Blood vessel recognition auxiliary blood sampling method and system for child blood sampling process and storage medium

Abstract

The invention discloses a blood vessel recognition auxiliary blood sampling method and system for a child blood sampling process and a storage medium, and relates to the technical field of auxiliary medical treatment. The method comprises the following steps: acquiring a blood vessel image of a current child patient by adopting wide-angle and non-wide-angle infrared image acquisition equipment, and matching and splicing the wide-angle and non-wide-angle blood vessel images to obtain a complete blood vessel image; the current blood flow direction is determined according to the pixel gray-time sequence corresponding relation corresponding to the infrared image in the complete blood vessel image, and meanwhile the thickness of the blood vessel wall is determined through edge detection and morphological analysis; dividing a blood vessel area in the complete blood vessel image into a first blood vessel and a second blood vessel by combining the blood flow direction and the blood vessel wall thickness, screening the first blood vessel and the second blood vessel again based on the blood vessel wall thickness, and selecting blood vessels within a preset threshold value; selecting blood vessels within a preset threshold value, and finally determining an optimal blood sampling blood vessel by combining the walking and the position of the current blood vessel and the current skin state of the corresponding child patient. The puncture success rate is increased.

Description

Technical Field

[0001] This invention relates to the field of assistive medical technology, and more specifically to a method, system, and storage medium for assistive blood collection by vascular identification in the process of blood collection in children. Background Technology

[0002] In pediatric medical diagnosis, blood collection is a crucial step in obtaining blood samples, and its efficiency and safety directly affect the progress of diagnosis and treatment and the child's medical experience. However, the physiological characteristics of children present significant challenges to blood collection procedures, becoming a pressing clinical problem that needs to be solved.

[0003] Children, especially infants, have thin blood vessels with delicate diameters and thin walls, and a thicker subcutaneous fat layer, making their blood vessels less visible and palpable on the skin. Traditional blood collection relies on the experience of medical staff, locating the blood collection point by visual observation or finger touch, which is prone to errors when dealing with children's hidden blood vessels. Furthermore, children's skin is delicate, and repeated punctures can cause tissue damage, leading to complications such as hematomas and infections, increasing pain and fear.

[0004] While clinical practice offers auxiliary tools such as infrared vascular imaging devices, their limitations are significant in pediatric blood collection scenarios. Infrared imaging is easily affected by uneven distribution of skin pigmentation and subcutaneous fat in children, resulting in insufficient image clarity and difficulty in accurately depicting the branches and directions of small blood vessels. Existing equipment is mostly fixed, lacking operational flexibility and unable to adapt to dynamic changes in children's limbs, thus limiting its auxiliary effect. Furthermore, some equipment is costly, making it difficult to popularize in primary healthcare institutions, and most pediatric blood collection still relies on traditional methods.

[0005] Therefore, developing auxiliary blood collection methods that accurately identify blood vessels, improve puncture success rates, and reduce pain, tailored to the characteristics of children's blood vessels and their blood collection needs, is an important topic for improving children's blood collection experience and enhancing clinical efficiency. Summary of the Invention

[0006] In view of this, the present invention provides a blood vessel identification-assisted blood collection method, system and storage medium for children's blood collection process, so as to improve the puncture success rate.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A blood collection method assisted by vascular identification for use in children's blood collection process includes the following steps:

[0009] Vascular images of the current child were acquired using wide-angle and non-wide-angle infrared image acquisition devices, and the wide-angle and non-wide-angle vascular images were matched and stitched together to obtain a complete vascular image of the current child.

[0010] The current blood flow direction is determined based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image, while the thickness of the blood vessel wall is determined by edge detection and morphological analysis.

[0011] By combining blood flow direction and vessel wall thickness, the blood vessels in the complete blood vessel image are divided into the first blood vessel and the second blood vessel. Based on the vessel wall thickness, the first blood vessel and the second blood vessel are further filtered to select the blood vessels within the preset threshold.

[0012] By selecting blood vessels within a preset threshold and combining the current course and location of the blood vessels with the corresponding skin condition of the child, the optimal blood collection vessel is finally determined.

[0013] Optionally, matching and stitching wide-angle and non-wide-angle vascular images includes:

[0014] The acquired wide-angle infrared vascular images and non-wide-angle infrared vascular images are preprocessed respectively. The preprocessing includes removing noise by using Gaussian filtering and enhancing the contrast between blood vessels and surrounding tissues by histogram equalization.

[0015] Feature tensors of various scales are extracted by a multimodal image matching network, fused to obtain coarse-grained descriptors of key points of blood vessels, and then fine-grained descriptors are generated by the descriptor head.

[0016] A confidence matrix is ​​generated based on coarse-grained descriptors, and the point pairs with the highest confidence scores are taken as coarse-grained matching point pairs.

[0017] For each coarse-grained matching point pair, the corresponding feature point block is clipped, the associated heatmap is calculated, and the offset is estimated.

[0018] After transforming the image based on the offset, the overlapping areas are merged, and the structural similarity index is used to determine whether the stitching is qualified.

[0019] Optionally, the current blood flow direction can be determined based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image, specifically including the following steps:

[0020] For an infrared video sequence of a complete blood vessel image, extract the sequence of grayscale values ​​of each pixel within the blood vessel region as a function of time.

[0021] The dense optical flow algorithm is used to calculate the instantaneous optical flow field between adjacent frames, and the average optical flow field is obtained by averaging all instantaneous optical flow fields in the sequence.

[0022] The optical flow continuity features of each pixel are extracted, including the similarity of optical flow vectors in eight neighborhoods and the similarity of optical flow vectors in circular neighborhoods. Combined with the fluctuation trend of pixel grayscale-time series, the main direction of blood flow is determined.

[0023] Optionally, the thickness of the vessel wall can be determined using edge detection and morphological analysis, specifically including the following steps:

[0024] Edge detection algorithms are used to extract the inner and outer edge contours of blood vessels, and morphological operations are combined to optimize edge accuracy.

[0025] Calculate the pixel distance between the inner and outer edges along the cross-sectional direction of the blood vessel, which is perpendicular to the blood flow direction, and convert it into actual physical thickness based on the image resolution.

[0026] Optionally, the first and second blood vessels correspond to veins and arteries, respectively, and venous blood or arterial blood can be collected depending on the purpose of the blood collection.

[0027] Optionally, the preset threshold can be adjusted based on vascular anatomy data for different age groups to ensure that the screening results meet the actual clinical needs.

[0028] A blood vessel identification-assisted blood collection system for children includes the following steps:

[0029] Complete vascular image acquisition module: used to acquire vascular images of the current child using wide-angle and non-wide-angle infrared image acquisition devices respectively, match and stitch the wide-angle and non-wide-angle vascular images to obtain a complete vascular image of the current child;

[0030] Blood vessel detection module: used to determine the current blood flow direction based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image, and to determine the thickness of the blood vessel wall using edge detection and morphological analysis.

[0031] The vessel screening module is used to distinguish the vessels in a complete vessel image into the first vessel and the second vessel by combining blood flow direction and vessel wall thickness. Based on the vessel wall thickness, the first vessel and the second vessel are further screened to select vessels within a preset threshold.

[0032] The optimal blood collection vessel determination module is used to select blood vessels within a preset threshold, and combine the current course and location of the blood vessel with the corresponding skin condition of the child to finally determine the optimal blood collection vessel.

[0033] A computer storage medium storing a computer program, which, when executed by a processor, implements the steps of any one of the methods for a blood vessel identification-assisted blood collection method for children's blood collection process.

[0034] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method, system and storage medium for vascular identification-assisted blood collection in the process of blood collection in children, which has the following beneficial effects:

[0035] 1. Complementary stitching of wide-angle and non-wide-angle infrared images not only preserves the global vascular distribution information of the wide-angle field of view, but also improves the clarity of local blood vessels through the high-resolution details of the non-wide-angle image, effectively avoiding missed detections or misjudgments caused by the limitation of angle or resolution of a single modality.

[0036] 2. By determining the blood flow direction through the gray-time series relationship, areas with abnormal blood flow caused by vascular malformation or compression (such as venous valves or thrombosis) can be excluded, ensuring that the selected blood vessels have normal physiological functions.

[0037] 3. By combining edge detection and morphological analysis, the thickness of the blood vessel wall can be accurately quantified, avoiding operation failures caused by blood vessel walls that are too thin (easily ruptured) or too thick (difficult to puncture). Attached Figure Description

[0038] To more clearly illustrate the technical solutions 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of the method flow provided by the present invention;

[0040] Figure 2 This is a schematic diagram of the system structure provided by the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] This invention discloses a blood vessel identification-assisted blood collection method for children's blood collection process, such as... Figure 1 As shown, it includes the following steps:

[0043] Step 1: Acquire vascular images of the current child using wide-angle and non-wide-angle infrared image acquisition devices respectively, match and stitch the wide-angle and non-wide-angle vascular images to obtain a complete vascular image of the current child;

[0044] Step 2: Determine the current blood flow direction based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image, and determine the thickness of the blood vessel wall using edge detection and morphological analysis.

[0045] Step 3: Combine blood flow direction and vessel wall thickness to distinguish the vessels in the complete vascular image into the first vessel and the second vessel. Based on the vessel wall thickness, the first vessel and the second vessel are further filtered to select vessels within the preset threshold.

[0046] Step 4: Select a blood vessel within the preset threshold, and determine the best blood collection vessel by combining the current course and location of the blood vessel with the corresponding skin condition of the child.

[0047] The course and location of the blood vessel are crucial: Choose a straight vessel with few branches, avoiding bends, bifurcations, or nodules. Bent vessels may cause the needle to accidentally enter a branch or penetrate the vessel wall, while bifurcations can lead to blood flow shunting and poor blood collection. The vessel should also be superficial and easily visible, avoiding deep vessels or those covered by fat or tendons (such as deep veins in obese patients) to reduce the difficulty of puncture. Furthermore, the vessel should be kept away from joints (such as the wrist and elbow) to prevent needle displacement or vessel damage due to patient movement.

[0048] Furthermore, in step one, matching and stitching the wide-angle and non-wide-angle vascular images includes:

[0049] Step 1.1: Preprocess the acquired wide-angle infrared vascular images and non-wide-angle infrared vascular images respectively. The preprocessing includes removing noise by using Gaussian filtering and enhancing the contrast between blood vessels and surrounding tissues by histogram equalization.

[0050] Step 1.2 Extracts features from the two preprocessed images using a multimodal image matching network. Specifically, the following steps are taken: feature tensors of various scales are extracted from the two images using an adaptive encoder; some feature tensors are upsampled bilinearly to the same size as the other feature tensor, and then summed element-wise to obtain coarse-grained descriptors of blood vessel key points; the coarse-grained descriptors are processed by a descriptor head composed of a multilayer perceptron to generate fine-grained descriptors.

[0051] Step 1.3: Generate coarse-grained matching point pairs: Input the coarse-grained descriptor into the detection head composed of convolutional layers and activation functions to obtain the key point detection score map; use non-maximum suppression to extract blood vessel key points with high scores, calculate the confidence matrix based on the coarse-grained descriptor, and take the point pairs with the highest confidence scores as coarse-grained matching point pairs;

[0052] Step 1.4, Subpixel-level offset estimation: For each coarse-grained matching point pair, crop the feature point blocks corresponding to the fine-grained descriptor with a given radius; calculate the correlation heatmap between the feature point blocks, and estimate the offset between the two images based on the correlation heatmap;

[0053] Step 1.5: Perform geometric transformation on the non-wide-angle image based on the offset, and use the weighted average method to fuse the overlapping areas to obtain a complete blood vessel image;

[0054] Step 1.6: Calculate the structural similarity index (SSIM) for the stitched image. If the SSIM is greater than the set threshold, it is considered qualified; otherwise, return to step 1.3 to rematch.

[0055] Furthermore, in step two, the current blood flow direction is determined based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image. This specifically includes the following steps:

[0056] For an infrared video sequence of a complete blood vessel image, extract the sequence of grayscale values ​​of each pixel within the blood vessel region as a function of time.

[0057] Dense optical flow algorithms (such as Farneback optical flow algorithm) are used to calculate the instantaneous optical flow field between any two adjacent frames in an infrared video sequence to characterize the motion vector of pixels at adjacent times. The average optical flow field is obtained by averaging all instantaneous optical flow fields in the video sequence to weaken instantaneous noise interference and highlight the overall trend of blood flow.

[0058] Optical flow continuity feature extraction includes:

[0059] Direct optical flow continuity feature: For each pixel, calculate the similarity between its optical flow vector in its eight neighborhoods and the optical flow vector of the blood vessel pixel (judged by the cosine of the vector angle and the velocity deviation threshold), and map it to a probability value; if the probability value exceeds the preset threshold, then the pixel is marked as having direct optical flow continuity.

[0060] Radial optical flow continuity feature: Perform the above similarity calculation within the circular neighborhood of the pixel, map it to a probability value and compare it with a threshold to mark the radial optical flow continuity.

[0061] By combining the gray-time series fluctuation trend of pixels (such as the periodic changes in gray-level caused by the periodic pulsation of arterial blood flow) with the vector direction of the average optical flow field, the blood flow direction is determined in the following way:

[0062] For a cluster of pixels with continuous optical flow characteristics within a vascular region, fit the principal direction of its optical flow vector;

[0063] The main direction of optical flow is verified by referring to the peak propagation direction in the gray-scale-time series (such as the gray-scale change delay from upstream to downstream);

[0064] Ultimately, the overall propagation direction of the vector field is taken as the direction of blood flow, with the direction away from the heart being the direction of arterial flow and the direction towards the heart being the direction of venous flow.

[0065] Furthermore, in step two, edge detection and morphological analysis are used to determine the thickness of the vessel wall, specifically including the following steps:

[0066] Based on complete vascular images, a DenseUNet neural network is used to perform preliminary segmentation of infrared vascular images, obtaining preliminary binary images of vascular structures. Simultaneously, background subtraction is used to process the infrared video sequence, and the results are accumulated to obtain a binary image of the vascular structure containing capillaries. The sum of these two images is used as a seed image, and a region growing algorithm (combining grayscale features, vascular continuity features, and optical flow continuity features) is applied to optimize the segmentation, resulting in a precise binary image of the vascular structure. Furthermore, an edge detection algorithm is used to extract the inner and outer edge contours of the blood vessels, preserving the boundary pixel coordinates of the vessel walls.

[0067] Morphological operations are performed on the extracted edge contours. Opening operations are used to remove edge burrs, and closing operations are used to fill edge gaps to ensure the continuity and integrity of the edges. For the cross-section of the blood vessel (perpendicular to the blood flow direction), clear edge segments are selected and abnormal edge points caused by noise or blood vessel branches are removed.

[0068] Along the cross-sectional direction of the blood vessel, the pixel distance between the inner and outer edges is calculated: for each cross-section, the pixel distance between multiple point pairs on the edge is taken, and the average value is used as the pixel-level wall thickness at that location; combined with the resolution of the infrared image (the actual physical size corresponding to each pixel), the pixel-level wall thickness is converted into the actual physical thickness (e.g., millimeters). At the same time, referring to the distribution characteristics of blood vessel wall thickness (e.g., the uniformity of arterial wall thickness is higher than that of veins, and it is thicker overall), outliers are eliminated through statistical analysis to obtain reliable blood vessel wall thickness data.

[0069] By combining blood flow characteristics (such as the centrifugal nature of arterial blood flow and the centripetal nature of venous blood flow), the calculated vessel wall thickness is cross-validated: if a vessel segment has a thicker wall and flows away from the heart, or a thinner wall and flows towards the heart, both are consistent with anatomical characteristics, and the wall thickness calculation result can be confirmed as valid; otherwise, the edge detection and segmentation steps need to be re-examined to optimize the accuracy of the wall thickness calculation.

[0070] Furthermore, the first and second blood vessels correspond to veins and arteries, respectively, and venous blood or arterial blood is selected for collection based on the intended use of the blood.

[0071] Furthermore, the preset threshold is adjusted based on vascular anatomy data for different age groups to ensure that the screening results meet the actual clinical needs.

[0072] and Figure 1 Corresponding to the method shown, the present invention also discloses a blood vessel identification-assisted blood collection system for children's blood collection process. Figure 1 The implementation of the method, specifically its structure, is as follows: Figure 2 As shown, it includes the following steps:

[0073] Complete vascular image acquisition module: used to acquire vascular images of the current child using wide-angle and non-wide-angle infrared image acquisition devices respectively, match and stitch the wide-angle and non-wide-angle vascular images to obtain a complete vascular image of the current child;

[0074] Blood vessel detection module: used to determine the current blood flow direction based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image, and to determine the thickness of the blood vessel wall using edge detection and morphological analysis.

[0075] The vessel screening module is used to distinguish the vessels in a complete vessel image into the first vessel and the second vessel by combining blood flow direction and vessel wall thickness. Based on the vessel wall thickness, the first vessel and the second vessel are further screened to select vessels within a preset threshold.

[0076] The optimal blood collection vessel determination module is used to select blood vessels within a preset threshold, and combine the current course and location of the blood vessel with the corresponding skin condition of the child to finally determine the optimal blood collection vessel.

[0077] Finally, this embodiment discloses a computer storage medium storing a computer program, which, when executed by a processor, implements the steps of any one of the methods for vascular identification-assisted blood collection in the process of blood collection in children.

[0078] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0079] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A blood collection method assisted by vascular identification for children's blood collection process, characterized in that, Includes the following steps: Vascular images of the current child were acquired using wide-angle and non-wide-angle infrared image acquisition devices, and the wide-angle and non-wide-angle vascular images were matched and stitched together to obtain a complete vascular image of the current child. The current blood flow direction is determined based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image, while the thickness of the blood vessel wall is determined by edge detection and morphological analysis. By combining blood flow direction and vessel wall thickness, the blood vessels in the complete blood vessel image are divided into the first blood vessel and the second blood vessel. Based on the vessel wall thickness, the first blood vessel and the second blood vessel are further filtered to select the blood vessels within the preset threshold. By selecting blood vessels within a preset threshold and combining the current course and location of the blood vessels with the corresponding skin condition of the child, the optimal blood collection vessel is finally determined.

2. The method for vascular identification-assisted blood collection in children's blood collection process according to claim 1, characterized in that, Matching and stitching wide-angle and non-wide-angle vascular images includes: The acquired wide-angle infrared vascular images and non-wide-angle infrared vascular images are preprocessed respectively. The preprocessing includes removing noise by using Gaussian filtering and enhancing the contrast between blood vessels and surrounding tissues by histogram equalization. Feature tensors of various scales are extracted by a multimodal image matching network, fused to obtain coarse-grained descriptors of key points of blood vessels, and then fine-grained descriptors are generated by the descriptor head. A confidence matrix is ​​generated based on coarse-grained descriptors, and the point pairs with the highest confidence scores are taken as coarse-grained matching point pairs. For each coarse-grained matching point pair, the corresponding feature point block is clipped, the associated heatmap is calculated, and the offset is estimated. After transforming the image based on the offset, the overlapping areas are merged, and the structural similarity index is used to determine whether the stitching is qualified.

3. The method for vascular identification-assisted blood collection in children's blood collection process according to claim 1, characterized in that, The current blood flow direction is determined based on the gray-time series correspondence of pixels in the infrared image of a complete blood vessel image, specifically including the following steps: For an infrared video sequence of a complete blood vessel image, extract the sequence of grayscale values ​​of each pixel within the blood vessel region as a function of time. The dense optical flow algorithm is used to calculate the instantaneous optical flow field between adjacent frames, and the average optical flow field is obtained by averaging all instantaneous optical flow fields in the sequence. The optical flow continuity features of each pixel are extracted, including the similarity of optical flow vectors in eight neighborhoods and the similarity of optical flow vectors in circular neighborhoods. Combined with the fluctuation trend of pixel grayscale-time series, the main direction of blood flow is determined.

4. The method for vascular identification-assisted blood collection in children's blood collection process according to claim 1, characterized in that, Determining the thickness of the vessel wall using edge detection and morphological analysis includes the following steps: Edge detection algorithms are used to extract the inner and outer edge contours of blood vessels, and morphological operations are combined to optimize edge accuracy. Calculate the pixel distance between the inner and outer edges along the cross-sectional direction of the blood vessel, which is perpendicular to the blood flow direction, and convert it into actual physical thickness based on the image resolution.

5. A method for vascular identification-assisted blood collection in children's blood collection process according to claim 1, characterized in that, The first and second blood vessels correspond to veins and arteries, respectively. Depending on the purpose of the blood collection, venous blood or arterial blood is selected for collection.

6. The method for vascular identification-assisted blood collection in children's blood collection process according to claim 1, characterized in that, The preset thresholds are adjusted based on vascular anatomy data for different age groups to ensure that the screening results meet the actual clinical needs.

7. A blood vessel identification-assisted blood collection system for children's blood collection process, characterized in that, Includes the following steps: Complete vascular image acquisition module: used to acquire vascular images of the current child using wide-angle and non-wide-angle infrared image acquisition devices respectively, match and stitch the wide-angle and non-wide-angle vascular images to obtain a complete vascular image of the current child; Blood vessel detection module: used to determine the current blood flow direction based on the gray-time series correspondence of pixels in the infrared image of the complete blood vessel image, and to determine the thickness of the blood vessel wall using edge detection and morphological analysis. The vessel screening module is used to distinguish the vessels in a complete vessel image into the first vessel and the second vessel by combining blood flow direction and vessel wall thickness. Based on the vessel wall thickness, the first vessel and the second vessel are further screened to select vessels within a preset threshold. The optimal blood collection vessel determination module is used to select blood vessels within a preset threshold, and combine the current course and location of the blood vessel with the corresponding skin condition of the child to finally determine the optimal blood collection vessel.

8. A computer storage medium, characterized in that, The computer storage medium stores a computer program, which, when executed by a processor, implements the steps of a vascular identification-assisted blood collection method for children's blood collection process as described in any one of claims 1-6.

Images