noise, calculations are performed only on the human skin region. The system generates a mask based on the aforementioned average infrared thermal image. The Otsu's method is used to automatically calculate the segmentation threshold, binarizing the image into foreground (human body) and background. The largest connected component in the foreground is retained, and a morphological closing operation (dilation followed by erosion) is performed on it to fill internal holes, ultimately obtaining an effective skin region mask. Pixels within the mask have a value of 1, while pixels in the background region have a value of 0. All subsequent calculations are performed only within regions with a pixel value of 1. The Otsu's method, threshold segmentation algorithm, and morphological closing operation are all well-known techniques and will not be elaborated further.

[0025] Considering that the acupuncture point is usually the center of the highest local heat source, the location of the pixel with the largest gray value within the effective skin area mask in the average infrared thermal image is taken as the acupuncture reference point, i.e., the acupuncture position. It should be noted that once the acupuncture reference point is determined, its coordinates remain unchanged in all subsequent processing steps, serving as the unique source point for graph theory analysis. Furthermore, if there are at least two pixels with the largest gray value in the average infrared thermal image, the geometric center of the region formed by all pixels with the largest gray values ​​is taken as the acupuncture reference point. In another embodiment, a pixel can be randomly selected from all pixels with the largest gray values ​​as the acupuncture reference point.

[0026] Step S2: Based on the spatial temperature gradient features and temporal temperature change features of each pixel in the infrared thermal image, obtain the diffusion feature value of each pixel, and then construct the diffusion feature matrix.

[0027] Specifically, compared to the continuous dynamic heating characteristics of the meridian sensing area, the temperature change rate of the traumatic hot spot is relatively low after entering the steady state, and its spatial gradient is significantly higher due to its point-like strong heat source. This algorithm utilizes this relative spatiotemporal characteristic difference for signal separation, which is significant in the temperature time-domain variation characteristics. In addition, the traumatic hot spot usually manifests as a local high-temperature island with a large edge gradient, while the meridian channel manifests as a smooth temperature rise band with a relatively small gradient, which is significant in the temperature spatial gradient characteristics. Therefore, this embodiment obtains the diffusion feature value of each pixel based on the spatial gradient and temporal variation features of temperature in the infrared thermal image, accurately reflecting the heat transfer resistance characteristics of each pixel. The larger the diffusion feature value, the more the corresponding pixel tends to have the characteristics of a traumatic hot spot or a static heat source, exhibiting high resistance to heat transfer; the smaller the diffusion feature value, the more the corresponding pixel tends to have the characteristics of a meridian channel, exhibiting low resistance to heat transfer. In order to comprehensively quantify the heat transfer characteristics of the entire target area and construct a diffusion feature matrix, it is beneficial to accurately separate the meridian trajectory and the traumatic hot spot using graph theory algorithms.

[0028] Preferably, in one feasible embodiment, the diffusion feature value is obtained as follows: for any pixel in the infrared thermal image at the current moment, the temperature sequence of that pixel is extracted within the preceding sliding time window of the current moment (arranged in chronological order); wherein, the end time of the preceding sliding time window of the current moment is the current moment. In this embodiment, the sliding time window is set to contain 100 frames of infrared thermal images (corresponding to a duration of 10 seconds) to ensure that the time window is long enough to smooth the random noise of the infrared thermal imager sensor, while being able to capture the second-level slow-changing characteristics of the thermal response of the human meridian. The implementer can set the size of the sliding time window according to the sampling frequency and the noise level of the thermal imager, which is not limited here. The elements in the temperature sequence are linearly fitted using the least squares method, and the slope of the resulting straight line is used as the temperature time-domain change analysis value of the pixel, accurately reflecting the degree of thermal activity of the pixel in the current stage (i.e., the rate of heating or cooling); wherein, if the slope is negative (i.e., cooling), it will be treated as an inactive area because it is less than the subsequently set positive threshold. The Sobel operator is used to obtain the temperature gradient of the pixel in the horizontal and vertical directions. Then, the magnitude of the vector formed by the two temperature gradients is calculated as the temperature spatial gradient value of the pixel, which accurately reflects the steepness of the temperature distribution around the pixel (i.e., whether it is at the edge of a heat source). To avoid erroneous low-resistance paths (i.e., short-circuiting) in non-thermally active areas (i.e., background areas or non-acupoint skin) due to near-zero temperature change rates, a threshold judgment is required. When the temperature time-domain change analysis value is greater than or equal to a preset thermal activity threshold, it indicates that the pixel is in a thermally active state. To construct differentiated thermal transmission resistance characteristics, the ratio of the temperature spatial gradient value to the sum of the temperature time-domain change analysis value and a preset smoothing factor is used as the diffusion characteristic value of the pixel. In this embodiment, the preset smoothing factor is set to 0.05 (unit consistent with the temperature time-domain change analysis value) to ensure that when the temperature time-domain change analysis value is small... The smoothing factor does not approach zero, suppressing noise amplification while maintaining numerical stability. The implementer can set the preset smoothing factor based on the thermal imager's temperature sensitivity or the median of the overall image change rate; this is not limited here. When the temperature time-domain change analysis value is less than the preset thermal activity threshold, it indicates that the pixel is in a non-thermally active state (background or extremely stable region). To prevent the shortest path search algorithm from passing through this invalid region, a preset maximum penalty value (e.g., 1000, which the implementer can set according to actual conditions, is used as the pixel's diffusion characteristic value, forcing it to be set as a high-resistivity region, compelling the path to detour to the thermally active region. In this embodiment, the preset thermal activity threshold is set to 0.01℃ / s to ensure that feature calculations are performed only on regions with actual physiological thermal responses. The implementer can set the preset thermal activity threshold based on the ambient temperature fluctuation level or sensor noise floor; this is not limited here. The least squares method for fitting straight lines, the Sobel operator, and the shortest path search algorithm are all well-known and will not be elaborated further.

[0029] By iterating through all pixels using the method for obtaining diffusion feature values, a complete diffusion feature matrix can be constructed.

[0030] Step S3: Map the diffusion feature matrix to a heat transport cost network, obtain the cumulative transport cost from the needle-piercing reference point to each pixel, and determine the topological connection relationship of each pixel in the minimum cost path tree; perform path backtracking statistics on each pixel based on the topological connection relationship, generate a path overlap feature field, and extract the heat transport skeleton line with high transport significance by combining the cumulative transport cost.

[0031] Specifically, in order to transform the physical feature values ​​into edge weights required for graph theory analysis and make the diffusion features of different pixels comparable, this embodiment maps the diffusion feature matrix into a heat transport cost network. This facilitates the unification of dimensions and the establishment of connectivity between pixels. In the heat transport cost network, the cumulative transport cost from the needle-puncture reference point to each pixel is obtained, accurately reflecting the total physical resistance overcome by the heat signal from the needle-puncture point to the location of each pixel. This is beneficial for the initial identification of low-resistance transport areas. At the same time, the topological connection relationship of each pixel in the minimum cost path tree is determined, accurately reflecting the optimal precursor path of heat flow transport. This is beneficial for subsequent path backtracking analysis and the construction of the heat transport topology of the entire graph.

[0032] To leverage the topological differences between the converging characteristics of meridian heat transmission and the radiative characteristics of traumatic heat diffusion for feature separation, ensuring the extraction of weak meridian signals even under strong traumatic heat spot interference, path backtracking statistics are performed on each pixel based on topological connectivity to generate a path overlap feature field. This accurately reflects the importance of each pixel as a main heat transmission channel (i.e., the number of optimal paths through that point), highlighting the skeletal features of the meridian channels. To integrate the dual features of low transmission resistance and high topological connectivity, artifacts are eliminated, and then high-significance heat transmission skeleton lines are extracted by combining cumulative transmission costs. This achieves precise localization and morphological description of the meridian trajectory, facilitating subsequent quantitative evaluation of the meridian transmission status.

[0033] Preferably, in one feasible method of this embodiment, the method of mapping the diffusion feature matrix to the heat transport cost network is as follows: Considering that the range of diffusion feature values ​​in the diffusion feature matrix may have a large dynamic range due to fluctuations in ambient temperature, directly using diffusion feature values ​​to obtain edge weights in the heat transport cost network will lead to instability of the shortest path search algorithm. Therefore, in this embodiment, the preset proportional quantile value of all diffusion feature values ​​in the diffusion feature matrix is ​​used as the boundary reference value. In this embodiment, the preset proportional quantile value is set to the 95th quantile value (that is, the value located at the 95th position after arranging all values ​​from smallest to largest), to ensure that the range of most normal diffusion feature values ​​is covered. The implementer can set the preset proportional quantile value according to the dispersion of the data or the noise ratio, which is not limited here. It should be noted that because the distribution of diffusion feature values ​​usually exhibits a long-tail feature, a very small number of noise points or extreme trauma points will produce values ​​far exceeding the normal range (outliers). If the maximum value is used directly for normalization, most effective feature values ​​will be compressed to a very small value range, thereby losing feature resolution. Using the preset proportional quantile value for truncation can effectively remove outlier interference and retain the distribution characteristics of the main data. Furthermore, diffusion feature values ​​in the diffusion feature matrix that are greater than the boundary reference value are truncated to the boundary reference value to ensure the elimination of the influence of extreme outliers on the overall data distribution and to prevent the compression of effective features during subsequent normalization. To unify the feature dimensions of different images and make the edge weights comparable across samples, a linear transformation is used to normalize the truncated diffusion feature matrix to a preset numerical range, ensuring that all feature values ​​fall within the standardized calculation range, which is beneficial for the convergence of the subsequent path search algorithm. The normalization process uses the following formula: In the formula, D is the normalized diffusion characteristic value; B is the original diffusion characteristic value; min is the minimum value function; M is the maximum value of a preset numerical range; in this embodiment, the preset numerical range is set to... M is 100, ensuring that the numerical accuracy is sufficient to distinguish subtle differences in resistance. Implementers can set a preset numerical range according to the calculation accuracy requirements, which is not limited here. To conform to the laws of geometry and physics—that is, the longer the transmission distance, the greater the physical resistance that heat conduction must overcome—the product of the mean of the normalized diffusion feature values ​​of two adjacent pixels and their Euclidean distance is used as the weight of the connection edge between the two pixels in the normalized diffusion feature matrix. This accurately reflects the comprehensive cost required for the heat signal to conduct between the two points. A larger connection edge weight indicates high resistance in the path segment, making it difficult for the heat signal to pass (preferably a wound or background); a smaller connection edge weight indicates low resistance in the path segment, making it easier for the heat signal to pass (preferably a meridian). To construct the heat transmission topology of the entire graph for use by the shortest path algorithm, a heat transmission cost network is constructed. The methods for obtaining the linear transformation and Euclidean distance are well-known techniques and will not be elaborated further.

[0034] Preferably, in one feasible embodiment of this invention, the method for obtaining the cumulative transmission cost is as follows: For any pixel in the heat transmission cost network, the optimal path from the needle-piercing reference point to the pixel is obtained through a shortest path search algorithm (e.g., Dijkstra's algorithm or A* algorithm; Dijkstra's algorithm and A* algorithm are well-known technologies and will not be described in detail here). This accurately determines the physical transmission route that minimizes the total resistance required for the heat signal to reach that location. Then, the sum of the weights of the connecting edges along the optimal path is used as the cumulative transmission cost from the needle-piercing reference point to the pixel. A larger cumulative transmission cost indicates that the pixel is farther from the effective transmission distance of the heat source, or is located in a high-barrier area where the thermal effect is difficult to reach.

[0035] Preferably, in one feasible implementation of this embodiment, the method for obtaining the topological connection relationship is as follows: the direct predecessor pixel (i.e., parent node) that minimizes the cumulative transmission cost of the pixel is taken as the predecessor related pixel of the pixel; the index mapping relationship between the pixel and its predecessor related pixel is taken as the topological connection relationship of the pixel in the minimum cost path tree. This relationship accurately records the unique optimal path chain backtracking from any position in the entire graph to the needle piercing reference point.

[0036] Preferably, in one feasible embodiment, the method for obtaining the path overlap feature field is as follows: In order to ensure that the path backtracking statistics strictly follow the flow direction from the far end (leaf node) to the near end (root node) and to ensure the connectivity and uniqueness of the backtracking path, this embodiment first arranges the pixels in the heat transfer cost network in descending order of cumulative transfer cost to obtain a reference sequence and accurately determines the traversal priority of the backtracking calculation; then, it initializes a counting matrix of the same size as the infrared thermal image, sets the initial count value of all pixels to 1, representing that each pixel generates one unit of backtracking flow, providing the basic flow for subsequent cumulative calculation; Then, each pixel in the reference sequence is traversed from left to right. Based on the topological connection relationship, the predecessor related pixels of each pixel are found. The count value of each pixel is accumulated into the count value of its predecessor related pixels to simulate the heat flow convergence process, that is, the parent node carries the flow of all the child nodes passing through it, until the traversal is completed, and the total number of times each pixel in the counting matrix is ​​traversed by the optimal path is determined. When a pixel is traversed by the optimal path more times, it means that the pixel is more likely to be on the main channel of heat transmission (i.e., meridian channel). To compress the dynamic range of the counting results (the count values ​​may grow exponentially) and improve the visual contrast of image features, a logarithmic transformation is performed on the final counting matrix, i.e., the calculation... Where m represents the count value in the counting matrix, e is the natural constant, and ln is the logarithmic function with the natural constant as the base. The path overlap of each pixel in the counting matrix is ​​then obtained, generating a path overlap feature field, which is beneficial for subsequent feature fusion calculations. The path overlap of pixels within a preset shielding radius from the acupuncture reference point is forcibly set to 0 to ensure the elimination of high-density convergence artifacts near the source point caused by geometric topological inevitability, highlighting the features of the distal meridian. In this embodiment, the preset shielding radius is set to 10 pixels to ensure coverage of the core high-temperature area of ​​the needle hole wound. The implementer can set the size of the preset shielding radius according to the size of the needle hole or the thermal image resolution; this is not limited here.

[0037] Preferably, in one feasible embodiment of this invention, the method for obtaining the heat transfer backbone line is as follows: In order to comprehensively measure the topological importance and transmission cost of each pixel, the ratio of the path overlap of each pixel to the sum of the cumulative transmission cost and a preset minimum positive number is used as the unit resistance overlap density of each pixel to quantify the path convergence degree under unit transmission resistance. The larger the unit resistance overlap density, the higher the path flow (i.e., the high-efficiency trunk road) the corresponding pixel carries under lower transmission resistance. In this embodiment, the preset minimum positive number is set to 0.001 to ensure that the denominator is not zero. The implementer can set the size of the preset minimum positive number according to the calculation accuracy, which is not limited here. To eliminate the attenuation effect of distance on saliency (generally, the greater the distance, the greater the cumulative cost), and to enhance the weight of weak meridian signals at the distal end, the product of the unit resistance overlap density and the theoretical reference cost of each pixel is used as the thermal transmission saliency value of each pixel, accurately reflecting the absolute transmission saliency after normalizing the distance effect. The larger the thermal transmission saliency value, the higher the probability that the corresponding pixel belongs to a strong meridian reaction zone. The theoretical reference cost is the product of the Euclidean distance from the corresponding pixel to the acupuncture reference point and the normalized average diffusion characteristic value of the whole image, representing the theoretical resistance expectation value for transmitting the same distance in a uniform medium. To construct a visualized meridian energy distribution map and generate a thermal transmission saliency map, which is beneficial for subsequent morphological segmentation, the weak signals at the distal end and the strong signals at the proximal end are comparable. To extract highly saliency meridian regions from the background and accurately identify the effective areas of meridian sensing, an adaptive threshold segmentation method using Otsu's (maximum inter-class variance) threshold segmentation is employed to obtain the foreground connected region, i.e., the preliminary meridian reaction zone. To achieve precise quantification of the meridian sensing "range" and address the technical challenge of defining the transmission length of planar hot spots, a morphological thinning algorithm (such as the Zhang-Suen thinning algorithm) is applied to the foreground connected region. This iteratively removes edge pixels until the foreground connected region is shrunk to a single-pixel-width connected curve, serving as the heat transmission skeleton line. This extracts the main path topology of heat transmission, achieving a linear description of the meridian trajectory, which is beneficial for subsequent geometric measurement and quantitative analysis. Both the Otsu threshold segmentation method and the Zhang-Suen thinning algorithm are well-known techniques and will not be elaborated further.

[0038] Step S4: Based on the heat transfer skeleton line, obtain the dynamic active status signal of the heat transfer channel.

[0039] Specifically, in order to quantitatively assess the smoothness of meridian sensation transmission and its evolution over time, and to achieve objective real-time monitoring of acupuncture effects, this study aims to obtain dynamic activity signals of heat transmission channels based on heat transmission skeleton lines, accurately reflect the physical processes of meridians in real time, and provide a scientific basis for deciding the needle retention time in clinical acupuncture.

[0040] Preferably, in one feasible embodiment of this invention, the method for obtaining the dynamic active state signal is as follows: for any pixel on the heat transmission skeleton line, the length of the trajectory line between the pixel and the acupuncture reference point is taken as the geodesic distance between the pixel and the acupuncture reference point, which accurately reflects the physical length of the actual transmission of the heat signal along the curved path of the meridian; the larger the geodesic distance, the wider the sensing range; considering that the range of meridian sensing is an important indicator for evaluating the therapeutic effect, the pixel on the heat transmission skeleton line corresponding to the largest geodesic distance is taken as the farthest endpoint of the skeleton, representing the limit position that the heat signal can effectively reach at the current moment; To comprehensively evaluate the overall thermal response intensity of the meridian channels, the sum of the thermal transmission saliency values ​​of all pixels on the thermal transmission skeleton line is obtained as the skeleton saliency sum, accurately reflecting the total thermal energy of the entire channel. The product of the skeleton saliency sum and the length of the thermal transmission skeleton line is used as the generalized transmission efficiency product, accurately reflecting the intensity-distance comprehensive efficiency of meridian transmission (i.e., the farther and stronger the transmission, the higher the efficiency). The ratio of the generalized transmission efficiency product to the cumulative transmission cost at the farthest endpoint of the skeleton and a preset minimum positive number is used as the thermal transmission efficiency index, accurately reflecting the transmission efficiency obtained under unit transmission resistance. The larger the thermal transmission efficiency index, the higher the meridian patency and the more active the Qi and blood circulation. The preset minimum positive number is 0.001 to prevent division by zero errors.

[0041] Considering that the absolute value at a single moment is greatly affected by individual differences, focusing on its dynamic change trend is more clinically significant. In order to capture the evolution of the meridian state, and then obtain the heat transfer efficiency change analysis value based on the change of the heat transfer efficiency index at the current moment compared with the heat transfer efficiency index at historical moments; in this embodiment, the historical moment is set as the previous sampling moment of the current moment to calculate the instantaneous change rate. The implementer can set the historical moment according to the sampling frequency, which is not limited here. The larger the change value of heat transfer efficiency, the faster the meridian patency improves. To classify the meridian state, this embodiment sets the first preset threshold to 0.05 / min and the second preset threshold to -0.03 / min to ensure that obvious upward, stable, and downward trends can be distinguished. The implementer can set the size of the first and second preset thresholds based on clinical experience data, which is not limited here. When the change value of heat transfer efficiency is greater than the first preset threshold, it indicates that the meridian resistance is continuously decreasing and the sensing range is expanding. At this time, the dynamic active state signal of the heat transfer channel is an excited state signal, indicating that it is in the optimal treatment rising period. When the change value of heat transfer efficiency is less than or equal to the first preset threshold and greater than the second preset threshold, it indicates that the meridian patency is maintained at a high stable state. At this time, the dynamic active state signal of the heat transfer channel is a saturated state signal, indicating that the qi reaches the disease site and the therapeutic effect is optimal. When the change value of heat transfer efficiency is less than or equal to the second preset threshold, it indicates that the body is experiencing adaptive fatigue and the transmission resistance is rising. At this time, the dynamic active state signal of the heat transfer channel is a declining state signal, indicating that the therapeutic effect is weakening.

[0042] In summary, this embodiment acquires time-series infrared thermographic images of the target area and acupuncture reference points; based on the spatial gradient and temporal temperature variation characteristics of pixels, a diffusion feature matrix is ​​constructed to accurately quantify heat transfer resistance; the diffusion feature matrix is ​​mapped to a heat transfer cost network, the cumulative transfer cost is calculated, and the topological connection relationship of the minimum cost path tree is determined; based on the topological connection relationship, path backtracking statistics are performed to generate a path overlap feature field; utilizing the difference between the convergence characteristics of meridian heat transfer and the radiation characteristics of traumatic heat diffusion, combined with the cumulative transfer cost, the heat transfer skeleton line is accurately extracted; based on the skeleton line, a dynamic active state signal reflecting the patency of the meridians is obtained. This invention effectively removes the interference of traumatic thermal spots, achieving accurate identification and quantitative evaluation of meridian sensing trajectories.

[0043] Example 2: This invention also proposes an intelligent recognition system for acupuncture point reaction zones combined with infrared thermal imaging. Please refer to [link to relevant documentation]. Figure 2 The diagram shows a structural diagram of an intelligent recognition system for acupuncture point reaction areas combined with infrared thermal imaging, provided by an embodiment of the present invention. The system includes: an image acquisition module 10, a diffusion feature matrix acquisition module 20, a heat transfer skeleton line acquisition module 30, and a state analysis module 40.

[0044] Image acquisition module 10 is used to acquire time-series infrared thermal images of the target area and acupuncture reference points.

[0045] The diffusion feature matrix acquisition module 20 is used to acquire the diffusion feature value of each pixel based on the temperature spatial gradient feature and temperature temporal change feature of each pixel in the infrared thermal image, and then construct the diffusion feature matrix.

[0046] The thermal transport skeleton line acquisition module 30 is used to map the diffusion feature matrix into a thermal transport cost network, obtain the cumulative transport cost from the needle-punching reference point to each pixel, and determine the topological connection relationship of each pixel in the minimum cost path tree; perform path backtracking statistics on each pixel based on the topological connection relationship, generate a path overlap feature field, and extract the thermal transport skeleton line with high transport significance by combining the cumulative transport cost.

[0047] The status analysis module 40 is used to acquire the dynamic active status signal of the heat transfer channel based on the heat transfer skeleton line.

[0048] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the intelligent recognition system for acupuncture point reaction areas combined with infrared thermal imaging and the intelligent recognition method for acupuncture point reaction areas combined with infrared thermal imaging provided in the above embodiments belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0049] Example 3: This invention also proposes an intelligent recognition device for acupuncture point reaction areas combined with infrared thermal imaging. The device includes a memory and a processor. The memory stores executable program code, and the processor is used to call and execute the executable program code to perform the intelligent recognition method for acupuncture point reaction areas combined with infrared thermal imaging provided in the embodiments of this application. Specifically, the device may be a chip, component, or module. The chip may include a connected processor and memory; the memory stores instructions, and when the processor calls and executes the instructions, the chip can perform the intelligent recognition method for acupuncture point reaction areas combined with infrared thermal imaging provided in the above embodiments.

[0050] Furthermore, this application also protects a computer device; please refer to [link to relevant documentation]. Figure 3 The computer device includes a memory 401, a processor 402, and a computer program 403 stored in the memory 401 and running on the processor 402. When the processor 402 executes the computer program 403, the computer device can execute any of the aforementioned intelligent recognition methods for acupuncture point reaction areas combined with infrared thermal imaging.

[0051] Example 4: This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the above-mentioned related method steps to realize the intelligent recognition method for acupuncture point reaction areas combined with infrared thermal imaging provided in the above embodiment.

[0052] Example 5: This embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related steps to realize the intelligent recognition method for acupuncture point reaction areas combined with infrared thermal imaging provided in the above embodiment.

[0053] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0054] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0055] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for intelligent identification of acupuncture point reaction zones combined with infrared thermal imaging, characterized in that, The method includes the following steps: Acquire time-series infrared thermal images and needle-puncture reference points of the target area; Based on the spatial temperature gradient features and temporal temperature variation features of each pixel in the infrared thermal image, the diffusion feature value of each pixel is obtained, and then a diffusion feature matrix is ​​constructed. The diffusion feature matrix is ​​mapped to a heat transport cost network to obtain the cumulative transport cost from the needle-punching reference point to each pixel and to determine the topological connection relationship of each pixel in the minimum cost path tree. Based on the topological connection relationship, path backtracking statistics are performed on each pixel to generate a path overlap feature field. The heat transport skeleton line with high transport significance is extracted by combining the cumulative transport cost. Based on the heat transfer skeleton line, the dynamic active status signal of the heat transfer channel is obtained; The method for obtaining the diffusion feature value is as follows: For any pixel in the infrared thermal image at the current moment, extract the temperature sequence of that pixel within the preceding sliding time window at the current moment; The slope of the straight line fitted to the element in the temperature sequence is used as the temperature time-domain change analysis value of that pixel. The Sobel operator is used to obtain the temperature gradient of the pixel in the horizontal and vertical directions, and the magnitude of the temperature gradient is used as the temperature spatial gradient value of the pixel. When the temperature time-domain change analysis value is greater than or equal to the preset thermal activity threshold, the ratio of the temperature spatial gradient value to the sum of the temperature time-domain change analysis value and the preset smoothing factor is used as the diffusion feature value of that pixel. When the temperature time-domain change analysis value is less than the preset thermal activity threshold, the preset maximum penalty value is used as the diffusion feature value of the pixel. The method for mapping the diffusion feature matrix to a heat transport cost network is as follows: The preset proportional quantile values ​​of all diffusion feature values ​​in the diffusion feature matrix are used as the boundary reference values; The diffusion feature values ​​in the diffusion feature matrix that are greater than the boundary reference value are truncated to the boundary reference value, and the truncated diffusion feature matrix is ​​normalized to the preset numerical range through linear transformation. In the normalized diffusion feature matrix, the product of the mean of the normalized diffusion feature values ​​of two adjacent pixels and their Euclidean distance is used as the weight of the connection edge between the two adjacent pixels, thereby constructing the heat transport cost network. The method for obtaining the cumulative transmission cost is as follows: For any pixel in the hot transport cost network, the optimal path from the needle-piercing reference point to the pixel is obtained by the shortest path search algorithm. The sum of the weights of the connecting edges traversed by the optimal path is used as the cumulative transport cost from the needle-piercing reference point to the pixel. The method for obtaining the topological connection relationship is as follows: The direct predecessor pixel that minimizes the cumulative transmission cost of this pixel will be used as the predecessor related pixel of this pixel. The index mapping relationship between the pixel and its predecessor related pixels is used as the topological connection relationship of the pixel in the minimum cost path tree; The method for obtaining the path overlap feature field is as follows: The pixels in the hot transport cost network are arranged in descending order of cumulative transport cost to obtain the reference sequence; Initialize a counting matrix of the same size as the infrared thermal image, and set the initial count value of all pixels to 1; Traverse each pixel in the reference sequence from left to right, find the predecessor related pixels of each pixel according to the topological connection relationship, and add the count value of each pixel to the count value of its predecessor related pixels until the traversal is completed. Logarithmic transformation is performed on the final counting matrix to obtain the path overlap of each pixel in the counting matrix and generate a path overlap feature field; wherein, the path overlap of pixels within a preset shielding radius from the needle puncture reference point is forcibly set to 0. The method for obtaining the heat transfer skeleton wire is as follows: The ratio of the path overlap of each pixel to the sum of the cumulative transmission cost and a preset minimum positive number is used as the unit resistance overlap density of each pixel. The product of the unit resistance overlap density and the theoretical reference cost of each pixel is used as the heat transfer significance value of each pixel, thereby generating a heat transfer significance map; the theoretical reference cost is the product of the Euclidean distance from the corresponding pixel to the needle reference point and the normalized average diffusion characteristic value of the whole map. Adaptive threshold segmentation is performed on the heat transfer saliency map to obtain the foreground connected regions; A morphological thinning algorithm is applied to the foreground connected region to iteratively peel off edge pixels until the foreground connected region is shrunk to a connected curve with a width of one pixel, which serves as a heat transfer skeleton line.

2. The intelligent identification method for acupuncture point reaction areas combined with infrared thermal imaging as described in claim 1, characterized in that, The method for obtaining the dynamic active state signal is as follows: For any pixel on the heat transfer skeleton line, the length of the trajectory line between the pixel and the needle reference point is taken as the geodesic distance between the pixel and the needle reference point. The pixel on the heat transfer skeleton line corresponding to the maximum geodesic distance is taken as the farthest endpoint of the skeleton. The sum of the heat transfer saliency values ​​of all pixels on the heat transfer skeleton line is taken as the skeleton saliency sum. The product of the sum of skeleton saliency and the length of the heat transport skeleton line is taken as the generalized transport efficiency product; The ratio of the generalized transmission efficiency product to the sum of the cumulative transmission cost at the farthest point of the skeleton and a preset minimum positive number is used as the heat transmission efficiency index. Based on the change in the heat transfer efficiency index at the current moment compared to the heat transfer efficiency index at historical moments, obtain the heat transfer efficiency change analysis value; When the heat transfer efficiency change analysis value is greater than the first preset threshold, the dynamic active state signal of the heat transfer channel is an excited state signal. When the heat transfer efficiency change analysis value is less than or equal to the first preset threshold and greater than the second preset threshold, the dynamic active state signal of the heat transfer channel is a saturated state signal. When the heat transfer efficiency change analysis value is less than or equal to the second preset threshold, the dynamic active state signal of the heat transfer channel is a decaying state signal.

3. The intelligent identification method for acupuncture point reaction areas combined with infrared thermal imaging as described in claim 1, characterized in that, The method for acquiring the time-series infrared thermal images is as follows: The original infrared thermal image sequence of the target area is acquired at a preset sampling frequency, and the first frame is used as the reference frame; Rigid registration based on feature points is performed on each frame of the original infrared thermal image sequence, except for the reference frame, to align it with the reference frame in spatial coordinates, thereby generating a time-series infrared thermal image.

4. The intelligent identification method for acupuncture point reaction areas combined with infrared thermal imaging as described in claim 3, characterized in that, The method for obtaining the acupuncture reference point is as follows: When the acquisition time exceeds the preset time threshold, the average infrared thermal image of all infrared thermal images within the preset time window is obtained. The location of the pixel with the highest gray value in the average infrared thermal image is used as the reference point for needle puncture.

Images