A detection method for assisting in evaluating the quality of in-surgery blood vessel anastomosis

Abstract

The present application relates to the field of artificial intelligence image processing technology, in particular to a detection method for assisting in evaluating the quality of in-surgery blood vessel anastomosis, comprising: acquiring a video stream representing in-surgery blood vessel patency, segmenting and establishing a blood vessel mask from the video stream; generating blood vessel geometric parameters according to the blood vessel mask; the blood vessel geometric parameters include blood vessel length, target position diameter and empty segment length; performing time sequence grayscale analysis according to the blood vessel geometric parameters to obtain blood vessel filling time and refilling velocity ratio. In view of the problem of insufficient identification ability of the existing technology for blood vessel anastomosis method for critical state anastomosis, by using in-surgery high-speed microscopic video, combining target detection, semantic segmentation, image skeletonization and time sequence analysis technology, the quantitative index of refilling velocity ratio (RVR) is automatically calculated, so that the current blood vessel state is more directly represented, and the doctor is facilitated to further judge the quality of blood vessel anastomosis.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence image processing technology, and more specifically to a detection method for assisting in the assessment of intraoperative vascular anastomosis quality. Background Technology

[0002] Intraoperative vascular anastomosis quality assessment is a crucial step in surgical procedures, evaluating whether the anastomosed vessels meet requirements for patency, tightness, alignment, and tension. It is widely used in various finger replantation, organ transplantation, flap transplantation, and cardiovascular surgeries. This assessment helps prevent postoperative complications such as thrombosis, vascular stenosis, and even tissue necrosis. Common methods for intraoperative vascular anastomosis quality assessment include visual observation, Doppler flow monitoring, and indocyanine green angiography. These methods primarily assess the postoperative vascular filling process, blood flow direction, and tissue perfusion extent through observation.

[0003] For example, patent application CN202411012432.4 discloses a quality assessment and monitoring system and method based on temperature and pressure at the vascular anastomosis site, relating to the field of medical device technology. It includes an implantable detection unit, a wireless receiver module, and a vascular anastomosis quality analysis platform. The implantable unit transmits the detected temperature / pressure data to the vascular anastomosis quality analysis platform via the wireless receiver module. This quality assessment and monitoring system and method based on temperature and pressure at the vascular anastomosis site can monitor and accurately analyze temperature / pressure differences at the surgical anastomosis site in real time, thereby providing doctors with crucial data to assess anastomosis quality. This not only improves the success rate of surgery and reduces the risk of postoperative complications but also effectively guides doctors to make timely adjustments during surgery to ensure optimal surgical results. Furthermore, the system can continuously monitor the recovery of the anastomosis site, providing a scientific basis for subsequent treatment, thus significantly improving the patient's treatment experience and rehabilitation quality.

[0004] For example, patent application CN202111492753.5 discloses a method, photoelectric forceps, and system for assessing the patency of vascular anastomosis. The method includes: selecting the distal location of the anastomotic vessel to be tested; acquiring the PPG signal of the artery at the distal location of the anastomotic vessel; processing the PPG signal to obtain the SpO2 value of blood oxygen saturation; monitoring the SpO2 value of the anastomotic vessel; and assessing the patency of the anastomotic vessel based on the SpO2 value. By monitoring the distal location of the anastomotic vessel, the PPG signal of the distal artery can be quickly acquired and converted into blood oxygen saturation. This allows for real-time monitoring of the blood oxygen saturation of the anastomotic vessel, providing a tool for rapidly and effectively assessing the patency of the anastomotic vessel during surgery. This enables early detection of abnormalities in reconstructed vascular anastomosis, facilitating timely correction and reversal of tissue ischemia or overperfusion injury, and improving the success rate of vascular anastomosis.

[0005] However, in practice, the inventors discovered that the most challenging aspect of clinical practice is not completely occluded anastomoses (such typical accidents can usually be detected through simple observation), but rather anastomoses in a "patent but not smooth" critical state. Such anastomoses may exhibit slow filling, localized eddies, or intermittent blood flow during surgery, which traditional observation and ICGA may misjudge as "patent." However, postoperatively, under the influence of factors such as coagulation factor activation, vasospasm, and tissue edema, they are highly prone to progressing to complete thrombotic occlusion. Studies have shown that ICGA has high sensitivity (approximately 95%) in identifying complete occlusion, but its sensitivity in identifying partial anastomotic stenosis (a reduction in lumen diameter of 50%–70%) is only about 60%–70%. Doppler ultrasound has even lower sensitivity in detecting low-velocity blood flow (approximately 55%–65%). Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, a detection method for assisting in the assessment of intraoperative vascular anastomosis quality is provided.

[0007] The specific technical solution is as follows: A detection method for assisting in the assessment of intraoperative vascular anastomosis quality, comprising: Step S1: Acquire a video stream characterizing intraoperative vascular patency, segment and construct a vascular mask from the video stream; Step S2: Generate vascular geometric parameters based on the vascular mask; The vascular geometric parameters include vascular length, target location diameter, and empty segment length; Step S3: Perform temporal grayscale analysis based on the aforementioned vascular geometric parameters to obtain the ratio of vascular filling time to refill rate.

[0008] On the other hand, step S1 includes: Step S11: Acquire the video stream and identify the blood vessel clip bounding box from the video stream; Step S12: Generate the region of interest based on the bounding box of the blood vessel clip; Step S13: Segment the region of interest to obtain the vascular region and establish the vascular mask.

[0009] On the other hand, step S12 includes: Step S121: Obtain the center point of the bounding box of the blood vessel clip and establish the center point connection line; Step S122: Determine the blood vessel direction based on the line connecting the center points, and generate the region width of the region of interest based on the length of the line connecting the center points; Step S123: Expand the bounding box of the blood vessel according to the direction of the blood vessel and the width of the region to obtain the region of interest.

[0010] On the other hand, step S13 includes: Step S131: Segment the region of interest to obtain a single-channel probability map corresponding to the blood vessel region; Step S132: Apply the Sigmoid activation function to the single-channel probability map to obtain a binarized image; Step S133: Perform morphological closing operation on the binarized image to obtain the blood vessel mask. On the other hand, step S2 includes: Step S21: Perform skeletonization processing on the blood vessel mask to obtain a single-pixel skeleton; Step S22: Generate a centerline coordinate sequence according to the single-pixel skeleton; Step S23: Measure the corresponding vascular geometric parameters according to the centerline coordinate sequence.

[0011] On the other hand, step S22 includes: Step S221: Obtain pixel coordinates from the single-pixel skeleton; Step S222: Perform 8-neighborhood discrimination on the pixel coordinates to obtain skeleton endpoints and branch points; Step S223: Establish an ordered sequence of centerline coordinates based on the skeleton endpoints and the branch points.

[0012] On the other hand, step S3 includes: Step S31: Perform time dimension analysis on the video stream to obtain the start and end times of blood filling; Step S32: Generate the vascular filling time based on the start and end times, and generate the refill rate ratio based on the empty segment length.

[0013] On the other hand, step S31 includes: Step S311: Calculate the average grayscale value of each image frame in the video stream based on the blood vessel mask; Step S312: Perform a 7-frame moving average smoothing on the average gray value to obtain a smoothed gray value sequence, and calculate the low threshold and high threshold respectively; Step S313: Classify each image frame in the smooth grayscale sequence according to the low threshold and the high threshold to obtain the fullness category; Step S314: Determine the start and end times according to the filling category.

[0014] On the other hand, in step S11, a pre-trained YOLOv10 is used to identify and output the blood vessel clip detection box.

[0015] A storage medium includes computer instructions that, when executed by a computer device, perform the detection method described above.

[0016] The above technical solution has the following advantages or beneficial effects: To address the issue of insufficient ability to identify critical anastomoses in existing vascular anastomosis methods, this paper utilizes intraoperative high-speed microscopic video, combined with target detection, semantic segmentation, image skeletonization, and temporal analysis techniques, to automatically calculate the refill rate ratio (RVR), a quantitative indicator that more intuitively represents the current vascular status and facilitates further assessment of vascular anastomosis quality by physicians. Attached Figure Description

[0017] Embodiments of the invention will be described more fully with reference to the accompanying drawings. However, the drawings are for illustration and explanation only and do not constitute a limitation on the scope of the invention.

[0018] Figure 1 This is an overall schematic diagram of an embodiment of the present invention; Figure 2 This is a schematic diagram of step S1 in an embodiment of the present invention; Figure 3 This is a schematic diagram of step S12 in an embodiment of the present invention; Figure 4 This is a schematic diagram of step S13 in an embodiment of the present invention; Figure 5 This is a schematic diagram of step S2 in an embodiment of the present invention; Figure 6 This is a schematic diagram of step S22 in an embodiment of the present invention; Figure 7 This is a schematic diagram of step S3 in an embodiment of the present invention; Figure 8 This is a schematic diagram of step S31 in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0022] This invention includes: A detection method for assisting in the assessment of intraoperative vascular anastomosis quality, such as... Figure 1 As shown, it includes: Step S1: Acquire the video stream characterizing intraoperative vascular patency, segment and construct a vascular mask from the video stream; Step S2: Generate vascular geometry parameters based on the vascular mask; Vascular geometry parameters include vessel length, target location diameter, and empty segment length; Step S3: Perform temporal grayscale analysis based on vascular geometric parameters to obtain the ratio of vascular filling time to refill rate.

[0023] Specifically, addressing the problem of insufficient ability to identify critical anastomoses in existing vascular anastomosis methods, this paper utilizes intraoperative high-speed microscopic video, combined with target detection, semantic segmentation, image skeletonization, and temporal analysis techniques, to automatically calculate the refill rate ratio (RVR), a quantitative indicator that more intuitively represents the current vascular status and facilitates further assessment of vascular anastomosis quality by physicians.

[0024] In practice, the above-mentioned detection method is mainly configured as a software implementation in computer equipment, such as a doctor's workstation or other equivalent computer equipment. This computer equipment is connected to an intraoperative high-speed microscopic imaging system, whose lens is pointed towards the surgical field, particularly focusing on the anastomosed blood vessels. Typical operating parameters are a frame rate of 1000fps, a resolution of 280×1024, and output of a real-time color or grayscale video stream. In some embodiments, this high-speed microscopic imaging system can also be implemented based on devices capable of outputting high-speed microscopic video streams, such as angiography systems, fluorescence imaging systems, ultrasound imaging systems, and intraoperative MRI systems. The main advantage is the ability to observe changes in filling at the anastomosis site and to output high-speed images to facilitate the assessment of the dynamic process of vascular filling.

[0025] Vascular anastomosis refers to the procedure of suturing blood vessels during certain surgical procedures, including finger replantation, organ transplantation, flap transplantation, and cardiovascular surgery. Taking finger replantation as an example, for each blood vessel requiring anastomosis, clamps are used to seal both sides of the vessel during the procedure. After the anastomosis is completed, the clamps are removed to allow blood flow to the anastomosis site and the severed finger area. Therefore, during video stream acquisition, the starting point is typically the time between the anastomosis and the removal of the clamps, and the entire process of blood flow into the anastomosis site after the clamps are removed is captured.

[0026] For multiple image frames in the video stream, a vascular mask is first obtained by processing the image through target detection and image segmentation to characterize the vascular anastomosis site to be identified. Then, the vascular geometric parameters, including vascular length, target position diameter and empty segment length, are calculated based on the vascular mask and the corresponding pixel length. These parameters are used to further calculate the vascular refilling speed.

[0027] After segmenting and obtaining the vascular mask corresponding to the vascular region, it can be returned to the video stream for temporal grayscale analysis to determine the time point when blood flow reperfusion in the vascular region causes the anastomosis site to fill. The ratio of vascular filling time to refill rate is further calculated. The above two indicators are finally presented on the display device so that doctors can further judge whether the blood flow perfusion at the vascular anastomosis site is normal and assess the quality of vascular anastomosis.

[0028] In one embodiment, such as Figure 2 As shown, step S1 includes: Step S11: Acquire the video stream and identify the bounding box of the blood vessel clip from the video stream; Step S12: Generate the region of interest based on the bounding box of the blood vessel clip; Step S13: Segment the region of interest to obtain the vascular region and create a vascular mask.

[0029] Specifically, to achieve better segmentation results, in this embodiment, after acquiring the corresponding video stream, when segmenting the blood vessel mask by extracting image frames, the image frames at the time when the blood vessel clips have not been removed are first selected. Then, a pre-trained YOLOv10 object detection network is used to detect the blood vessel clips in each image frame within that time interval in real time, generating the bounding box coordinates and confidence scores of the blood vessel clip bounding boxes. Typically, the microscopic video stream is acquired directly above the surgical field, coinciding with the projection direction of the blood vessel clips. Therefore, the line connecting the center points of the blood vessel clip bounding boxes can be considered equivalent to the direction of the blood vessel.

[0030] Based on this, the bounding box of the identified blood vessel clip is expanded to define the corresponding region of interest and then segmented, thereby excluding blood vessels in other irrelevant tissues, obtaining a more accurate blood vessel region, and establishing a binarized blood vessel mask.

[0031] In one embodiment, such as Figure 3 As shown, step S12 includes: Step S121: Obtain the center point of the blood vessel clip bounding box and establish the connection between the center points; Step S122: Determine the direction of the blood vessels based on the line connecting the center points, and generate the region width of the region of interest based on the length of the line connecting the center points; Step S123: Expand the bounding box of the blood vessel according to the direction of the blood vessel and the width of the region to obtain the region of interest.

[0032] Specifically, in the process of determining the region of interest, the center point of the bounding box of the blood vessel can be obtained first, and the center point line can be established to determine the direction of the blood vessel.

[0033] Subsequently, using the two center points as a reference, a rectangular region of interest is generated by expanding outward along the direction of the blood vessels. Specifically, the height of the rectangle is 1.5 times the distance between the two center points, and the width of the rectangle is appropriately extended perpendicular to the direction of the blood vessels to ensure that the region of interest completely includes the blood vessel segment between the two blood vessels.

[0034] Finally, the position of the blood vessel clip and the coordinates of the corresponding region of interest are recorded in each frame of the image to generate a frame-by-frame detection record file, which is used for dynamic localization of blood vessel regions in subsequent time-series analysis.

[0035] In one embodiment, such as Figure 4 As shown, step S13 includes: Step S131: Segment the region of interest to obtain a single-channel probability map corresponding to the blood vessel region; Step S132: Apply the Sigmoid activation function to the single-channel probability map to obtain a binarized image; Step S133: Perform morphological closing operation on the binarized image to obtain the blood vessel mask.

[0036] Specifically, after determining the region of interest, pixel-level semantic segmentation of the blood vessel region is required. Here, a U-Net network trained on a dataset pre-constructed from blood vessel microscopic images is used as the recognition model. The image of the region of interest is taken as input, and a single-channel probability map with the same size as the input image is output. The Sigmoid activation function is then applied to the single-channel probability map to obtain a binarized image.

[0037] Finally, morphological closing operations are performed on the binarized image, including first dilating the portion of the binarized image corresponding to the blood vessel region to fill the tiny holes in the segmented blood vessel region, and then performing erosion to remove the portion of the edge added during the dilution process, so that the blood vessel mask has spatial continuity.

[0038] In one embodiment, such as Figure 5 As shown, step S2 includes: Step S21: Perform skeletonization processing on the blood vessel mask to obtain a single-pixel skeleton; Step S22: Generate a centerline coordinate sequence based on the single-pixel skeleton; Step S23: Measure the corresponding vascular geometric parameters based on the centerline coordinate sequence.

[0039] Specifically, to achieve the measurement of vascular geometric parameters, this embodiment first needs to establish a refined vascular skeleton for the segmented vascular region to indicate the direction of the blood vessels. This includes using the skimage.morphology.skeletonize algorithm to refine the vascular region into a skeleton with a width of one pixel. If there are breakpoints in the skeleton, morphological dilation is performed and the skeleton is re-skeletonized, with a maximum of two iterations. Furthermore, it is determined whether there are branching parts in the skeleton. If there are small branches with a length of less than 15 pixels, the small branches are pruned, and the centerline of the main blood vessel is retained.

[0040] Then, the system traverses the path according to the single-pixel skeleton, saves the coordinates of each pixel along the path, generates a centerline coordinate sequence, and finally measures the corresponding blood vessel geometric parameters based on the centerline coordinate sequence.

[0041] In one embodiment, such as Figure 6 As shown, step S22 includes: Step S221: Obtain pixel coordinates from the single-pixel skeleton; Step S222: Perform 8-neighborhood discrimination on the pixel coordinates to obtain the skeleton endpoints and branch points; Step S223: Establish an ordered sequence of centerline coordinates based on the skeleton endpoints and branch points.

[0042] For the centerline coordinate sequence, firstly, the coordinates of all pixels on the skeleton are obtained. Then, an 8-neighborhood discrimination is performed on each pixel coordinate, and the number of its neighbors is recorded. Pixels with 1 neighbor are marked as skeleton endpoints, and pixels with 3 or more neighbors are marked as branch points. For simple blood vessel segments without branches, the 8-neighborhood is traced from the upstream endpoint to the downstream endpoint to generate an ordered list of centerline coordinates.

[0043] When a branch point appears, a directed search is performed on each branch point, the longest path is retained, and the order of the pixel coordinates recorded on the path is used as the center line coordinate sequence.

[0044] Subsequently, the vascular geometric parameters were calculated based on the centerline coordinate sequence.

[0045] Specifically, for the blood vessel length, for the centerline coordinate sequence, the Euclidean distance between adjacent points is calculated point by point and accumulated to obtain the pixel length of the blood vessel as the blood vessel length, with the unit being pixels.

[0046] The target location diameter is determined by taking a target point P at a position 10% of the arc length from the proximal starting point (i.e., a 1:9 ratio point at the proximal end) based on the blood vessel length. At point P, the local tangent direction is fitted using three centerline points before and after the point P, and the perpendicular normal direction is calculated. The target location diameter is then calculated by scanning along the normal direction in the original mask and recording the two intersection points between the normal and the boundary of the blood vessel mask (the position where the mask value changes from 1 to 0). The Euclidean distance between the intersection points is then calculated. To reduce the influence of noise, the diameter is also measured at the equivalent positions two pixels before and after point P, and the median of the five measurements is taken as the final target location diameter.

[0047] In one embodiment, such as Figure 7 As shown, step S3 includes: Step S31: Perform time-dimensional analysis on the video stream to obtain the start and end times of blood filling; Step S32: Generate the vascular filling time based on the start and end times, and generate the refill rate ratio based on the empty segment length.

[0048] Specifically, in order to determine the refill rate ratio, in this embodiment, the video stream is first analyzed in terms of time dimension to obtain the start and end times of blood filling, the blood vessel filling time is calculated based on the start and end times, and the refill rate ratio is generated by combining the empty segment length.

[0049] In one embodiment, such as Figure 8 As shown, step S31 includes: Step S311: Calculate the average gray value of each image frame in the video stream based on the blood vessel mask; Step S312: Smooth the average gray value by performing a 7-frame moving average to obtain a smoothed gray value sequence, and calculate the low threshold and high threshold respectively; Step S313: Classify each image frame in the smooth grayscale sequence according to the low threshold and the high threshold to obtain the fullness category; Step S314: Determine the start and end times based on the filling category.

[0050] Specifically, to accurately determine the start and end points, in this embodiment, the vascular mask is returned to the video stream, and then the average gray value is calculated for each image frame. This average gray value corresponds to the average gray value of the pixels covered by the vascular mask. The vascular masks applied to different image frames are the different vascular masks obtained through their respective processing.

[0051] Subsequently, the average grayscale value is added to the average grayscale sequence according to the image frame number, and the average grayscale sequence is smoothed according to the average grayscale of the preceding and following 7 frames. The smoothed values ​​are then added to the smoothed grayscale sequence in sequence.

[0052] Subsequently, low and high thresholds are calculated for the smoothed grayscale sequence. The low threshold is the 25th percentile of the corresponding grayscale distribution in the smoothed grayscale sequence, and the high threshold is the 75th percentile of the corresponding grayscale distribution.

[0053] Then, for each smooth gray value in the smooth gray-level sequence, high and low thresholds are applied for discrimination, thereby classifying the filling state of the vascular anastomosis site in the corresponding image frame into specific categories, including: Incomplete filling: The smooth gray value corresponding to the image frame is less than the low threshold; In fullness: The smooth grayscale value corresponding to the image frame is greater than or equal to the low threshold and less than the high threshold; Fully filled: The smooth gray value corresponding to the image frame is greater than or equal to the high threshold.

[0054] The classification process, which is based on grayscale thresholds for state classification, can be replaced by a time-series state prediction model based on Hidden Markov Models (HMMs) or Long Short-Term Memory Networks (LSTMs).

[0055] Then, the first image frame classified as being in the process of filling is marked as the start point of the start and end times, and the first image frame classified as being fully filled is marked as the end point of the start and end times. The time length between the start and end points is the vascular filling time.

[0056] As an optional implementation, when determining the start and end times, when the first image frame corresponding to the classification state is determined, the next two or more consecutive image frames are also jointly determined. When multiple consecutive image frames are in the corresponding classification state, the corresponding start and end times are retained; otherwise, they are determined as noise and discarded.

[0057] After obtaining the vascular filling time, the refill rate V can be calculated: V = empty_length / T; Where V is the refill rate, empty_length is the pre-calculated length of the empty segment of the blood vessel, and T is the blood vessel filling time.

[0058] The refilling rates of each vessel site during the anastomosis process are calculated according to this procedure, including the upstream refilling rate V_proximal and the downstream refilling rate V_distal.

[0059] The refill rate ratio RVR was then calculated: RVR = V_distal / V_proximal; Where RVR is the recharge rate ratio, V_distal is the downstream recharge rate, and V_proximal is the upstream recharge rate.

[0060] In the actual assessment process, it may be necessary to assess the patient's total RVR. In this case, RVRa is calculated from the arterial video and RVRv_single is calculated from the venous video. The sum of RVRv_single from all venous videos of the same patient is the patient's total venous RVR, RVRv.

[0061] This information, along with the vascular filling time, is then provided to doctors to help them assess the quality of the vascular anastomosis.

[0062] A storage medium includes computer instructions that, when executed by a computer device, perform the aforementioned detection method.

[0063] Those skilled in the art will understand that various aspects, or possible implementations of various aspects, of the present invention can be embodied as systems, methods, or computer program products. Therefore, various aspects, or possible implementations of various aspects, of the present invention can take the form of entirely hardware embodiments, entirely software embodiments (including firmware, resident software, etc.), or embodiments combining software and hardware aspects, all collectively referred to herein as "circuit," "module," or "system." Furthermore, various aspects, or possible implementations of various aspects, of the present invention can take the form of computer program products, which are computer instructions stored in memory.

[0064] The memory can be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or apparatuses, or any suitable combination thereof, such as random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, and portable read-only memory (CD-ROM).

[0065] A processor in a computer reads computer instructions stored in memory, enabling the processor to execute the functional actions specified in each step or combination of steps in a flowchart; and to generate means for implementing the functional actions specified in each block or combination of blocks in a flowchart.

[0066] It should be understood that a processor in a computer can be understood as one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components used to execute the aforementioned computer instructions.

[0067] Computer instructions may be executed entirely on the user's local computer, partially on the user's local computer, as a separate software package, partially on the user's local computer and partially on a remote computer, or entirely on a remote computer or server. It should also be noted that in some alternative implementations, the functions indicated by the steps in the flowchart or the blocks in the block diagram may not occur in the order shown in the diagram. For example, depending on the functions involved, two consecutive steps or blocks may actually be executed approximately simultaneously, or these blocks may sometimes be executed in reverse order.

[0068] Of course, in practical applications, the various components of a computer system are coupled together through a bus system. The bus system is used to enable communication and connection between these components. In addition to the data bus, the bus system also includes a power bus, a control bus, and a status signal bus.

[0069] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A detection method for assisting in evaluating the quality of anastomosis of blood vessels in surgery, characterized in that, include: Step S1: Acquire a video stream characterizing intraoperative vascular patency, segment and construct a vascular mask from the video stream; Step S2: Generate vascular geometric parameters based on the vascular mask; The vascular geometric parameters include vascular length, target location diameter, and empty segment length; Step S3: Perform temporal grayscale analysis based on the aforementioned vascular geometric parameters to obtain the ratio of vascular filling time to refill rate.

2. The detection method according to claim 1, characterized in that, Step S1 includes: Step S11: Acquire the video stream and identify the blood vessel clip bounding box from the video stream; Step S12: Generate the region of interest based on the bounding box of the blood vessel clip; Step S13: Segment the region of interest to obtain the vascular region and establish the vascular mask.

3. The detection method according to claim 2, characterized in that, Step S12 includes: Step S121: Obtain the center point of the bounding box of the blood vessel clip and establish the center point connection line; Step S122: Determine the blood vessel direction based on the line connecting the center points, and generate the region width of the region of interest based on the length of the line connecting the center points; Step S123: Expand the bounding box of the blood vessel according to the direction of the blood vessel and the width of the region to obtain the region of interest.

4. The method of claim 2, wherein Step S13 includes: Step S131: Segment the region of interest to obtain a single-channel probability map corresponding to the blood vessel region; Step S132: Apply the Sigmoid activation function to the single-channel probability map to obtain a binarized image; Step S133: Perform morphological closing operation on the binarized image to obtain the blood vessel mask.

5. The method of claim 1, wherein, Step S2 includes: Step S21: Perform skeletonization processing on the blood vessel mask to obtain a single-pixel skeleton; Step S22: Generate a centerline coordinate sequence according to the single-pixel skeleton; Step S23: Measure the corresponding vascular geometric parameters according to the centerline coordinate sequence.

6. The detection method according to claim 5, characterized in that, Step S22 includes: Step S221: Obtain pixel coordinates from the single-pixel skeleton; Step S222: Perform 8-neighborhood discrimination on the pixel coordinates to obtain skeleton endpoints and branch points; Step S223: Establish an ordered sequence of centerline coordinates based on the skeleton endpoints and the branch points.

7. The method of claim 1, wherein, Step S3 includes: Step S31: Perform time dimension analysis on the video stream to obtain the start and end times of blood filling; Step S32: Generate the vascular filling time based on the start and end times, and generate the refill rate ratio based on the empty segment length.

8. The detection method according to claim 7, characterized in that, Step S31 includes: Step S311: Calculate the average grayscale value of each image frame in the video stream based on the blood vessel mask; Step S312: Perform a 7-frame moving average smoothing on the average gray value to obtain a smoothed gray value sequence, and calculate the low threshold and high threshold respectively; Step S313: Classify each image frame in the smooth grayscale sequence according to the low threshold and the high threshold to obtain the fullness category; Step S314: Determine the start and end times according to the filling category.

9. The method of claim 2, wherein, In step S11, a pre-trained YOLOv10 is used to identify and output the blood vessel clip detection box.

10. A storage medium comprising computer instructions, characterized in that, When the computer device executes the computer instructions, it performs the detection method as described in any one of claims 1-9.

Images