Image generation method and apparatus, electronic device, computer-readable storage medium, and computer program product

Abstract

The present application provides an image generation method and apparatus, an electronic device, a computer-readable storage medium, and a computer program product. The method comprises: constructing a three-dimensional vascular distribution of a preset biological site; projecting the three-dimensional vascular distribution to obtain a two-dimensional vascular distribution corresponding to the three-dimensional vascular distribution; and generating an image of the preset biological site on the basis of the two-dimensional vascular distribution.

Description

Image generation methods, apparatuses, electronic devices, computer-readable storage media, and computer program products

[0001] Cross-references to related applications

[0002] This application is based on and claims priority to Chinese Patent Application No. 202510127717.0, filed on January 27, 2025, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of computer technology, and in particular to a method, apparatus, electronic device, computer-readable storage medium, and computer program product for generating an image. Background Technology

[0004] Palm vein recognition is a biometric identification technology that uses the pattern of blood vessels in the palm of the hand for identity authentication. Unlike externally visible biometrics such as faces, fingerprints, and irises, palm veins are located beneath the skin, and their pattern is formed by the blood vessels inside the palm. Because the palm vein pattern is not exposed on the skin surface, its acquisition requires the use of light of a specific wavelength (usually infrared light) to penetrate the skin, and an image sensor to capture an image of the palm veins inside the hand.

[0005] In related technologies, the recognition model is usually trained based on the palm vein data of the directly collected object. Due to the limitations of data collection cost and data privacy protection, the collection of images to generate feature images is difficult, which affects the efficiency of image generation. Summary of the Invention

[0006] This application provides an image generation method, apparatus, electronic device, computer-readable storage medium, and computer program product, which can effectively improve image generation efficiency.

[0007] The technical solution of this application embodiment is implemented as follows:

[0008] This application provides an image generation method, including:

[0009] A three-dimensional blood vessel distribution of a preset biological site is constructed, wherein the three-dimensional blood vessel distribution is used to represent the three-dimensional blood vessel distribution morphology of the preset biological site;

[0010] Projecting the three-dimensional blood vessel distribution yields a two-dimensional blood vessel distribution corresponding to the three-dimensional blood vessel distribution, which is used to represent the two-dimensional blood vessel distribution morphology of the preset biological site;

[0011] Based on the two-dimensional blood vessel distribution, an image of the preset biological site is generated.

[0012] This application provides an image generation apparatus, comprising:

[0013] The acquisition module is configured to construct a three-dimensional blood vessel distribution of a preset biological site, wherein the three-dimensional blood vessel distribution is used to represent the three-dimensional blood vessel distribution morphology of the preset biological site.

[0014] The projection module is configured to project the three-dimensional blood vessel distribution to obtain a two-dimensional blood vessel distribution corresponding to the three-dimensional blood vessel distribution. The two-dimensional blood vessel distribution is used to represent the two-dimensional blood vessel distribution morphology of the preset biological site.

[0015] The generation module is configured to generate an image of the preset biological site based on the two-dimensional blood vessel distribution.

[0016] This application provides a device for constructing a three-dimensional blood vessel distribution, comprising:

[0017] The identification module is configured to determine the coordinates of multiple key points used to generate the three-dimensional main blood vessel.

[0018] The generation module is configured to generate the three-dimensional main blood vessel based on the coordinates of the multiple key points;

[0019] The construction module is configured to construct the branch vessels of the three-dimensional main blood vessel to obtain the three-dimensional blood vessel distribution of the preset biological site.

[0020] This application provides an electronic device, including:

[0021] Memory, configured to store computer-executable instructions or computer programs;

[0022] When the processor is configured to execute computer-executable instructions or computer programs stored in the memory, it implements the image generation method and the three-dimensional blood vessel distribution construction method provided in the embodiments of this application.

[0023] This application provides a computer-readable storage medium storing computer-executable instructions configured to, when executed by a processor, implement the image generation method and the three-dimensional blood vessel distribution construction method provided in this application.

[0024] This application provides a computer program product, which includes a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the electronic device to perform the image generation method and the three-dimensional blood vessel distribution construction method described in this application embodiment.

[0025] The embodiments of this application have the following beneficial effects:

[0026] A three-dimensional blood vessel distribution of a pre-defined biological site is constructed, and this distribution is projected to obtain a two-dimensional blood vessel distribution. Based on this two-dimensional distribution, an image corresponding to the two-dimensional blood vessel distribution is generated. Thus, since the three-dimensional blood vessel distribution is three-dimensional, projecting it to obtain a two-dimensional blood vessel distribution transforms the three-dimensional distribution into a two-dimensional form, ensuring that the dimension of the two-dimensional blood vessel distribution matches the dimension of the image. This allows for the simulation of the actual image acquisition process through projection, eliminating the need to acquire a large number of actual images. Simultaneously, projection achieves dimensionality reduction of the three-dimensional blood vessel distribution, resulting in a two-dimensional blood vessel distribution. The corresponding image is then generated based on this two-dimensional distribution. Since the generated image is also two-dimensional, dimensionality reduction through projection makes the generation process a conversion from a two-dimensional blood vessel distribution to a two-dimensional image, rather than directly generating a two-dimensional image from a three-dimensional blood vessel distribution. This effectively saves the computational cost of directly generating a two-dimensional image from a three-dimensional blood vessel distribution, thereby significantly improving image generation efficiency. Attached Figure Description

[0027] Figure 1 is a schematic diagram of the architecture of the image generation system provided in an embodiment of this application;

[0028] Figure 2 is a schematic diagram of the structure of an electronic device for generating images provided in an embodiment of this application;

[0029] Figure 3 is a schematic diagram of the structure of an electronic device for constructing three-dimensional blood vessel distribution according to an embodiment of this application;

[0030] Figure 4 is a schematic flowchart of the image generation method provided in an embodiment of this application;

[0031] Figure 5 is a schematic flowchart of the image generation method provided in the embodiment of this application (II).

[0032] Figure 6 is a schematic flowchart of the image generation method provided in the embodiment of this application;

[0033] Figure 7 is a schematic flowchart of the image generation method provided in the embodiments of this application;

[0034] Figure 8 is a schematic flowchart of the image generation method provided in the embodiments of this application;

[0035] Figure 9 is a flowchart illustrating the method for constructing a three-dimensional blood vessel distribution according to an embodiment of this application.

[0036] Figure 10 is a schematic diagram of the palm vein recognition process provided in an embodiment of this application;

[0037] Figure 11 is a schematic diagram of the anatomical structure of the palm blood vessels and the palm vein image under near-infrared light provided in the embodiments of this application;

[0038] Figure 12 is a schematic diagram of the structural types of blood vessels in the palm provided in an embodiment of this application;

[0039] Figure 13 is a schematic diagram illustrating the principle of the image generation method provided in the embodiments of this application;

[0040] Figure 14 is a schematic diagram of the branch generation principle provided in the embodiment of this application;

[0041] Figure 15 is a schematic diagram illustrating the principle of training the palm vein recognition model provided in the embodiments of this application. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0043] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0044] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0045] 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 application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0046] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0047] 1) Palm vein recognition: This is a biometric identification technology that uses the pattern of blood vessels in the palm of the hand for identity authentication. Unlike externally visible biometrics such as faces, fingerprints, and irises, palm veins are located beneath the skin, and their pattern is formed by blood vessels inside the palm. Because the pattern of blood vessels in the palm (e.g., the vein pattern or artery pattern of the palm) is not exposed on the skin surface, its acquisition requires the use of light of a specific wavelength (usually infrared light) to penetrate the skin and capture images of the blood vessels inside the palm using an image sensor.

[0048] 2) Palmar Blood Vessels: These are the blood vessel networks distributed along the palmar side of the hand, including arteries, veins, and capillaries. These vessels are responsible for supplying blood to the palm and fingers to maintain normal hand function and vitality. Palmar blood vessels include palmar arteries, palmar veins, and palmar capillaries. The palmar arteries are responsible for transporting oxygen-rich blood from the heart to the hand. Major arteries include: Radial artery: extending downwards from the wrist along the thumb side. Ulnar artery: extending downwards from the wrist along the little finger side. Superficial palmar arch: formed by branches of the radial and ulnar arteries, located superficially in the palm. Deep palmar arch: formed by branches of the radial and ulnar arteries, located deep in the palm. Palmar veins are responsible for transporting blood containing metabolic waste back to the heart from the hand. Major veins include: Radial vein: running parallel to the radial artery, responsible for collecting blood from the area supplied by the radial artery. Ulnar vein: Runs parallel to the ulnar artery, it collects blood from the area supplied by the ulnar artery. Dorsal venous network: Located on the back of the hand, it consists of multiple veins that eventually converge into larger veins, such as the cephalic vein and basilic vein. Capillaries are tiny blood vessels connecting arteries and veins. They form an extensive network among the tissues and cells of the hand, responsible for delivering oxygen and nutrients to the cells while collecting waste products produced by the cells. The health of the blood vessels in the palm is crucial for hand function. Any damage or disease to blood vessels, such as blockage, inflammation, or aneurysm, can affect the blood supply to the hand, leading to pain, swelling, or other functional impairments.

[0049] 3) Deep Neural Networks: Deep Neural Networks (DNNs) are complex machine learning models that mimic the neuronal connections in the human brain. They learn and extract deep features from data through multiple layers of non-linear processing units. A deep neural network is a multi-layered neural network, where each layer contains multiple neurons connected by weighted sums and processed using non-linear activation functions. Deep neural networks are powerful machine learning models that, through their multi-layered structure and complex non-linear processing capabilities, can achieve high-performance prediction and classification on various types of data.

[0050] 4) Branching vessels: Branching vessels are smaller vessels that branch off from major blood vessels (such as arteries or veins). In the arterial system, branching vessels are responsible for transporting blood from the main arteries to tissues and organs throughout the body. In the venous system, branching vessels collect blood from these tissues and organs and return it to the heart. Branching vessels typically taper gradually; they branch off from larger vessels to form smaller vessels, such as capillaries. Characteristics of branching vessels include: they are part of a vascular network, responsible for increasing the contact area of ​​the vascular system for more efficient blood distribution. The diameter of branching vessels is usually smaller than that of the main vessels to which they belong. They may have specific functions, such as coronary artery branches supplying blood to the heart muscle.

[0051] 5) Main Blood Vessels: Also known as trunk vessels, the main blood vessels refer to the major arteries or veins in the vascular system, which are the main channels for blood flow. Main blood vessels typically have a large diameter and are a major component of the vascular network. In the human body, the main blood vessels include: Main arteries: such as the aorta, the most important artery originating from the heart, responsible for transporting blood to all parts of the body. Main veins: such as the superior vena cava and inferior vena cava, responsible for returning blood from various parts of the body to the heart. Characteristics of the main blood vessels include: They are the main channels for blood flow, undertaking the task of transporting or returning large amounts of blood. They are typically large in diameter and have thick walls to accommodate high-pressure blood flow. They play a central or core role in the vascular network, providing the source or confluence point for branch vessels.

[0052] 6) Near-infrared (NIR) light: This refers to a spectral region in the electromagnetic spectrum immediately beyond the red end of the visible light spectrum, with wavelengths roughly between 700 nanometers (nm) and 2500 nanometers. NIR light is a form of electromagnetic radiation, with wavelengths falling between red visible light (approximately 620-750 nm) and far-infrared light (approximately 2500 nm and above). Although called light, it is beyond the range of human vision, i.e., invisible light. In biological tissues, NIR light has good penetrating power and causes less damage, thus finding wide application in medical imaging and phototherapy. Light within the NIR spectral range can be used to acquire images of vein distribution in an individual's palm. NIR light sources are used because this light wave can penetrate the skin surface and is moderately absorbed by hemoglobin, making blood vessels more prominent in the image. When NIR light shines on a user's palm, due to the absorption characteristics of hemoglobin, the hemoglobin in the blood absorbs some of the light energy, making the veins appear darker relative to the surrounding tissue. A highly sensitive camera or image sensor captures infrared light penetrating the skin, recording an image of the distribution of subcutaneous veins in the palm. Specific algorithms extract vein features from the image, such as vein direction, diameter, and branching patterns—features unique to each individual. The extracted feature data is compared with palm vein features stored in a database to verify the individual's identity.

[0053] During the implementation of the embodiments of this application, the applicant discovered the following problems with the related technology:

[0054] In related technologies, the recognition model is usually trained based on the palm vein data of the directly collected object. However, due to the limitations of data collection costs and data privacy protection, the collection of images to generate feature images is quite difficult.

[0055] This application provides an image generation method, apparatus, electronic device, computer-readable storage medium, and computer program product, which can improve the efficiency of image generation. The following describes an exemplary application of the image sample construction system provided in this application.

[0056] Referring to Figure 1, which is a schematic diagram of the architecture of the image sample construction system 100 provided in the embodiment of this application, the terminal (exemplarily shown as terminal 400) connects to the server 200 through the network 300, which can be a wide area network or a local area network, or a combination of both.

[0057] Terminal 400 is configured for users to use client 410 to display a set of vascular image samples on a graphical interface 410-1 (graphical interface 410-1 is shown as an example). Terminal 400 and server 200 are interconnected via wired or wireless network.

[0058] In some embodiments, server 200 can be a standalone physical server, a server cluster or business system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Terminal 400 can be a smartphone, tablet, laptop, desktop computer, smart speaker, smart TV, smartwatch, in-vehicle terminal, etc., but is not limited to these. The electronic device provided in this application embodiment can be implemented as a terminal or a server. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application embodiment.

[0059] In some embodiments, the server 200 constructs a three-dimensional blood vessel distribution of a preset biological site, projects the three-dimensional blood vessel distribution to obtain a two-dimensional blood vessel distribution of the three-dimensional blood vessel distribution, generates an image corresponding to the two-dimensional blood vessel distribution based on the two-dimensional blood vessel distribution, and sends the image to the terminal 400.

[0060] In other embodiments, the terminal 400 constructs a three-dimensional blood vessel distribution of a preset biological site, projects the three-dimensional blood vessel distribution to obtain a two-dimensional blood vessel distribution of the three-dimensional blood vessel distribution, generates an image corresponding to the two-dimensional blood vessel distribution based on the two-dimensional blood vessel distribution, and sends the image to the server 200.

[0061] Referring to Figure 2, which is a schematic diagram of the structure of an electronic device 500 for generating images according to an embodiment of this application, the electronic device 500 shown in Figure 2 can be the server 200 or terminal 400 in Figure 1. The electronic device 500 shown in Figure 2 includes at least one processor 430, a memory 450, and at least one network interface 420. The various components in the electronic device 500 are coupled together through a bus system 440. It is understood that the bus system 440 is configured to enable communication between these components. In addition to a data bus, the bus system 440 also includes a power bus, a control bus, and a status signal bus. However, for clarity, all buses are labeled as bus system 440 in Figure 2.

[0062] Processor 430 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0063] The memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 450 may optionally include one or more storage devices physically located away from the processor 430.

[0064] The memory 450 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 450 described in this application embodiment is intended to include any suitable type of memory.

[0065] In some embodiments, memory 450 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0066] Operating system 451 includes system programs configured to handle various basic system services and perform hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0067] The network communication module 452 is configured to reach other electronic devices via one or more (wired or wireless) network interfaces 420, such as Bluetooth, WiFi, and Universal Serial Bus.

[0068] In some embodiments, the image generation apparatus provided in this application can be implemented in software. Figure 2 shows an image generation apparatus 455 stored in memory 450, which can be software in the form of programs and plug-ins, including the following software modules: acquisition module 4551, projection module 4552, and generation module 4553. These modules are logically related and can therefore be arbitrarily combined or further split according to their implemented functions. The functions of each module will be described below.

[0069] Referring to Figure 3, which is a schematic diagram of an electronic device for constructing three-dimensional blood vessel distribution according to an embodiment of this application, the electronic device 600 shown in Figure 3 can be the server 200 or terminal 400 in Figure 1. The electronic device 600 shown in Figure 3 includes at least one processor 530, a memory 550, and at least one network interface 520. The various components in the electronic device 600 are coupled together via a bus system 540. It is understood that the bus system 540 is configured to enable communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, all buses are labeled as bus system 540 in Figure 3.

[0070] Processor 530 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or any conventional processor.

[0071] The memory 550 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 550 may optionally include one or more storage devices physically located away from the processor 530.

[0072] The memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this application embodiment is intended to include any suitable type of memory.

[0073] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0074] Operating system 551 includes system programs configured to handle various basic system services and perform hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., configured to implement various basic business functions and handle hardware-based tasks;

[0075] The network communication module 552 is configured to reach other electronic devices via one or more (wired or wireless) network interfaces 520, such as Bluetooth, WiFi, and Universal Serial Bus.

[0076] In some embodiments, the three-dimensional blood vessel distribution construction apparatus provided in this application can be implemented in software. Figure 3 shows a three-dimensional blood vessel distribution construction apparatus 555 stored in memory 550, which can be software in the form of programs and plug-ins, including the following software modules: identification module 5551, generation module 5552, and construction module 5553. These modules are logically related and can therefore be arbitrarily combined or further split according to the functions they implement. The functions of each module will be described below.

[0077] In other embodiments, the image generation apparatus provided in this application can be implemented in hardware. As an example, the image generation apparatus provided in this application can be a processor in the form of a hardware decoding processor, which is programmed to execute the image generation method provided in this application. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0078] In some embodiments, the terminal or server can implement the image generation method provided in this application by running a computer program or computer-executable instructions. For example, the computer program can be a native program (e.g., a dedicated build program) or a software module in an operating system; for instance, it can be a build module embedded in any program (such as an instant messaging client, photo album program, electronic map client, or navigation client); or it can be a native application (APP), i.e., a program that needs to be installed in the operating system to run. In summary, the above-mentioned computer program can be any form of application, module, or plugin.

[0079] The image generation method provided in this application will be described in conjunction with exemplary applications and implementations of the server or terminal provided in the embodiments of this application.

[0080] Referring to Figure 4, which is a flowchart of the image generation method provided in this application embodiment, the method will be described in conjunction with steps 101 to 103 shown in Figure 4. The image generation method provided in this application embodiment can be implemented by the server or the terminal alone, or by the server and the terminal working together. The following description will take the implementation by the server alone as an example.

[0081] In step 101, a three-dimensional blood vessel distribution of a preset biological site is constructed.

[0082] In some embodiments, a predefined biological site refers to a specific anatomical region or organ predefined or designated in medical image processing, 3D modeling, or biomechanical analysis, where a 3D vascular distribution is used to represent the morphology of the blood vessels within that site. These sites typically have well-defined anatomical features and functions, such as the heart, liver, brain, and lungs. When constructing a 3D vascular distribution, the predefined biological site serves as the carrier of the vascular distribution, within which the vascular network branches, extends, and forms a complex topology. In medical image analysis, a predefined biological site refers to a specific body region or tissue of particular interest to researchers or physicians. This region may become the focus of research due to its pathological state, anatomical structure, or functional characteristics. In vascular imaging, a predefined biological site could be the heart, brain, palm, or any other organ or tissue containing blood vessels. Researchers may focus on these areas to diagnose diseases, plan surgeries, or conduct scientific research.

[0083] In some embodiments, a three-dimensional vascular distribution, also known as a three-dimensional vascular model or a three-dimensional vascular tree, is a computer-generated three-dimensional structure used to simulate and represent the vascular network of a specific location within a living organism (such as the human hand, head, or other parts). The three-dimensional vascular distribution represents the three-dimensional vascular distribution morphology of the predetermined biological location. By analyzing a first vascular image of the predetermined biological location, the two-dimensional coordinates of key vascular points in the image are identified. These key points can be vascular bifurcation points, endpoints, or specific locations on the vascular centerline. Using the identified two-dimensional key point coordinates, the main trunk of the vascular system is reconstructed in three-dimensional space using certain algorithms and modeling techniques. Based on the three-dimensional main trunk, branch vessels are further constructed, typically formed by extending and refining the main trunk. This ultimately forms a complete three-dimensional vascular network model with multiple levels and branches. The three-dimensional vascular distribution can reflect the spatial structure and morphology of the vessels in the predetermined biological location in detail.

[0084] In some embodiments, the above-mentioned construction of the three-dimensional blood vessel distribution of the preset biological site can be achieved by constructing the three-dimensional blood vessel distribution of the preset biological site based on a first blood vessel image of the preset biological site.

[0085] In some embodiments, a vascular image refers to an image acquired using medical imaging techniques (such as X-ray, CT scan, MRI, or ultrasound) that displays vascular structure and blood flow. These images can be used to observe the morphology, distribution, stenosis, or obstruction of blood vessels. A first vascular image of a preset biological site refers to an image showing the vascular structure acquired for a specific region of interest (preset biological site).

[0086] As an example, refer to Figure 11, which is a schematic diagram of the anatomical structure of the palm blood vessels and the palm vein image under near-infrared light provided in the embodiments of this application. The first blood vessel image of the aforementioned preset biological site can be the palm vein image under near-infrared light shown in Figure 11.

[0087] As an example, the palm or sole can be chosen as the preset biometric site because these sites typically have abundant vascular structures, making them easy to capture and analyze through images. Vascular images of the palm and sole have wide applications in medical diagnosis and biometrics, such as palmprint recognition and angiography. The blood vessels in the palm and sole are relatively superficial, making them easy to locate and identify through images. The vascular patterns in these sites are usually unique, providing excellent features for biometrics.

[0088] As an example, in the application scenario of palm blood vessel recognition, near-infrared imaging technology (such as NIR imaging) or optical coherence tomography (OCT) equipment is used to image the palm and obtain an image of the blood vessel distribution in the palm region. The palm is placed under the imaging device, ensuring that the preset biological parts (such as the central area of ​​the palm) are clearly visible. The device generates a first blood vessel image, which shows the blood vessel distribution in the palm region. The first blood vessel image (2D image) contains the vascular structure information of the palm region. Using the vascular information in the first blood vessel image, a three-dimensional blood vessel distribution model of the palm region is constructed. The first blood vessel image is preprocessed to extract the contour and branching information of the blood vessels. The two-dimensional blood vessel information is extended to three-dimensional space to generate a three-dimensional blood vessel distribution model. The three-dimensional blood vessel distribution model contains the branching, direction, and spatial distribution information of the blood vessels. To obtain blood vessel information from more perspectives, the three-dimensional blood vessel distribution is projected from multiple angles to generate multiple two-dimensional blood vessel distribution images. Multiple perspectives (such as front, side, oblique, etc.) are selected to project the three-dimensional blood vessel distribution, generating a two-dimensional blood vessel distribution image for each perspective. The two-dimensional blood vessel distribution image corresponds to the three-dimensional blood vessel distribution. Texture extraction and enhancement are performed on each 2D blood vessel distribution image to generate a clearer blood vessel texture image. The extracted blood vessel texture is then combined with the original 2D blood vessel distribution image to generate a new image.

[0089] In some embodiments, a 3D Vascular Tree is a three-dimensional model constructed based on a two-dimensional vascular image (such as a first vascular image) to represent the spatial distribution, branching structure, and topological relationships of blood vessels in a predetermined biological site (such as the palm, brain, or other organs). The 3D Vascular Tree is a vascular network model with a three-dimensional structure generated by extending vascular information from a two-dimensional vascular image into three-dimensional space. It can intuitively display the branching, direction, diameter variations, and spatial distribution of blood vessels.

[0090] In some embodiments, the first vascular image serves as the foundational data for constructing the three-dimensional vascular distribution, which is a three-dimensional representation of the vascular information in the first vascular image. The first vascular image is the data source for the three-dimensional vascular distribution, providing two-dimensional distribution information of the blood vessels. The three-dimensional vascular distribution, being a three-dimensional representation of the vascular information in the first vascular image, can more intuitively display the spatial distribution and branching structure of the blood vessels.

[0091] As an example, in the application scenario of palm blood vessel recognition, a two-dimensional image of the palm blood vessels acquired through near-infrared imaging displays the distribution of blood vessels in the palm region. Based on the first blood vessel image, a three-dimensional model of the palm blood vessels is generated using a three-dimensional reconstruction algorithm, showing the spatial distribution and branching structure of the blood vessels.

[0092] In some embodiments, the above-mentioned three-dimensional blood vessel distribution includes three-dimensional main blood vessels and three-dimensional branch blood vessels. Referring to Figure 5, Figure 5 is a schematic flowchart of the image generation method provided in the embodiments of this application. Step 101 shown in Figure 4 can be implemented by steps 1011 to 1013 shown in Figure 5.

[0093] In step 1011, the coordinates of multiple key points used to generate the three-dimensional trunk blood vessel are determined.

[0094] In some embodiments, key points refer to points with significant features in the first vascular image, such as bifurcation points, intersection points, endpoints, or curvature change points of the blood vessel. By identifying these key points, the topological structure of the blood vessel can be better described, providing basic data for subsequent 3D reconstruction and path planning.

[0095] In some embodiments, the determination of the coordinates of multiple key points for generating the three-dimensional trunk blood vessel can be achieved by: selecting one type as the target distribution type from multiple distribution types of blood vessel distribution in the preset biological site; and determining the coordinates of multiple key points for generating the three-dimensional trunk blood vessel based on the target distribution type.

[0096] In some embodiments, the vascular distribution of a pre-defined biological site (such as the liver, brain, etc.) may have multiple types, such as tree-like distribution, network distribution, and hierarchical distribution. Each distribution type reflects the topological structure and functional characteristics of the blood vessels. Depending on specific needs (such as surgical planning, disease diagnosis, etc.), one of the multiple distribution types is selected as the target distribution type. This step ensures that the generated three-dimensional trunk blood vessels conform to the actual anatomical structure and functional requirements. Based on the selected target distribution type, the coordinates of key points required to generate the three-dimensional trunk blood vessels are determined. These key points may include: Starting point: the origin of the trunk blood vessel (such as the heart exit, the origin of the aorta). Branching point: the location where the blood vessel bifurcates. Ending point: the end of the trunk blood vessel (such as the capillary inlet).

[0097] As an example, referring to Figure 12, from the various distribution types shown in Figure 12(a) to Figure 12(d), one type is selected as the target distribution type; based on the target distribution type, the coordinates of multiple key points used to generate the three-dimensional trunk blood vessel are determined.

[0098] In some embodiments, the coordinates of the plurality of key points are three-dimensional coordinates. The determination of the coordinates of the plurality of key points used to generate the three-dimensional trunk blood vessel according to the target distribution type can be achieved by: identifying the two-dimensional coordinates of the plurality of key points used to generate the three-dimensional trunk blood vessel from the image of the target distribution type; and converting the two-dimensional coordinates of the plurality of key points into three-dimensional coordinates.

[0099] In some embodiments, the target distribution type of the image is typically a two-dimensional medical image (such as a CT or MRI slice) or a two-dimensional projection image of a predefined biological site (such as the liver, brain, etc.). Key points for generating the three-dimensional main blood vessel are identified from the image. These key points include: Starting point: the origin of the main blood vessel; Branching point: the location where the blood vessel bifurcates; Ending point: the end of the main blood vessel.

[0100] As an example, the expression for the three-dimensional coordinates of the above key points can be:

[0101] Where, x 3d y 3d and z 3d Used to indicate the x, y, and vertices of a three-dimensional coordinate system. 2d and y 2d Used to indicate the x and y coordinates of a two-dimensional coordinate system. The function Uniform(40-D / 2, 40+D / 2) represents a uniformly distributed random number generator that generates random numbers within a specified interval. 40-D / 2 is the left boundary of the interval, and 40+D / 2 is the right boundary. D is a given parameter that determines the size of the interval.

[0102] Continuing the previous example, the two-dimensional coordinates include the x-coordinate and y-coordinate of the two-dimensional coordinates, and the three-dimensional coordinates include the x-coordinate, y-coordinate, and vertical coordinate of the three-dimensional coordinates; the conversion of the two-dimensional coordinates of the multiple key points into three-dimensional coordinates includes: determining the x-coordinate of the two-dimensional coordinates as the x-coordinate of the three-dimensional coordinates, and determining the y-coordinate of the two-dimensional coordinates as the y-coordinate of the three-dimensional coordinates; generating a random number within a preset interval using a random number generation function, and determining the random number as the vertical coordinate of the three-dimensional coordinates.

[0103] In some embodiments, taking the three-dimensional reconstruction of human blood vessels as an application background, the compositional relationship between two-dimensional and three-dimensional coordinates and the process of converting two-dimensional coordinates into three-dimensional coordinates are described in detail. This conversion process can map key point information in two-dimensional blood vessel images to three-dimensional space, providing basic coordinate data for the subsequent construction of three-dimensional blood vessel distribution. The coordinate compositional relationship is clarified: the two-dimensional coordinates are used to characterize the position of key points in planar space, containing parameters of two dimensions: horizontal and vertical coordinates. These two parameters uniquely determine the specific position of the key point in the two-dimensional plane. The three-dimensional coordinates are used to characterize the position of key points in three-dimensional space, adding a vertical coordinate dimension to the two-dimensional coordinates, i.e., containing parameters of three dimensions: horizontal, vertical, and vertical coordinates. These three parameters uniquely determine the specific position of the key point in three-dimensional space.

[0104] In some embodiments, the specific process of converting the two-dimensional coordinates of multiple key points into three-dimensional coordinates is as follows: First, a direct mapping of the horizontal and vertical coordinates is performed, directly determining the horizontal coordinate in the two-dimensional coordinates as the horizontal coordinate in the three-dimensional coordinates, and simultaneously determining the vertical coordinate in the two-dimensional coordinates as the vertical coordinate in the three-dimensional coordinates. This mapping method ensures that the planar projection position of the key points in the three-dimensional space is consistent with the original position in the two-dimensional plane, avoiding the offset of the relative planar position of the key points due to coordinate mapping, and ensuring the accurate correspondence between the subsequently constructed three-dimensional structure and the two-dimensional image.

[0105] In some embodiments, the vertical coordinate is generated and determined by generating random numbers within a preset interval using a random number generation function, and then directly determining these random numbers as the vertical coordinates of the three-dimensional coordinate system. The preset interval needs to be determined in conjunction with the physiological structural characteristics of the preset biological part. For example, for a human upper limb blood vessel scenario, the preset interval can be set according to the physiological thickness range of the upper limb, ensuring that the generated vertical coordinates can reasonably represent the depth position of key points in three-dimensional space and conform to the actual physiological structural scale of that part of the human body. The random number generation function is used to generate random numbers that conform to a uniform or normal distribution, so that the generated vertical coordinates have reasonable discrete characteristics, simulating the natural depth distribution of key points of blood vessels in three-dimensional space, avoiding the vertical coordinate values ​​from being too concentrated, which would cause the subsequently constructed three-dimensional blood vessel structure to be too flat and unable to realistically reproduce the three-dimensional spatial morphology of the blood vessels. Through the above-described horizontal and vertical coordinate mapping and vertical coordinate generation method, the conversion from two-dimensional coordinates to three-dimensional coordinates of a single key point can be completed. By performing the above processing on all key points in sequence, the three-dimensional coordinates corresponding to all key points can be obtained. These three-dimensional coordinates can be directly used in the subsequent generation of three-dimensional main blood vessels and the construction of three-dimensional blood vessel distribution, providing accurate spatial location data support for three-dimensional blood vessel reconstruction.

[0106] In step 1012, the three-dimensional main blood vessel is generated based on the coordinates of the multiple key points.

[0107] In some embodiments, the three-dimensional trunk vessel is a vascular network model generated based on key point information from two-dimensional vascular images. It displays the distribution, branching structure, and topological relationships of blood vessels in three-dimensional space. The three-dimensional trunk vessel can intuitively show the direction and distribution of blood vessels in three-dimensional space. It clearly shows the bifurcation points and connections of blood vessels, reflecting the complexity of the vascular network. The three-dimensional trunk vessel preserves the connection methods between blood vessels and the direction of blood flow, facilitating the analysis of the functional characteristics of blood vessels.

[0108] In some embodiments, the two-dimensional coordinates of a keypoint refer to the position of a point with significant features (such as bifurcation points, intersections, or endpoints) extracted from a two-dimensional vascular image on the image plane, typically denoted by (x, y). Keypoints are usually bifurcation points, intersections, or endpoints of blood vessels, capable of describing the topological structure of the blood vessels. The coordinates of keypoints only contain planar information (x, y) and lack depth information (z).

[0109] As an example, see Figure 13. In the application scenario of palm blood vessel recognition, based on the two-dimensional coordinates of the multiple key points, the three-dimensional main blood vessel corresponding to the blood vessel is generated (as shown in Figure 13, the main distribution of palm blood vessels).

[0110] In some embodiments, the coordinates of the plurality of key points are three-dimensional coordinates, and step 1012 can be implemented as follows: for the three-dimensional coordinates of the plurality of key points, the different three-dimensional coordinates are connected to obtain the initial three-dimensional trunk blood vessel of the blood vessel, the initial three-dimensional trunk blood vessel includes a plurality of blood vessel segments, the blood vessel segments are with the key points as endpoints; the radius of each blood vessel segment in the initial three-dimensional trunk blood vessel is determined to obtain the three-dimensional trunk blood vessel.

[0111] In some embodiments, the two-dimensional coordinates (x, y) of keypoints are extended to three-dimensional coordinates (x, y, z), which can be estimated using the following methods: Multi-view images: Calculating depth values ​​using multi-view images (e.g., stereo vision). Prior knowledge: Estimating depth values ​​based on the anatomical structure or statistical model of blood vessels. Deep learning: Predicting depth information using a deep learning model. Combining the two-dimensional coordinates (x, y) with the estimated depth value (z) generates the three-dimensional coordinates (x, y, z). Three-dimensional coordinates of multiple keypoints, such as: Keypoint 1: (x1, y1, z1); Keypoint 2: (x2, y2, z2); Keypoint 3: (x3, y3, z3).

[0112] As an example, key point 1: (x1, y1), key point 2: (x2, y2), key point 3: (x3, y3); Processing: estimate depth values ​​and generate 3D coordinates: key point 1: (x1, y1, z1); key point 2: (x2, y2, z2); key point 3: (x3, y3, z3).

[0113] Thus, by connecting different 3D coordinates based on the type of keypoints, an initial 3D main blood vessel can be constructed. This main vessel consists of multiple vessel segments with keypoints as endpoints. By adjusting the radius of each vessel segment based on its parameters (such as length and curvature), a more realistic and accurate 3D main blood vessel can be generated. This not only improves the geometric accuracy of the blood vessel model but also enhances the biological plausibility of the blood vessel structure.

[0114] In some embodiments, the vascular segments in the initial three-dimensional trunk blood vessel have a flow direction; the determination of the radius of each vascular segment in the initial three-dimensional trunk blood vessel to obtain the three-dimensional trunk blood vessel can be achieved by performing the following processing on each vascular segment to obtain the three-dimensional trunk blood vessel: obtaining the initial radius of the vascular segment in the initial three-dimensional trunk blood vessel, and the number of terminal key points flowing out of the vascular segment according to the flow direction; determining the radius of the vascular segment based on the initial radius and the number, and obtaining the three-dimensional trunk blood vessel.

[0115] In some embodiments, the terminal keypoints are keypoints in the initial 3D trunk vessel that are not connected to the root node of the initial 3D trunk vessel. The initial radius refers to the initial radius value of the vessel segment in the initial 3D trunk vessel. It is usually initially set based on the length, position, or branch level of the vessel segment. The initial radius is the initial value before adjustment, which may be set based on simple rules or default values. The initial radius is a basic parameter, serving as the starting point for adjusting the radius of the vessel segment. The radius refers to the final radius value of the vessel segment in the adjusted 3D trunk vessel. It is calculated based on the initial radius and the number of keypoints connected to the endpoints of the vessel segment. The radius is an optimization result of the initial radius, which better conforms to the morphology of real blood vessels. A vessel segment refers to the portion of the blood vessel between two keypoints in the 3D trunk vessel. It is the basic building block of the trunk vessel. The endpoints are the two endpoints of the vessel segment, which are two keypoints respectively. The parameters of the vessel segment include length, curvature, radius, etc.

[0116] In some embodiments, the number of terminal keypoints connected to the endpoints of a vessel segment refers to the number of other terminal keypoints connected to both endpoints of the vessel segment. For example, if the endpoint of a vessel segment is a bifurcation point, there are more connected keypoints. If the endpoint of a vessel segment is a single endpoint, there are fewer connected keypoints. The number of terminal keypoints connected to the endpoints of a vessel segment can reflect the complexity of the branching: the more connected keypoints, the more complex the branching at that endpoint. Impact on radius adjustment: The number of connected keypoints is an important basis for adjusting the radius of a vessel segment.

[0117] As an example, if the endpoints of a vessel segment connect to a large number of critical points (such as bifurcation points), and the vessels at bifurcation points are typically thinner, then the radius of the vessel segment should be reduced. If the endpoints of a vessel segment connect to a small number of critical points (such as the endpoints), and since the main vessels are typically thicker, then the radius of the vessel segment should be maintained or appropriately increased.

[0118] As an example, the target radius is calculated using a formula or rule based on the initial radius and the number of connected keypoints. For example: if the number of connected keypoints > 2, then the target radius = initial radius × 0.8. If the number of connected keypoints ≤ 2, then the target radius = initial radius × 1.2. The radius of the vessel segment is adjusted from the initial radius to the target radius. The adjusted vessel segment is then smoothed to ensure the continuity and naturalness of the main vessel. The adjusted 3D main vessel contains the updated vessel segment radii.

[0119] As an example, vessel segment 1: initial radius: 2.0 mm; number of keypoints connecting the endpoints: 3; vessel segment 2: initial radius: 2.0 mm; number of keypoints connecting the endpoints: 1. Calculate the target radius: vessel segment 1: target radius; 2.0 × 0.8 = 1.6 mm; vessel segment 2: target radius = 2.0 × 1.2 = 2.4 mm. The radius of vessel segment 1 is adjusted from 2.0 mm to 1.6 mm. The radius of vessel segment 2 is adjusted from 2.0 mm to 2.4 mm. Output: The adjusted 3D main vessel. By adjusting the vessel segment radii, the 3D main vessel is made to better resemble the morphology of a real blood vessel.

[0120] Thus, by adjusting the radii of vessel segments in the initial 3D trunk vessel based on initial vessel segment parameters (such as the initial radius and the number of keypoints connecting the endpoints), the accuracy and realism of the 3D trunk vessel can be significantly improved. For each vessel segment, a target radius that better reflects the actual morphology of the vessel is calculated based on its initial radius and the number of keypoints connecting the endpoints, and the radius of the vessel segment is adjusted from the initial radius to the target radius. This adjustment process not only better reflects the branching complexity and hemodynamic characteristics of the vessel, but also makes the 3D trunk vessel closer to the actual anatomical structure, thereby providing more reliable basic data for medical diagnosis, surgical planning, and scientific research analysis.

[0121] In some embodiments, determining the radius of the blood vessel segment based on the initial radius and the quantity can be achieved by: obtaining the weight of the blood vessel segment, the weight being used to indicate the importance of the blood vessel segment; multiplying the quantity by the weight to obtain a multiplication result, and adding the multiplication result to the initial radius of the blood vessel segment to obtain the radius of the blood vessel segment.

[0122] In some embodiments, weights are parameters used to indicate the importance of a vessel segment within the initial three-dimensional trunk vessel. Weights can be determined based on factors such as: branching level (main vessels have higher weights, branch vessels have lower weights), length (longer segments have higher weights), and curvature (segments with lower curvature have higher weights). Weights are calculated using predefined rules or algorithms. For example, the weight of a main vessel might be 1.0, a first-order branch vessel 0.8, and a second-order branch vessel 0.6.

[0123] As an example, the expression for the radius of the above-mentioned blood vessel segment can be: r i =r roi +n i *ratioE (2)

[0124] Where, r i Used to indicate the radius of a blood vessel segment, r roi Used to indicate the initial radius, n iThe ratioE is used to indicate the weight, while the ratioE is used to indicate the quantity.

[0125] Thus, by acquiring the weights of vascular segments and determining the target radius based on the initial radius and quantity, the modeling accuracy and realism of 3D trunk vessels can be significantly improved. Weights, as parameters reflecting the importance of vascular segments in the initial 3D trunk vessel, allow for more detailed adjustment of the segment radius, making it more consistent with actual vascular anatomy. The introduction of weight parameters allows for personalized radius adjustments for each vascular segment, considering factors such as branching level, length, and curvature, thereby more accurately reflecting the morphology of the vessel. By multiplying the quantity by the weight and adding the initial radius, the resulting target radius is closer to the diameter of the actual vessel, improving the modeling accuracy of 3D trunk vessels. The use of weights helps to better simulate hemodynamic characteristics, providing more valuable information for the diagnosis and treatment of vascular diseases.

[0126] In some embodiments, the types of key points mentioned above include trunk type, bifurcation type, and leaf type. The initial three-dimensional trunk vessel of the blood vessel is obtained by connecting the different three-dimensional coordinates of the multiple key points as follows: connecting the three-dimensional coordinates of the key points of each trunk type to obtain a first trunk vessel; connecting the three-dimensional coordinates of the key points of each bifurcation type with the three-dimensional coordinates of at least one key point of a leaf type to obtain a second trunk vessel corresponding to each key point of the bifurcation type; and connecting the first trunk vessel with each of the second trunk vessels to obtain the initial three-dimensional trunk vessel.

[0127] In some embodiments, the trunk type refers to the main vessels in a vascular network, vessels with a diameter greater than a diameter threshold and a number of branches less than a branch number threshold. In 3D vascular reconstruction, the key points of the trunk type are typically the center points connecting multiple branches. These key points play a connecting and supporting role in the vascular network and are the core of the vascular structure.

[0128] In some embodiments, a bifurcation type refers to a key point at a branching point in a vascular network, which is the starting point where a vessel branches off from its trunk. Key points of a bifurcation type typically connect two or more vessel segments, reflecting the complexity of the vascular network.

[0129] In some embodiments, a leaf type refers to a terminal vessel in a vascular network, typically the most distal endpoint of a branching vessel that branches no further. The key feature of a leaf type is that it is usually small in diameter and represents the terminal structure of the vascular network.

[0130] In some embodiments, key points for all trunk types are identified. The 3D coordinates of these key points are connected to form a first vascular trunk. Key points for all bifurcation and leaf types are identified. The 3D coordinates of each bifurcation type key point are connected to the 3D coordinates of at least one leaf type key point to form a second vascular trunk. The first vascular trunk is then connected to each of the second vascular trunks to obtain a complete initial 3D vascular trunk.

[0131] As an example, key points for trunk type: A, B, C; key points for bifurcation type: D, E; key points for leaf type: F, G, H. Connecting trunk type key points: A→B→C; connecting bifurcation and leaf type key points: D→FE→GE→H. Connecting the first and second main vascular trunks: A→D→FA→D→HB→E→GB→E→H. The initial three-dimensional main vascular system includes trunks, branches, and leaf vessels. By distinguishing the key points of trunk, bifurcation, and leaf types, the structure of the vascular network can be clearly constructed. Connection methods based on key point types can more accurately reflect anatomical characteristics.

[0132] As an example, suppose there is a vascular network containing the following keypoints: trunk type keypoints: A, B, C; bifurcation type keypoints: D, E; leaf type keypoints: F, G, H. Connecting the trunk type keypoints, all trunk type keypoints are connected to form the first vascular trunk. This process can be represented as: A→B→C. Connecting the bifurcation and leaf type keypoints, next, each bifurcation type keypoint is connected to at least one leaf type keypoint to form the second vascular trunk. This process can be represented as: D→FD→HE→G. D and E are bifurcation type keypoints, which are connected to leaf type keypoints F, H, and G, respectively. Connecting the first and second vascular trunks, finally, the first vascular trunk is connected to each second vascular trunk to obtain the complete initial three-dimensional vascular trunk. This process can be represented as: A→D→FA→D→HB→E→G.

[0133] Thus, by classifying keypoint types and connecting 3D coordinates based on these types, the initial 3D structure of blood vessels can be effectively constructed. Distinguishing between trunk, branch, and leaf types of keypoints clearly represents the structural hierarchy of blood vessels, aiding in the understanding and analysis of their anatomical characteristics. A connection strategy based on keypoint types ensures that the vascular model more accurately reflects the actual vascular network. Automated connection based on keypoint type significantly improves the efficiency of vascular reconstruction and reduces the need for manual intervention. Adaptable to vascular networks of varying complexity, from simple linear vessels to complex branching structures, keypoint types can be effectively modeled.

[0134] In step 1013, the three-dimensional branch vessels are generated based on the three-dimensional trunk vessels to obtain the three-dimensional vessel distribution of the preset biological site.

[0135] In some embodiments, a three-dimensional trunk vessel refers to a main vascular channel in three-dimensional space, formed by a series of key points connected together. It is the core part of the vascular network, typically with a large diameter and few branches, responsible for delivering blood to all parts of the body. The three-dimensional trunk vessel is usually located at the center of the vascular network and is the main path for blood flow. Due to its large diameter, the three-dimensional trunk vessel is relatively stable and not easily affected by external factors. The three-dimensional trunk vessel connects with other vessels (such as branch vessels) to form a complex vascular network.

[0136] In some embodiments, a three-dimensional trunk vessel refers to a main vascular channel in three-dimensional space, formed by a series of key points connected together. It is the core part of the vascular network, typically with a large diameter and few branches, responsible for delivering blood to all parts of the body. The three-dimensional trunk vessel is usually located at the center of the vascular network and is the main path for blood flow. Due to its large diameter, the three-dimensional trunk vessel is relatively stable and less susceptible to external factors.

[0137] In some embodiments, branch vessels refer to smaller vessels that branch off from the main vessels in a three-dimensional vascular network, responsible for delivering blood to more specific parts of the body. Branch vessels are typically smaller in diameter than the main vessels, making them more flexible and adaptable to the complex internal environment of the body. Branch vessels usually appear at the bifurcation points of the main vessels, contributing to the complexity of the vascular network. Branch vessels have diverse functions, some responsible for nutrient supply, others for oxygen delivery, and so on.

[0138] In some embodiments, the three-dimensional vascular distribution refers to the overall vascular network structure composed of three-dimensional trunk vessels and their branches. It is similar to the structure of a tree in nature, exhibiting a hierarchical structure with branches and sub-branches. The three-dimensional vascular distribution includes all vessels in the vascular network, from the starting point of the trunk to the end of the branch vessels. The three-dimensional vascular distribution is not only an anatomical structure but also a functional network responsible for blood transport and distribution.

[0139] In some embodiments, step 1013 above can be implemented as follows: selecting multiple three-dimensional coordinate points that are not on the three-dimensional main blood vessel; constructing the branch blood vessels on the three-dimensional main blood vessel based on the multiple three-dimensional coordinate points to obtain the initial three-dimensional blood vessel distribution of the preset biological site; adjusting the shape of each three-dimensional branch blood vessel in the initial three-dimensional blood vessel distribution to obtain the three-dimensional blood vessel distribution of the preset biological site.

[0140] In some embodiments, multiple 3D coordinate points not located on the 3D trunk vessel are selected in the 3D coordinate system; these points will be used to construct 3D branch vessels. Based on the selected 3D coordinate points, 3D branch vessels are constructed on the 3D trunk vessel. Points on the 3D trunk vessel closest to the selected 3D coordinate points are found; these points will become the connection points for the 3D branch vessels. The shapes of each branch vessel in the initial 3D vascular distribution are adjusted to obtain a 3D vascular distribution that better conforms to the actual anatomical structure. If the direction of a branch vessel does not conform to the actual anatomical structure, its direction can be adjusted by changing the position of the connection point or the curvature of the branch. The radius of the branch vessels is adjusted according to their function and blood flow requirements to better conform to physiological conditions. A smooth transition is ensured at the connection points between the branches and the trunk vessel, avoiding abrupt corners or unnatural shapes. A 3D vascular distribution containing the 3D trunk vessel and branch vessels is obtained. This vascular distribution model more accurately reflects the anatomical structure and functional characteristics of the vascular network.

[0141] In some embodiments, the initial three-dimensional vascular distribution refers to the vascular network model formed by selecting multiple three-dimensional coordinate points not located on the three-dimensional trunk vessels in the three-dimensional coordinate system, and constructing branch vessels on the three-dimensional trunk vessels. This model includes the three-dimensional trunk vessels and their branch vessels, and is an intermediate product in the three-dimensional vascular reconstruction process. It consists of the three-dimensional trunk vessels and branch vessels constructed based on the three-dimensional coordinate points. All operations are performed in the three-dimensional coordinate system, ensuring the model's three-dimensionality and accuracy. The initial three-dimensional vascular distribution is a stage in the vascular reconstruction process and usually requires adjustment and optimization to achieve the final three-dimensional vascular distribution model. By adjusting the shape of the branch vessels in the initial three-dimensional vascular distribution, the model can better conform to the actual vascular anatomy and functional requirements.

[0142] As an example, taking the arteries of the human lower limbs as the preset biological site, this paper describes in detail the process of selecting three-dimensional coordinate points, constructing branch vessels to obtain the initial three-dimensional vascular distribution, and adjusting the shape of the branch vessels to obtain the final three-dimensional vascular distribution for the application scenario of three-dimensional image reconstruction in the diagnosis of lower limb arteriosclerosis obliterans. This process can accurately restore the main trunk and branch structure of the lower limb arteries, providing reliable vascular morphology data support for lesion localization, stenosis assessment and treatment planning.

[0143] Continuing from the previous example, when selecting multiple 3D coordinate points not on the main 3D blood vessels, the lower limb region of the target object is first scanned using a lower limb computed tomography (CT) angiography device to obtain raw vascular image data. After preliminary processing, the 3D main blood vessels of the lower limb arteries are reconstructed. This 3D main blood vessel includes the spatial orientation and key endpoint positions of core main blood vessels such as the femoral artery and popliteal artery. Combining the physiological and anatomical structure of the human lower limb arteries, feature points of branch vessels not on the aforementioned 3D main blood vessels are extracted from the raw image data as 3D coordinate points. The selected 3D coordinate points correspond to key positions such as the starting point, bifurcation, and distal endpoint of branch vessels such as the anterior tibial artery, posterior tibial artery, and peroneal artery. Each 3D coordinate point is converted from image pixel coordinates to spatial coordinates to obtain accurate spatial 3D coordinates, ensuring that each coordinate point can accurately represent the true physiological position of the corresponding branch vessel.

[0144] Continuing the previous example, branch vessels are constructed on the three-dimensional main blood vessel based on the aforementioned multiple three-dimensional coordinate points to obtain an initial three-dimensional vascular distribution. The specific process is as follows: For each three-dimensional coordinate point, the closest vessel segment on the three-dimensional main blood vessel to that coordinate point is determined through spatial distance calculation as the attachment segment of the branch vessel. The initial attachment point of the branch vessel is then determined on this attachment segment, and this attachment point must conform to the physiological origin of lower limb arterial branches. Based on the spatial relationship between the three-dimensional coordinate points and the initial attachment point, a vessel segment connecting the initial attachment point and the corresponding three-dimensional coordinate point is constructed. During the construction process, the diameter of the branch vessels is set with reference to the diameter parameters of normal lower limb arteries in the same region to ensure that the diameter of the branch vessels matches the main blood vessel in accordance with physiological laws. All constructed branch vessels are integrated with the initial three-dimensional main blood vessel to obtain an initial three-dimensional vascular distribution of the preset biological site, including the main blood vessel and the preliminary shapes of each branch vessel.

[0145] Continuing from the previous example, the shapes of the three-dimensional branch vessels in the initial three-dimensional vascular distribution are adjusted to obtain the final three-dimensional vascular distribution. The specific processing method is as follows: For each branch vessel in the initial three-dimensional vascular distribution, considering the natural curvature of the lower limb arteries, several auxiliary three-dimensional coordinate points are selected on one or both sides of the branch vessel. The trajectory formed by the selected auxiliary coordinate points conforms to the curvature of the actual branch vessel. The number of auxiliary coordinate points is determined according to the length of the branch vessel; longer branches use 5 to 7 auxiliary coordinate points, and shorter branches use 2 to 3. The spacing between each auxiliary coordinate point is uniform, and the distance from the branch vessel is controlled within a preset range to avoid excessive distortion or deviation from the physiological position of the adjusted vascular shape. The two endpoints of the branch vessel are then connected to the selected auxiliary three-dimensional coordinate points sequentially along the physiological direction using a smooth curve fitting technique to form the adjusted branch vessel. Compared to the initially constructed straight or zigzag branch vessels, the adjusted branch vessel better conforms to the natural physiological morphology of the lower limb arteries. After all branch vessels have completed shape adjustment, the adjusted branch vessels are integrated with the three-dimensional main vessels to obtain a three-dimensional vascular distribution of the preset biological location that can accurately reproduce the real spatial morphology and branch structure of the human lower limb arteries. This three-dimensional vascular distribution can be directly used in clinical scenarios such as lesion location of lower limb arteriosclerosis obliterans, measurement of vascular stenosis degree and interventional treatment path planning.

[0146] Thus, by selecting multiple 3D coordinate points not on the main trunk within the 3D coordinate system, and constructing branch vessels based on these points, an initial 3D vascular distribution can be generated. Adjusting the shape of each branch vessel in the initial 3D vascular distribution can optimize the geometric structure and biological plausibility of the vessels, resulting in a more realistic and accurate 3D vascular distribution, thus improving the integrity and detail of the vascular model.

[0147] In this way, by precisely selecting coordinate points, branch vessels that correspond to the actual vascular network can be constructed on the three-dimensional main blood vessel, resulting in a complete vascular distribution model that includes the main trunk. Subsequently, adjusting the shape of the branch vessels not only optimizes the morphology of the vascular model, making it more realistic, but also enhances the model's practicality and accuracy in medical diagnosis, surgical planning, and scientific research analysis. It also improves the efficiency of vascular reconstruction, reduces the need for manual intervention, and provides strong support for clinical decision-making.

[0148] In some embodiments, the multiple three-dimensional coordinate points include the i-th three-dimensional coordinate point, where 1 < i ≤ N. Based on the multiple three-dimensional coordinate points, constructing the branch vessels on the three-dimensional main vessel to obtain the initial three-dimensional vessel distribution of the preset biological part can be achieved as follows: determining the three-dimensional main vessel as the first three-dimensional vessel distribution, determining the vessel segment on the (i - 1)-th three-dimensional vessel distribution that is closest to the i-th three-dimensional coordinate point as the i-th vessel segment, and generating the i-th three-dimensional vessel distribution based on the i-th vessel segment; traversing i to obtain the N-th three-dimensional vessel distribution, and determining the N-th three-dimensional vessel distribution as the initial three-dimensional vessel distribution.

[0149] In some embodiments, the first three-dimensional vessel distribution refers to the already constructed three-dimensional main vessel, which serves as the starting point for the subsequent construction of branch vessels. The i-th coordinate point is determined in the three-dimensional coordinate system and serves as the basis for constructing the branch vessel. On the (i - 1)-th three-dimensional vessel distribution, the vessel segment closest to the i-th three-dimensional coordinate point is found. This vessel segment will serve as the basis for constructing the i-th branch vessel. The newly constructed branch vessel is integrated with the (i - 1)-th three-dimensional vessel distribution to form the i-th three-dimensional vessel distribution. The above process is repeated for all three-dimensional coordinate points (i.e., i ranges from 2 to N), and each time a new branch vessel is constructed based on the previous vessel distribution. After the traversal is completed, the N-th three-dimensional vessel distribution containing all branch vessels is obtained. The N-th three-dimensional vessel distribution is determined as the initial three-dimensional vessel distribution of the object.

[0150] In some embodiments, the multiple three-dimensional coordinate points are brain branch vessel feature points obtained by pre-scanning with a brain magnetic resonance angiography device and processed through data. Each three-dimensional coordinate point corresponds to a key position of a real brain branch vessel. The multiple three-dimensional coordinate points include the i-th three-dimensional coordinate point, where i is a positive integer greater than 1 and less than or equal to N, and N is the total number of three-dimensional coordinate points. That is, the three-dimensional coordinate points are sequentially numbered as the first three-dimensional coordinate point to the N-th three-dimensional coordinate point in a preset order, and the numbering order can be determined in combination with the physiological distribution law of the brain vessels, for example, numbered sequentially from the brain main vessel to the distal branch vessels.

[0151] In some embodiments, the specific process of constructing branch vessels on the three-dimensional main blood vessels based on the above-mentioned multiple three-dimensional coordinate points to obtain an initial three-dimensional blood vessel distribution is as follows: First, the pre-reconstructed three-dimensional main blood vessels of the brain are determined as the first three-dimensional blood vessel distribution. For the i-th three-dimensional coordinate point, the (i-1)-th three-dimensional blood vessel distribution is first obtained. The (i-1)-th three-dimensional blood vessel distribution is the blood vessel distribution morphology after the construction of the branch vessels corresponding to the (i-1)-th three-dimensional coordinate point is completed. It includes the initial three-dimensional main blood vessels and the branch vessel structures corresponding to the first (i-1)-th three-dimensional coordinate points. Based on the spatial distance calculation method, all blood vessel segments in the (i-1)-th three-dimensional blood vessel distribution are traversed, and the spatial distance between each blood vessel segment and the i-th three-dimensional coordinate point is calculated. The blood vessel segment with the smallest spatial distance is determined as the i-th blood vessel segment. This i-th blood vessel segment is the attachment blood vessel segment for the subsequent construction of the branch vessels corresponding to the i-th three-dimensional coordinate point.

[0152] In some embodiments, the i-th three-dimensional vascular distribution is generated based on the i-th vascular segment. Specifically, the starting position of the branch vessel is determined on the i-th vascular segment according to the spatial position of the i-th three-dimensional coordinate point and the physiological morphological characteristics of the cerebral blood vessels. The starting position is the point on the i-th vascular segment that is closest to the i-th three-dimensional coordinate point. Then, a branch vessel extending from the starting position to the i-th three-dimensional coordinate point is constructed. During the construction process, the diameter and curvature of the branch vessel are ensured to conform to the physiological parameters of the branch vessels in the corresponding brain region. The branch vessel is then integrated into the (i-1)-th three-dimensional vascular distribution to obtain the i-th three-dimensional vascular distribution.

[0153] In some embodiments, i is traversed in the manner described above, that is, the operations of determining the i-th vascular segment and generating the i-th three-dimensional vascular distribution are performed sequentially on the second to the Nth three-dimensional coordinate points until the construction of the branch vessels corresponding to the Nth three-dimensional coordinate point is completed, resulting in the Nth three-dimensional vascular distribution. At this point, the Nth three-dimensional vascular distribution includes the initial three-dimensional trunk vessels and all the branch vessels corresponding to the multiple three-dimensional coordinate points, and each branch vessel is precisely attached to its corresponding vascular segment, which conforms to the natural branching structure of cerebral vessels. Therefore, the Nth three-dimensional vascular distribution is determined as the initial three-dimensional vascular distribution of the preset biological site.

[0154] As an example, this section illustrates how to construct branch vessels on a three-dimensional trunk vessel based on multiple three-dimensional coordinate points to obtain an initial three-dimensional vessel distribution: Assume there is a three-dimensional trunk vessel that has been constructed using some method (such as medical imaging data), and the following three-dimensional coordinate points will be used to construct branch vessels: 1st three-dimensional coordinate point: P1; 2nd three-dimensional coordinate point: P2; 3rd three-dimensional coordinate point: P3; Nth three-dimensional coordinate point: PN. For the 2nd three-dimensional coordinate point P2, find the vessel segment closest to P2 on T1, denoted as S2. For the 3rd three-dimensional coordinate point P3, find the vessel segment closest to P3 on the second three-dimensional vessel distribution T2 constructed based on T1 and S2, denoted as S3. And so on, for each subsequent three-dimensional coordinate point Pi, we find the vessel segment Si closest to Pi on the previous three-dimensional vessel distribution Ti. Based on S2, construct a new vessel segment from the end of S2, connecting it to P2, forming the second three-dimensional vessel distribution T2. Based on S3, a new vascular segment is constructed from the end of S3 and connected to P3, forming the third three-dimensional vascular distribution T3. For each Pi, a new vascular segment is constructed based on Si and connected to Pi, forming the i-th three-dimensional vascular distribution Ti. This process continues until all three-dimensional coordinate points have been processed. After processing all coordinate points, we obtain the N-th three-dimensional vascular distribution TN. TN is then defined as the initial three-dimensional vascular distribution.

[0155] As an example, T1 (the first three-dimensional vascular distribution) already exists. Find the vascular segment S2 closest to P2, construct a branch vessel connecting it to P2, and obtain T2 (the second three-dimensional vascular distribution). Find the vascular segment S3 closest to P3 (within T2), construct a branch vessel connecting it to P3, and obtain T3 (the third three-dimensional vascular distribution). Repeat this process until the last coordinate point PN, obtaining TN (the Nth three-dimensional vascular distribution). TN is the desired initial three-dimensional vascular distribution, which includes the three-dimensional main vessels and all branch vessels constructed based on the three-dimensional coordinate points.

[0156] This ensures that each branch vessel is closely connected to its corresponding vessel segment, thereby improving the accuracy of the vascular model and the realism of the anatomical structure. By traversing all coordinate points and gradually constructing the vessel distribution, complex vascular networks can be effectively handled, improving the comprehensiveness and systematicness of the modeling. The incremental construction method also helps to optimize the use of computing resources, as it avoids the computational burden of processing the entire vascular network at once.

[0157] In some embodiments, the above-mentioned generation of the i-th three-dimensional blood vessel distribution based on the i-th blood vessel segment can be achieved by: determining the region formed by the endpoint of the i-th blood vessel segment and the i-th three-dimensional coordinate point, and determining the bifurcation point of the i-th three-dimensional blood vessel distribution in the region; deleting the i-th blood vessel segment from the (i-1)-th three-dimensional blood vessel distribution, and connecting the endpoint of the i-th blood vessel segment, the i-th three-dimensional coordinate point, and the bifurcation point to obtain the i-th three-dimensional blood vessel distribution.

[0158] In some embodiments, the vascular segment closest to the i-th 3D coordinate point on the (i-1)-th 3D vascular distribution is identified, i.e., the i-th vascular segment. The endpoint of the i-th vascular segment and the region formed by this endpoint and the i-th 3D coordinate point are determined. This region forms the basis for subsequent determination of the bifurcation point. The bifurcation point is determined within the region formed by the i-th vascular segment and the i-th 3D coordinate point. The bifurcation point is the starting point of the new branch vessel and is typically located near the endpoint of the i-th vascular segment. The i-th vascular segment is deleted from the (i-1)-th 3D vascular distribution to prepare for the construction of the new branch vessel. The endpoint of the i-th vascular segment, the i-th 3D coordinate point, and the bifurcation point are connected. This process typically involves creating a new vascular segment in 3D space that starts from the endpoint of the i-th vascular segment, passes through the bifurcation point, and finally connects to the i-th 3D coordinate point. Through the above connection, a new branch vessel is formed, which will be integrated into the i-th 3D vascular distribution.

[0159] In some embodiments, connecting the endpoint of the i-th vascular segment, the i-th three-dimensional coordinate point, and the bifurcation point means connecting the endpoint of the i-th vascular segment on the (i-1)-th three-dimensional vascular distribution to the i-th three-dimensional coordinate point, connecting the endpoint of the i-th vascular segment to the bifurcation point, and connecting the i-th three-dimensional coordinate point and the bifurcation point.

[0160] In some embodiments, the coordinates of the bifurcation point include the x-coordinate, y-coordinate, and d-coordinate of the bifurcation point. The determination of the bifurcation point of the i-th three-dimensional blood vessel distribution in the region can be achieved as follows: the x-coordinate of the bifurcation point is determined based on the x-coordinate of the endpoint of the i-th blood vessel segment and the x-coordinate of the i-th three-dimensional coordinate point; the y-coordinate of the bifurcation point is determined based on the y-coordinate of the endpoint of the i-th blood vessel segment and the y-coordinate of the i-th three-dimensional coordinate point; and the d-coordinate of the bifurcation point is determined based on the d-coordinate of the endpoint of the i-th blood vessel segment and the d-coordinate of the i-th three-dimensional coordinate point.

[0161] As an example, the x-coordinate of the bifurcation point p above can be:

[0162] Where, x i The x-coordinates of the endpoints of the i-th blood vessel segment and the x-coordinates of the i-th three-dimensional coordinate point are used to indicate the coordinates of the endpoints. i ,ri Segmentation of blood vessels i Length and radius.

[0163] As an example, the ordinate of the bifurcation point p above can be:

[0164] Among them, y i The ordinates used to indicate the endpoint of the i-th blood vessel segment and the ordinates of the i-th three-dimensional coordinate point, l i ,r i Segmentation of blood vessels i Length and radius.

[0165] As an example, the vertical coordinate of the bifurcation point p above can be:

[0166] Among them, z i The vertical coordinates of the endpoint of the i-th blood vessel segment and the vertical coordinates of the i-th three-dimensional coordinate point are used to indicate the vertical coordinates of the endpoints. i ,r i Segmentation of blood vessels i Length and radius.

[0167] As an example, referring to Figure 14, the endpoints of the i-th blood vessel segment are P0 and P1, and the i-th three-dimensional coordinate point is P2. i Let s0, s1, and s2 be the connections between P0P, s1 between P1P, and s2 between P2P.

[0168] As an example, referring to Figure 14, the region (P0P1P2) formed by the endpoints (P0 and P1) of the i-th blood vessel segment and the i-th three-dimensional coordinate point (P2) is determined, and the bifurcation point (p) of the i-th three-dimensional blood vessel distribution is determined in the region (P0P1P2); the i-th blood vessel segment is deleted from the (i-1)-th three-dimensional blood vessel distribution, and the endpoints (P0 and P1) of the i-th blood vessel segment, the i-th three-dimensional coordinate point (P2) and the bifurcation point (p) are connected to obtain the i-th three-dimensional blood vessel distribution.

[0169] Thus, by generating the i-th three-dimensional vascular distribution based on the i-th vascular segment, the region formed by the endpoints of the vascular segment and the i-th three-dimensional coordinate point is determined, thereby accurately determining the bifurcation point. This not only ensures a smooth transition between branch vessels and main vessels, improving the anatomical accuracy and realism of the model, but also optimizes the structure of the vascular distribution by deleting the original vascular segment and constructing new connections, making it more consistent with the actual vascular network characteristics.

[0170] In some embodiments, the above-mentioned adjustment of the shape of each branch vessel in the initial three-dimensional vascular distribution to obtain the three-dimensional vascular distribution of the preset biological site can be achieved by the following method: performing the following processing on each branch vessel in the initial three-dimensional vascular distribution to obtain the three-dimensional vascular distribution: selecting multiple three-dimensional coordinate points on at least one side of the branch vessel, and sequentially connecting the endpoint of the branch vessel with the multiple three-dimensional coordinate points to obtain the connected branch vessel, wherein the connected vessel is different from the branch vessel.

[0171] In some embodiments, multiple three-dimensional coordinate points are selected on at least one side of the branch vessel. These points are used to adjust the shape of the branch vessel to make it more realistic. The endpoints of the branch vessel are sequentially connected to the selected multiple three-dimensional coordinate points to form new vessel segments. These new vessel segments, together with the original branch vessel, constitute the adjusted branch vessel. By connecting the endpoints of the branch vessel to the newly selected three-dimensional coordinate points, the shape of the branch vessel can be changed. This adjustment can be a change in curvature, length, or diameter to better simulate the morphology of actual blood vessels. The newly connected vessel segments need to have a smooth transition with the original branch vessel to avoid unnatural corners or protrusions, ensuring the overall aesthetics and practicality of the vessel distribution. By adjusting the shape of the vessels, the three-dimensional vessel distribution model can be made closer to the anatomical structure of real blood vessels, improving the realism and reliability of the model. The adjusted branch vessel shape is more natural, improving the visual acceptability and interpretability of the model.

[0172] In some embodiments, the connected branch vessel is a new vascular segment formed by selecting multiple three-dimensional coordinate points on at least one side of the original branch vessel and connecting the endpoint of the branch vessel to these coordinate points. Its connection to the original branch vessel lies in the fact that they share a common endpoint and together constitute part of the three-dimensional vascular distribution. The connected branch vessel represents an adjustment and optimization of the shape of the original branch vessel; it inherits the function of the original vessel while providing a morphology that more closely conforms to the actual anatomical structure.

[0173] As an example, taking the human coronary arteries as the preset biological site, this paper details the process of obtaining the target three-dimensional vascular distribution by adjusting the shape of each branch vessel in the initial three-dimensional vascular distribution, targeting the clinical application scenario of three-dimensional reconstruction of the coronary arteries. This process can improve the matching degree between the three-dimensional reconstruction results of the coronary arteries and the actual morphology of the human coronary arteries, and provide accurate vascular morphology data support for the diagnosis and interventional treatment planning of coronary heart disease.

[0174] Continuing from the previous example, the cardiac region of the target object is scanned using a coronary computed tomography angiography (CT) angiography device to obtain raw vascular image data. After preliminary three-dimensional reconstruction, an initial three-dimensional vascular distribution is obtained. This initial three-dimensional vascular distribution includes the approximate spatial location and direction of major branch vessels such as the left main coronary artery, anterior descending artery, circumflex artery, and right coronary artery. However, the curvature and smoothness of some branch vessels deviate from the actual morphology of human coronary vessels and need to be adjusted.

[0175] Following the previous example, shape adjustment processing is performed on each branch vessel in the initial three-dimensional vascular distribution. Taking a branch vessel of the left anterior descending artery as an example, the initial shape of this branch vessel is a zigzag line, which does not match the natural curvature of the actual human coronary artery. The specific adjustment process is as follows: Based on the physiological morphological data of normal human coronary arteries, multiple three-dimensional coordinate points are selected on one side of this branch vessel. The selection is based on the natural curvature of the vessel, ensuring that the trajectory formed by the selected three-dimensional coordinate points conforms to the curvature of the actual coronary artery. The number of three-dimensional coordinate points selected is determined according to the length of the branch vessel. For longer branches, 6 to 8 three-dimensional coordinate points are selected, and for shorter branches, 3 to 4 three-dimensional coordinate points are selected. The spacing between each three-dimensional coordinate point is uniform, and the distance from the branch vessel is controlled within a preset range to avoid excessive distortion or deviation from the physiological position of the adjusted vessel shape.

[0176] Continuing from the previous example, the two endpoints of the branch vessel are sequentially connected to multiple selected three-dimensional coordinate points in a preset order. During the connection process, a smooth curve fitting method is used to connect adjacent coordinate points to form the connected branch vessel. The connected branch vessel has a natural, smooth, curved shape, which is significantly different from the initial zigzag-shaped branch vessel. Its curvature and direction are more in line with the actual physiological morphology of the left anterior descending artery branch vessel in the human body.

[0177] Continuing from the previous example, other branch vessels in the initial three-dimensional vascular distribution, such as branches of the circumflex artery and branches of the right coronary artery, are processed separately in the same way. If some branch vessels require physiological structural adjustments, multiple three-dimensional coordinate points can be selected on both sides of the vessel. The endpoints are then sequentially connected to the three-dimensional coordinate points on both sides and fused to obtain connected branch vessels that conform to physiological morphology. After all branch vessels have undergone shape adjustment, the connected branch vessels are integrated to obtain the three-dimensional vascular distribution of the human coronary arteries. This three-dimensional vascular distribution can accurately reproduce the true spatial morphology and branching structure of the coronary arteries and can be directly used in clinical scenarios such as locating coronary lesions, assessing the degree of vascular stenosis, and planning interventional treatment pathways.

[0178] Thus, by adjusting the shape of each branch vessel in the initial three-dimensional vascular distribution, that is, selecting multiple three-dimensional coordinate points on at least one side of the branch vessel and connecting the endpoints of the branch vessel to these coordinate points to form a new connected branch vessel, not only does the adjusted branch vessel have a different shape from the original branch vessel and is closer to the real vascular anatomical structure, but also the accuracy and functionality of the vascular model are enhanced. The shape adjustment improves the visual realism of the three-dimensional vascular distribution.

[0179] In some embodiments, the multiple three-dimensional coordinate points include the j-th three-dimensional coordinate point, where 1 < j ≤ M, and the endpoints of the branch vessel include the starting point of the branch vessel.

[0180] In some embodiments, the above-mentioned selection of multiple three-dimensional coordinate points on at least one side of the branch vessel can be achieved by the following method: traverse j and perform the following processing: when j = 1, superimpose the random perturbation step corresponding to the first three-dimensional coordinate point on the starting point of the branch vessel in the random perturbation direction corresponding to the first three-dimensional coordinate point to obtain the first three-dimensional coordinate point; when j > 1, superimpose the random perturbation step corresponding to the j-th three-dimensional coordinate point on the (j - 1)-th three-dimensional coordinate point in the random perturbation direction corresponding to the j-th three-dimensional coordinate point to obtain the j-th three-dimensional coordinate point, and the random perturbation direction points to at least one side of the branch vessel.

[0181] In some embodiments, the random perturbation step refers to a randomly varying vector added on the basis of the selected three-dimensional coordinate points when constructing a branch vessel in three-dimensional space to simulate the variability and uncertainty of actual branch vessels. This vector is used to fine-tune the positions of the coordinate points during the construction of the branch vessel, making the shape and position of the branch vessel closer to the complexity and naturalness of real blood vessels.

[0182] In some embodiments, the random perturbation direction refers to a specific direction used to introduce changes on one side (inside or outside) of the blood vessel when adjusting the shape of the branch vessel in three-dimensional space. This direction is randomly selected, aiming to simulate the natural changes that may occur during the actual growth of blood vessels or to increase the diversity and adaptability of the model. The selection of the random perturbation direction is based on a certain random algorithm or random process to ensure that each direction is unique and not predefined. The random perturbation direction is a vector starting from the current position (starting point or previous coordinate point) of the branch vessel and pointing to at least one side of the blood vessel. Its function is to add a small offset to the original blood vessel path, so that the new coordinate points are not strictly on the original path but have a certain deviation. The random perturbation direction is directional and always points to at least one side of the branch vessel, which means that the perturbation does not occur inside the blood vessel but outside it or along the surface of the blood vessel.

[0183] In some embodiments, the endpoints of a branch vessel include its origin and its end point. The origin typically refers to the point where the branch vessel bifurcates from the main vessel, while the end point is the distal end of the branch vessel. Multiple three-dimensional coordinate points are used to define new coordinate positions on at least one side of the branch vessel to adjust its shape. When j = 1, i.e., the first three-dimensional coordinate point is processed, a random perturbation step size is superimposed on the origin (end point) of the branch vessel to obtain the first adjusted three-dimensional coordinate point. This random perturbation step size simulates the natural variation of the vessel at its origin. When j is greater than 1, i.e., subsequent three-dimensional coordinate points are processed, a random perturbation step size is superimposed on the previous three-dimensional coordinate point (the (j-1)th point) to obtain the current three-dimensional coordinate point (the j-th point). This process is repeated until all three-dimensional coordinate points have been processed.

[0184] In some embodiments, the origin of the branch vessel is perturbed along the random perturbation direction corresponding to the first 3D coordinate point. Specifically, the coordinates of the origin are added to the random perturbation step size corresponding to the first 3D coordinate point to obtain the first perturbed 3D coordinate point. This random perturbation direction is predefined, ensuring that the perturbation originating from the origin occurs on one side of the vessel. The (j-1)th coordinate point is perturbed along the random perturbation direction corresponding to the j-th 3D coordinate point. This is achieved by adding the coordinates of the (j-1)-th coordinate point to the random perturbation step size corresponding to the j-th coordinate point. This iterative process generates a series of continuous perturbed coordinate points, and the path formed by these points simulates the natural growth and changes of the vessel.

[0185] As an example, let's illustrate how to select multiple 3D coordinate points on at least one side of a branch vessel and use a random perturbation step size to adjust the positions of these points: Suppose there is a branch vessel that bifurcates from a point on the main vessel. The goal is to create a series of 3D coordinate points along the path of this branch vessel to adjust its shape. The starting point of the branch vessel: Assume the starting coordinates of the branch vessel are (x0, y0, z0). The number of 3D coordinate points to be created: M (i.e., a total of M points, including the starting and ending points). The random perturbation step size is a randomly generated vector whose magnitude and direction are randomized to simulate the natural variation of the vessel. When j=1, a random perturbation step size is superimposed on the starting point (x0, y0, z0) of the branch vessel. Assume the random perturbation step size is a vector (dx1, dy1, dz1). The first 3D coordinate point P1 = (x0 + dx1, y0 + dy1, z0 + dz1). For cases where j ranges from 2 to M, repeat the following steps: For the (j-1)th 3D coordinate point P(j-1) = (x(j-1), y(j-1), z(j-1)), superimpose a new random perturbation step size, assuming it is (dxj, dyj, dzj). The jth 3D coordinate point Pj = (x(j-1) + dxj, y(j-1) + dyj, z(j-1) + dzj).

[0186] As an example, assume the origin of the branch vessel is (0, 0, 0). To create 5 three-dimensional coordinate points (including the origin), M = 5. For j = 1, a random perturbation step size is obtained (0.2, 0.1, 0.3), therefore, P1 = (0.2, 0.1, 0.3). For j = 2, assuming the random perturbation step size is (-0.1, 0.3, 0.1), therefore, P2 = (0.1, 0.4, 0.4). Repeat this process to obtain the subsequent coordinate points P3, P4, and P5.

[0187] Thus, by selecting multiple three-dimensional coordinate points on at least one side of a branching blood vessel and applying random perturbation step sizes, the natural variations and individual differences in blood vessels can be effectively simulated. This not only improves the realism and accuracy of the three-dimensional blood vessel distribution model, but also makes the three-dimensional blood vessel distribution more consistent with the actual vascular structure of biological organisms.

[0188] In step 102, the three-dimensional blood vessel distribution is projected to obtain the two-dimensional blood vessel distribution corresponding to the three-dimensional blood vessel distribution.

[0189] In some embodiments, a two-dimensional vascular distribution refers to a graphical representation of a three-dimensional vascular distribution projected onto a two-dimensional plane. This two-dimensional vascular distribution is used to represent the two-dimensional vascular distribution morphology of the predetermined biological site. This representation is typically used to display and analyze vascular structure, but it loses the depth information found in three dimensions. Two-dimensional vascular distributions can be generated in various ways, such as perspective projection, orthographic projection, or other projection techniques in computer graphics. In a two-dimensional vascular distribution, the branches, directions, and connections of blood vessels are clearly shown, but the thickness of the blood vessels or their specific location in three-dimensional space cannot be directly represented.

[0190] In some embodiments, projection is a graphics technique that involves projecting a 3D scene from multiple different angles to obtain multiple 2D views. In each view, certain aspects of the 3D scene are emphasized or highlighted, while others may be occluded or reduced. Projection allows for a comprehensive observation and analysis of the 3D structure, leading to a deeper understanding.

[0191] In some embodiments, the three-dimensional blood vessel distribution (source domain) is mapped to a two-dimensional plane (target domain). Orthogonal projection or perspective projection is typically used. Orthogonal projection preserves the length and proportions of the blood vessels, while perspective projection simulates the visual effect of the human eye, exhibiting the characteristic of objects appearing larger when closer and smaller when farther away. The source domain of the projection is the blood vessel distribution in three-dimensional space. Three-dimensional space is typically defined based on a Cartesian coordinate system (x, y, z), where x, y, and z represent the three dimensions of space. The target domain of the projection is the blood vessel distribution in a two-dimensional plane. Two-dimensional space is typically defined based on a planar coordinate system (u, v), where u and v represent the two dimensions of the plane. The three-dimensional space, based on the Cartesian coordinate system (x, y, z), is used to describe the position and shape of blood vessels in three-dimensional space. The two-dimensional space can be based on a planar coordinate system (u, v), used to describe the projected blood vessel distribution. Orthogonal projection can be used, which projects the three-dimensional blood vessel distribution along a certain direction (such as the z-axis) onto a two-dimensional plane (x, y), ignoring depth information. It can be achieved through perspective projection, which simulates the perspective of the human eye or a camera, projecting the three-dimensional blood vessel distribution onto a two-dimensional plane while preserving depth information.

[0192] In some embodiments, referring to FIG6, FIG6 is a schematic flowchart of the image generation method provided in the embodiments of this application. Step 102 shown in FIG4 can be implemented by obtaining at least one projection viewpoint and performing steps 1021 to 1022 shown in FIG6 for each of the aforementioned viewpoints.

[0193] In step 1021, the projection matrix corresponding to the viewpoint is obtained. The projection matrix is ​​used to indicate the relationship between the three-dimensional blood vessel distribution and the two-dimensional blood vessel distribution under the viewpoint.

[0194] In some embodiments, a projection matrix is ​​used to map points in three-dimensional space onto a two-dimensional screen. In computer graphics, this mapping is typically achieved through perspective projection or orthographic projection. The projection matrix defines how points in a three-dimensional scene are represented on a two-dimensional plane based on the observer's viewpoint, the camera's position, and orientation.

[0195] In some embodiments, the projection matrix described above can be obtained as follows: First, the parameters of the viewpoint need to be defined, which typically include the observer's position (camera position), viewing direction (camera orientation), the width of the viewpoint (field of view), and the positions of the near and far planes. The view matrix defines the camera's position and orientation, transforming points in the 3D scene into a view coordinate system centered on the camera. If perspective projection is used, the projection matrix is ​​constructed based on the field of view (FOV), aspect ratio, and the distances between the near and far planes. Perspective projection makes objects in the scene appear smaller as distance increases, thus creating a sense of depth. If orthographic projection is used, the projection matrix projects objects in the scene onto a 2D plane according to their actual size, without producing a perspective effect. The projection matrix maps each point of the 3D blood vessel distribution onto a 2D plane, forming a 2D blood vessel distribution. This mapping preserves the structure of the blood vessel distribution but changes its dimension, mapping from 3D space to a 2D screen. Different viewpoint parameters result in different projection matrices, thus producing different 2D blood vessel distribution views on the 2D screen. By changing the viewing angle parameters, the three-dimensional blood vessel distribution can be observed from different angles.

[0196] In step 1022, the three-dimensional blood vessel distribution is projected from the viewpoint based on the projection matrix to obtain the two-dimensional blood vessel distribution from the viewpoint.

[0197] In some embodiments, a projection matrix is ​​applied to each coordinate point of the three-dimensional blood vessel distribution. This process involves multiplication of the matrix with the three-dimensional coordinate points, transforming each three-dimensional coordinate point into a two-dimensional screen coordinate system. After transformation by the projection matrix, the resulting two-dimensional coordinate points form the framework of the two-dimensional blood vessel distribution. These points need to be connected on the two-dimensional plane to form continuous blood vessel paths. By connecting adjacent two-dimensional coordinate points, the blood vessel paths can be drawn, forming the two-dimensional blood vessel distribution. This process may require consideration of the connectivity of the blood vessels to ensure that the two-dimensional blood vessel distribution correctly reflects the structure of the three-dimensional blood vessel distribution.

[0198] As an example, taking human brain blood vessels as the research object, this paper provides a detailed explanation of the process of obtaining the projection matrix corresponding to the viewpoint and projecting the three-dimensional blood vessel distribution based on the matrix to obtain the two-dimensional blood vessel distribution under the corresponding viewpoint in the medical imaging diagnosis scenario of brain vascular lesions. This process can provide clinicians with two-dimensional images of brain blood vessels from multiple perspectives, and help improve the accuracy of lesion detection.

[0199] Continuing the previous example, a head magnetic resonance angiography (MRI) device is used to scan the target subject's brain, acquiring three-dimensional data of the brain's blood vessels. After data reconstruction, a three-dimensional vascular distribution is obtained, including information on the branching, luminal morphology, and spatial location of cerebral arteries and veins. Based on clinical diagnostic needs, the projection angle is determined to be the coronal view of the brain. This view clearly presents the anteroposterior vascular distribution layers of the brain, facilitating doctors' observation of the symmetry of blood vessels on both sides of the brain's midline.

[0200] Continuing the previous example, the projection matrix corresponding to the coronal viewpoint is obtained. This projection matrix is ​​used to clarify the mapping relationship between each blood vessel pixel in the three-dimensional vascular distribution and its corresponding pixel in the two-dimensional vascular distribution under the coronal viewpoint. The projection matrix is ​​obtained based on a preset imaging model and coronal viewpoint parameters. The imaging model adopts a clinically commonly used orthogonal projection model, and the viewpoint parameters include the projection direction, the relative position of the imaging plane and the three-dimensional space of the brain, and the imaging scaling ratio. By inputting the coronal viewpoint parameters into the orthogonal projection model, a projection matrix containing spatial coordinate transformation, scaling, and projection mapping relationships is calculated. This matrix can accurately characterize how blood vessel pixels in three-dimensional space are mapped to the coronal two-dimensional imaging plane.

[0201] Continuing from the previous example, based on the aforementioned projection matrix, the three-dimensional vascular distribution of the brain is projected from a coronal perspective. During projection, the three-dimensional coordinates of each vascular pixel in the three-dimensional vascular distribution are extracted. These coordinates are then substituted into the projection matrix for calculation, yielding the two-dimensional coordinates of each vascular pixel on the coronal two-dimensional imaging plane. All obtained two-dimensional coordinates are then processed, duplicate coordinates are removed, and coordinate information of vascular edge pixels is added. Simultaneously, the grayscale and boundary features of the vessels in the three-dimensional vascular distribution are preserved, ultimately resulting in a two-dimensional vascular distribution from a coronal perspective. This two-dimensional vascular distribution clearly presents the vascular distribution morphology in the coronal direction of the brain and can be directly used for clinical imaging diagnosis, assisting doctors in screening for cerebral vascular stenosis, malformations, and other lesions. If it is necessary to obtain two-dimensional vascular distributions from other perspectives, the corresponding projection matrix can be obtained and projection processing completed in the same manner to meet the needs of different clinical diagnostic scenarios.

[0202] In some embodiments, step 1022 above can be implemented as follows: multiply the coordinates of each three-dimensional coordinate point in the three-dimensional blood vessel distribution by the projection matrix to obtain the two-dimensional coordinate points corresponding to each three-dimensional coordinate point; determine the two-dimensional blood vessel distribution based on the two-dimensional graphic formed by each two-dimensional coordinate point.

[0203] In some embodiments, the above-mentioned determination of the two-dimensional blood vessel distribution based on the two-dimensional graphic formed by each two-dimensional coordinate point can be achieved in the following way: the two-dimensional graphic formed by each two-dimensional coordinate point is determined as the initial two-dimensional blood vessel distribution, the initial two-dimensional blood vessel distribution includes multiple initial blood vessel segments; the gray value of each initial blood vessel segment in the initial two-dimensional blood vessel distribution is adjusted to obtain the two-dimensional blood vessel distribution under the viewpoint, and at least two blood vessel segments with different gray values ​​exist in the two-dimensional blood vessel distribution.

[0204] In some embodiments, each 3D coordinate point (typically represented as (x, y, z)) in the 3D vascular distribution needs to be multiplied by a projection matrix to map it from 3D space to a 2D screen coordinate system. This mapping is achieved through matrix multiplication: the projection matrix can be a 4x4 matrix, and the 3D coordinate point needs to be converted to homogeneous coordinates for matrix multiplication, i.e., adding a dimension to become (x, y, z, 1). The result of the multiplication is a 2D coordinate point (typically represented as (x', y')), which represents the projected position of the 3D coordinate point on the 2D screen. The obtained 2D coordinate points are connected according to their connection relationships in 3D space to form vascular segments. These vascular segments together constitute the initial 2D vascular distribution. The initial 2D vascular distribution consists of multiple vascular segments, each of which is a line segment connecting two or more 2D coordinate points. In the 2D vascular distribution, each vascular segment can be assigned a different grayscale value to represent different attributes, such as vascular flow, density, or other medically relevant parameters. Adjusting the grayscale values ​​of the vascular segments in the initial 2D vascular distribution can modify their grayscale values ​​according to the characteristics of the vascular segments or specific visualization requirements. The adjusted two-dimensional vascular distribution contains at least two vascular segments with different grayscale values. This difference in grayscale values ​​can help distinguish different vascular segments or highlight specific vascular characteristics.

[0205] As an example, the following illustration demonstrates how to project coordinate points from a 3D vascular distribution onto a 2D plane and adjust the grayscale values ​​to obtain the final 2D vascular distribution. Assume a simple 3D vascular distribution model consisting of four 3D coordinate points representing two vascular segments: A(1, 2, 3); B(4, 5, 6); C(7, 8, 9); D(10, 11, 12). These points are connected sequentially to form two vascular segments: AB and BC. Multiplying the 3D coordinate points by the projection matrix yields the corresponding 2D coordinate points: A'(x1', y1'); B'(x2', y2'); C'(x3', y3'); D'(x4', y4'). These points are now located on a 2D plane and form the initial 2D vascular distribution according to their connection relationships in 3D space. Connecting the 2D coordinate points A' to B' and B' to C' yields the two vascular segments of the initial 2D vascular distribution. The grayscale values ​​of these two vascular segments are then adjusted. Suppose that different grayscale values ​​are assigned to vessel segments based on certain characteristics, such as flow rate or diameter: vessel segment AB' is assigned a grayscale value of 100 (brighter), and vessel segment BC' is assigned a grayscale value of 150 (darker). The adjusted grayscale values ​​result in a two-dimensional vessel distribution with different grayscale levels, making the two vessel segments visually distinguishable. The two-dimensional vessel distribution is as follows: vessel segment AB' appears as a line segment with a brightness of 100. Vessel segment BC' appears as a line segment with a brightness of 150. In the two-dimensional vessel distribution image, the observer can clearly see the two vessel segments with different grayscale values, which helps in identifying and analyzing the characteristics of the vessels.

[0206] In this way, by multiplying the coordinates of each three-dimensional coordinate point in the three-dimensional blood vessel distribution with the projection matrix, the corresponding two-dimensional coordinate points are obtained, and an initial two-dimensional blood vessel distribution is constructed. Then, the gray values ​​of each blood vessel segment are adjusted to obtain the two-dimensional blood vessel distribution under the viewpoint. This not only achieves accurate mapping of the three-dimensional blood vessel structure to the two-dimensional plane and preserves the key anatomical information of the blood vessels, but also enhances the visual contrast of blood vessel segments through the difference in gray values, making different blood vessel segments clearly distinguishable in the two-dimensional image, thereby greatly improving the accuracy and efficiency of blood vessel analysis and diagnosis.

[0207] In some embodiments, the above-mentioned adjustment of the grayscale values ​​of each initial blood vessel segment in the initial two-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution under the viewpoint can be achieved in the following manner: Perform the following processing on each initial blood vessel segment in the initial two-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution under the viewpoint: obtain the depth value of the initial blood vessel segment, and subtract the minimum depth value of the blood vessel segment in the three-dimensional blood vessel distribution from the depth value to obtain a first depth value; obtain the difference between the maximum depth value and the minimum depth value of the blood vessel segment in the three-dimensional blood vessel distribution, and determine the first grayscale value corresponding to the initial blood vessel segment based on the difference and the first depth value; adjust the second grayscale value of the initial blood vessel segment in the initial two-dimensional blood vessel distribution to the first grayscale value.

[0208] In some embodiments, a grayscale value refers to a numerical value used to represent pixel brightness in a two-dimensional image. In an 8-bit grayscale image, the grayscale value typically ranges from 0 to 255, where 0 represents pure black, 255 represents pure white, and intermediate values ​​represent different shades of shadow. The first depth value is the normalized value of the vessel segment depth in the three-dimensional vessel distribution; specifically, it is obtained by subtracting the minimum depth value of all vessel segments in the three-dimensional vessel distribution from the original depth value of the vessel segment. The second depth value refers to the original grayscale value of the vessel segment in the initial two-dimensional vessel distribution, which already exists before grayscale adjustment. The maximum depth value is the maximum value among all vessel segment depth values ​​in the three-dimensional vessel distribution; this value is used to determine the range of depth values ​​for normalization. The minimum depth value is the minimum value among all vessel segment depth values ​​in the three-dimensional vessel distribution; it is the reference point for normalizing the depth value and is used to calculate the first depth value.

[0209] In some embodiments, the depth value refers to the depth information of a blood vessel segment in three-dimensional space, typically representing the distance of the segment from the observer. For each blood vessel segment, its depth value is obtained. The depth value can be calculated from coordinate points in the three-dimensional blood vessel distribution, for example, by taking the average depth of the two endpoints of the segment. The first depth value = the depth value of the blood vessel segment, which is the minimum depth value of the segment in the three-dimensional blood vessel distribution. By subtracting the minimum depth value, the depth value is normalized to a relative value based on the minimum depth value, facilitating subsequent calculations. The difference = the maximum depth value of the blood vessel segment in the three-dimensional blood vessel distribution - the minimum depth value. The difference represents the depth range of the blood vessel segment in the three-dimensional blood vessel distribution and is used to map the first depth value to a grayscale value range. The first grayscale value = (first depth value / difference) × maximum grayscale value. By mapping the first depth value proportionally to the grayscale value range, the first grayscale value corresponding to the blood vessel segment is obtained. The maximum grayscale value is typically 255 (8-bit grayscale image), but can be adjusted according to specific needs. The second grayscale value is the original grayscale value of the blood vessel segment in the initial two-dimensional blood vessel distribution. By adjusting the grayscale value to the first grayscale value, the grayscale value of each blood vessel segment is correlated with its depth value, thus reflecting the depth information of the blood vessels in the two-dimensional image. In the adjusted two-dimensional blood vessel distribution, the grayscale value of each blood vessel segment is related to its depth value; blood vessel segments with greater depth have higher grayscale values ​​(brighter) and blood vessel segments with less depth have lower grayscale values ​​(darker). Grayscale value adjustment allows the two-dimensional blood vessel distribution to more intuitively reflect the depth information of the three-dimensional blood vessel distribution, enhancing the image's stereoscopic effect and readability.

[0210] As an example, the expression for the first depth value mentioned above can be:

[0211] Where G indicates the first depth value, and y indicates the depth value of the vessel segment. min Used to indicate the minimum depth value of a vascular segment in a three-dimensional vascular distribution, y max w is used to indicate the maximum depth value of a vascular segment in a three-dimensional vascular distribution. random This is used to indicate a random perturbation that causes a slight change in grayscale value with each projection, corresponding to the effect of depth variation on veins.

[0212] Continuing with the previous example, the determination of the first grayscale value corresponding to the initial blood vessel segment based on the difference and the first depth value can be achieved as follows: randomly perturb the first depth value to obtain a perturbed first depth value; divide the perturbed first depth value by the difference to obtain a candidate first depth value; and multiply the candidate first depth value by a preset value to determine the first depth value corresponding to the initial blood vessel segment.

[0213] As an example, the following illustration shows how to adjust the grayscale values ​​of initial vessel segments in an initial two-dimensional vessel distribution. Assume there are four initial vessel segments with the following depth values ​​(Z-coordinate): initial vessel segment AB has a depth value of 10; initial vessel segment BC has a depth value of 20; initial vessel segment CD has a depth value of 30; and initial vessel segment DE has a depth value of 40. Find the minimum depth value of the initial vessel segments in the three-dimensional vessel distribution. In this example, the minimum depth value is 10 (the depth value of initial vessel segment AB). For each initial vessel segment, calculate the first depth value: first depth value of initial vessel segment AB = 10 - 10 = 0; first depth value of initial vessel segment BC = 20 - 10 = 10; first depth value of initial vessel segment CD = 30 - 10 = 20; first depth value of initial vessel segment DE = 40 - 10 = 30. Calculate the difference between the maximum and minimum depth values: maximum depth value = 40 (the depth value of initial vessel segment DE); minimum depth value = 10 (the depth value of initial vessel segment AB), difference = 40 - 10 = 30.

[0214] Continuing the previous example, the first depth value and the difference are used to determine the first grayscale value of each initial vessel segment. Assuming the maximum grayscale value is 255: the first grayscale value of initial vessel segment AB = (0 / 30)*255 = 0. The first grayscale value of initial vessel segment BC = (10 / 30)*255 ≈ 85. The first grayscale value of initial vessel segment CD = (20 / 30)*255 ≈ 170, and the first grayscale value of initial vessel segment DE = (30 / 30)*255 = 255. Assuming the initial vessel segments in the initial two-dimensional vessel distribution have the following second grayscale values ​​(original grayscale values): the second grayscale value of initial vessel segment AB = 128, the second grayscale value of initial vessel segment BC = 128, the second grayscale value of initial vessel segment CD = 128, and the second grayscale value of initial vessel segment DE = 128. The second grayscale value of each initial vascular segment is adjusted to its first grayscale value: the new grayscale value of initial vascular segment AB = 0; the new grayscale value of initial vascular segment BC ≈ 85; the new grayscale value of initial vascular segment CD ≈ 170; and the new grayscale value of initial vascular segment DE = 255. In the adjusted two-dimensional vascular distribution, the grayscale value of each initial vascular segment now reflects its depth information in three-dimensional space. Deeper initial vascular segments (such as DE) have higher grayscale values ​​(brighter), while shallower initial vascular segments (such as AB) have lower grayscale values ​​(darker). In this way, the two-dimensional vascular distribution not only shows the structure of the blood vessels but also provides visual cues about the depth of the blood vessels.

[0215] Thus, through refined depth value processing and grayscale mapping, the visual expressiveness and information content of two-dimensional blood vessel distribution are significantly improved. By normalizing the depth values ​​of blood vessel segments, it is ensured that blood vessel segments at different depths have clearly distinguishable grayscale differences in two-dimensional images. This not only enhances the sense of layering and three-dimensionality of the image, but also helps to highlight the structural features of blood vessels and potential lesion areas.

[0216] In step 103, an image of a preset biological site is generated based on the two-dimensional blood vessel distribution.

[0217] In some embodiments, the images are simulated images of a predetermined biological site acquired through medical imaging techniques such as CT, MRI, and ultrasound. These images are a visual representation of the internal structure of the predetermined biological site, including a two-dimensional blood vessel distribution. The images are two-dimensional, representing the cross-section or projection of the predetermined biological site through a pixel array. These pixel values ​​reflect the optical properties or density of the biological tissue. The two-dimensional blood vessel distribution includes features such as the shape, size, direction, and branching of blood vessels, which are manifested as specific textures and patterns in the image. Based on skin texture and the two-dimensional blood vessel distribution, blood vessel textures can be generated using image processing techniques such as texture mapping and pattern synthesis. This process involves combining the visual features of skin texture with the geometric features of the blood vessel distribution. The generated blood vessel texture simulates the appearance and texture of real blood vessels, making the two-dimensional blood vessel distribution image more realistic. The generated blood vessel texture is applied to the two-dimensional blood vessel distribution image, and the final image is formed through synthesis techniques such as alpha channel blending and layer overlay.

[0218] In some embodiments, referring to FIG7, FIG7 is a schematic flowchart of the image generation method provided in the embodiments of this application. Step 103 shown in FIG4 can be implemented by executing steps 1031 to 1033 shown in FIG7.

[0219] In step 1031, the skin texture of the preset biological site is obtained.

[0220] In some embodiments, skin texture refers to the microstructure of the skin surface, including features such as skin texture, color, spots, and wrinkles. In medical imaging or computer vision, skin texture is often used to identify individuals, analyze skin conditions, or detect the development of lesions.

[0221] In some embodiments, a predefined bio-site refers to an entity from which skin texture needs to be acquired. This can be a part of the human body (such as an arm, leg, or face), animal skin, or any object with skin texture. A predefined bio-site refers to a specific area on an object from which texture information needs to be acquired. This area may be selected due to the needs of medical, biological, or other scientific research; for example, it may be because the area has specific lesions, scars, or other distinctive features. Skin texture refers to the microstructure of the skin surface, including skin texture patterns, color, spots, wrinkles, etc. Skin texture is a unique feature of the skin and can be used to identify individuals, analyze skin conditions, or perform biometric identification.

[0222] In step 1032, based on the skin texture and the two-dimensional blood vessel distribution, a blood vessel texture corresponding to the two-dimensional blood vessel distribution is generated.

[0223] In some embodiments, the above-mentioned generation of the vascular texture corresponding to the two-dimensional vascular distribution based on the skin texture and the two-dimensional vascular distribution can be achieved by superimposing the skin texture and the two-dimensional vascular distribution to obtain the vascular texture corresponding to the two-dimensional vascular distribution.

[0224] In some embodiments, the superposition of the skin texture with the two-dimensional blood vessel distribution to obtain the blood vessel texture corresponding to the two-dimensional blood vessel distribution can be achieved as follows: when the coordinate system of the skin texture coincides with the coordinate system of the two-dimensional blood vessel distribution, the skin texture is superimposed on each of the two-dimensional blood vessel distributions to obtain the blood vessel texture corresponding to each of the two-dimensional blood vessel distributions; when the coordinate system of the skin texture does not coincide with the coordinate system of the two-dimensional blood vessel distribution, a transformation matrix between the coordinate system of the skin texture and the coordinate system of the two-dimensional blood vessel distribution is obtained, and based on the transformation matrix, the skin texture is transformed to the coordinate system of the two-dimensional blood vessel distribution to obtain the transformed skin texture; the transformed skin texture is superimposed on each of the two-dimensional blood vessel distributions to obtain the blood vessel texture corresponding to each of the two-dimensional blood vessel distributions.

[0225] In some embodiments, coordinate system coincidence means that the skin texture image and the 2D vascular distribution image are aligned in spatial position and orientation. This alignment can be achieved by transforming the images through rotation, translation, scaling, etc. Ensure that both the skin texture image and the 2D vascular distribution image have been digitized and are in a processable format. Before overlaying, confirm that the coordinate systems of the two images are calibrated to the same reference frame, i.e., their coordinate axes are aligned and their origins are in the same position. Overlay the skin texture image onto the 2D vascular distribution image. This can be achieved by adjusting the transparency of the skin texture image to make it semi-transparently overlay the 2D vascular distribution image, thereby visually blending the information from both. Use layer blending functions in image editing software or programming libraries to merge the skin texture as a new layer with the 2D vascular distribution layer.

[0226] In some embodiments, the coordinate system of skin texture refers to a reference frame used to describe the position and relationship of skin texture features in space. In image processing, this coordinate system typically corresponds to the pixel coordinate system of an image, where each pixel has specific coordinates representing its position in the image. The coordinate system of a two-dimensional blood vessel distribution refers to a reference frame used to describe the position and relationship of various points (such as the endpoints of blood vessel segments) in a two-dimensional blood vessel distribution. This coordinate system can be an image coordinate system, where the x-axis and y-axis represent the width and height of the image, respectively, or a custom coordinate system defined according to specific application requirements.

[0227] In some embodiments, the coincidence of the coordinate system of the skin texture and the coordinate system of the two-dimensional blood vessel distribution can be determined by confirming whether the directions of the x-axis and y-axis in the two coordinate systems are consistent, i.e., whether they point to the same angle and direction. It can also be confirmed whether the origins of the two coordinate systems are the same, i.e., whether their reference points coincide. If the directions of the x-axis and y-axis in the two coordinate systems are consistent and the origins of the two coordinate systems are the same, then it can be determined that the coordinate system of the skin texture and the coordinate system of the two-dimensional blood vessel distribution coincide. If the directions of the x-axis and y-axis in the two coordinate systems are inconsistent or the origins of the two coordinate systems are the same, then it can be determined that the coordinate system of the skin texture and the coordinate system of the two-dimensional blood vessel distribution do not coincide.

[0228] In some embodiments, when the coordinate system of the skin texture does not coincide with the coordinate system of the two-dimensional blood vessel distribution, a transformation matrix between the coordinate system of the skin texture and the coordinate system of the two-dimensional blood vessel distribution is obtained, and based on the transformation matrix, the skin texture is transformed to the coordinate system of the two-dimensional blood vessel distribution to obtain the transformed skin texture.

[0229] In some embodiments, the transformed skin texture refers to an image whose coordinate system matches that of the two-dimensional blood vessel distribution after coordinate transformation of the original skin texture image. This means that each point in the skin texture image has been repositioned according to the transformation matrix to correspond to a point in the two-dimensional blood vessel distribution image. The transformed skin texture and the two-dimensional blood vessel distribution have the same coordinate system, allowing them to be compared and analyzed in the same space. Feature points on the skin texture (such as the starting points and branch points of blood vessels) are spatially consistent with their corresponding points on the two-dimensional blood vessel distribution. The transformed skin texture can be overlaid on the two-dimensional blood vessel distribution image, thereby visually fusing information from the two images and enhancing image readability and analytical capabilities.

[0230] In some embodiments, a transformation matrix is ​​a mathematical matrix that describes the transformation relationship between two coordinate systems. It contains information about rotation, translation, and scaling, mapping points in one coordinate system to another. If the transformation involves only rotation, translation, and scaling, without perspective transformation, then the transformation matrix is ​​an affine transformation matrix. If the transformation involves perspective effects, then the transformation matrix is ​​a perspective transformation matrix, which is more complex than an affine transformation matrix.

[0231] In some embodiments, when the coordinate system of the skin texture does not coincide with the coordinate system of the two-dimensional blood vessel distribution, the relative position and rotation angle between the two images are determined by identifying and matching common feature points in both the skin texture and two-dimensional blood vessel distribution images. The matching information of the feature points is used to calculate a transformation matrix. This matrix can be an affine transformation matrix or a perspective transformation matrix, depending on the relationship between the two coordinate systems. The calculated transformation matrix is ​​applied to each pixel of the skin texture image, thereby transforming the image from the original coordinate system to the coordinate system of the two-dimensional blood vessel distribution. The transformed skin texture can now be analyzed in the coordinate system of the two-dimensional blood vessel distribution and can be overlaid with the blood vessel distribution image to obtain richer image information.

[0232] In some embodiments, the skin texture is transformed to the coordinate system of the two-dimensional blood vessel distribution based on the transformation matrix to obtain the transformed skin texture. The transformation matrix can be applied to each pixel of the skin texture image to achieve coordinate transformation, thereby obtaining the transformed skin texture.

[0233] As an example, taking the human forearm as the preset biological site, this paper details the process of generating vascular texture based on coordinate system adaptation for medical image preprocessing scenarios for forearm vein puncture assistance. This process can provide accurate vascular texture images for puncture assistance devices and improve the accuracy of puncture positioning.

[0234] Continuing from the previous example, skin texture and multiple two-dimensional blood vessel distribution data of a specified area of ​​the human forearm were acquired. The skin texture was captured using a high-resolution industrial camera on an area 3 to 5 cm below the elbow. During acquisition, the acquisition range was set based on the medial axis of the forearm, and the pixel coordinate system parameters of the acquired images were recorded simultaneously. Multiple two-dimensional blood vessel distribution data were generated by scanning the same forearm area using ultrasound imaging equipment, containing information on vein branches, diameters, and directions at different depth levels. Similarly, the pixel coordinate system parameters corresponding to each two-dimensional blood vessel distribution were recorded. The acquired skin texture images were preprocessed: median filtering was used to remove environmental noise, and the image grayscale range was adjusted to highlight skin texture details and ensure clarity. Contour enhancement processing was performed on the two-dimensional blood vessel distribution data to strengthen the distinction between the blood vessel area and surrounding tissues.

[0235] Continuing from the previous example, we compare the coordinate system of the skin texture with the coordinate system parameters of each two-dimensional blood vessel distribution to determine if they overlap. If the skin texture and a certain two-dimensional blood vessel distribution use the same acquisition reference and pixel scale, i.e., their coordinate systems overlap, then we directly overlay the skin texture and the two-dimensional blood vessel distribution. During the overlay process, we maintain the skin texture as the background layer and the two-dimensional blood vessel distribution as the foreground layer, ensuring that the blood vessel area can accurately cover the corresponding physiological position of the skin texture, thereby obtaining the blood vessel texture corresponding to the two-dimensional blood vessel distribution. This blood vessel texture can clearly present the blood vessel distribution pattern against the background of the skin texture.

[0236] Continuing the previous example, if the coordinate system of the skin texture does not coincide with the coordinate system of a two-dimensional blood vessel distribution—for example, the skin texture uses a coordinate system with a pixel scale of 0.01 mm per pixel, while the two-dimensional blood vessel distribution uses a coordinate system with a pixel scale of 0.02 mm per pixel, and there is an offset between their acquisition references—then a transformation matrix between the two is first obtained. This transformation matrix is ​​obtained by selecting three common feature points, which are fixed markers on the forearm skin. The coordinate values ​​of each feature point in both the skin texture coordinate system and the two-dimensional blood vessel distribution coordinate system are read, and a transformation matrix including scaling and translation parameters is calculated based on these coordinate values. Based on this transformation matrix, the skin texture undergoes coordinate transformation, adjusting its pixel scale to match the two-dimensional blood vessel distribution and correcting the acquisition reference offset, resulting in the transformed skin texture. The transformed skin texture is then superimposed on the two-dimensional blood vessel distribution, ensuring accurate correspondence between the physiological positions of the blood vessel regions and the skin texture during superposition, ultimately yielding the blood vessel texture corresponding to the two-dimensional blood vessel distribution.

[0237] Thus, a differentiated processing strategy is adopted to address whether the coordinate systems of skin texture and 2D blood vessel distribution overlap. By directly superimposing when the coordinate systems overlap and performing coordinate transformation based on a transformation matrix before superimposing when they do not, the problem of misaligned superposition of skin texture and 2D blood vessel distribution caused by coordinate system deviation can be fundamentally avoided. This ensures that the generated blood vessel texture accurately matches the real physiological structure of the preset biological site, improving the realism and reliability of the blood vessel texture. The image generated based on this accurately generated blood vessel texture can more realistically restore the natural distribution of blood vessels and skin in the preset biological site, providing high-quality data support for subsequent applications such as lesion diagnosis model training, medical teaching demonstrations, and clinical auxiliary diagnosis based on this type of image, effectively improving the accuracy and effectiveness of related applications. At the same time, this differentiated coordinate system processing method can adapt to skin texture data and 2D blood vessel distribution data obtained from different sources and different acquisition devices, without the need for additional unified coordinate system preprocessing of the original data, thus lowering the data processing threshold.

[0238] In step 1033, an image corresponding to the two-dimensional blood vessel distribution is generated based on the blood vessel texture.

[0239] In some embodiments, overlay parameters, such as transparency, can be adjusted as needed. Transparency determines the visibility of the skin texture image during the overlay process and usually needs to be adjusted according to specific circumstances to achieve the best visual effect. In image processing software, the skin texture image is placed as a new layer on top of the two-dimensional blood vessel distribution image. The layer's transparency is adjusted to make the skin texture semi-transparent so that the underlying blood vessel distribution image can be seen. A suitable layer blending mode (such as "Normal," "Overlay," "Soft Light," etc.) is selected to optimize the fusion effect of the two images. In the overlay image, the structure of the blood vessel distribution is clearly displayed on the skin texture background. In this way, the position, direction, and branching of blood vessels can be observed and analyzed against the skin texture background.

[0240] In some embodiments, vascular texture refers to the unique pattern or structure that blood vessels present on the surface of an image or a physical object. In medical imaging, vascular texture usually refers to the fine structure of the blood vessel wall or the visible features of blood vessels on a two-dimensional image, such as the direction, branching and morphology of blood vessels. These texture features are of great significance for the identification and analysis of vascular lesions. Different two-dimensional blood vessel distributions correspond to different vascular textures.

[0241] As an example, taking the back of the human hand as a preset biological site, this paper provides a detailed explanation of the process of acquiring skin texture, generating vascular texture by combining two-dimensional vascular distribution, and generating corresponding images for medical imaging-assisted diagnosis scenarios related to vascular lesions on the back of the hand. This process can directly provide technical support for expanding image data for screening vascular lesions on the back of the hand.

[0242] Continuing the previous example, when acquiring skin texture from a predetermined biological location, a high-resolution image acquisition device is used to capture images of a designated area on the back of the hand. The acquisition area is selected from the middle palmar side of the back of the hand, where there are no obvious scars or pigmentation to ensure the integrity and authenticity of the skin texture. During the acquisition process, the ambient light intensity is kept uniform to avoid shadows interfering with the accuracy of skin texture acquisition. The acquired raw back of the hand skin image is preprocessed, including image denoising, texture enhancement, and size standardization. Image denoising uses a mean filtering algorithm to remove random noise generated during acquisition. Texture enhancement highlights the details of the skin texture by adjusting the image contrast. Size standardization uniformly adjusts the image to a preset pixel size to ensure consistency in subsequent processing. After the above processing, skin texture data that clearly characterizes the direction, density, and depth of the skin texture on the back of the hand is obtained.

[0243] Continuing the previous example, when generating the vascular texture corresponding to the two-dimensional vascular distribution based on the skin texture and the two-dimensional vascular distribution, the two-dimensional vascular distribution data of the back of the human hand is first acquired. This two-dimensional vascular distribution data is obtained by scanning the same area of ​​the back of the hand using medical imaging equipment, and includes core information such as the branching direction of blood vessels, changes in blood vessel diameter, and the location of blood vessel nodes. The preprocessed skin texture data and the two-dimensional vascular distribution data are then fused. During the fusion process, the skin texture is used as a background texture constraint to ensure that the generated vascular texture conforms to the natural growth pattern of the skin on the back of the hand. Simultaneously, the two-dimensional vascular distribution data serves as the core structural basis to ensure that key features such as the branching and diameter of the vascular texture are consistent with the actual vascular distribution. Specifically, by matching the direction of the skin texture with the branching direction of the vascular distribution, the vascular texture is distributed along the gaps in the skin texture, avoiding unreasonable overlap between the vascular texture and the skin texture, ultimately resulting in a vascular texture that combines the characteristics of the skin background and the features of the real vascular structure.

[0244] Continuing the previous example, when generating the image corresponding to the two-dimensional blood vessel distribution based on the aforementioned blood vessel texture, the generated blood vessel texture is input into a preset image generation model. This image generation model is trained on a large number of human hand dorsum blood vessel image samples and can simulate the imaging effect of commonly used clinical medical imaging equipment. During processing, the image generation model restores the gray-level difference between the blood vessel area and the surrounding skin area according to the gray-level distribution characteristics of the blood vessel texture, while simulating the imaging noise and contrast characteristics of medical images, so that the generated image has the same visual effect and detail features as real medical images. The generated image clearly presents the distribution pattern of blood vessels on the back of the hand and the skin texture background, which can be directly used to expand the training data of subsequent hand dorsum blood vessel lesion screening models, or as an auxiliary reference image for medical imaging diagnosis, meeting the actual needs of clinical applications.

[0245] Thus, by acquiring the skin texture of a predefined biological site and, when the skin texture coordinate system coincides with or is adjusted by a transformation matrix to coincide with the two-dimensional blood vessel distribution coordinate system, the skin texture and the two-dimensional blood vessel distribution are superimposed. The resulting blood vessel texture not only provides rich visual information for medical image analysis but also enhances the intuitive representation of the relationship between blood vessels and skin. When the coordinate systems coincide, the superimposed image can clearly show the direction and distribution of blood vessels on the skin, helping to detect potential lesions and abnormalities. When the coordinate systems do not coincide, precise adjustment of the transformation matrix can ensure that the skin texture and blood vessel distribution are correctly aligned in space, avoiding errors caused by positional deviations.

[0246] In some embodiments, an image refers to a two-dimensional image extracted and processed from the original vascular texture. It is generated based on the vascular texture using specific image processing algorithms or techniques to highlight or analyze the structural features of blood vessels. The image is derived from the original vascular texture and typically contains key information related to vascular structure from the texture. Images are usually customized for specific medical or research needs to meet specific requirements for vascular feature analysis. Images are often designed to facilitate quantitative analysis or visualization, making the morphology, orientation, and distribution of blood vessels more apparent.

[0247] In some embodiments, referring to FIG8, FIG8 is a schematic flowchart of the image generation method provided in the embodiments of this application. Step 1033 shown in FIG7 can be implemented by executing steps 10331 to 10332 shown in FIG8.

[0248] In step 10331, multiple equally spaced noise features are obtained, and step 10332 is performed for each of the two-dimensional blood vessel distributions.

[0249] In some embodiments, noise features are not randomly selected, but rather according to certain rules (such as equidistant distribution). This helps control how noise is introduced and maintains the similarity between samples. The L2 distance (Euclidean distance) constraint means that the distance between each noise feature point in the noise feature space is equal. This constraint helps maintain the relative positional relationship of the generated sample points in the feature space.

[0250] In some embodiments, the multiple equally distributed noise features can be l₂ distance (Euclidean distance) constrained noise. l₂ distance constrained noise means that the noise added to the original palm vein image maintains a certain distance constraint, ensuring that the noise's influence is within a certain range and does not lead to excessive image distortion. Different noises are used to generate palm vein images for the same palm vascular texture. For the same palm vascular texture, multiple new palm vein images are generated using different noise features. The degree of variation between these images is stable; that is, the introduction of noise does not significantly change the basic features of the vascular texture. These noise features are equally distributed in the feature space, ensuring that each noise feature affects the original image in a similar way, thus obtaining uniformly enhanced samples.

[0251] In step 10332, based on the vascular texture corresponding to the two-dimensional vascular distribution and each of the noise features, the image corresponding to the two-dimensional vascular distribution is predicted to obtain multiple images corresponding to the two-dimensional vascular distribution, and the images correspond one-to-one with the noise features.

[0252] In some embodiments, multiple equally distributed noise features are introduced and combined with the vascular texture of a two-dimensional blood vessel distribution to generate multiple images, thereby achieving sample enhancement of the blood vessel image. The noise features, constrained by l_2 distance, ensure stable variation in the generated images, avoiding image distortion or excessive deviation. The noise features are uniformly distributed in the feature space, ensuring coverage of different noise intensities and types. For example, five noise features are uniformly sampled in the feature space, each representing a specific noise pattern. Each two-dimensional blood vessel distribution corresponds to a vascular texture, describing the structural and morphological features of the blood vessel. For example, five vascular textures represent five different blood vessel distribution patterns. For each two-dimensional blood vessel distribution, a corresponding image is generated based on its vascular texture and each noise feature. Due to the combination of noise features and vascular textures, the number of generated images is the product of the number of vascular textures and the number of noise features (e.g., 5×5 = 25 images). The l_2 distance constraint ensures that the noise features are uniformly distributed and their variation is controllable in the feature space. For the same vascular texture, the difference between images generated by different noise features remains within a reasonable range, avoiding abrupt changes or distortion in the generated images. By introducing noise features, the generated images retain the core features of the vascular texture while adding subtle variations. By combining noise features with vascular texture, the dataset size was significantly expanded (e.g., 25 images). The enhanced dataset can be used to train a more robust vascular image analysis model, improving the model's generalization ability. The l_2 distance-constrained noise features ensured the stability of the generated images' variability, avoiding image quality degradation.

[0253] In some embodiments, multiple images can be generated by combining the vascular texture of a two-dimensional blood vessel distribution with equidistantly distributed noise features through model prediction, thereby achieving sample enhancement of blood vessel images. Pre-trained generative models (such as GANs, VAEs, etc.) can be used, taking the vascular texture and noise features of the two-dimensional blood vessel distribution as input, to predict and generate corresponding images. For each two-dimensional blood vessel distribution, the model generates one image based on its vascular texture and each noise feature. Due to the combination of noise features and vascular texture, the number of generated images is the product of the number of vascular textures and the number of noise features (e.g., 5×5=25 images).

[0254] As an example, we take medical image processing of the blood vessels in the human fundus as a specific application scenario. The core requirement in this scenario is to improve the generalization ability of the subsequent fundus vascular lesion detection model by expanding the image samples corresponding to the two-dimensional distribution of blood vessels in the fundus. The following details the process of obtaining multiple equally distributed noise features and subsequent image prediction processing.

[0255] Continuing the previous example, when acquiring multiple equally spaced noise features, the first step is to determine the type of noise features as Gaussian noise based on the imaging characteristics of fundus vascular images. This type of noise is consistent with the noise characteristics generated during fundus camera imaging. The value range of the noise features is set to 0 to 0.1. This range covers the common noise intensity range in clinical fundus imaging while avoiding excessively high noise intensity that could damage vascular texture information. Based on the above value range, noise features are selected at equal intervals. If the interval is set to 0.01, a total of 11 equally spaced noise features with values ​​from 0, 0.01, 0.02 to 0.1 can be obtained. The differences between adjacent noise features are uniform, ensuring that the subsequent predicted image can comprehensively cover different noise interference scenarios.

[0256] Following the previous example, the following processing is performed on each two-dimensional vascular distribution in the fundus: First, the vascular texture information corresponding to the current two-dimensional vascular distribution is extracted. This vascular texture information includes features such as the branching direction of the fundus arteries, the texture density of the veins, the texture morphology of the vascular intersections, and the texture gradient of the vascular edges. The above texture information is obtained by performing vascular segmentation processing on the original fundus images acquired clinically, which can accurately characterize the actual vascular structure corresponding to the two-dimensional vascular distribution.

[0257] Continuing with the previous example, the extracted vascular texture information is correlated with the aforementioned equidistant noise features, and then input into a preset image prediction model to predict the image corresponding to the two-dimensional vascular distribution. The core function of the image prediction model is to simulate the fundus vascular imaging effect under different noise interferences. During the prediction process, it determines the pixel distribution pattern of the vascular region based on the input vascular texture information, and adjusts the pixel fluctuation amplitude of non-vascular regions based on the correlated noise features, ensuring that the generated image retains the complete vascular texture structure while accurately reflecting the image quality under the corresponding noise intensity.

[0258] Continuing with the previous example, through the above prediction process, multiple images are generated for each two-dimensional retinal vessel distribution, and each image corresponds one-to-one with a noise feature. For example, when the input noise feature is 0.03, the generated image corresponds to the retinal imaging effect simulating a noise intensity of 0.03, with clear vessel texture and slight pixel fluctuations. When the input noise feature is 0.08, the generated image corresponds to the retinal imaging effect simulating a noise intensity of 0.08, with vessel texture still discernible but increased pixel fluctuations, consistent with the characteristics of retinal images under high noise interference in clinical settings. The multiple images generated in this way effectively expand the retinal vessel image sample set, providing richer scene data support for the training of subsequent lesion detection models.

[0259] Thus, by acquiring multiple equally distributed noise features and predicting each two-dimensional blood vessel distribution based on its vessel texture and noise features to generate corresponding images, effective sample enhancement of blood vessel images can be achieved. For example, a combination of 5 vessel textures and 5 noise features can generate 25 images, significantly expanding the dataset size. Simultaneously, the noise features with l_2 distance constraints ensure that the degree of variation between images generated from the same vessel texture by different noises remains stable, avoiding excessive deviation or distortion of the generated images. This stable variation not only improves the diversity and controllability of the generated images but also provides richer and higher-quality samples for subsequent blood vessel image analysis, enhancing the model's generalization ability and robustness.

[0260] In some embodiments, after step 103 above, the following processing may also be performed: using images corresponding to multiple two-dimensional blood vessel distributions as blood vessel image samples, and using the preset biological site as the sample label of the blood vessel image samples, a blood vessel image sample set of the preset biological site is constructed.

[0261] In some embodiments, the vascular image sample set is used to train a recognition model to obtain an object recognition model, which is used to recognize the object based on vascular images of a preset biological part of the object.

[0262] In some embodiments, the vascular image sample set is a collection of images corresponding to multiple two-dimensional vascular distributions. These images are generated by combining vascular texture and equidistantly distributed noise features, exhibiting both diversity and stability. For each two-dimensional vascular distribution, a corresponding image is generated based on its vascular texture and multiple noise features.

[0263] In some embodiments, the object recognition model is a machine learning model that identifies objects based on vascular images of preset biological sites (such as the palm, retina, etc.). This model learns the features of vascular images to identify and distinguish different objects. Using a set of vascular image samples as training data, the model learns the feature representation of vascular images. Through supervised or unsupervised learning, the model can identify objects corresponding to vascular images of preset biological sites. A set of images corresponding to multiple two-dimensional vascular distributions, possessing diversity and stability, is used to train the object recognition model. The machine learning model for object recognition based on vascular images of preset biological sites achieves high-precision and robust recognition by learning the features of vascular images.

[0264] In some embodiments, a vascular image sample set is constructed by combining a first vascular image and multiple images corresponding to two-dimensional vascular distributions. Objects are used as sample labels to form supervised learning training data for training an object recognition model. The first vascular image is a raw, noise-free vascular image, typically of high quality and clarity. It serves as the foundational data for the sample set, providing core information about vascular texture and structure. The images are generated based on vascular texture from two-dimensional vascular distributions and equidistant noise features. By introducing noise features, the images exhibit diversity in detail, while l_2 distance constraints ensure stable variability. Each vascular image sample (including the first vascular image and the images) corresponds to an object label. The object label identifies the object to which the vascular image belongs (e.g., individual identity, lesion type, etc.). Using the first vascular image and the images as sample data, and the objects as sample labels, a vascular image sample set is formed. For example, one first vascular image and 25 images can constitute a sample set of 26 samples.

[0265] In some embodiments, after constructing the vascular image sample set of the object based on the images corresponding to the multiple two-dimensional vascular distributions, the following processing can also be performed: training the recognition model based on the vascular image sample set of the preset biological site to obtain an initial object recognition model; obtaining a real image sample set obtained by image acquisition of the preset biological site, and training the initial object recognition model based on the real image sample set to obtain an object recognition model.

[0266] In some embodiments, the preset biological site is a pre-defined biological tissue location that requires blood vessel recognition, such as specific locations like the human brain, heart, liver, or limbs. A pre-constructed blood vessel image sample set for the preset biological site is built. This sample set contains several blood vessel images of the preset biological site under different acquisition conditions and physiological states. Each blood vessel image is labeled with corresponding blood vessel region information or blood vessel-related target object information. This labeling information provides supervision for the training of the recognition model. During the construction of the blood vessel image sample set, the acquired raw blood vessel images can be preprocessed. Preprocessing operations include image denoising, image enhancement, image size standardization, or image grayscale adjustment to improve the quality of the sample images and reduce the impact of interference factors on the model training effect.

[0267] In some embodiments, an initially selected recognition model is trained based on a pre-constructed set of vascular image samples from predetermined biological sites. The recognition model includes a feature extraction layer and an object recognition layer. The feature extraction layer employs a convolutional neural network structure, containing several convolutional and pooling layers, to extract multi-dimensional features from the input vascular images. Specifically, it extracts texture features, edge features, grayscale distribution features, and vascular morphology-related features from the vascular images. The object recognition layer, connected after the feature extraction layer, is constructed using a fully connected layer combined with an attention mechanism. It filters and fuses the multi-dimensional features output by the feature extraction layer, thereby completing the classification and recognition of vascular regions or vascular-related target objects, and outputting the corresponding recognition results. During training, vascular images from the vascular image sample set are input into the recognition model. Feature extraction and recognition are completed through the collaborative work of the feature extraction layer and the object recognition layer. The output recognition results are compared with the corresponding annotation information of the sample images, and the difference loss between the two is calculated. The internal parameters of the feature extraction layer and the object recognition layer are adjusted synchronously based on the difference loss. This process is iterated repeatedly until the difference loss reaches a preset threshold or the number of iterations meets a preset requirement. At this point, training stops, and the initial object recognition model is obtained. The initial object recognition model has the preliminary ability to identify blood vessels in preset biological sites. Its feature extraction layer can effectively extract basic blood vessel features, and the object recognition layer can complete preliminary recognition based on the extracted features.

[0268] In some embodiments, a real image sample set is obtained by actually acquiring images of a preset biological site. The acquisition process uses image acquisition equipment and parameters consistent with the subsequent actual application scenario to ensure the adaptability of the real image sample set to the actual application scenario. The real image sample set contains several real blood vessel images of the preset biological site from different individuals at different acquisition times. Each real blood vessel image is also labeled with corresponding blood vessel region information or blood vessel-related target object information, and the labeling rules are consistent with the labeling rules of the aforementioned blood vessel image sample set to ensure the continuity and consistency of the training process.

[0269] In some embodiments, an initial object recognition model is trained based on the aforementioned real image sample set. During training, real blood vessel images from the real image sample set are successively input into the initial object recognition model. The initial object recognition model uses its feature extraction layer to perform deep feature extraction and recognition on the real blood vessel images, enhancing its ability to capture blood vessel features in real-world application scenarios. The object recognition layer then accurately filters, fuses, and recognizes the deeply extracted features, outputting the recognition result. This recognition result is compared with the corresponding annotation information to calculate the recognition error. Based on the recognition error, the internal parameters of the initial object recognition model are continuously adjusted, specifically targeting the convolution kernel parameters and pooling strategy of the feature extraction layer and the fully connected weights and attention allocation coefficients of the object recognition layer. This continuously optimizes the model's feature extraction accuracy and recognition logic. The above training iteration process is repeated until the recognition error decreases to a preset acceptable threshold and the model performance stabilizes. Training is then stopped, and the final object recognition model is obtained. This object recognition model can accurately adapt to the image acquisition conditions in real-world application scenarios. Through the synergistic effect of the feature extraction layer and the object recognition layer, it can accurately identify blood vessel regions or blood vessel-related target objects in real blood vessel images of preset biological sites, meeting the recognition accuracy requirements in practical applications.

[0270] In some embodiments, an initial object recognition model is trained using a vascular image sample set, and then optimized using a real image sample set to improve the model's recognition accuracy and generalization ability. The vascular image sample set includes a first vascular image and multiple images corresponding to two-dimensional vascular distributions. Each vascular image sample corresponds to an object label used to identify the object to which the vascular image belongs. The recognition model is trained using the vascular image sample set, and the model learns the feature representation of the vascular images. Through supervised learning, the model can recognize objects corresponding to vascular images of preset biological sites. Real vascular images are obtained by acquiring images of preset biological sites (such as the palm, retina, etc.). Each real image sample corresponds to an object label used to identify the object to which the real image belongs. The initial object recognition model is trained using the real image sample set, and the model learns the feature representation of real vascular images. Through transfer learning or fine-tuning, the model can better adapt to real image data, improving recognition accuracy and generalization ability. The optimized object recognition model has higher recognition accuracy and robustness.

[0271] Thus, the process involves two stages: First, training an initial object recognition model using a set of vascular image samples, where the model learns the basic features of vascular images. Second, optimizing the initial model using a set of real image samples, where the model learns the feature representations of real images. The vascular image sample set, by introducing noisy features, increases the diversity of samples and helps improve the model's generalization ability. Optimizing the initial model using a set of real image samples, through transfer learning, allows the model to better adapt to real image data. Through these two stages of training, the model learns richer feature representations, improving recognition accuracy and generalization ability.

[0272] As an example, referring to Figure 15, Figure 15 is a schematic diagram of the principle of training the palm vein recognition model provided in the embodiment of this application. Based on the vascular image sample set of the object (the large batch synthetic palm vein dataset shown in Figure 15), the object recognition model (the pre-trained model shown in Figure 15) is trained to obtain the initial object recognition model (the fine-tuned model shown in Figure 15); the real image sample set obtained by image acquisition of the preset biological parts (the publicly available real palm vein dataset shown in Figure 15) is obtained, and the initial object recognition model (the fine-tuned model shown in Figure 15) is trained based on the real image sample set.

[0273] Thus, by acquiring a first blood vessel image of a pre-defined biological part of the object, and constructing a three-dimensional blood vessel distribution based on the first blood vessel image, and adjusting the three-dimensional blood vessel distribution from multiple perspectives, a two-dimensional blood vessel distribution of the three-dimensional blood vessel distribution is obtained, thereby expanding the two-dimensional representation of the three-dimensional blood vessel distribution. By acquiring the blood vessel texture corresponding to each two-dimensional blood vessel distribution, and generating an image corresponding to the two-dimensional blood vessel distribution based on the blood vessel texture, the generated different images reflect different aspects of the pre-defined biological part of the object, thereby enhancing a single first blood vessel image into an image of the three-dimensional blood vessel distribution. By constructing a blood vessel image sample set of the object based on multiple images corresponding to two-dimensional blood vessel distributions, the sample size of the constructed image sample set is effectively improved.

[0274] Thus, a three-dimensional blood vessel distribution of a pre-defined biological site is constructed, and this three-dimensional blood vessel distribution is projected to obtain a two-dimensional blood vessel distribution. Based on this two-dimensional blood vessel distribution, an image corresponding to the two-dimensional blood vessel distribution is generated. Since the three-dimensional blood vessel distribution is three-dimensional, by projecting it, a two-dimensional blood vessel distribution is obtained, thus converting the three-dimensional blood vessel distribution into a two-dimensional form. This ensures that the dimension of the two-dimensional blood vessel distribution matches the dimension of the image. Therefore, projection simulates the actual image acquisition process without requiring the acquisition of a large number of actual images. Simultaneously, projection achieves dimensionality reduction of the three-dimensional blood vessel distribution, resulting in a two-dimensional blood vessel distribution. The image corresponding to this two-dimensional blood vessel distribution is then generated. Since the generated image is also two-dimensional, dimensionality reduction through projection makes the generation process a conversion from a two-dimensional blood vessel distribution to a two-dimensional image, rather than directly generating a two-dimensional image from a three-dimensional blood vessel distribution. This effectively saves the computational cost of directly generating a two-dimensional image from a three-dimensional blood vessel distribution, thereby significantly improving image generation efficiency.

[0275] Referring to Figure 9, which is a flowchart of the method for constructing a three-dimensional blood vessel distribution provided in this application embodiment, the method will be described in conjunction with steps 201 to 203 shown in Figure 9. The image generation method provided in this application embodiment can be implemented by the server or the terminal alone, or by the server and the terminal working together. The following description will take the implementation by the server alone as an example.

[0276] In step 201, the coordinates of multiple key points used to generate the three-dimensional main blood vessel are determined.

[0277] In some embodiments, step 201 above can also be implemented as follows: select one type as the target distribution type from multiple distribution types of blood vessel distribution in the preset biological site; determine the coordinates of multiple key points for generating the three-dimensional trunk blood vessel according to the target distribution type.

[0278] In some embodiments, the coordinates of the plurality of key points are three-dimensional coordinates. The determination of the coordinates of the plurality of key points used to generate the three-dimensional trunk blood vessel according to the target distribution type can be achieved by: identifying the two-dimensional coordinates of the plurality of key points used to generate the three-dimensional trunk blood vessel from the image of the target distribution type; and converting the two-dimensional coordinates of the plurality of key points into three-dimensional coordinates.

[0279] In step 202, a three-dimensional main blood vessel corresponding to the blood vessel is generated based on the coordinates of the multiple key points.

[0280] In some embodiments, the three-dimensional trunk vessel is a vascular network model generated based on key point information from two-dimensional vascular images. It displays the distribution, branching structure, and topological relationships of blood vessels in three-dimensional space. The three-dimensional trunk vessel can intuitively show the direction and distribution of blood vessels in three-dimensional space. It clearly shows the bifurcation points and connections of blood vessels, reflecting the complexity of the vascular network. The three-dimensional trunk vessel preserves the connection methods between blood vessels and the direction of blood flow, facilitating the analysis of the functional characteristics of blood vessels.

[0281] In some embodiments, the coordinates of the plurality of key points are three-dimensional coordinates. The generation of the three-dimensional trunk blood vessel based on the coordinates of the plurality of key points can be achieved in the following manner: For the three-dimensional coordinates of the plurality of key points, the different three-dimensional coordinates are connected to obtain the initial three-dimensional trunk blood vessel of the blood vessel. The initial three-dimensional trunk blood vessel includes a plurality of blood vessel segments, and the blood vessel segments are connected to the key points as endpoints. The radius of each blood vessel segment in the initial three-dimensional trunk blood vessel is determined, and the initial radius of each blood vessel segment in the initial three-dimensional trunk blood vessel is adjusted to the corresponding radius to obtain the three-dimensional trunk blood vessel.

[0282] In some embodiments, the vascular segments in the initial three-dimensional trunk blood vessel have a flow direction. The determination of the radius of each vascular segment in the initial three-dimensional trunk blood vessel can be achieved by performing the following processing on each vascular segment: obtaining the initial radius of the vascular segment in the initial three-dimensional trunk blood vessel, and the number of terminal key points flowing out of the vascular segment according to the flow direction; determining the radius of the vascular segment based on the initial radius and the number, thereby obtaining the three-dimensional trunk blood vessel.

[0283] In some embodiments, determining the radius of the blood vessel segment based on the initial radius and the quantity can be achieved by: obtaining the weight of the blood vessel segment, the weight being used to indicate the importance of the blood vessel segment; multiplying the quantity by the weight to obtain a multiplication result, and adding the multiplication result to the initial radius of the blood vessel segment to obtain the radius of the blood vessel segment.

[0284] In some embodiments, the types of the above-mentioned key points include trunk type, bifurcation type, and leaf type; for the three-dimensional coordinates of the multiple key points, different three-dimensional coordinates are connected to obtain the initial three-dimensional trunk blood vessel of the blood vessel, which can be achieved in the following manner: Connect the three-dimensional coordinates of the key points of each trunk type to obtain a first trunk blood vessel, and connect the three-dimensional coordinates of the key points of each bifurcation type to the three-dimensional coordinates of at least one leaf type key point respectively to obtain a second trunk blood vessel corresponding to each key point of the bifurcation type; Connect the first trunk blood vessel to each of the second trunk blood vessels to obtain the initial three-dimensional trunk blood vessel.

[0285] In some embodiments, the coordinates of the key points refer to the positions of points with significant features (such as bifurcation points, intersection points, and end points) extracted from the two-dimensional blood vessel image on the image plane, usually represented by (x, y). The key points are usually the bifurcation points, intersection points, or end points of the blood vessel, which can describe the topological structure of the blood vessel. The coordinates of the key points only contain plane information (x, y) and lack depth information (z).

[0286] In step 203, three-dimensional branch blood vessels are generated on the three-dimensional trunk blood vessel to obtain the three-dimensional blood vessel distribution of the preset biological part.

[0287] In some embodiments, the above step 203 can also be achieved in the following manner: Select multiple three-dimensional coordinate points not on the three-dimensional trunk blood vessel; Based on the multiple three-dimensional coordinate points, construct three-dimensional branch blood vessels on the three-dimensional trunk blood vessel to obtain the initial three-dimensional blood vessel distribution of the preset biological part; Adjust the shapes of the three-dimensional branch blood vessels in the initial three-dimensional blood vessel distribution to obtain the three-dimensional blood vessel distribution of the preset biological part.

[0288] In some embodiments, the above multiple three-dimensional coordinate points include the i-th three-dimensional coordinate point, 1 < i ≤ N. Based on the multiple three-dimensional coordinate points, constructing three-dimensional branch blood vessels on the three-dimensional trunk blood vessel to obtain the initial three-dimensional blood vessel distribution of the preset biological part can be achieved in the following manner: Determine the three-dimensional trunk blood vessel as the first three-dimensional blood vessel distribution, and determine the blood vessel segment on the (i - 1)-th three-dimensional blood vessel distribution that is closest to the i-th three-dimensional coordinate point as the i-th blood vessel segment, and generate the i-th three-dimensional blood vessel distribution based on the i-th blood vessel segment; Traverse i to obtain the N-th three-dimensional blood vessel distribution, and determine the N-th three-dimensional blood vessel distribution as the initial three-dimensional blood vessel distribution.

[0289] In some embodiments, the generation of the i-th three-dimensional blood vessel distribution based on the i-th blood vessel segment may be achieved as follows: Determine the region formed by the endpoints of the i-th blood vessel segment and the i-th three-dimensional coordinate points, and determine the bifurcation points of the i-th three-dimensional blood vessel distribution in the region; Delete the i-th blood vessel segment from the (i - 1)-th three-dimensional blood vessel distribution, and connect the endpoints of the i-th blood vessel segment, the i-th three-dimensional coordinate points, and the bifurcation points to obtain the i-th three-dimensional blood vessel distribution.

[0290] In some embodiments, the adjustment of the shapes of the three-dimensional branch blood vessels in the initial three-dimensional blood vessel distribution to obtain the three-dimensional blood vessel distribution of the preset biological site may be achieved as follows: Perform the following processing on each of the three-dimensional branch blood vessels in the initial three-dimensional blood vessel distribution to obtain the three-dimensional blood vessel distribution: Select a plurality of three-dimensional coordinate points on at least one side of the three-dimensional branch blood vessel, and connect the endpoints of the three-dimensional branch blood vessel and the plurality of three-dimensional coordinate points in sequence to obtain a reference three-dimensional branch blood vessel, which is different from the three-dimensional branch blood vessel; Replace the three-dimensional branch blood vessel in the initial three-dimensional blood vessel distribution with the reference three-dimensional branch blood vessel.

[0291] In some embodiments, the plurality of three-dimensional coordinate points include the j-th three-dimensional coordinate point, where 1 < j ≤ M, and the endpoints of the three-dimensional branch blood vessel include the starting point of the three-dimensional branch blood vessel; The selection of a plurality of three-dimensional coordinate points on at least one side of the three-dimensional branch blood vessel may be achieved as follows: Traverse j and perform the following processing: When j = 1, superimpose the random perturbation step corresponding to the first three-dimensional coordinate point on the starting point of the three-dimensional branch blood vessel in the random perturbation direction corresponding to the first three-dimensional coordinate point to obtain the first three-dimensional coordinate point; When j > 1, superimpose the random perturbation step corresponding to the j-th three-dimensional coordinate point on the (j - 1)-th three-dimensional coordinate point in the random perturbation direction corresponding to the j-th three-dimensional coordinate point to obtain the j-th three-dimensional coordinate point, and the random perturbation direction points to at least one side of the three-dimensional branch blood vessel.

[0292] In some embodiments, the constructed three-dimensional blood vessel distribution can be used to generate an image corresponding to the three-dimensional blood vessel distribution (e.g., palm blood vessel image, etc.), and can also be used in multiple fields such as biomechanics, medical education, virtual reality, etc. The subsequent application process of the constructed three-dimensional blood vessel distribution will be illustrated by examples below.

[0293] In some embodiments, the constructed three-dimensional blood vessel distribution can be used to generate an image corresponding to the three-dimensional blood vessel distribution (e.g., palm blood vessel image, etc.). After the above step 203, the following processing may also be performed: Project the three-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution of the three-dimensional blood vessel distribution; Generate an image corresponding to the two-dimensional blood vessel distribution according to the two-dimensional blood vessel distribution.

[0294] In some embodiments, a two-dimensional vascular distribution refers to a graphical representation of a three-dimensional vascular distribution projected onto a two-dimensional plane. This representation is typically used to visualize and analyze vascular structure, but it loses the depth information found in three dimensions. Two-dimensional vascular distributions can be generated in various ways, such as perspective projection, orthographic projection, or other projection techniques in computer graphics. In a two-dimensional vascular distribution, the branches, directions, and connections of blood vessels are clearly shown, but the thickness of the vessels or their specific location in three-dimensional space cannot be directly represented.

[0295] In some embodiments, projection is a graphics technique that involves projecting a 3D scene from multiple different angles to obtain multiple 2D views. In each view, certain aspects of the 3D scene are emphasized or highlighted, while others may be occluded or reduced. Projection allows for a comprehensive observation and analysis of the 3D structure, leading to a deeper understanding.

[0296] In some embodiments, the above-described projection of the three-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution can also be achieved by: obtaining the projection matrix corresponding to the viewpoint, the projection matrix being used to indicate the relationship between the three-dimensional blood vessel distribution and the two-dimensional blood vessel distribution under the viewpoint; and based on the projection matrix, projecting the three-dimensional blood vessel distribution under the viewpoint to obtain the two-dimensional blood vessel distribution under the viewpoint.

[0297] In some embodiments, the above-mentioned generation of an image corresponding to the two-dimensional blood vessel distribution based on the two-dimensional blood vessel distribution can be achieved by: obtaining the skin texture of the preset biological site; superimposing the skin texture with the two-dimensional blood vessel distribution to obtain the blood vessel texture corresponding to the two-dimensional blood vessel distribution; and generating an image corresponding to the two-dimensional blood vessel distribution based on the blood vessel texture.

[0298] In some embodiments, the constructed three-dimensional vascular distribution can be used for medical diagnosis and disease analysis. Through the three-dimensional vascular distribution, doctors can intuitively observe the morphology and structure of blood vessels, which helps in the diagnosis of vascular diseases such as aneurysms, stenosis, and thrombosis. By detecting changes in the three-dimensional vascular distribution, disease progression and treatment effectiveness can be assessed.

[0299] In some embodiments, the constructed three-dimensional vascular distribution can be used to plan the path of vascular interventional surgery, ensuring the safety and effectiveness of the procedure. During the surgery, the three-dimensional vascular distribution can serve as part of a navigation system, assisting surgeons in performing accurate surgical operations. Based on the three-dimensional vascular distribution, hemodynamic simulation studies can be conducted to analyze blood flow within the vessels. By simulating the pressure and stress distribution within the vessels, the mechanical properties of the vessel walls can be studied.

[0300] In some embodiments, the constructed three-dimensional vascular distribution can be used for hemodynamic simulation studies to analyze blood flow within the vessels. By simulating the pressure and stress distribution within the vessels, the mechanical properties of the vessel walls can be studied.

[0301] In some embodiments, the constructed three-dimensional vascular distribution can be used for medical education, providing students with an intuitive anatomical teaching model. Through virtual reality technology, students can conduct simulated surgical training on the three-dimensional vascular distribution model.

[0302] Thus, by using keypoint recognition to process the first vascular image of the preset biological site, multiple key points within the vascular region can be effectively identified and their two-dimensional coordinates obtained. This process significantly reduces the complexity and computational load of subsequent three-dimensional vascular reconstruction. Based on these precise keypoint coordinates, a three-dimensional trunk vascular model corresponding to the vascular site can be generated efficiently, ensuring the accuracy and morphological integrity of the three-dimensional trunk vascular model. Constructing the branch vessels of the three-dimensional trunk vascular system ultimately yields the three-dimensional vascular distribution of the preset biological site, which not only improves the detail of the model but also enhances the practicality of the vascular distribution model.

[0303] The following will describe an exemplary application of the embodiments of this application in a real-world palm vein recognition scenario.

[0304] Palm vein recognition is an emerging biometric identification technology. Unlike external biometric technologies such as facial recognition, fingerprint recognition, and iris recognition, palm veins are located beneath the skin. Therefore, palm vein images are more difficult to capture and less likely to be illegally obtained by others. This means that palm vein recognition technology has unique advantages in protecting user privacy and providing higher security.

[0305] However, obtaining sufficient palm vein data to train deep learning recognition models is challenging due to high data collection costs and privacy restrictions. This has sparked interest in generating pseudo-palm vein data using generative models. Related techniques often produce unrealistic palm vein patterns or struggle with controlling identity and style attributes. To address these issues...

[0306] In this embodiment, palm vein identity is defined by a 3D palm blood vessel distribution generated by an improved constraint construction optimization algorithm. Palm vein textures with the same identity but without intra-class variations (variations within the same category) are obtained by projecting the same 3D blood vessel distribution onto a 2D image from different perspectives. Therefore, this embodiment satisfies the requirements of identity consistency and intra-class diversity. Extensive experiments on multiple public datasets demonstrate that this embodiment outperforms existing methods and achieves higher recognition accuracy under a 1:1 open set protocol.

[0307] With increasing public concern about privacy, palm vein recognition will see wider application in real-world scenarios such as payments and identity verification. This application's embodiment improves model performance by using a large-scale synthetic palm vein dataset, combined with palm detection technology, to ultimately achieve high-precision user identity verification.

[0308] In some embodiments, referring to Figure 10, which is a schematic flowchart of palm vein recognition provided in this application embodiment, the palm vein recognition process can be implemented through steps 301 to 307 shown in Figure 10. Before each user starts recognition, the user's palm veins need to be registered in the background registration database. In each subsequent recognition, the user's recognition photo will be compared with each photo in the registration database to identify the palm veins, and finally the user's identity can be determined.

[0309] In step 301, an identification photo is taken.

[0310] In some embodiments, taking an identification photo refers to the process of capturing an image during palm vein recognition to verify user identity. When a user needs to authenticate, they place their palm on the recognition device again, and the system takes a new image of the palm veins, the so-called identification photo. This identification photo is taken in real time and is used to compare with palm vein images stored in a registry. During registration, the user is required to place their palm on a specific palm vein recognition device, which captures an image of the user's palm veins. When the user needs to authenticate, they place their palm on the recognition device again, and a new image of the palm veins, the so-called identification photo, is taken in real time and is used to compare with palm vein images stored in a registry.

[0311] In step 302, a registration photo is taken.

[0312] In some embodiments, users must complete a registration process before they can authenticate themselves using the palm vein recognition system. The purpose of the registration phase is to collect and store unique images of the user's palm veins, which serve as a biometric identifier. The user is asked to place their palm on the recognition device, typically a specially designed palm vein scanner. The device captures images of the veins in the user's palm, clearly showing the network of veins within the hand. This captured image is known as the registration photo, a digital record of the user's palm veins. The registration photo is stored in the system's backend registry.

[0313] In step 303, the palm veins are detected and extracted.

[0314] In some embodiments, during user registration or identification, the position and outline of the user's palm are first detected. This is typically achieved through image processing techniques, such as using information like color, texture, and shape from images captured by a camera to identify and locate the palm. The purpose of palm detection is to ensure accurate identification of the user's palm and exclude other irrelevant parts of the image. Once the palm is detected, the image of the palm region is analyzed to extract detailed features of the palm veins. The image contrast and brightness are adjusted to more clearly display the veins.

[0315] In step 304, the data is transmitted to the backend.

[0316] In some embodiments, transmission to the backend refers to the process of sending a user's palm vein image or related data to a server or database for storage, processing, and analysis. During user registration, their palm vein image is captured using a palm vein recognition device. The image data is then transmitted to a backend system, typically a central server or database. Once the image data arrives at the backend, it is stored in a registry, a dedicated data storage solution used to store the user's biometric data and associated identity information.

[0317] In step 305, it is added to the registry.

[0318] In some embodiments, adding to the registry refers to the process of storing a user's palm vein image and its related feature information in a central database or registry during palm vein recognition. Users must complete the registration process before they can use palm vein recognition for authentication. At the start of the registration process, the user's palm vein image is captured using a dedicated scanning or recognition device. After capturing the vein image of the user's palm, the device processes these images to extract the features of the palm veins. The extracted palm vein feature data and the original image are transmitted to the backend system. The data arriving at the backend system is then added to the registry. The registry is a specially designed data storage solution for storing users' biometric features and identity information. In the registry, each user's palm vein data is associated with their identity identifier (such as name, ID number, etc.), forming a record.

[0319] In step 306, palm vein recognition is performed.

[0320] In some embodiments, palm vein recognition is a biometric technology that uses each person's unique palm vein pattern for identity verification. When a user needs to verify their identity, they place their palm on the palm vein recognition device again, and the device captures a new palm vein image, also known as a recognition photo. The palm vein features extracted in real time are compared with palm vein features stored in a registry. If the palm vein features extracted in real time match the features of a user in the registry, the user's identity is confirmed.

[0321] In step 307, the recognition result is returned to the front end.

[0322] In some embodiments, returning the recognition result to the front end refers to the process in a palm vein recognition system of sending the authentication result from the backend server back to the user interface. Once the backend system has completed the palm vein recognition comparison and determined the user's identity, it prepares to send this result back to the front end. The front end, i.e., the user interface, such as a mobile application, webpage, or desktop software, receives the recognition result returned from the backend. The front end application typically designs an interface to process this returned data and display it to the user. The user will receive feedback based on the recognition result displayed on the front end. For example, if the recognition is successful, the user may see a verification success message and be allowed to proceed to or perform the next step; if the recognition fails, the user may receive a verification failure message and be asked to retry.

[0323] In some embodiments, referring to Figure 11, which is a schematic diagram of the anatomical structure of the palm blood vessels and the palm vein image under near-infrared light provided in the embodiments of this application, the significant features of the palm blood vessels can be summarized as follows: The blood vessels of the palm are mainly supplied by two arteries from the wrist: the ulnar artery (near the little finger) and the radial artery (near the thumb); these two arteries are interconnected to form the superficial palmar arch and the deep palmar arch, forming a complex vascular network in the palm area; from these two palmar arches, many small branch vessels extend out to supply the fingers and other parts of the palm. Each finger has a pair of arteries that supply blood to at least one side of the finger and the fingertip.

[0324] In some embodiments, referring to Figure 12, which is a schematic diagram of the structural types of palm blood vessels provided in this application embodiment, and combining the patterns summarized in Figure 11 above, and observing a large amount of real palm vein data, the four types shown in Figures 12(a) to 12(d) are finally identified. Figure 12(a) represents the type with a palmar arch, and Figures 12(c), 12(b), and 12(d) represent the type without a palmar arch. Among them, Figures 12(c) and 12(d) represent the type where only a single artery extends outwards, and Figure 12(b) represents the type where both the ulnar and radial arteries extend outwards. These patterns can better serve the subsequent modeling of palm vein texture.

[0325] In some embodiments, referring to Figure 12, to ensure the realism of blood vessel modeling, the main veins are derived from the distribution of real images. Figure 12(e) illustrates the main vein generation process of Figure 12(a). Palm vein images of the corresponding type are selected from the real database and first cropped to fit the palm region corresponding to Ω, as shown in the box in Figure 12. Then, the key points of the palm veins are divided into four types based on their distribution characteristics, namely the root node v... r Palm arch bifurcation point v rb Digital vein bifurcation point v fb and digital vein nodes v f Where v r This is the inflow point of the radial and ulnar arteries, which form the palmar arch and extend to other areas. rb The fork point where the palm arch extends in all directions, v fb This is the inflow point of the digital artery, v f By v fb Starting from the five fingers, these key points ensure the inflow and outflow direction of each blood vessel segment, as shown by the arrows in Figure 12(e).

[0326] In some embodiments, referring to Figure 13, which is a schematic diagram of the image generation method provided in this application embodiment, the main trunk of the palm blood vessel distribution is first generated based on the distribution of real images. Figure 12(e) shows the trunk generation process of Figure 12(a). Palm vein images of the corresponding type are selected from the real database and cropped to satisfy the palm region corresponding to Ω, as shown in the box in Figure 12. Then, the key points of the palm veins are divided into four types according to their distribution characteristics, namely the root node v... r Palm arch bifurcation point v rb Digital vein bifurcation point v fb and digital vein nodes v f Where v r This is the inflow point of the radial and ulnar arteries, which form the palmar arch and extend to other areas. rb The fork point where the palm arch extends in all directions, v fb This is the inflow point of the digital artery, v f By v fb Starting from the five fingers, these key points ensure the inflow and outflow direction of each blood vessel segment, as shown by the arrows in Figure 12(e). Then, each key point v(x) needs to be... 2d y 2d Its transfer to Ω space yields a three-dimensional representation v(x) 3d y 3d , z 3d ), which is defined as:

[0327] The function `Uniform(40-D / 2, 40+D / 2)` represents a uniformly distributed random number generator that generates random numbers within a specified interval. `40-D / 2` is the left boundary of the interval, and `40+D / 2` is the right boundary. `D` is a given parameter that determines the size of the interval. This function generates a random number within the interval [40-D / 2, 40+D / 2] (inclusive). For example, if `D = 20`, the interval will be [40-20 / 2, 40+20 / 2], which is [30, 50]. If `D = 10`, the interval will be [40-10 / 2, 40+10 / 2], which is [35, 45]. The x-axis of the two-dimensional real image corresponds to the x-axis of the three-dimensional blood vessel distribution, and the y-axis of the two-dimensional real image corresponds to the z-axis of the three-dimensional blood vessel distribution. The y-axis of the three-dimensional blood vessel distribution is generated by a uniformly distributed random tree generation function. The range (40-D / 2, 40+D / 2) is only used to center the three-dimensional blood vessel distribution.

[0328] In some embodiments, referring to Figure 13, the generated palm blood vessel distribution backbone is constrained and optimized. For the radius optimization of each blood vessel segment, the relationship between the radius and the outflow of the blood vessel segment is established as follows: r i =r roi +n i *ratioE (8)

[0329] Where r roi To initialize the radius, n i To segment blood vessels i The number of outflowing terminal nodes is represented by ratioE, which is their respective weights (the weights can be determined according to the anatomical structures corresponding to the vessel segments; the radial and ulnar arteries have a greater weight than the digital arteries). This completes the construction of the main vascular distribution trunk for type (a). The construction of the trunks for types (b), (c), and (d) is similar.

[0330] In some embodiments, referring to Figure 13, branch generation is performed on the constrained optimized vascular distribution trunk to obtain a three-dimensional palm vascular distribution. Iteratively adding N points within Ω, connecting them to the trunk to form vascular distribution branches, allows finding the optimal location of new bifurcation points under the constraint of minimizing the total volume of the vascular distribution. Specifically, referring to the branch vascular generation in Figure 14 or Figure 13, Figure 14 is a schematic diagram of the branch generation principle provided in this embodiment. p2 is a newly added node, which needs to be connected to the nearest vascular segment s1 to form a new vein branch. The bifurcation point p is located within the triangle formed by the two endpoints p0 and p1 of s1 and the newly added point p2. Subsequently, p2 and p need to be connected to the existing vascular distribution. During this process, vascular segment s1 is removed and reconnected to p and p1, while bifurcation point p is connected to p0 and p2 to form new vascular segments s0 and s2. With the optimization objective being to minimize the total vascular volume, the coordinates (x, y, z) of the bifurcation point p can be calculated as follows:

[0331] Among them, (x i y i , z i ) is p i The coordinates, l i r i Segmentation of blood vessels i The length and radius of the newly formed blood vessel segment are set according to the radius optimization method described above.

[0332] In some embodiments, referring to Figure 13, a simulated 3D palm blood vessel distribution is obtained according to a trajectory simulation mechanism. Since each blood vessel segment is modeled as a cylinder in 3D space, its projection is a line segment. However, the real palm blood vessel pattern is usually composed of curves. To better simulate this, we propose a trajectory simulation mechanism. The overall process is shown in the trajectory simulation mechanism in Figure 13. Assume T is the generated 3D palm blood vessel distribution, s is a blood vessel segment, and p1 and p2 are the inflow and outflow points of this segment, respectively. The inflow point p1 is regarded as a moving node p move Its target direction is:

[0333] d target =p move -p2 (10)

[0334] The step size for each movement is l step During each movement, we apply a random perturbation to its main direction d. Therefore, we can obtain the position P after the i-th movement. i for:

[0335] Pi =P i-1 +l step *(d target +w rand *d rand (11)

[0336] Among them, w rand The weights are randomly perturbed, and when the moving step size is l step When the value is small enough, an approximate curve trajectory from p1 to p2 can be obtained.

[0337] In some embodiments, referring to Figure 13, a two-dimensional palm vein texture is generated based on the simulated three-dimensional palm blood vessel distribution. The simulated three-dimensional palm blood vessel distribution is projected to obtain a two-dimensional palm blood vessel distribution. Then, the two-dimensional palm blood vessel distribution is cropped to obtain a cropped two-dimensional palm blood vessel distribution. Then, the cropped blood vessel distribution and texture curve are image-blended and image enhancement is performed to obtain the palm blood vessel texture. Based on the noise collected from N(0,1) with l2 distance constraint, the generator is called to generate a vein image of the palm blood vessel texture to obtain a palm vein image.

[0338] In some embodiments, referring to Figure 13, during the acquisition of palm vein datasets, images of the same identity are typically affected by factors such as acquisition viewpoint, vein depth, palm pose, acquisition environment, and palm print. To simulate the influence of palm print, we combine the Bezier curve (the texture curve shown in Figure 13) with the vascular texture projected in 3D. For other influencing factors, several control components are introduced to simulate them. Interference in 3D projection: 3D palm vein distribution has a natural advantage in simulating acquisition viewpoint and vein depth. By rotating the 3D palm vein distribution around the z-axis, we obtained projected images from different viewpoints. It is worth noting that large rotation angles may cause significant changes in the vein texture representing identity, which is detrimental to maintaining the uniqueness of identity. After multiple tests, we found that the optimal rotation range for the vein distribution is between -3 and 3 degrees.

[0339] Furthermore, for the y-axis value representing the depth of the blood vessel, its relationship with the grayscale value G of the projection is established as follows:

[0340] Among them, y max and y min w represents the maximum and minimum depths of vessel segments in the 3D vessel distribution. random It is a random perturbation that causes a slight change in grayscale value with each projection, corresponding to the effect of depth changes on veins.

[0341] In some embodiments, interference with palm vein texture: To simulate the influence of hand pose, we employed several image enhancement techniques. Specifically, we applied slight random scaling, rotation, distortion, and cropping to the palm vein texture to obtain variations of the same identity under different hand poses. Constraints on the sampling noise distance: For other intra-class variations, noise with L2 distance constraints was used instead of random noise from a normal distribution. By adjusting the L2 distance, vein images with different degrees of intra-class variation were obtained. After obtaining the palm vein texture, rendering it into a realistic vein image is crucial. To this end, we used a graph-to-image transformation model, and to better apply it to the task of this application, we improved its inference stage. In the inference stage, we removed the step of transferring lines to the PCE domain in the method. This is because real vein datasets often exhibit local blurring due to varying vein depths; therefore, the lines extracted using PCE M represent the clearly imaged parts of the vein rather than the complete vein texture. A generative model trained using these non-depth lines, i.e., pure black lines, will give the synthesized image an indication that the areas of dark lines correspond to the clearly visible vein areas in the synthesized image. The vein lines we generate are complete vein textures containing depth information. Due to the aforementioned characteristics, the generated images obtained by directly using these lines for inference correspond to areas with prominent veins and high vein depth information. This preserves the depth information of the veins well and can better simulate the local blurring characteristics of real veins.

[0342] In some embodiments, referring to Figure 15, which is a schematic diagram of the principle of training a palm vein recognition model provided in this application embodiment, the pre-trained model is trained using the palm vein images (large-batch synthetic palm vein dataset) obtained above to obtain a fine-tuned model. Then, the fine-tuned model is trained using a publicly available real palm vein dataset to obtain a palm vein recognition model. Synthetic palm vein datasets can help the model pay attention to subtle changes in texture, prompting the model to learn more discriminative features. However, there are still certain differences between synthetic and real palm veins. Therefore, training the model with both synthetic and real data improves the model's recognition performance in real-world scenarios.

[0343] In some embodiments, refer to Table 1 below. Table 1 is a performance comparison table of the embodiments of this application and related technologies. Table 1 shows the recognition performance of the recognition model pre-trained on the synthetic datasets of the present invention and other synthetic datasets in the field of biometric generation on three public datasets. The evaluation indicators are evaluation indicator A and evaluation indicator B, respectively. The smaller the evaluation indicator A is, the better, and the higher the evaluation indicator B is, the better. It can be seen that the performance of the recognition model trained on the dataset generated by the embodiments of this application far exceeds that of the prior art.

[0344] Table 1. Performance Comparison of Embodiments of This Application and Related Technologies

[0345] Thus, the texture modeling approach effectively improves the palm vein recognition performance of the model, enabling the network to learn more discriminative features. It does not incur additional computational or training burdens, is simple to implement, and is highly effective. By improving palm vein recognition performance, the user experience of using palm recognition for payments, identity verification, access control, and public transportation can be enhanced.

[0346] It is understood that in the embodiments of this application, data related to vascular images are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0347] The following description further illustrates the exemplary structure of the image generation apparatus 455 provided in this application embodiment as a software module. In some embodiments, as shown in FIG2, the software module stored in the image generation apparatus 455 in the memory 450 may include: an acquisition module configured to construct a three-dimensional blood vessel distribution of a preset biological part, the three-dimensional blood vessel distribution being used to represent the three-dimensional distribution pattern of the preset biological part; a projection module configured to project the three-dimensional blood vessel distribution to obtain a two-dimensional blood vessel distribution of the three-dimensional blood vessel distribution, the two-dimensional blood vessel distribution being used to represent the two-dimensional distribution pattern of the preset biological part; and a generation module configured to generate an image of the preset biological part based on the two-dimensional blood vessel distribution.

[0348] In some embodiments, the three-dimensional vascular distribution includes three-dimensional trunk vessels and three-dimensional branch vessels; the above-mentioned construction module is further configured to determine the coordinates of multiple key points for generating the three-dimensional trunk vessels; generate the three-dimensional trunk vessels based on the coordinates of the multiple key points; and generate the three-dimensional branch vessels based on the three-dimensional trunk vessels to obtain the three-dimensional vascular distribution of the preset biological site.

[0349] In some embodiments, the coordinates of the plurality of key points are three-dimensional coordinates; the above-described construction module is further configured to identify the two-dimensional coordinates of the plurality of key points used to generate the three-dimensional trunk blood vessel from the image of the target distribution type; and convert the two-dimensional coordinates of the plurality of key points into three-dimensional coordinates.

[0350] In some embodiments, the coordinates of the multiple key points are three-dimensional coordinates; the above construction module is further configured to connect different three-dimensional coordinates of the multiple key points to obtain an initial three-dimensional main blood vessel of the blood vessel, where the initial three-dimensional main blood vessel includes multiple blood vessel segments, and the blood vessel segments have the key points as endpoints; determine the radius of each blood vessel segment in the initial three-dimensional main blood vessel to obtain the three-dimensional main blood vessel.

[0351] In some embodiments, the blood vessel segments in the initial three-dimensional main blood vessel have flow directions; the above construction module is further configured to perform the following processing for each blood vessel segment respectively to obtain the three-dimensional main blood vessel: obtain the initial radius of the blood vessel segment in the initial three-dimensional main blood vessel and the number of end key points flowing out of the blood vessel segment determined according to the flow direction; determine the radius of the blood vessel segment based on the initial radius and the number to obtain the three-dimensional main blood vessel.

[0352] In some embodiments, the above construction module is further configured to obtain the weight of the blood vessel segment, where the weight is used to indicate the importance degree of the blood vessel segment; multiply the number by the weight to obtain a multiplication result, and add the multiplication result to the initial radius of the blood vessel segment to obtain the radius of the blood vessel segment.

[0353] In some embodiments, the types of the key points include a main trunk type, a bifurcation type, and a leaf type; the above construction module is further configured to connect the three-dimensional coordinates of each key point of the main trunk type to obtain a first main blood vessel, and connect the three-dimensional coordinates of each key point of the bifurcation type to the three-dimensional coordinates of at least one key point of the leaf type respectively to obtain a second main blood vessel corresponding to each key point of the bifurcation type; connect the first main blood vessel to each second main blood vessel respectively to obtain the initial three-dimensional main blood vessel.

[0354] In some embodiments, the multiple three-dimensional coordinate points include the i-th three-dimensional coordinate point, 1 < i ≤ N, the above construction module is further configured to determine the three-dimensional main blood vessel as the first three-dimensional blood vessel distribution, determine the blood vessel segment on the (i - 1)-th three-dimensional blood vessel distribution that is closest to the i-th three-dimensional coordinate point as the i-th blood vessel segment, and generate the i-th three-dimensional blood vessel distribution based on the i-th blood vessel segment; traverse i to obtain the N-th three-dimensional blood vessel distribution, and determine the N-th three-dimensional blood vessel distribution as the initial three-dimensional blood vessel distribution.

[0355] In some embodiments, the above construction module is further configured to determine the region formed by the endpoints of the ith blood vessel segment and the ith three-dimensional coordinate point, and determine the bifurcation point of the ith three-dimensional blood vessel distribution in the region; delete the ith blood vessel segment from the (i - 1)th three-dimensional blood vessel distribution, and connect the endpoints of the ith blood vessel segment, the ith three-dimensional coordinate point and the bifurcation point to obtain the ith three-dimensional blood vessel distribution.

[0356] In some embodiments, the above construction module is further configured to perform the following processing for each of the branch blood vessels in the initial three-dimensional blood vessel distribution to obtain the three-dimensional blood vessel distribution: select a plurality of three-dimensional coordinate points on at least one side of the branch blood vessel, and sequentially connect the endpoints of the branch blood vessel and the plurality of three-dimensional coordinate points to obtain the connected branch blood vessel, which is different from the branch blood vessel; replace the branch blood vessel in the initial three-dimensional blood vessel distribution with the connected branch blood vessel.

[0357] In some embodiments, the plurality of three-dimensional coordinate points include the jth three-dimensional coordinate point, where 1 < j ≤ M. The above construction module is further configured to traverse j and perform the following processing: when j = 1, superimpose the random perturbation step corresponding to the 1st three-dimensional coordinate point on the starting point of the branch blood vessel in the random perturbation direction corresponding to the 1st three-dimensional coordinate point to obtain the 1st three-dimensional coordinate point; when j > 1, superimpose the random perturbation step corresponding to the jth three-dimensional coordinate point on the (j - 1)th three-dimensional coordinate point in the random perturbation direction corresponding to the jth three-dimensional coordinate point to obtain the jth three-dimensional coordinate point, and the random perturbation direction points to at least one side of the branch blood vessel.

[0358] In some embodiments, the above construction module is further configured to obtain at least one projection view angle, and perform the following processing for each of the view angles: obtain the projection matrix corresponding to the view angle, which is used to indicate the relationship between the three-dimensional blood vessel distribution and the two-dimensional blood vessel distribution under the view angle; project the three-dimensional blood vessel distribution under the view angle based on the projection matrix to obtain the two-dimensional blood vessel distribution under the view angle.

[0359] In some embodiments, the above construction module is further configured to multiply the coordinates of each three-dimensional coordinate point in the three-dimensional blood vessel distribution by the projection matrix respectively to obtain the two-dimensional coordinate points corresponding to each three-dimensional coordinate point; determine the two-dimensional blood vessel distribution based on the two-dimensional graph formed by the two-dimensional coordinate points.

[0360] In some embodiments, the above-mentioned construction module is further configured to determine the two-dimensional graphic formed by each of the two-dimensional coordinate points as an initial two-dimensional blood vessel distribution, the initial two-dimensional blood vessel distribution including multiple initial blood vessel segments; and to adjust the grayscale value of each initial blood vessel segment in the initial two-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution under the viewpoint, wherein there are at least two blood vessel segments with different grayscale values ​​in the two-dimensional blood vessel distribution.

[0361] In some embodiments, the above-described construction module is further configured to perform the following processing on each initial blood vessel segment in the initial two-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution from the perspective: obtaining the depth value of the initial blood vessel segment and subtracting the minimum depth value of the blood vessel segment in the three-dimensional blood vessel distribution from the depth value to obtain a first depth value; obtaining the difference between the maximum depth value and the minimum depth value of the blood vessel segment in the three-dimensional blood vessel distribution, and determining a first grayscale value corresponding to the initial blood vessel segment based on the difference and the first depth value; adjusting the second grayscale value of the initial blood vessel segment in the initial two-dimensional blood vessel distribution to the first grayscale value.

[0362] In some embodiments, the above-mentioned generation module is further configured to acquire the skin texture of the preset biological site, superimpose the skin texture on the two-dimensional blood vessel distribution to obtain the blood vessel texture corresponding to the two-dimensional blood vessel distribution, and generate an image corresponding to the two-dimensional blood vessel distribution based on the blood vessel texture.

[0363] In some embodiments, the above-mentioned generation module is further configured to acquire multiple equally distributed noise features, and perform the following processing on each of the two-dimensional blood vessel distributions: based on the blood vessel texture corresponding to the two-dimensional blood vessel distribution and each of the noise features, predict the image corresponding to the two-dimensional blood vessel distribution to obtain multiple images corresponding to the two-dimensional blood vessel distribution, wherein the images correspond one-to-one with the noise features.

[0364] In some embodiments, the above-described construction module is further configured to construct a vascular image sample set of the preset biological site using images corresponding to the multiple two-dimensional vascular distributions as vascular image samples; train the recognition model based on the vascular image sample set of the preset biological site to obtain an initial object recognition model; obtain a real image sample set obtained by image acquisition of the preset biological site, and train the initial object recognition model based on the real image sample set.

[0365] This application provides a device for constructing a three-dimensional blood vessel distribution, comprising:

[0366] The identification module is configured to determine the coordinates of multiple key points used to generate the three-dimensional main blood vessel.

[0367] The generation module is configured to generate the three-dimensional main blood vessel based on the coordinates of the multiple key points;

[0368] The construction module is configured to construct the branch vessels of the three-dimensional main blood vessel to obtain the three-dimensional blood vessel distribution of the preset biological site.

[0369] In some embodiments, the identification module is further configured to select one type as a target distribution type from multiple distribution types of blood vessel distribution in the preset biological site; and determine the coordinates of multiple key points for generating the three-dimensional trunk blood vessel based on the target distribution type.

[0370] In some embodiments, the coordinates of the plurality of key points are three-dimensional coordinates; the above-mentioned identification module is further configured to connect the different three-dimensional coordinates of the plurality of key points to obtain the initial three-dimensional trunk blood vessel of the blood vessel, the initial three-dimensional trunk blood vessel including a plurality of blood vessel segments, the blood vessel segments having the key points as endpoints; and to determine the radius of each blood vessel segment in the initial three-dimensional trunk blood vessel to obtain the three-dimensional trunk blood vessel.

[0371] This application provides a computer program product, which includes a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the electronic device to perform the image generation method and the three-dimensional blood vessel distribution construction method described in this application embodiment.

[0372] This application provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor will execute the image generation method and the three-dimensional blood vessel distribution construction method provided in this application, such as the image generation method shown in FIG4.

[0373] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of electronic devices including one or any combination of the above-mentioned memories.

[0374] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0375] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file specifically configured for the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0376] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0377] In summary, the embodiments of this application have the following beneficial effects:

[0378] (1) Construct a three-dimensional blood vessel distribution of a pre-defined biological site, and project the three-dimensional blood vessel distribution to obtain a two-dimensional blood vessel distribution. Based on the two-dimensional blood vessel distribution, generate an image corresponding to the two-dimensional blood vessel distribution. Thus, since the three-dimensional blood vessel distribution is three-dimensional, by projecting the three-dimensional blood vessel distribution, a two-dimensional blood vessel distribution is obtained, thereby converting the three-dimensional blood vessel distribution into a two-dimensional blood vessel distribution in two-dimensional form. This makes the dimension of the two-dimensional blood vessel distribution consistent with the dimension of the image. Thus, the actual image acquisition process is simulated through projection, without the need to acquire a large number of actual images. At the same time, the dimension reduction of the three-dimensional blood vessel distribution is achieved through projection, resulting in a two-dimensional blood vessel distribution. Based on the two-dimensional blood vessel distribution, an image corresponding to the two-dimensional blood vessel distribution is generated. Since the generated image is also two-dimensional, dimension reduction is achieved through projection, making the generation process a conversion process from a two-dimensional blood vessel distribution to a two-dimensional image, instead of directly generating a two-dimensional image from a three-dimensional blood vessel distribution. This effectively saves the computational cost of directly generating a two-dimensional image from a three-dimensional blood vessel distribution, thereby effectively improving the image generation efficiency.

[0379] (2) By acquiring the first blood vessel image of the preset biological part of the object, and constructing the three-dimensional blood vessel distribution of the object based on the first blood vessel image, the three-dimensional blood vessel distribution is adjusted from multiple perspectives to obtain the two-dimensional blood vessel distribution of the three-dimensional blood vessel distribution, thereby realizing the extension of the two-dimensional representation of the three-dimensional blood vessel distribution. By acquiring the blood vessel texture corresponding to each two-dimensional blood vessel distribution, and generating the image corresponding to the two-dimensional blood vessel distribution based on the blood vessel texture, the generated different images reflect different aspects of the preset biological part of the object, thereby enhancing the single first blood vessel image into the image of the three-dimensional blood vessel distribution. By constructing the blood vessel image sample set of the object based on the images corresponding to multiple two-dimensional blood vessel distributions, the sample size of the constructed image sample set is effectively improved.

[0380] (3) By adjusting the radius of the vessel segments in the initial 3D trunk vessels based on initial vessel segment parameters (such as the initial radius and the number of key points connecting the endpoints), the accuracy and realism of the 3D trunk vessels can be significantly improved. For each vessel segment, a target radius that better reflects the actual morphology of the vessel is calculated based on its initial radius and the number of key points connecting the endpoints, and the radius of the vessel segment is adjusted from the initial radius to the target radius. This adjustment process not only better reflects the branching complexity and hemodynamic characteristics of the vessels, but also makes the 3D trunk vessels closer to the actual anatomical structure, thereby providing more reliable basic data for medical diagnosis, surgical planning, and scientific research analysis.

[0381] (4) By obtaining the weights of vascular segments and determining the target radius based on the initial radius and quantity, the modeling accuracy and realism of 3D trunk vessels can be significantly improved. Weights, as parameters reflecting the importance of vascular segments in the initial 3D trunk vessel, allow for more detailed adjustment of the segment radius, making it more consistent with the actual vascular anatomy. The introduction of weight parameters allows for personalized radius adjustments for each vascular segment, considering factors such as branching level, length, and curvature, thus more accurately reflecting the morphology of the vessel. By multiplying the quantity by the weights and adding the initial radius, the resulting target radius is closer to the diameter of the actual vessel, improving the modeling accuracy of 3D trunk vessels. The use of weights helps to better simulate hemodynamic characteristics, providing more valuable information for the diagnosis and treatment of vascular diseases.

[0382] (5) By classifying keypoint types and connecting 3D coordinates based on these types, the initial 3D structure of blood vessels can be effectively constructed. Distinguishing between trunk, branch, and leaf types of keypoints clearly represents the structural hierarchy of blood vessels, aiding in the understanding and analysis of their anatomical characteristics. A connection strategy based on keypoint types ensures that the blood vessel model more accurately reflects the actual vascular network. Automated connection based on keypoint types significantly improves the efficiency of blood vessel reconstruction and reduces the need for manual operations. Adaptable to vascular networks of varying complexity, from simple linear vessels to complex branching structures, keypoint types can be effectively used for modeling.

[0383] (6) By precisely selecting coordinate points, branch vessels that correspond to the actual vascular network can be constructed on the three-dimensional main blood vessel, thus obtaining a complete vascular distribution model including the main vessel. Subsequently, the shape of the branch vessels is adjusted, which not only optimizes the morphology of the vascular model, making it more realistic, but also enhances the practicality and accuracy of the model in medical diagnosis, surgical planning, and scientific research analysis. It also improves the efficiency of vascular reconstruction, reduces the need for manual intervention, and provides strong support for clinical decision-making.

[0384] (7) It can ensure that each branch vessel is closely connected to its corresponding vessel segment, thereby improving the accuracy of the vascular model and the realism of the anatomical structure; by traversing all coordinate points and gradually building the vascular distribution, it can effectively handle complex vascular networks and improve the comprehensiveness and systematicness of modeling; the incremental construction method also helps to optimize the use of computing resources, because it can avoid the computational burden caused by processing the entire vascular network at once.

[0385] (8) The method of generating the i-th three-dimensional blood vessel distribution based on the i-th blood vessel segment is to determine the region formed by the endpoint of the blood vessel segment and the i-th three-dimensional coordinate point, and then accurately determine the bifurcation point. This not only ensures the smooth transition between the branch blood vessels and the main blood vessels, improving the anatomical accuracy and realism of the model, but also optimizes the structure of the blood vessel distribution by deleting the original blood vessel segments and constructing new connections, making it more in line with the actual blood vessel network characteristics.

[0386] (9) By adjusting the shape of each branch vessel in the initial three-dimensional vascular distribution, that is, selecting multiple three-dimensional coordinate points on at least one side of the branch vessel and connecting the endpoints of the branch vessel with these coordinate points to form a new connected branch vessel, the adjusted branch vessel not only differs from the original branch vessel in shape and is closer to the real vascular anatomy, but also enhances the accuracy and functionality of the vascular model. The shape adjustment improves the visual realism of the three-dimensional vascular distribution.

[0387] (10) By selecting multiple three-dimensional coordinate points on at least one side of the branch vessels and applying random perturbation step sizes, the natural variations and individual differences of blood vessels can be effectively simulated. This not only improves the realism and accuracy of the three-dimensional blood vessel distribution model, but also makes the three-dimensional blood vessel distribution more consistent with the actual vascular structure of organisms.

[0388] (11) By multiplying the coordinates of each three-dimensional coordinate point in the three-dimensional blood vessel distribution with the projection matrix, the corresponding two-dimensional coordinate points are obtained and an initial two-dimensional blood vessel distribution is constructed. Then, the gray values ​​of each blood vessel segment are adjusted to obtain the two-dimensional blood vessel distribution under the viewpoint. This not only realizes the accurate mapping of the three-dimensional blood vessel structure to the two-dimensional plane and preserves the key anatomical information of the blood vessels, but also strengthens the visual contrast of the blood vessel segments through the difference in gray values, making different blood vessel segments clearly distinguishable in the two-dimensional image, thereby greatly improving the accuracy and efficiency of blood vessel analysis and diagnosis.

[0389] (12) Through refined depth value processing and grayscale mapping, the visual expressiveness and information content of two-dimensional blood vessel distribution are significantly improved. By normalizing the depth values ​​of blood vessel segments, it is ensured that blood vessel segments of different depths have clear and distinguishable grayscale differences in two-dimensional images. This not only enhances the sense of layering and three-dimensionality of the image, but also helps to highlight the structural features of blood vessels and potential lesion areas.

[0390] (13) By acquiring the skin texture of a pre-defined biological site, and when the skin texture coordinate system coincides with or coincides with the two-dimensional blood vessel distribution coordinate system after adjustment by a transformation matrix, the skin texture and the two-dimensional blood vessel distribution are superimposed. The resulting blood vessel texture not only provides rich visual information for medical image analysis but also enhances the intuitive representation of the relationship between blood vessels and skin. When the coordinate systems coincide, the superimposed image can clearly show the direction and distribution of blood vessels on the skin, which helps to discover potential lesions and abnormalities. When the coordinate systems do not coincide, the precise adjustment of the transformation matrix can ensure that the skin texture and blood vessel distribution are correctly aligned in space, avoiding the error offset caused by positional deviation.

[0391] (14) By acquiring multiple equally distributed noise features and predicting each two-dimensional blood vessel distribution based on its blood vessel texture and noise features, corresponding images can be generated, effectively enhancing the samples of blood vessel images. For example, a combination of 5 blood vessel textures and 5 noise features can generate 25 images, significantly expanding the dataset size. At the same time, the noise features with l_2 distance constraints ensure that the degree of variation between images generated by different noises for the same blood vessel texture remains stable, avoiding excessive deviation or distortion of the generated images. This stable variation not only improves the diversity and controllability of the generated images, but also provides richer and higher-quality samples for subsequent blood vessel image analysis, enhancing the model's generalization ability and robustness.

[0392] (15) The texture modeling approach effectively improves the palm vein recognition performance of the model, enabling the network to learn more discriminative features. It does not incur additional computational or training burdens, is simple to operate, and is effective. By improving palm vein recognition performance, the user experience of swiping their palm can be enhanced in services such as payment, identity verification, access control, and public transportation.

[0393] (16) First stage: Train the initial object recognition model using a vascular image sample set. The model learns the basic features of vascular images. Second stage: Optimize the initial model using a real image sample set. The model learns the feature representations of real images. The vascular image sample set increases sample diversity by introducing noise features, which helps improve the model's generalization ability. Optimizing the initial model using a real image sample set, and utilizing transfer learning, allows the model to better adapt to real image data. Through two-stage training, the model can learn richer feature representations, improving recognition accuracy and generalization ability.

[0394] (17) By converting two-dimensional coordinates to three-dimensional coordinates and connecting different three-dimensional coordinates based on the type of key points, an initial three-dimensional trunk blood vessel can be constructed. This trunk consists of multiple blood vessel segments with key points as endpoints. By adjusting the radius of each blood vessel segment based on its parameters (such as length, curvature, etc.), a more realistic and accurate three-dimensional trunk blood vessel can be generated. This not only improves the geometric accuracy of the blood vessel model but also enhances the biological rationality of the blood vessel structure.

[0395] (18) By selecting multiple three-dimensional coordinate points that are not on the main trunk in the three-dimensional coordinate system where the main trunk blood vessel is located, and constructing branch blood vessels based on these points, an initial three-dimensional blood vessel distribution can be generated. Adjusting the shape of each branch blood vessel in the initial three-dimensional blood vessel distribution can optimize the geometric structure and biological rationality of the blood vessels, thereby obtaining a more realistic and accurate three-dimensional blood vessel distribution, and improving the integrity and detail of the blood vessel model.

[0396] (19) By using key point recognition to process the first blood vessel image of the preset biological site, multiple key points within the blood vessel region can be effectively identified and their two-dimensional coordinates obtained. This process significantly reduces the complexity and computational load of subsequent three-dimensional blood vessel reconstruction. Based on these accurate key point coordinates, a three-dimensional trunk blood vessel model corresponding to the blood vessel can be generated efficiently, ensuring the accuracy and morphological integrity of the three-dimensional trunk blood vessel model. Constructing the branch blood vessels of the three-dimensional trunk blood vessel, the final three-dimensional blood vessel distribution of the preset biological site is obtained, which not only improves the detail of the model, but also helps to enhance the practicality of the blood vessel distribution model.

[0397] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A method for generating an image, the method comprising: A three-dimensional blood vessel distribution of a preset biological site is constructed, wherein the three-dimensional blood vessel distribution is used to represent the three-dimensional blood vessel distribution morphology of the preset biological site; Projecting the three-dimensional blood vessel distribution yields a two-dimensional blood vessel distribution corresponding to the three-dimensional blood vessel distribution, which is used to represent the two-dimensional blood vessel distribution morphology of the preset biological site; Based on the two-dimensional blood vessel distribution, an image of the preset biological site is generated.

2. The method according to claim 1, wherein, The three-dimensional vascular distribution includes three-dimensional trunk vessels and three-dimensional branch vessels; The construction of the three-dimensional blood vessel distribution of the preset biological site includes: Determine the coordinates of multiple key points used to generate the three-dimensional main blood vessel; The three-dimensional main blood vessel is generated based on the coordinates of the multiple key points; The three-dimensional branch vessels are generated on the three-dimensional main blood vessel to obtain the three-dimensional blood vessel distribution of the preset biological site.

3. The method according to claim 2, wherein, Determining the coordinates of multiple key points used to generate the three-dimensional main blood vessel includes: From multiple distribution types of blood vessel distribution in the preset biological site, select one type as the target distribution type; Based on the target distribution type, the coordinates of multiple key points used to generate the three-dimensional trunk blood vessel are determined.

4. The method according to claim 3, wherein, The coordinates of the multiple key points are three-dimensional coordinates; The step of determining the coordinates of multiple key points used to generate the three-dimensional main blood vessel based on the target distribution type includes: From the image of the target distribution type, identify the two-dimensional coordinates of multiple key points used to generate the three-dimensional main blood vessel; The two-dimensional coordinates of the multiple key points are converted into three-dimensional coordinates.

5. The method according to claim 4, wherein, The two-dimensional coordinates include the horizontal coordinate and the vertical coordinate of the two-dimensional coordinates, and the three-dimensional coordinates include the horizontal coordinate, the vertical coordinate, and the vertical coordinate of the three-dimensional coordinates. The process of converting the two-dimensional coordinates of the multiple key points into three-dimensional coordinates includes: The abscissa of the two-dimensional coordinate is determined as the abscissa of the three-dimensional coordinate, and the ordinate of the two-dimensional coordinate is determined as the ordinate of the three-dimensional coordinate; A random number is generated within a preset interval using a random number generation function, and the random number is then used as the vertical coordinate of the three-dimensional coordinate system.

6. The method according to any one of claims 2 to 5, wherein, The coordinates of the multiple key points are three-dimensional coordinates; The generation of the three-dimensional main blood vessel based on the coordinates of the multiple key points includes: For the three-dimensional coordinates of the multiple key points, the different three-dimensional coordinates are connected to obtain the initial three-dimensional trunk blood vessel of the blood vessel. The initial three-dimensional trunk blood vessel includes multiple blood vessel segments, and the blood vessel segments are terminated at the key points. The radius of each segment in the initial three-dimensional trunk blood vessel is determined, and the initial radius of each segment in the initial three-dimensional trunk blood vessel is adjusted to the corresponding radius to obtain the three-dimensional trunk blood vessel.

7. The method according to claim 6, wherein, The vascular segments in the initial three-dimensional main blood vessel have a flow direction; Determining the radius of each of the vessel segments in the initial three-dimensional main blood vessel includes: The following treatments were performed on each of the aforementioned vascular segments: Obtain the initial radius of the blood vessel segment in the initial three-dimensional main blood vessel and the number of end key points flowing out of the blood vessel segment determined according to the flow direction; Based on the initial radius and the number, determine the radius of the blood vessel segment.

8. The method according to claim 9, wherein, The determining the radius of the blood vessel segment based on the initial radius and the number includes: Obtain the weight of the blood vessel segment, and the weight is used to indicate the importance degree of the blood vessel segment; Multiply the number by the weight to obtain a multiplication result, and add the multiplication result to the initial radius of the blood vessel segment to obtain the radius of the blood vessel segment.

9. The method according to any one of claims 6 to 8, wherein, The types of the key points include the main trunk type, the bifurcation type and the leaf type; For the three-dimensional coordinates of the multiple key points, connecting the different three-dimensional coordinates to obtain the initial three-dimensional main blood vessel of the blood vessel includes: Connect the three-dimensional coordinates of the key points of each main trunk type to obtain a first main blood vessel, and connect the three-dimensional coordinates of the key points of each bifurcation type with the three-dimensional coordinates of at least one key point of the leaf type respectively to obtain a second main blood vessel corresponding to the key points of each bifurcation type; Connect the first main blood vessel with each of the second main blood vessels to obtain the initial three-dimensional main blood vessel.

10. The method according to any one of claims 2 to 9, wherein, The generating the three-dimensional branch blood vessels on the three-dimensional main blood vessel to obtain the three-dimensional blood vessel distribution of the preset biological part includes: Select multiple three-dimensional coordinate points not on the three-dimensional main blood vessel; Based on the multiple three-dimensional coordinate points, construct the three-dimensional branch blood vessels on the three-dimensional main blood vessel to obtain the initial three-dimensional blood vessel distribution of the preset biological part; Adjust the shapes of the three-dimensional branch blood vessels in the initial three-dimensional blood vessel distribution to obtain the three-dimensional blood vessel distribution of the preset biological part.

11. The method according to claim 10, wherein, The multiple three-dimensional coordinate points include the i-th three-dimensional coordinate point, 1 < i ≤ N, and the constructing the three-dimensional branch blood vessels on the three-dimensional main blood vessel based on the multiple three-dimensional coordinate points to obtain the initial three-dimensional blood vessel distribution of the preset biological part includes: Determine the three-dimensional main blood vessel as the first three-dimensional blood vessel distribution, determine the blood vessel segment on the (i - 1)-th three-dimensional blood vessel distribution that is closest to the i-th three-dimensional coordinate point as the i-th blood vessel segment, and generate the i-th three-dimensional blood vessel distribution based on the i-th blood vessel segment; Traverse i to obtain the N-th three-dimensional blood vessel distribution, and determine the N-th three-dimensional blood vessel distribution as the initial three-dimensional blood vessel distribution.

12. The method according to claim 11, wherein, The generating the i-th three-dimensional blood vessel distribution based on the i-th blood vessel segment includes: Determine the region formed by the endpoints of the i-th blood vessel segment and the i-th three-dimensional coordinate point, and determine the bifurcation point of the i-th three-dimensional blood vessel distribution in the region; Delete the i-th blood vessel segment from the (i - 1)-th three-dimensional blood vessel distribution, and connect the endpoints of the i-th blood vessel segment, the i-th three-dimensional coordinate point and the bifurcation point to obtain the i-th three-dimensional blood vessel distribution.

13. The method according to any one of claims 10 to 12, wherein, The adjusting the shapes of the three-dimensional branch blood vessels in the initial three-dimensional blood vessel distribution to obtain the three-dimensional blood vessel distribution of the preset biological part includes: For each of the three-dimensional branch vessels in the initial three-dimensional blood vessel distribution, perform the following processing to obtain the three-dimensional blood vessel distribution: Select a plurality of three-dimensional coordinate points on at least one side of the three-dimensional branch vessel, and sequentially connect the end points of the three-dimensional branch vessel and the plurality of three-dimensional coordinate points to obtain a reference three-dimensional branch vessel, which is different from the three-dimensional branch vessel; Replace the three-dimensional branch vessel in the initial three-dimensional blood vessel distribution with the reference three-dimensional branch vessel.

14. The method according to claim 13, wherein, The plurality of three-dimensional coordinate points include the j-th three-dimensional coordinate point, where 1 < j ≤ M, and the end point of the three-dimensional branch vessel includes the starting point of the three-dimensional branch vessel; The selecting a plurality of three-dimensional coordinate points on at least one side of the three-dimensional branch vessel includes: Traverse j and perform the following processing: When j = 1, superimpose the random perturbation step corresponding to the first three-dimensional coordinate point on the starting point of the three-dimensional branch vessel in the random perturbation direction corresponding to the first three-dimensional coordinate point to obtain the first three-dimensional coordinate point; When j is greater than 1, superimpose the random perturbation step corresponding to the j-th three-dimensional coordinate point on the (j - 1)-th three-dimensional coordinate point in the random perturbation direction corresponding to the j-th three-dimensional coordinate point to obtain the j-th three-dimensional coordinate point, and the random perturbation direction points to at least one side of the three-dimensional branch vessel.

15. The method according to any one of claims 1 to 14, wherein, The projecting the three-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution corresponding to the three-dimensional blood vessel distribution includes: Obtain at least one projection perspective, and for each of the perspectives, perform the following processing: Obtain the projection matrix corresponding to the perspective, and the projection matrix is used to indicate the relationship between the three-dimensional blood vessel distribution and the two-dimensional blood vessel distribution under the perspective; Based on the projection matrix, perform projection of the three-dimensional blood vessel distribution under the perspective to obtain the two-dimensional blood vessel distribution under the perspective.

16. The method according to claim 15, wherein, The performing projection of the three-dimensional blood vessel distribution under the perspective based on the projection matrix to obtain the two-dimensional blood vessel distribution under the perspective includes: Multiply the coordinates of each three-dimensional coordinate point in the three-dimensional blood vessel distribution by the projection matrix respectively to obtain the two-dimensional coordinate points corresponding to each three-dimensional coordinate point; Determine the two-dimensional blood vessel distribution based on the two-dimensional graph formed by each of the two-dimensional coordinate points.

17. The method according to claim 15, wherein, The determining the two-dimensional blood vessel distribution based on the two-dimensional graph formed by each of the two-dimensional coordinate points includes: Determine the two-dimensional graph formed by each of the two-dimensional coordinate points as the initial two-dimensional blood vessel distribution, and the initial two-dimensional blood vessel distribution includes a plurality of initial blood vessel segments; Adjust the gray value of each initial blood vessel segment in the initial two-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution under the perspective, and there are at least two blood vessel segments with different gray values in the two-dimensional blood vessel distribution.

18. The method according to claim 17, wherein, The adjusting the gray value of each initial blood vessel segment in the initial two-dimensional blood vessel distribution to obtain the two-dimensional blood vessel distribution under the perspective includes: For each initial blood vessel segment in the initial two-dimensional blood vessel distribution, perform the following processing to obtain the two-dimensional blood vessel distribution under the perspective: Obtain the depth value of the blood vessel segment corresponding to the initial blood vessel segment in the three-dimensional blood vessel distribution, and subtract the minimum depth value of the blood vessel segment in the three-dimensional blood vessel distribution from the depth value to obtain the first depth value; Obtain the difference between the maximum depth value and the minimum depth value of the blood vessel segment in the three-dimensional blood vessel distribution, and determine the first gray value corresponding to the initial blood vessel segment based on the difference and the first depth value; The second gray value of the initial blood vessel segment in the initial two-dimensional blood vessel distribution is adjusted to the first gray value.

19. The method according to claim 18, wherein, Determining the first grayscale value corresponding to the initial blood vessel segment based on the difference and the first depth value includes: The first depth value is randomly perturbed to obtain a perturbed first depth value. The perturbed first depth value is divided by the difference to obtain a candidate first depth value. The product of the candidate first depth value and the preset value is determined as the first depth value corresponding to the initial blood vessel segment.

20. The method according to any one of claims 1 to 18, wherein, The step of generating an image of the preset biological site based on the two-dimensional blood vessel distribution includes: Obtain the skin texture of the preset biological site, and generate a blood vessel texture corresponding to the two-dimensional blood vessel distribution based on the skin texture and the two-dimensional blood vessel distribution; Based on the blood vessel texture, an image corresponding to the two-dimensional blood vessel distribution is generated.

21. The method according to claim 20, wherein, The step of generating a vascular texture corresponding to the two-dimensional vascular distribution based on the skin texture and the two-dimensional vascular distribution includes: When the coordinate system of the skin texture coincides with the coordinate system of the two-dimensional blood vessel distribution, the skin texture is superimposed on each of the two-dimensional blood vessel distributions to obtain the blood vessel texture corresponding to each of the two-dimensional blood vessel distributions. When the coordinate system of the skin texture does not coincide with the coordinate system of the two-dimensional blood vessel distribution, obtain the transformation matrix between the coordinate system of the skin texture and the coordinate system of the two-dimensional blood vessel distribution; Based on the transformation matrix, the skin texture is transformed to the coordinate system of the two-dimensional blood vessel distribution to obtain the transformed skin texture; The converted skin texture is superimposed on each of the two-dimensional blood vessel distributions to obtain the blood vessel texture corresponding to each of the two-dimensional blood vessel distributions.

22. The method according to claim 20, wherein, The step of generating an image corresponding to the two-dimensional blood vessel distribution based on the blood vessel texture includes: Multiple equally spaced noise features are acquired, and the following processing is performed on the two-dimensional blood vessel distribution respectively: Based on the vascular texture corresponding to the two-dimensional vascular distribution and each of the noise features, the image corresponding to the two-dimensional vascular distribution is predicted to obtain multiple images corresponding to the two-dimensional vascular distribution, and the images correspond one-to-one with the noise features.

23. The method according to any one of claims 1 to 22, wherein, After generating an image corresponding to the two-dimensional blood vessel distribution based on the two-dimensional blood vessel distribution, the method further includes: Using images corresponding to multiple two-dimensional blood vessel distributions as blood vessel image samples, a blood vessel image sample set for the preset biological site is constructed. Based on the blood vessel image sample set of the preset biological parts, the recognition model is trained to obtain an initial object recognition model; A set of real image samples obtained by capturing images of the preset biological parts is acquired, and the initial object recognition model is trained based on the set of real image samples to obtain the object recognition model.

24. A method for constructing a three-dimensional blood vessel distribution, wherein the three-dimensional blood vessel distribution includes three-dimensional trunk vessels and three-dimensional branch vessels, the method comprising: Determine the coordinates of multiple key points used to generate the three-dimensional main blood vessel; The three-dimensional main blood vessel is generated based on the coordinates of the multiple key points; The three-dimensional branch vessels are generated on the three-dimensional main blood vessel to obtain the three-dimensional blood vessel distribution of the preset biological site.

25. The method according to claim 24, wherein, Determining the coordinates of multiple key points used to generate the three-dimensional main blood vessel includes: From multiple distribution types of blood vessel distribution in the preset biological site, select one type as the target distribution type; Based on the target distribution type, the coordinates of multiple key points used to generate the three-dimensional trunk blood vessel are determined.

26. The method according to claim 24 or 25, wherein, The coordinates of the multiple key points are three-dimensional coordinates; The generation of the three-dimensional main blood vessel based on the coordinates of the multiple key points includes: For the three-dimensional coordinates of the multiple key points, the different three-dimensional coordinates are connected to obtain the initial three-dimensional trunk blood vessel of the blood vessel. The initial three-dimensional trunk blood vessel includes multiple blood vessel segments, and the blood vessel segments are terminated at the key points. The radius of each segment in the initial three-dimensional trunk blood vessel is determined, and the initial radius of each segment in the initial three-dimensional trunk blood vessel is adjusted to the corresponding radius to obtain the three-dimensional trunk blood vessel.

27. The method according to claim 24, wherein, The process of generating the three-dimensional branch vessels on the three-dimensional main blood vessel to obtain the three-dimensional blood vessel distribution of the preset biological site includes: Select multiple three-dimensional coordinate points that are not on the three-dimensional main blood vessel; Based on the multiple three-dimensional coordinate points, the three-dimensional branch vessels are constructed on the three-dimensional main blood vessel to obtain the initial three-dimensional blood vessel distribution of the preset biological site. The shape of each of the three-dimensional branch vessels in the initial three-dimensional vascular distribution is adjusted to obtain the three-dimensional vascular distribution of the preset biological site.

28. An image generation apparatus, the apparatus comprising: The acquisition module is configured to construct a three-dimensional blood vessel distribution of a preset biological site, wherein the three-dimensional blood vessel distribution is used to represent the three-dimensional blood vessel distribution morphology of the preset biological site. The projection module is configured to project the three-dimensional blood vessel distribution to obtain a two-dimensional blood vessel distribution corresponding to the three-dimensional blood vessel distribution. The two-dimensional blood vessel distribution is used to represent the two-dimensional blood vessel distribution morphology of the preset biological site. The generation module is configured to generate an image of the preset biological site based on the two-dimensional blood vessel distribution.

29. A device for constructing a three-dimensional blood vessel distribution, the device comprising: The identification module is configured to determine the coordinates of multiple key points used to generate the three-dimensional main blood vessel. The generation module is configured to generate the three-dimensional main blood vessel corresponding to the blood vessel based on the coordinates of the multiple key points; The construction module is configured to generate the three-dimensional branch vessels based on the three-dimensional main blood vessels, thereby obtaining the three-dimensional blood vessel distribution of a preset biological site.

30. An electronic device, the electronic device comprising: Memory, configured to store computer-executable instructions or computer programs; When a processor is configured to execute computer-executable instructions or computer programs stored in the memory, it implements the method of any one of claims 1 to 27.

31. A computer-readable storage medium storing computer-executable instructions or a computer program that, when executed by a processor, implements the method of any one of claims 1 to 27.

32. A computer program product comprising a computer program or computer-executable instructions, wherein the computer program or computer-executable instructions, when executed by a processor, implement the method of any one of claims 1 to 27.

Images