Machine learning system and method for blood vessel imaging and 3D reconstruction for assited intrusion

Abstract

The present invention provides an automated system (100) and a method (300) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques comprising an imaging module (102) incorporating plurality of monochrome infrared stereo camera (104) with Near-Infrared Light (NIR) illumination (106); a robotic arm (108) to perform assisted intrusion with the feedback of a 3D reconstructed blood vessel; and a control unit (110) incorporating plurality of microprocessors with neural network to employ machine learning to control and automate 3D reconstruction and intrusion of blood vessel using machine learning techniques in real-time, wherein the neural network comprises of an image registration module, image library construction module, a training module, and a 3D reconstruction module.

Description

[0001] MACHINE LEARNING SYSTEM AND METHOD FOR BLOOD VESSEL IMAGING AND 3D RECONSTRUCTION FOR ASSITED INTRUSION

[0002] FIELD OF THE INVENTION

[0003] The present invention relates to a system and method for blood vessel imaging and 3D reconstruction for intrusion via an automatic vein cannulation system. More particularly, the present invention provides an advanced automated imaging system and method for performing blood vessel imaging, and 3D reconstruction by utilizing machine learning techniques to increase the accuracy and quality of the system and process for detecting blood vessels for intrusion by an automatic vein cannulation.

[0004] BACKGROUND OF THE INVENTION

[0005] Intravenous (IV) cannulation is a critical procedure for providing venous access in medical settings, essential for various medical purposes such as blood sampling, administering fluids, medications, nutrition and blood products. Every patient who visits a healthcare setting has a potential for intravenous cannulation procedure and related pain, fear, and distress, as access to the blood stream via a needle is perhaps the most common procedure administered by healthcare professionals.

[0006] Establishing intravenous (IV) access in younger patient populations via the traditional cannulation technique is often fraught with challenges, particularly in pediatric patients. Traditionally performed manually by healthcare professionals, the procedure, despite its ubiquity, is fraught with significant challenges and a notable error rate. Studies indicate error rates as high as 34% in general populations, with even higher rates reported in pediatric and geriatric patients, where the failure rate can reach up to 60% (African Health Science, 2020). These high failure rates not only compromise patient safety but also contribute to increased discomfort and potential complications.

[0007] The IV procedure involves the needle puncturing the vein at an angle, followed by advancing the needle-cannula combo until the cannula tip is fully within the vein. However, if the needle's angle is not adjusted during advancement, there is a risk of puncturing the back wall of the vein, resulting in the cannula becoming interstitial.

[0008] Veins consist of three layers: an inner endothelium, a layer of muscle fibers, and an outer connective tissue layer. Venous valves within veins ensure blood flow direction and prevent pooling in extremities, although they can pose challenges during catheter insertion, particularly where veins join or in lower extremities. These challenges underscore the critical need for improved techniques and technologies to enhance the success rates of vein cannulation, minimize patient discomfort, and reduce the risk of complications associated with multiple insertion attempts.

[0009] To combat the issues that arise during cannulation many medical practitioners use simple methods, such as utilizing a warm compress, tourniquets, and mild agitation (slapping) of the vein to increase the likelihood of a successful first pass. While these methods are fairly simple to initiate, cost effective, and unobtrusive to the patient, the low first pass rates indicate other methods need to be explored. As such, technologically enhanced methods are currently in use for better vein visualization including near infrared (NIR), ultrasound, and transillumination technologies. These technologically advanced methods use either sound, in the case of ultrasound, or light, in the case of NIR and transillumination to visualize the vein.

[0010] Near-Infrared Vein Visualization devices allow for non-invasive identification of veins for multiple uses. They function by illuminating the skin with near-infrared light (700-900 nm), which penetrates the skin and subcutaneous tissues to a depth of approximately 3 mm. The light is differentially absorbed by the underlying tissues, with increased absorption by deoxygenated haemoglobin. The difference in absorption is detected by the device camera, and the reproduction is projected onto the subject’s skin in real-time. Infrared (IR) vein visualization is a modality that aids venous cannulation; however, few reports of this technique exist in the infant and toddler population. Therefore, there is a technological gap wherein there is a requirement of an advanced vein detection and guidance systems, utilizing optics, cameras, and imaging technologies, be optimized to accurately locate veins in patients with diverse anatomical challenges, including different hair coverage, skin types, and complexions, particularly in pediatric and geriatric populations. Further, an advanced vein intrusion mechanism is needed to ensure minimal discomfort and high success rates in IV cannulation, particularly in challenging patient populations such as pediatric and geriatric patients. Moreover, implementation of robotics and automated systems in vein cannulation impact the precision and success rates compared to traditional manual techniques.

[0011] OBJECTIVE OF THE INVENTION

[0012] Accordingly, to overcome the drawbacks of the prior art, the main object of the present invention is to provide an advanced automated system and method for blood vessel imaging, 3D reconstruction and an automatic vein cannulation utilizing machine learning techniques to increase the accuracy and quality of the system and process for detecting blood vessels for machine assisted intrusion.

[0013] Yet another object of the present invention is to provide an automated system and method for blood vessel imaging, 3D reconstruction and an automatic vein cannulation for intrusion using machine learning techniques for pediatric and geriatric patients.

[0014] Yet another object of the present invention is to provide an automated system and method for blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques to facilitate venous access for various medical purposes such as blood sampling and administering fluids, medications, nutrition and blood products.

[0015] Yet another object of the present invention is to provide an automated system and method for blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques to reduce patient discomfort, and minimize complications associated with multiple insertion attempts.

[0016] Yet another object of the present invention is to provide an automated system and method for automated system and method for blood vessel imaging, 3D reconstruction and an automatic vein cannulation with an advanced vein detection and guidance system that utilizes optics, cameras and imaging technologies to accurately locate blood vessels, taking into consideration the anatomical challenges such as hair, skin types, skin complexions.

[0017] SUMMARY OF THE INVENTION

[0018] In carrying out the above objects of the present invention, the present invention provides an automated system and method for blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques to increase the accuracy and quality of the process for detecting blood vessel for blood vessel intrusion. The advanced automated imaging system may utilize robotics to improve precision and accuracy in locating veins, thereby reducing the reliance on manual techniques prone to human error and act as an automated vein cannulation system, thereby enhancing patient safety and procedural efficacy in clinical.

[0019] In an embodiment of the present invention, the present invention provides automated system and method for blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques comprising of an imaging module incorporating plurality of monochrome infrared stereo camera with Near-Infrared Light (NIR) illumination, a robotic arm to perform the intrusion with the feedback of a 3D reconstructed blood vessel from the imaging module and a control unit incorporating plurality of microprocessors with neural network to employ machine learning to control and automate 3D reconstruction and facilitate intrusion of blood vessel in real-time via the robotic arm.

[0020] In another embodiment of the present invention, the invention provides a method for performing for blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques in real-time. The method comprises step of: capturing images with Near-Infrared Light (NIR) illuminated on different coordinates through plurality of monochrome infrared stereo camera; processing the captured image using polarized filters, using varied wavelengths of light and adjusting illumination area to identifying distinctive points or features; organizing, categorizing, and annotating captured images to make them easily searchable and retrievable; cropping the region of interest from the processed image to perform machine learning techniques for image enhancement; generating a 3D reconstruction of blood vessel from the processed image using annotated depth maps using Fourier Domain - Optical Coherence Tomography (FD-OCT) setup for depth estimation of blood vessel; using precise anatomical landmarks and region of interest to guide the robotic arm for accurate and efficient blood vessel intrusion; and employing neural network to control robotic arm and needle for accurate and efficient blood vessel intrusion.

[0021] In preferred embodiment, the neural network to employ machine learning to control and automate 3D reconstruction and facilitate intrusion of blood vessel in real-time via the robotic arm comprises of an image registration module, image library construction module, a training module, a 3D reconstruction module, a guided insertion module and a needle handling module.

[0022] In preferred embodiment of the present invention, the invention provides a method for performing blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques in real-time wherein the cropped region of interest images have noise reduction, enhanced signal-to-noise ratio, effectively processing hair details, and have minimal vignetting effects.

[0023] In preferred embodiment of the present invention, the invention provides a system and method for performing for blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques in real-time wherein stereo image registration includes the process of aligning and matching corresponding points or features between two or more stereo images taken from slightly different viewpoints for tasks such as stereo matching, 3D reconstruction, and depth estimation.

[0024] In preferred embodiment of the present invention, the invention provides an automated method for performing blood vessel imaging, 3D reconstruction and an automatic vein cannulation using machine learning techniques in real-time wherein the calibration of the imaging module for performing blood vessel imaging and 3D reconstruction follows camera intrinsic parameters, camera extrinsic parameters and rotation translation matrix using machine learning techniques.

[0025] In preferred embodiment of the present invention, the invention provides an automated system and method to perform blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques wherein image enhancement is performed through neural network consisting of an encoder network that compresses input data into a latent-space representation, and a decoder network that reconstructs the original input from this representation.

[0026] BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The object of the invention may be understood in more details and more particularly description of the invention briefly summarized above by reference to certain embodiments thereof which are illustrated in the appended drawings, which drawings form a part of this specification. It is to be noted, however, that the appended drawings illustrate preferred embodiments of the invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective equivalent embodiments.

[0028] Figure 1 illustrates schematic of the preferred embodiment of an automated system for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention;

[0029] Figure 2 illustrates schematic of the automated system for performing blood vessel imaging, 3D reconstruction and intrusion using handheld vein viewer according to the embodiment of the present invention; Figure 3 illustrates a flow diagram of method for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention;

[0030] Figure 4 illustrates comparison of processed image achieved through the system for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention;

[0031] Figure 5 illustrates image registration of processed image achieved through the system for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention; and

[0032] Figure 6 illustrates depth estimation of processed image achieved through the system (100) for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention.

[0033] DETAIL DESCRIPTION OF THE INVENTION

[0034] Various features and embodiments of the present invention here will be discernible from the following further description thereof, set out hereunder. The illustrative embodiments described in the detailed description are not meant to be limiting. Other embodiments may be used, and other changes may be made, without departing from the spirit or scope of the subject matter presented herein. It will be readily understood that the aspects of the present disclosure, as generally described herein, and illustrated in the Figures, can be arranged, substituted, combined, separated, and designed in a wide variety of different configurations, all of which are explicitly contemplated herein.

[0035] The term “blood vessel” herein refer to any blood vessel such as artery and / or vein.

[0036] The term “intrinsic parameters” herein refers to focal length, principal point, lens distortion coefficients, pixel size or similar. The term “extrinsic parameters” herein refers to camera position (x, y, and z coordinates of the camera) and the camera orientation (rotation).

[0037] Referring to Fig, 1 illustrates schematic representation of the most preferred embodiment of the automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques according to the embodiment of the present invention.

[0038] The automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques comprising of an imaging module (102) incorporating plurality of monochrome infrared stereo camera (104) with Near-Infrared Light (NIR) illumination (106). In a preferred embodiment the stereo camera (102) is a monochrome camera made from silicon metal (CMOS) and includes NIR illumination (106) within the range of 700Nm - lOOONm. The calibration of the imaging module (102) follows camera intrinsic parameters, camera extrinsic parameters and rotation translation matrix using machine learning techniques. The imaging module (102) is calibrated to recalculate the positions of the objects in plurality of cameras (104) to triangulate 3D coordinates in the real-time.

[0039] The automated system (100) also comprises of a robotic arm (108) to perform assisted intrusion with the feedback of a 3D reconstructed blood vessel. In preferred embodiment the robotic arm (108) has 5-axis control, a head to attach needle and has organ arrest bed to arrest the motion assists an intrusion. The robotic arm (108) also incorporates sensors such as but not limited to pressure sensor, blood pressure sensor, temperature sensor, IR sensor or similar to assist in functioning of automated intrusion using machine learning techniques. Furthermore, the robotic arm (108) follows a strict sterility protocol wherein the robotic arm (108) is prepped in a sterile environment, and all tools are sterilized to prevent infection.

[0040] The automated system (100) also comprises of a control unit (110) incorporating plurality of microprocessors to control and automate 3D reconstruction and intrusion of blood vessel using machine learning techniques in real-time. The controller unit (110) further comprises of multiple processors configured to receive and process images acquired by the one or more cameras (104), and configured to maneuver and position the automated robotic arm (108). Further, the controller unit (110) is configured to receive the information from the multiple processors about potential insertion sites, and selects a region of interest and directs the automated robotic arm (108) to insert the needle into the region of interest.

[0041] The controller unit (110) incorporates a neural network configured to process the image recognition and 3D reconstruction of blood vessel and position the needle proximate a targeted location on a patient's skin surface, and guide the automated robotic arm (108) to cause the distal end of the needle to puncture the skin surface and penetrate a desired depth into the blood vessel or tissue at the targeted location in order to avoid, as much as possible, damage to surrounding tissue, blood vessels, and the like.

[0042] The neural network employs machine learning to control and automate 3D reconstruction and facilitate intrusion of blood vessel in real-time via the robotic arm comprises of an image registration module, image library construction module, a training module, a 3D reconstruction module, a guided insertion module and a needle handling module.

[0043] The image registration module is responsible for identifying distinctive points or features in each image that can be matched between the monochrome infrared stereo camera (104) pairs; finding correspondences between the descriptors of features detected in the left and right images; and estimating the disparity or depth map that represents the geometric relationship (translation, rotation, and possibly scale) between the images. The distinctive point or feature incorporates parameters such as but not limited to camera intrinsic parameters, camera extrinsic parameters and rotation translation matrix.

[0044] The image library construction module builds a structured collection of images captured by the monochrome infrared stereo camera (104). It involves organizing, categorizing, and potentially annotating captured images to make them easily searchable and retrievable. Furthermore, it helps to create a valuable and easily accessible resource for medical professionals, enabling them to quickly find the information they need, improve efficiency, and ensure quality of the system (100).

[0045] The training module incorporates autoencoders that performs unsupervised learning tasks, particularly in data compression and feature learning. The autoencoder consist of an encoder network that compresses input data into a latent-space representation, and a decoder network that reconstructs the original input from this representation. By training on unlabelled data, the autoencoders can learn efficient representations of the input data, capture important features and reduce noise. The training module facilitates in identifying distinctive points or features in each image captured by the monochrome infrared stereo camera (104) even if the image is scaled, or if it has noise and different illumination. Furthermore, the training module helps in establishing region of interest from the processed images for image enhancement.

[0046] The 3D reconstruction module uses annotated depth maps to generate a 3D reconstruction for any input image captured by the monochrome infrared stereo camera (104). The 3D reconstruction module employs using Fourier Domain - Optical Coherence Tomography (FD-OCT) setup or similar for depth estimation of blood vessel.

[0047] The guided insertion module enables robotic arm (108) to use precise anatomical landmarks and region of interest to guide the needle into the target vein. The module uses sensors and real-time imaging to confirm proper needle placement and integrates feedback from these sensors to adjust the needle placement dynamically. Once the needle is in place, the module uses the robotic arm’s sensors for continuously monitoring blood return to ensure consistent and pulsatile flow, indicating successful venous access. In case the module founds any irregularities, it triggers automatic recalibration or adjustment procedures.

[0048] The needle handling module ensures steady pressure during needle insertion and withdrawal. The module helps the robotic arm (108) to maintain precise control over needle handling, minimizing patient discomfort and reducing the risk of complications. The module also manages the timing of needle ejection with high precision to avoid dislodgement or damage to the vein.

[0049] Altogether, each of the module helps in the calibration of the imaging module (102) follows camera intrinsic parameters, camera extrinsic parameters and rotation translation matrix using machine learning techniques.

[0050] Referring to Figure 2 illustrates schematic of the automated system (100) for performing blood vessel imaging, 3D reconstruction and intrusion using handheld vein viewer (200) according to the embodiment of the present invention

[0051] The handheld vein viewer (200) is a compact unit that utilizes the principles of nearinfrared (NIR) and infrared (IR) imaging to visualize superficial veins in real time. The handheld vein viewer (200) is designed to be portable, cost-effective, and user- friendly, making it highly suitable for clinical, emergency, or home-care environments where quick and non-invasive vein detection is required.

[0052] The hardware components of the handheld vein viewer (200) comprise of an infrared camera module (202), an infrared illumination module (204), a processing unit (206), a display module (208) and an enclosure (210).

[0053] The infrared camera module (202) incorporates a high-sensitivity IR camera compatible with Raspberry Pi to capture real-time images of the veins beneath the skin. The camera is optimized for near-infrared wavelengths (typically between 740 nm to 940 nm) where hemoglobin absorption is high, making veins appear darker than surrounding tissue.

[0054] The infrared illumination module (204) incorporates two IR LEDs positioned adjacent to the camera lens to uniformly illuminate the skin surface. This setup of two IR LEDs ensures consistent near-infrared (NIR) lighting, enhancing vein contrast without visible light interference, allowing for clear vein visualization even in low-light environments.

[0055] The processing unit (206) incorporates a Raspberry Pi board serves as the processing hub, handling image acquisition, filtering, and display tasks. The processing unit (206) utilizes the custom image processing algorithms (OpenCV- based) to enhance vein contrast and suppress background noise. In preferred embodiment, the processing unit (206) can easily establish a communication channel with the control unit (110) incorporating plurality of microprocessors with neural network to employ machine learning to control and automate 3D reconstruction and intrusion of blood vessel using machine learning techniques in real-time.

[0056] The display module (208) incorporates a mini-Raspberry Pi-compatible display is integrated for real-time monocular visualization of the veins. The display module (208) is compact, lightweight, and optimized for high-contrast IR image rendering. The processed image is displayed instantly on the display module (208) for the user to view.

[0057] The enclosure (210) incorporates a custom-designed, ergonomic case which houses all the components. The enclosure (210) is lightweight, portable, and durable, providing ease of use during handheld operation. Furthermore, the enclosure (210) has thoughtful design considerations including heat ventilation, proper IR LED alignment, and camera stabilization.

[0058] In preferred embodiment, the enclosure (210) can be a molded structure or a 3D printed structure.

[0059] The handheld vein viewer (200) has ability to provide a low power consumption and cost-effective solution for imaging module (102) for the automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion.

[0060] Referring to Fig. 3 illustrates a flow diagram of method (300) for performing blood vessel imaging, 3D reconstruction and intrusion using the automated system (100) according to the embodiment of the present invention. The method (300) begins at step (302), where Near-Infrared Light (NIR) illuminated images on different coordinates through plurality of monochrome infrared stereo camera (104,202) are captured. At step (304) the image registration module, processes the captured images using polarized filters, varied wavelengths of light and adjusting illumination area to identifying distinctive points or features. At step (306) the image library construction module organizes, categorize, and annotates captured images to make them easily searchable and retrievable. At step (308) the training module, cropping region of interest from the processed images to perform image enhancement. At step (310) the 3D reconstruction module uses annotated depth maps to generate a 3D reconstruction of blood vessel from the processed image. Fourier Domain - Optical Coherence Tomography (FD-OCT) setup is used for depth estimation of blood vessel. At step (312) the guided insertion uses precise anatomical landmarks and region of interest to guide the robotic arm for accurate and efficient blood vessel intrusion. At step (314) the needle handling module ensures steady pressure will be applied during needle insertion and withdrawal.

[0061] Referring to Fig. 4 illustrates comparison of original image and the processed image achieved through the automated system (100) for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention. The original acquired image through monochrome infrared stereo camera (104) is shown on the left and on the right, the processed image is highlighted. The automated system (100) employed diverse image processing techniques incorporating polarized filters, varied wavelengths of light, adjusting illumination areas and cropping region of interest using neural network. The processing enables cropped region of interest images to have noise reduction, enhanced signal-to-noise ratio, effectively processing hair details, and have minimal vignetting effects.

[0062] Further, the processed images are enhanced, which is performed through neural network consisting of an encoder network that compresses input data into a latent- space representation, and a decoder network that reconstructs the original input from this representation.

[0063] Referring to Fig. 5 illustrates image registration of processed image achieved through the automated system (100) for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention. During image registration aligning and matching of corresponding points or features between two or more stereo images taken from slightly different viewpoints or axis. In preferred embodiment image registration is performed over the processed enhanced image, wherein distinctive points or features in each image that can be matched between the identified stereo pair, matched and used for geometric transformation. Further, algorithms like scaleinvariant feature transform (SIFT) are employed for feature detection on processed enhanced images from the autoencoders.

[0064] Referring to Fig. 6 illustrates depth estimation of processed image achieved through the automated system (100) for performing blood vessel imaging, 3D reconstruction and intrusion using machine learning techniques according to the embodiment of the present invention. In a preferred embodiment, techniques such as, but not limited to Fourier Domain - Optical Coherence Tomography (FD-OCT) are incorporated that allows depth estimation in sub millimeter range up to a few nanometers error difference. Optical Coherence Tomography (OCT) are paired with machine learning techniques that gives very accurate depth estimation for blood vessel intrusion systems, significantly decreases the error rate and blood vessel penetration will be much smoother and less likely to get into complications.

[0065] While certain exemplary embodiments have been described and shown in the accompanying drawings, it is to be understood that such embodiments are merely illustrative of, and not restrictive on, the broad invention, and that this invention not be limited to the specific constructions and arrangements shown and described, since various other changes, combinations, omissions, modifications and substitutions, in addition to those set forth in the above paragraphs, are possible. Those skilled in the art will appreciate that various adaptations and modifications of the just described embodiments can be configured without departing from the scope and spirit of the invention.

Claims

CLAIMSWe Claim1. An automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques comprising: an imaging module (102) incorporating plurality of monochrome infrared stereo camera (104) with Near-Infrared Light (NIR) illumination (106); a robotic arm (108) to perform assisted intrusion with the feedback of a 3D reconstructed blood vessel; and a control unit (110) incorporating plurality of microprocessors with neural network to employ machine learning to control and automate 3D reconstruction and intrusion of blood vessel using machine learning techniques in real-time; wherein, the neural network comprises of an image registration module, image library construction module, a training module, and a 3D reconstruction module; the image registration module identifies distinctive points or features in each image that can be matched between the monochrome infrared stereo camera (104) pairs; the image registration module finds correspondences between the descriptors of features detected in the left and right images captured by the monochrome infrared stereo camera (104); the image registration module estimates the disparity or depth map that represents the geometric relationship (translation, rotation, and possibly scale) between the images; the image library construction module organizes, categorize, and annotating captured images by the monochrome infrared stereo camera (104) to make them easily searchable and retrievable; the training module identifies distinctive points or features in each image captured by the monochrome infrared stereo camera (104) even if the image is scaled, or if it has noise and different illumination;the 3D reconstruction module uses annotated depth maps to generate a 3D reconstruction for any input image captured by the monochrome infrared stereo camera (104); the guided insertion module enables robotic arm (108) to use precise anatomical landmarks and imaging data to guide the needle into the target vein; and the needle handling module ensures steady pressure during needle insertion and withdrawal.

2. The automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques as claimed in claim 1, wherein the imaging module (102) can also be a handheld vein viewer (200) comprises of an infrared camera module (202), an infrared illumination module (204), a processing unit (206), a display module (208) and an enclosure (210).

3. The automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques as claimed in claim 1, wherein the stereo camera (102) is a monochrome camera made from silicon metal (CMOS) and includes NIR illumination (106) within the range of 740Nm - 940Nm.

4. The automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques as claimed in claim 1 , wherein the calibration of the imaging module (102) follows camera intrinsic parameters, camera extrinsic parameters and rotation translation matrix using machine learning techniques.

5. The automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques as claimed in claim 1, wherein the robotic arm (108) has 5-axis control, a head to attach needle and has organ arrest bed to arrest the motion assists an intrusion.

6. The automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques as claimed in Claim 1, wherein the method enables cropped region of interest images to have noise reduction, enhanced signal-to-noise ratio, effectively processing hair details, and have minimal vignetting effects.

7. The automated system (100) for performing blood vessel imaging, 3D reconstruction and assisted intrusion using machine learning techniques as claimed in Claim 1, wherein the handheld vein viewer (200) is designed to be portable, cost-effective, and user-friendly, making it highly suitable for clinical, emergency, or home-care environments.

8. A method (300) for performing blood vessel imaging, 3D reconstruction and intrusion using the automated system (100). The method comprises of: capturing Near-Infrared Light (NIR) illuminated images on different coordinates through plurality of monochrome infrared stereo camera (104,202); identifying distinctive points or features on the captured images using polarized fdters, varied wavelengths of light and adjusting illumination area using neural network; organizing, categorizing, and annotating captured images to make them easily searchable and retrievable using neural network. performing image enhancement by cropping region of interest from the processed images using neural network; performing 3D reconstruction uses annotated depth maps to generate a 3D reconstruction of blood vessel from the processed image using neural network; guiding the robotic arm (108) using precise anatomical landmarks and region of interest for accurate and efficient blood vessel intrusion using neural network; andensuring steady pressure will be applied during needle insertion and withdrawal using neural network; wherein, the neural network comprises of an image registration module, image library construction module, a training module, and a 3D reconstruction module; the image registration module identifies distinctive points or features in each image that can be matched between the monochrome infrared stereo camera (104, 202) pairs; the image registration module finds correspondences between the descriptors of features detected in the left and right images captured by the monochrome infrared stereo camera (104, 202); the image registration module estimates the disparity or depth map that represents the geometric relationship (translation, rotation, and possibly scale) between the images; the image library construction module organizes, categorize, and annotating captured images by the monochrome infrared stereo camera (104, 202) to make them easily searchable and retrievable; the training module identifies distinctive points or features in each image captured by the monochrome infrared stereo camera (104, 202) even if the image is scaled, or if it has noise and different illumination; the 3D reconstruction module uses annotated depth maps to generate a 3D reconstruction for any input image captured by the monochrome infrared stereo camera (104, 202); the guided insertion module enables robotic arm (108) to use precise anatomical landmarks and imaging data to guide the needle into the target vein; and the needle handling module ensures steady pressure during needle insertion and withdrawal.

9. The method (300) for performing blood vessel imaging, 3D reconstruction and intrusion using the automated system (100) as claimed in Claim 8, wherein the infrared stereo camera (104, 202) and is optimized for near-infrared wavelengths within the range of 740 nm to 940 nm.

10. The method (300) for performing blood vessel imaging, 3D reconstruction and intrusion using the automated system (100) as claimed in Claim 8, wherein the robotic arm (108) has 5-axis control, a head to attach needle and has organ arrest bed to arrest the motion assists an intrusion.

11. The method (300) for performing blood vessel imaging, 3D reconstruction and intrusion using the automated system (100) as claimed in Claim 8, wherein the method enables cropped region of interest images to have noise reduction, enhanced signal-to-noise ratio, effectively processing hair details, and have minimal vignetting effects.

Images