CN-122004855-A - Automatic subcutaneous vein blood sampling method based on infrared imaging

CN122004855ACN 122004855 ACN122004855 ACN 122004855ACN-122004855-A

Abstract

The invention discloses an automatic subcutaneous vein blood sampling method based on infrared imaging, which relates to the technical field of vein blood sampling and comprises the steps of obtaining infrared thermal imaging original data of a skin area on the surface of a limb to be sampled, generating a two-dimensional temperature matrix reflecting subcutaneous tissue thermal distribution through temperature field reconstruction, extracting the temperature gradient direction and the thermal diffusion rate of vascular tissues, inputting a preset thermodynamic causal reasoning model, inverting subcutaneous vein three-dimensional space coordinates according to a heat transfer physical rule, calculating vein diameter and depth estimated values by combining temperature values, geometrically correcting the three-dimensional coordinates to obtain vein center coordinates, and mapping the vein center coordinates through a coordinate system to generate vein position data under a world coordinate system. The method can accurately capture the three-dimensional distribution of subcutaneous veins, compensate positioning errors, improve the accuracy of vein position data, solve the problem of positioning deviation of the conventional technology and realize more accurate vein positioning.

Inventors

QIU XUEDE
JIANG YING
QIN QIN
ZHAO SHUANG
YAO HUAN
WANG XUELEI
CUI XIAOJIE
LI YAN
YANG JIAJIA

Assignees

中国人民解放军总医院第一医学中心

Dates

Publication Date: 20260512
Application Date: 20260319

Claims (10)

1. An infrared imaging-based automatic subcutaneous vein blood collection method, comprising: acquiring infrared thermal imaging original data of a target area, wherein the target area is a surface skin area of a limb to be sampled; Carrying out temperature field reconstruction processing on the infrared thermal imaging original data to generate a two-dimensional temperature matrix reflecting subcutaneous tissue thermal distribution; Extracting heat conduction characteristics of the vascular tissue based on the two-dimensional temperature matrix, wherein the heat conduction characteristics of the vascular tissue comprise a temperature gradient direction and a heat diffusion rate; Inputting the temperature gradient direction and the thermal diffusion rate into a preset thermodynamic causal reasoning model, and inverting the three-dimensional space coordinate of the subcutaneous vein by the thermodynamic causal reasoning model according to the physical rule of heat transfer from the in-vivo core area to the body surface; According to the three-dimensional space coordinates, calculating a diameter estimated value and a depth estimated value of the subcutaneous vein by combining temperature values in the two-dimensional temperature matrix; Performing geometric correction on the three-dimensional space coordinate of the subcutaneous vein based on the diameter estimated value and the depth estimated value to obtain corrected vein center coordinate; and carrying out mapping conversion between an infrared image coordinate system and a physical space coordinate system by using the corrected vein center coordinate, and generating vein position data under a world coordinate system.
2. The method for automatic subcutaneous vein blood sampling based on infrared imaging according to claim 1, wherein performing a temperature field reconstruction process on the infrared thermal imaging raw data to generate a two-dimensional temperature matrix reflecting a subcutaneous tissue thermal distribution comprises: carrying out non-uniformity correction on the infrared thermal imaging original data, and eliminating fixed pattern noise introduced by the self characteristics of a detector in the infrared thermal imaging original data; adopting a self-adaptive Gaussian filtering algorithm to carry out smooth denoising on corrected infrared thermal imaging original data, and reserving temperature mutation characteristics of the blood vessel edge; according to the Planck blackbody radiation law, converting the gray image data subjected to smooth denoising into an absolute temperature value to obtain an initial temperature distribution map; Detecting abnormal high-temperature or low-temperature pixel points in the initial temperature distribution diagram, and replacing the abnormal high-temperature or low-temperature pixel points with the temperature median value of surrounding neighborhood pixel points to finish the repair of a temperature field; resampling the repaired initial temperature distribution map to a preset resolution to generate the two-dimensional temperature matrix.
3. An infrared imaging-based automatic subcutaneous vein collection method as in claim 2, wherein extracting thermal conduction characteristics of vascular tissue based on the two-dimensional temperature matrix comprises: calculating the first temperature derivative of the two-dimensional temperature matrix in the horizontal direction and the vertical direction to obtain a temperature gradient vector field; Vector synthesis is carried out on the temperature gradient vector field, and the heat flow direction of each pixel point is calculated; Carrying out convolution operation on the two-dimensional temperature matrix by using a Laplacian operator, and calculating a thermal diffusion coefficient at each pixel point, wherein the thermal diffusion coefficient reflects the difficulty degree of heat diffusion at the pixel point; selecting continuous pixel areas with consistent heat flow direction and thermal diffusion coefficient lower than a preset threshold value, and marking the continuous pixel areas as suspected blood vessel areas; And extracting the average temperature gradient direction and the average thermal diffusion rate of all pixel points in the suspected blood vessel region as the heat conduction characteristics of the blood vessel tissue.
4. An infrared imaging-based automatic subcutaneous vein blood sampling method as claimed in claim 3, wherein said thermodynamic causal inference model inverts the three-dimensional space coordinates of the subcutaneous vein according to the physical law of heat transfer from the core area to the body surface in the body, comprising: Setting a constant heat source boundary condition representing the core temperature of a human body in the thermodynamic causal inference model; based on a Fourier heat conduction equation, establishing a nonlinear mapping relation between the surface temperature and the subcutaneous tissue depth in the two-dimensional temperature matrix; carrying out iterative solution on the nonlinear mapping relation by using a numerical solution method, and calculating theoretical temperature values of different depths below the suspected blood vessel region; Comparing the calculated theoretical temperature value with an actual measured temperature value obtained from the two-dimensional temperature matrix, and calculating a residual error between the theoretical temperature value and the actual measured temperature value; The residual error is minimized by adjusting depth parameters and radius parameters of the subcutaneous vein, and the depth parameters and the radius parameters corresponding to the residual error are depth estimated values and diameter estimated values of the subcutaneous vein at the moment, so that the three-dimensional space coordinates are determined; The thermodynamic causal reasoning model is constructed by: Acquiring a training data set, wherein the training data set comprises a plurality of groups of marked infrared thermal imaging sample data and real subcutaneous vein three-dimensional space position data which are matched with the marked infrared thermal imaging sample data and are measured by medical imaging equipment; defining a model architecture, wherein the thermodynamic causal reasoning model adopts a nonlinear regression model based on a multilayer neural network, an input layer is a heat conduction characteristic of the vascular tissue and comprises a temperature gradient direction and a heat diffusion rate, an output layer is a predicted vein depth parameter and a predicted vein radius parameter, at least one hidden layer is arranged between the input layer and the output layer, and a nonlinear activation function is applied in the hidden layer; Establishing a physical constraint loss function, wherein the physical constraint loss function consists of two parts, the first part is the mean square error loss between the predicted output and the real labeling value in the training data set, the second part is the physical consistency loss based on the Fourier heat conduction law, and the physical consistency loss penalizes the deviation between the theoretical body surface temperature distribution caused by the predicted output of the model and the actual temperature distribution observed by the infrared image; Performing iterative training on the thermodynamic causal inference model by using the training data set, inputting sample data into the model to obtain prediction output in each iteration, calculating total loss according to the physical constraint loss function, and updating weight parameters of the thermodynamic causal inference model through a back propagation algorithm until the total loss is converged below a preset threshold; After model training is completed, performing performance evaluation on the thermodynamic causal reasoning model on an independent verification data set, and verifying whether average errors of vein depth and radius of the reverse performance of the thermodynamic causal reasoning model and a true value are lower than a preset precision threshold; solidifying the thermodynamic causal reasoning model passing through the evaluation and deploying the thermodynamic causal reasoning model as the preset thermodynamic causal reasoning model.
5. The method of infrared imaging-based automatic subcutaneous vein blood collection according to claim 4, wherein calculating the diameter estimate and the depth estimate of the subcutaneous vein based on the three-dimensional space coordinates in combination with the temperature values in the two-dimensional temperature matrix comprises: taking the three-dimensional space coordinate as a center, and extracting a cross-section temperature profile curve which passes through the three-dimensional space coordinate as a center from the two-dimensional temperature matrix; Analyzing the half-width of the cross section temperature profile curve, wherein the half-width refers to the width when the temperature is reduced to half of the peak value; Calculating the physical diameter of the subcutaneous vein according to the half-width and the attenuation coefficient of the pre-calibrated infrared light in the tissue, and taking the physical diameter as the diameter estimated value; Measuring a temperature difference between a surface temperature value at the three-dimensional space coordinates and an average temperature value of the peripheral adipose tissue region; And according to the temperature difference and the typical temperature difference range of the arterial blood and the venous blood of the human body, the heat conduction model is combined, and the path length of heat transfer is reversely deduced to be used as the depth estimated value.
6. An infrared imaging-based automatic subcutaneous vein collection method as in claim 5, wherein geometrically correcting three-dimensional space coordinates of said subcutaneous vein based on said diameter estimate and said depth estimate to obtain corrected vein center coordinates comprises: Judging whether the diameter estimated value is smaller than a preset minimum puncturable diameter threshold value or not; If the blood vessel is smaller than the candidate list, judging that the blood vessel which is currently identified is a capillary vessel or a micro vein, and removing the blood vessel from the candidate list; if not smaller than the three-dimensional space coordinate, adjusting the vertical component of the three-dimensional space coordinate according to the depth estimation value, and correcting the projection error caused by uneven skin surface; Re-fitting an elliptic curve in the local neighborhood by using the diameter estimation value as a constraint condition, wherein the elliptic curve replaces the cross-section projection of the subcuticular vein; And updating the center coordinates of the elliptic curve into the corrected vein center coordinates.
7. The method of infrared imaging-based automatic subcutaneous vein collection as claimed in claim 6, wherein using the corrected vein center coordinates, performing mapping conversion between an infrared image coordinate system and a physical space coordinate system, generating vein position data in a world coordinate system, comprises: reading a three-dimensional coordinate point set of a calibration plate arranged below the infrared camera under a world coordinate system; Extracting a two-dimensional pixel coordinate point set corresponding to the calibration plate in the infrared image; Calculating a conversion matrix from an infrared image coordinate system to a world coordinate system through a perspective transformation algorithm; Substituting the corrected vein center coordinates into the transformation matrix, and calculating to obtain three-dimensional coordinate values of the corrected vein center coordinates under a world coordinate system; and packaging the three-dimensional coordinate values and the corresponding attitude angles into vein position data under the world coordinate system.
8. An infrared imaging-based automatic subcutaneous vein collection method as in claim 7, further comprising: Performing time sequence smoothing filtering on vein position data under the world coordinate system to eliminate coordinate jitter generated by limb micro motion; Registering and superposing the filtered vein position data under the world coordinate system with visible light image data of a pre-acquired target area; According to the data after registration superposition, outputting a positioning result containing the space position, trend angle and subcutaneous depth information of the subcutaneous vein; The step of performing time sequence smoothing filtering on vein position data in the world coordinate system to eliminate coordinate shake caused by limb inching comprises the following steps: Creating a fixed-length data buffer for storing vein position data in the world coordinate system in a last several sampling periods; calculating Euclidean distance between vein position data under the world coordinate system at the current moment and vein position data under the world coordinate system at the last moment; if the Euclidean distance is larger than a preset maximum displacement threshold, judging that the limb movement occurs to a large extent, emptying the data buffer area and performing repositioning; If the Euclidean distance is not greater than a preset maximum displacement threshold value, extracting all historical coordinate points from the data buffer area; and predicting an optimal coordinate estimated value by adopting a Kalman filtering algorithm and combining the historical coordinate point and the measured value at the current moment, and replacing vein position data under the current world coordinate system by using the optimal coordinate estimated value.
9. The method of infrared imaging-based automatic subcutaneous vein collection as set forth in claim 8, wherein registering and superimposing the filtered vein position data in the world coordinate system with the pre-acquired visible image data of the target area comprises: Detecting feature points of the visible light image data, and extracting blood vessel texture feature points; detecting the same characteristic points of the infrared image data, and extracting temperature edge characteristic points; using a feature matching algorithm to find feature point pairs matched in the visible light image data and the infrared image data; According to the characteristic point pairs, affine transformation parameters between the visible light image data and the infrared image data are calculated; And mapping the filtered vein position data in the world coordinate system into the coordinate system of the visible light image by utilizing the affine transformation parameters, so as to realize visual superposition of the vein position data and the visible light image data.
10. The method for automatic subcutaneous vein blood sampling based on infrared imaging according to claim 9, wherein the outputting of the positioning result including the spatial position, the trend angle and the subcutaneous depth information of the subcutaneous vein according to the data after registration and superposition comprises: Reading a specific numerical value of the corrected vein center coordinate in a world coordinate system from the data after registration superposition, and taking the specific numerical value as a space position of the subcutaneous vein; selecting a plurality of continuous sampling points along the trend of the vein trunk where the corrected vein central coordinate is located; calculating the included angle between the connecting line between two adjacent sampling points and the horizontal axis, and averaging all the included angles to obtain the trend angle; reading the depth estimation value used for generating the corrected vein center coordinates as the subcutaneous depth information; and packaging the spatial position, the trend angle and the subcutaneous depth information into a structured data packet, and sending the data packet to a designated display terminal for display.

Description

Automatic subcutaneous vein blood sampling method based on infrared imaging Technical Field The invention belongs to the technical field of vein blood sampling, and particularly relates to an automatic subcutaneous vein blood sampling method based on infrared imaging. Background At present, in automatic subcutaneous vein blood sampling technology, infrared imaging technology is widely used for vein recognition, and in the prior art, the approximate position of a vein is recognized by acquiring infrared images on the surface of a limb to be sampled and relying on image gray level difference, and part of technology can carry out simple filtering and noise reduction treatment on infrared original data, and only can output two-dimensional projection information of the vein on the body surface, so that the three-dimensional space distribution condition of the vein under the skin cannot be accurately reflected. The prior art does not fully consider the difference of heat conduction characteristics of vascular tissues and surrounding subcutaneous tissues, can not accurately capture key parameters such as depth, diameter and the like of veins only through image gray scale characteristics, can not invert three-dimensional space coordinates of veins, does not carry out targeted geometric correction on the positions of the recognized veins, causes deviation of mapping between an infrared image coordinate system and a physical space coordinate system, and has insufficient vein positioning precision. The prior art is difficult to accurately acquire the three-dimensional space position of subcutaneous veins through infrared imaging data, and can not optimize vein position accuracy through reasonable correction means, leads to automatic blood sampling in-process positioning accuracy not enough, is difficult to satisfy automatic blood sampling's accurate positioning demand, influences blood sampling efficiency and security. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art; therefore, the invention provides an automatic subcutaneous vein blood sampling method based on infrared imaging, which comprises the following steps: acquiring infrared thermal imaging original data of a target area, wherein the target area is a surface skin area of a limb to be sampled; Carrying out temperature field reconstruction processing on the infrared thermal imaging original data to generate a two-dimensional temperature matrix reflecting subcutaneous tissue thermal distribution; Extracting heat conduction characteristics of the vascular tissue based on the two-dimensional temperature matrix, wherein the heat conduction characteristics of the vascular tissue comprise a temperature gradient direction and a heat diffusion rate; Inputting the temperature gradient direction and the thermal diffusion rate into a preset thermodynamic causal reasoning model, and inverting the three-dimensional space coordinate of the subcutaneous vein by the thermodynamic causal reasoning model according to the physical rule of heat transfer from the in-vivo core area to the body surface; According to the three-dimensional space coordinates, calculating a diameter estimated value and a depth estimated value of the subcutaneous vein by combining temperature values in the two-dimensional temperature matrix; Performing geometric correction on the three-dimensional space coordinate of the subcutaneous vein based on the diameter estimated value and the depth estimated value to obtain corrected vein center coordinate; and carrying out mapping conversion between an infrared image coordinate system and a physical space coordinate system by using the corrected vein center coordinate, and generating vein position data under a world coordinate system. Further, performing temperature field reconstruction processing on the infrared thermal imaging raw data to generate a two-dimensional temperature matrix reflecting subcutaneous tissue thermal distribution, including: carrying out non-uniformity correction on the infrared thermal imaging original data, and eliminating fixed pattern noise introduced by the self characteristics of a detector in the infrared thermal imaging original data; adopting a self-adaptive Gaussian filtering algorithm to carry out smooth denoising on corrected infrared thermal imaging original data, and reserving temperature mutation characteristics of the blood vessel edge; according to the Planck blackbody radiation law, converting the gray image data subjected to smooth denoising into an absolute temperature value to obtain an initial temperature distribution map; Detecting abnormal high-temperature or low-temperature pixel points in the initial temperature distribution diagram, and replacing the abnormal high-temperature or low-temperature pixel points with the temperature median value of surrounding neighborhood pixel points to finish the repair of a temperature field; resampling t