WO2023130663A1 - 基于血管造影视频进行冠脉造影定量分析的方法和装置 - Google Patents

基于血管造影视频进行冠脉造影定量分析的方法和装置 Download PDF

Info

Publication number
WO2023130663A1
WO2023130663A1 PCT/CN2022/097245 CN2022097245W WO2023130663A1 WO 2023130663 A1 WO2023130663 A1 WO 2023130663A1 CN 2022097245 W CN2022097245 W CN 2022097245W WO 2023130663 A1 WO2023130663 A1 WO 2023130663A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
blood vessel
video
centerline
pixel point
Prior art date
Application number
PCT/CN2022/097245
Other languages
English (en)
French (fr)
Inventor
张碧莹
吴泽剑
曹君
Original Assignee
乐普（北京）医疗器械股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 乐普（北京）医疗器械股份有限公司 filed Critical 乐普（北京）医疗器械股份有限公司
Publication of WO2023130663A1 publication Critical patent/WO2023130663A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30101—Blood vessel; Artery; Vein; Vascular
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30172—Centreline of tubular or elongated structure

Definitions

  • the invention relates to the technical field of data processing, in particular to a method and device for quantitative analysis of coronary angiography based on angiography video.
  • Coronary artery stenosis will lead to insufficient blood supply to the heart, causing myocardial dysfunction and (or) pathological changes.
  • Angiography technology is based on the principle that X-rays cannot penetrate the contrast agent.
  • the contrast agent is injected into the blood vessel of the detection object, and the process of the contrast agent passing through the blood vessel under X-ray is imaged to output an angiography video.
  • the angiographic video of coronary vessels is first obtained based on angiographic technology, and then the doctor screens out the video images with stenotic blood vessels from the angiographic video based on personal experience for angiographic analysis.
  • the quantitative analysis of stenosis rate is also called qualitative comparative analysis (Qualitative Comparative Analysis, QCA) to calculate the corresponding vascular stenosis rate.
  • QCA quantitative Comparative Analysis
  • the purpose of the present invention is to address the defects of the prior art, to provide a method, device, electronic equipment and computer-readable storage medium for quantitative analysis of coronary angiography based on angiography video, and to perform video analysis on angiography video of coronary vessels.
  • Interception and video frame image extraction processing based on the image target detection and semantic segmentation model, the extracted image sequence is subjected to target detection and semantic segmentation processing of stenotic blood vessels, and the extracted image is optimized based on the confidence of target recognition.
  • Quantitative analysis of coronary angiography was performed on each stenotic segment of blood vessel to generate the corresponding vascular stenosis rate.
  • the invention can get rid of the excessive dependence on manual experience in the traditional practice, and improve the accuracy of image extraction and calculation accuracy of stenosis rate.
  • the first aspect of the embodiment of the present invention provides a method for quantitative analysis of coronary angiography based on angiography video, the method comprising:
  • each target detection frame corresponds to a detection frame confidence
  • a coronary angiography quantitative analysis is performed on the stenotic segment blood vessel mask image in each target detection frame to generate a corresponding blood vessel stenosis rate.
  • performing video frame image extraction on the angiography video to generate a corresponding first image sequence specifically includes:
  • the video frame image sequence includes a plurality of video frame images
  • each of the video frame images is used as the corresponding first image, and all the obtained first images are sorted in chronological order to generate the the first sequence of images;
  • the image target detection and semantic segmentation model includes a Mask R-CNN model; when the image target detection and semantic segmentation model is specifically a Mask R-CNN model, the residual network ResNet50 is used as its feature extraction backbone network.
  • performing image optimization processing on the first image sequence to obtain a specified number of optimal images specifically includes:
  • All the first images are sorted in descending order of the corresponding average confidence of the first images, and a specified number of first images that are ranked higher are used as the preferred images.
  • the coronary angiography quantitative analysis is performed on each of the preferred images for the stenotic vessel mask image in each target detection frame to generate a corresponding vascular stenosis rate, which specifically includes:
  • the first centerline includes a plurality of centerline pixels Point P i , where the first centerline pixel point P 1 is the closest point to the coronary artery entrance according to the blood flow direction, and the last centerline pixel point P N is the farthest point from the coronary artery entrance according to the blood flow direction, 1 ⁇ i ⁇ N, N is the total number of centerline pixels of the first centerline;
  • the edge of the first blood vessel analyze the length of the blood vessel diameter corresponding to each centerline pixel point P i on the first centerline to generate a corresponding first blood vessel diameter d i ;
  • a maximum value is selected as the blood vessel stenosis rate corresponding to the blood vessel mask image of the stenosis segment within the current target detection frame.
  • analyzing the blood vessel diameter length corresponding to each centerline pixel point P i on the first centerline to generate a corresponding first blood vessel diameter d i specifically includes:
  • a straight line passes through the upper left adjacent pixel point of the centerline pixel point P i , the center line pixel point P i and the lower right adjacent pixel point of the center line pixel point P i ;
  • the second straight line passes through The upper adjacent pixel of the center line pixel P i , the center line pixel P i and the lower adjacent pixel of the center line pixel P i ;
  • the third straight line passes through the center line pixel
  • the third straight line passes through the right side of the center line pixel point P i Adjacent pixels on the side, the centerline pixel P i and the left adjacent pixel of the center
  • the line segments where the first, second, third and fourth straight lines intersect with the edge of the first blood vessel are respectively recorded as corresponding first, second, third and fourth line segments; and for the first, The line segment lengths of the second, third and fourth line segments are calculated to generate corresponding first, second, third and fourth line segment lengths; and from the first, second, third and fourth line segment lengths, A minimum value is selected as the first blood vessel diameter d i corresponding to the central line pixel point P i .
  • the second aspect of the embodiment of the present invention provides a device for implementing the method described in the first aspect above, including: an acquisition module, an image preprocessing module, a stenotic segment blood vessel processing module, an image optimization module, and a quantitative analysis module;
  • the acquiring module is used to acquire angiography video of coronary angiography
  • the image preprocessing module is used to perform video frame image extraction on the angiography video to generate a corresponding first image sequence
  • the stenotic blood vessel processing module is used to perform target detection and semantic segmentation processing on each first image in the first image sequence based on a preset image target detection and semantic segmentation model, so that each Obtain one or more target detection frames for marking stenotic blood vessels on the first image and a section of stenotic blood vessel mask image in each target detection frame; each target detection frame corresponds to a detection frame Confidence;
  • the image optimization module is used to perform image optimization processing on the first image sequence according to the confidence of the detection frame to obtain a specified number of optimal images
  • the quantitative analysis module is configured to perform coronary angiography quantitative analysis on the mask image of the stenotic segment blood vessel in each target detection frame on each of the preferred images to generate a corresponding vascular stenosis rate.
  • the third aspect of the embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;
  • the processor is configured to be coupled with the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above;
  • the transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.
  • the fourth aspect of the embodiments of the present invention provides a computer-readable storage medium, the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer executes the above-mentioned first aspect. method directive.
  • An embodiment of the present invention provides a method, device, electronic device, and computer-readable storage medium for quantitative analysis of coronary angiography based on angiography video, which performs video interception and video frame image extraction processing on angiography video of coronary vessels, Based on the image target detection and semantic segmentation model, the extracted image sequence is subjected to target detection and semantic segmentation of stenotic vessels, and the extracted images are optimized based on the confidence of target recognition, and each stenotic vessel is crowned on the optimized image. Quantitative analysis of the angiography yields a corresponding vessel stenosis rate. Through the invention, the excessive dependence on manual experience in the traditional practice is eliminated, and the accuracy of image extraction and the calculation accuracy of stenosis rate are improved.
  • FIG. 1 is a schematic diagram of a method for quantitative analysis of coronary angiography based on angiography video provided by Embodiment 1 of the present invention
  • FIG. 2 is a block diagram of a device for quantitative analysis of coronary angiography based on angiography video provided by Embodiment 2 of the present invention
  • FIG. 3 is a schematic structural diagram of an electronic device provided by Embodiment 3 of the present invention.
  • Embodiment 1 of the present invention provides a method for quantitative analysis of coronary angiography based on angiography video, as shown in Figure 1, a schematic diagram of a method for quantitative analysis of coronary angiography based on angiography video provided in Embodiment 1 of the present invention , this method mainly includes the following steps:
  • Step 1 obtain angiographic video of coronary angiography.
  • the coronary angiography is to inject a contrast agent into the blood vessel of the detection object and take an image of the process of the contrast agent passing through the coronary artery under X-ray, and the angiography video is the video data obtained from the image shooting.
  • Step 2 performing video frame image extraction on the angiography video to generate a corresponding first image sequence
  • step 21 performing video interception on the angiography video, retaining the video content of the contrast agent filling coronary artery stage to generate a corresponding intercepted angiography video;
  • the embodiment of the present invention focuses on the video content after the contrast agent reaches the coronary artery, in order to improve Data analysis efficiency Therefore, it is necessary to intercept the angiographic video in advance.
  • intercepting There are many ways of intercepting; one of them is to set a relative time threshold based on the implementation experience of coronary angiography and place the angiographic video at this relative time. The video data before the threshold is cut off, and the video data after the relative time threshold is retained as the video content of the contrast agent filling coronary artery stage to generate a corresponding intercepted contrast video;
  • Step 22 in chronological order, perform video frame image extraction processing on the intercepted contrast video to generate a corresponding video frame image sequence, and perform statistics on the number of video frame images in the video frame image sequence to generate a corresponding first total number;
  • the video frame The image sequence includes a plurality of video frame images;
  • each type of video data has a corresponding video sampling frame rate parameter by default.
  • the video sampling frame rate parameter corresponding to the angiography video the angiography video is intercepted and extracted into the video frame image, and each frame image extracted is the video frame image.
  • the sequence of video frame images can be obtained by sorting the video frame images in chronological order;
  • Step 23 when the first total number does not exceed the preset threshold of the total number of images, use each video frame image as the corresponding first image, and sort all the obtained first images in chronological order to generate the first image sequence; when When the first total number exceeds the threshold of the total number of images, extract the video frame images whose sorting indexes are all odd-numbered indexes from the video frame image sequence as the corresponding first image, or extract the video frame images whose sorting indexes are all even-numbered indexes as the corresponding first images; and sort all the obtained first images in chronological order to generate a first image sequence.
  • the control method is to pre-set an image Total threshold, and identify whether the total number of images in the video frame image sequence, that is, the first total number, exceeds the threshold; if it does not exceed the threshold, it means that there is no need to reduce the number of images in the video frame image sequence, and directly use each video frame image as The first image, and the first image sequence composed of the first image is sent to the subsequent steps for processing; if it exceeds the threshold, it means that the video frame image sequence needs to be reduced in number of images.
  • the method of frame extraction and reduction of adjacent images is used to achieve the purpose of not losing valid data.
  • the current step extracts all odd frames from the video frame image sequence or extracts all even frames to form the first image sequence.
  • the image is subjected to frame reduction processing.
  • Step 3 Based on the preset image target detection and semantic segmentation model, perform target detection and semantic segmentation processing on each first image in the first image sequence to obtain one or more A target detection frame for marking a stenotic blood vessel and a mask image of a stenotic blood vessel in each target detection frame;
  • each target detection frame corresponds to a detection frame confidence
  • the image target detection and semantic segmentation model includes the Mask R-CNN model
  • the residual network ResNet50 is used as Its feature extraction backbone network.
  • the image target detection and semantic segmentation model is used to detect the stenotic blood vessel target on the first input image, so as to obtain one or more target detection frames for marking the stenotic blood vessels, and the detection frame confidence of each target detection frame The degree is used to identify the credibility of the image in the frame as a stenotic blood vessel image; the image target detection and semantic segmentation model is also used to perform image semantic segmentation on the recognized target, that is, the recognized stenotic blood vessel in each target detection frame , and use the segmented mask image as the stenotic blood vessel mask image;
  • the network structure can refer to the article "Mask R-CNN" published by the authors Kaiming He, Georgia Gkioxari, Piotr Doll'ar and Ross Girshick, including: feature extraction network layer, region candidate network (Region Proposal Network, RPN) layer, region alignment (Region Of Interest Align, ROI Align) network layer and region head (ROI HEAD) network layer; the feature extraction network layer is connected to the region candidate network layer, and the region candidate network layer is connected to the region alignment network layer; the region alignment network layer is connected to the region The head network layer is connected; the regional head network layer includes; two sub-networks are the target detection branch network and the target segmentation branch network respectively; the target detection branch network is used to output the target detection frame and the detection frame confidence of the narrow blood vessel,
  • the feature extraction network layer in the embodiment of the present invention is specifically composed of a five-level residual network (Residual Network, ResNet) and a corresponding five-level feature pyramid network (Feature Pyramid Networks, FPN);
  • the region candidate network layer includes a five-level region candidate network, Corresponding to the five-level feature pyramid network; when implementing the five-level residual network, the embodiment of the present invention uses the ResNet-50 network structure for implementation and uses it as the backbone network for feature extraction.
  • Step 4 according to the confidence of the detection frame, perform image optimization processing on the first image sequence to obtain a specified number of optimal images;
  • the first image sequence calculates the mean value of the detection frame confidences of all target detection frames on each first image to generate the corresponding first image average confidence; To the smallest order, sort all the first images, and use the specified number of first images that are ranked higher as the preferred images.
  • the three first images with the most obvious features of segmental blood vessels were selected as the preferred images.
  • Step 5 on each preferred image, carry out coronary angiography quantitative analysis to the stenotic segment blood vessel mask image in each target detection frame to generate the corresponding blood vessel stenosis rate;
  • step 51 traverse each target detection frame on the current preferred image, and mark the currently traversed target detection frame as the current target detection frame;
  • Step 52 performing vessel edge and vessel centerline identification on the mask image of the stenotic segment blood vessel within the current target detection frame to generate a corresponding first vessel edge and first centerline;
  • the first centerline includes a plurality of centerline pixel points P i , wherein the first centerline pixel point P 1 is the point closest to the entrance of the coronary artery according to the blood flow direction, and the last centerline pixel point P N is the point according to the blood flow direction.
  • the point farthest from the entrance of the coronary artery in the flow direction 1 ⁇ i ⁇ N, where N is the total number of centerline pixels of the first centerline;
  • the recognition of the blood vessel centerline on the stenotic blood vessel mask image there are many ways to realize the recognition of the blood vessel centerline on the stenotic blood vessel mask image; one of them is to perform binary processing on the image content of the current target detection frame to obtain the second binary image, the second binary image
  • the pixel values of all pixels in the original stenotic blood vessel mask image on the map will be converted to the preset foreground pixel value A, and the pixel values of all pixel points outside the stenotic blood vessel mask image will be converted to the preset
  • the background pixel value B of the vascular image on the premise of not changing the topological properties of the vascular image, based on the topology thinning method, the centerline extraction process is performed on the stenotic segment blood vessel mask image of the first binary image to generate the first centerline;
  • the blood vessel Image topological properties mainly refer to the connectivity of blood vessels;
  • the point closest to the entrance of the coronary artery that is, the entry point of the stenosis segment
  • the distance from the coronary artery is taken as the last centerline pixel point P N of the first centerline
  • Step 53 according to the edge of the first blood vessel, analyze the length of the blood vessel diameter corresponding to each centerline pixel point P i on the first centerline to generate a corresponding first blood vessel diameter d i ;
  • step 531 according to the directional relationship between the central line pixel point P i and its adjacent eight-field pixel points, four straight lines are made across the central line pixel point P i and recorded as the first, second, third and fourth straight lines respectively;
  • the first straight line passes through the upper left adjacent pixel point of the central line pixel point P i , the central line pixel point P i and the lower right adjacent pixel point of the central line pixel point P i ;
  • the second straight line passes through the central line pixel point P The upper adjacent pixel of i , the centerline pixel P i and the lower adjacent pixel of the centerline pixel P i ;
  • the third straight line passes through the upper right adjacent pixel of the centerline pixel P i , the centerline pixel P i and the lower left adjacent pixel of the central line pixel P i ;
  • the third straight line passes through the right adjacent pixel of the central line pixel P i , the central line pixel P i and the left adjacent pixel of the central line pixel P i ;
  • Step 532 record the line segments where the first, second, third and fourth straight lines intersect with the edge of the first blood vessel as the corresponding first, second, third and fourth line segments respectively; and for the first, second, The line segment lengths of the third and fourth line segments are calculated to generate corresponding first, second, third and fourth line segment lengths; and from the first, second, third and fourth line segment lengths, the minimum value is selected as the The first blood vessel diameter d i corresponding to the centerline pixel point P i ;
  • the diameter of the blood vessel passing through the centerline pixel point P i to the edge of the first blood vessel should be within the range of all straight lines passing through the centerline pixel point P i ; and there are actually only 4 straight lines passing through any pixel point on the image
  • the diameter of the blood vessel passing through the centerline pixel point P i to the edge of the first blood vessel can only be the first, second, third and fourth One of the four line segments; after the selection range of the vessel diameter is determined, the length of the shortest line segment is used as the first vessel diameter d i corresponding to the central line pixel point P i ;
  • Step 54 according to the blood vessel linear change relationship from the centerline pixel point P 1 to the centerline pixel point P N , and the first blood vessel diameter d i of each centerline pixel point P i , the stenosis corresponding to each centerline pixel point P i Analyze the rate to generate the corresponding first stenosis rate r i ;
  • Step 542 calculate the linear change diameter length corresponding to each centerline pixel point P i to generate the corresponding first reference diameter d' i ;
  • the vascular stenosis rate at the location is the first stenosis rate r i ;
  • Step 55 from all obtained first stenosis rates r i , select the maximum value as the blood vessel stenosis rate corresponding to the stenosis segment blood vessel mask image in the current target detection frame;
  • the maximum stenosis rate in a section of blood vessel is taken as the stenosis rate of the blood vessel in the mask image of the narrowed section blood vessel in the current target detection frame;
  • Step 56 take the next unprocessed target detection frame as the current target detection frame, and go to step 52 to continue processing until the blood vessel stenosis ratios of the stenotic segment blood vessel mask images in all target detection frames on the current preferred image are confirmed. .
  • multiple preferred images can be extracted from an angiographic video, and the vascular stenosis rate of one or more stenotic vessels on each preferred image can be obtained through quantitative analysis of coronary angiography.
  • multiple optimal images with stenotic blood vessel target detection frame and blood vessel stenosis rate can be provided to the doctor as parameter data at the same time; image fusion can also be performed on multiple optimal images, and Calculate the mean value of the stenosis rate of the stenotic blood vessels at the same position, and finally provide a fused image with a stenotic blood vessel target detection frame and the mean value of the stenosis rate to the doctor as parameter data.
  • Fig. 2 is a module structure diagram of a device for quantitative analysis of coronary angiography based on angiography video provided in Embodiment 2 of the present invention.
  • the device may be a terminal device or a server implementing the method of the embodiment of the present invention, or it may be the same as the above-mentioned
  • the device includes: an acquisition module 201 , an image preprocessing module 202 , a stenotic segment blood vessel processing module 203 , an image optimization module 204 and a quantitative analysis module 205 .
  • the acquiring module 201 is used for acquiring angiography video of coronary angiography.
  • the image preprocessing module 202 is used to extract video frame images from the angiography video to generate a corresponding first image sequence.
  • the stenotic blood vessel processing module 203 is used to perform target detection and semantic segmentation processing on each first image in the first image sequence based on a preset image target detection and semantic segmentation model, so that on each first image Obtain one or more target detection frames for marking stenotic blood vessels and a mask image of a segment of stenotic blood vessels in each target detection frame; each target detection frame corresponds to a detection frame confidence.
  • the image optimization module 204 is configured to perform image optimization processing on the first image sequence to obtain a specified number of optimal images according to the confidence of the detection frame.
  • the quantitative analysis module 205 is used to perform coronary angiography quantitative analysis on the stenotic segment blood vessel mask image in each target detection frame on each preferred image to generate a corresponding blood vessel stenosis rate.
  • An apparatus for performing quantitative analysis of coronary angiography based on angiography video provided by an embodiment of the present invention can perform the method steps in the above method embodiment, and its implementation principle and technical effect are similar, and will not be repeated here.
  • each module of the above device is only a division of logical functions, and may be fully or partially integrated into one physical entity or physically separated during actual implementation.
  • these modules can all be implemented in the form of calling software through processing elements; they can also be implemented in the form of hardware; some modules can also be implemented in the form of calling software through processing elements, and some modules can be implemented in the form of hardware.
  • the acquisition module can be a separate processing element, or it can be integrated into a chip of the above-mentioned device.
  • it can also be stored in the memory of the above-mentioned device in the form of program code, and a certain processing element of the above-mentioned device can Call and execute the functions of the modules identified above.
  • each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or an instruction in the form of software.
  • the above modules may be one or more integrated circuits configured to implement the above method, for example: one or more specific integrated circuits (Application Specific Integrated Circuit, ASIC), or, one or more digital signal processors ( Digital Signal Processor, DSP), or, one or more Field Programmable Gate Arrays (Field Programmable Gate Array, FPGA), etc.
  • ASIC Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • FPGA Field Programmable Gate Array
  • the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processors that can call program codes.
  • these modules can be integrated together and implemented in the form of a System-on-a-chip (SOC).
  • SOC System-on-a-chip
  • all or part of them may be implemented by software, hardware, firmware or any combination thereof.
  • software When implemented using software, it may be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer instructions.
  • the above-mentioned computers may be general-purpose computers, special-purpose computers, computer networks, or other programmable devices.
  • the above-mentioned computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • the above-mentioned computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or a data center integrated with one or more available media.
  • the above-mentioned usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, or a magnetic tape), an optical medium (for example, DVD), or a semiconductor medium (for example, a solid state disk (solid state disk, SSD)) and the like.
  • FIG. 3 is a schematic structural diagram of an electronic device provided by Embodiment 3 of the present invention.
  • the electronic device may be the aforementioned terminal device or server, or may be a terminal device or server connected to the aforementioned terminal device or server to implement the method of the embodiment of the present invention.
  • the electronic device may include: a processor 301 (such as a CPU), a memory 302 , and a transceiver 303 ;
  • Various instructions may be stored in the memory 302 for completing various processing functions and realizing the methods and processing procedures provided in the above-mentioned embodiments of the present invention.
  • the electronic device involved in this embodiment of the present invention further includes: a power supply 304 , a system bus 305 and a communication port 306 .
  • the system bus 305 is used to realize the communication connection among the components.
  • the above-mentioned communication port 306 is used for connection and communication between the electronic device and other peripheral devices.
  • the system bus mentioned in FIG. 3 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus or the like.
  • PCI Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the system bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.
  • the communication interface is used to realize the communication between the database access device and other devices (such as client, read-write library and read-only library).
  • the memory may include random access memory (Random Access Memory, RAM), and may also include non-volatile memory (Non-Volatile Memory), such as at least one disk memory.
  • processor can be general-purpose processor, comprises central processing unit CPU, network processor (Network Processor, NP) etc.; Can also be digital signal processor DSP, application-specific integrated circuit ASIC, field programmable gate array FPGA or other available Program logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • CPU central processing unit
  • NP Network Processor
  • DSP digital signal processor
  • ASIC application-specific integrated circuit
  • FPGA field programmable gate array
  • embodiments of the present invention also provide a computer-readable storage medium, and instructions are stored in the storage medium, and when the storage medium is run on a computer, the computer executes the methods and processing procedures provided in the above-mentioned embodiments.
  • the embodiment of the present invention also provides a chip for running instructions, and the chip is used for executing the method and the processing procedure provided in the foregoing embodiments.
  • An embodiment of the present invention provides a method, device, electronic device, and computer-readable storage medium for quantitative analysis of coronary angiography based on angiography video, which performs video interception and video frame image extraction processing on angiography video of coronary vessels, Based on the image target detection and semantic segmentation model, the extracted image sequence is subjected to target detection and semantic segmentation of stenotic vessels, and the extracted images are optimized based on the confidence of target recognition, and each stenotic vessel is crowned on the optimized image. Quantitative analysis of the angiography yields a corresponding vessel stenosis rate. Through the invention, the excessive dependence on manual experience in the traditional practice is eliminated, and the accuracy of image extraction and the calculation accuracy of stenosis rate are improved.
  • RAM random access memory
  • ROM read-only memory
  • EEPROM electrically programmable ROM
  • EEPROM electrically erasable programmable ROM
  • registers hard disk, removable disk, CD-ROM, or any other Any other known storage medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Geometry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

本发明实施例涉及一种基于血管造影视频进行冠脉造影定量分析的方法和装置，所述方法包括：获取血管造影视频；进行视频帧图像提取生成第一图像序列；基于图像目标检测与语义分割模型对第一图像序列中各个第一图像进行狭窄段血管的目标检测和语义分割处理从而在每个第一图像上得到一个或多个用于标记狭窄段血管的目标检测框以及在每个目标检测框中的一段狭窄段血管掩膜图像；对第一图像序列进行图像优选处理得到指定数量的优选图像；在各个优选图像上对每个目标检测框内的狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率。通过本发明可摆脱传统做法中对人工经验的过度依赖，提高图像提取准确度和狭窄率计算精度。

Description

基于血管造影视频进行冠脉造影定量分析的方法和装置
本申请要求于2022年1月7日提交中国专利局、申请号为202210018415.6、发明名称为“基于血管造影视频进行冠脉造影定量分析的方法和装置”的中国专利申请的优先权。
技术领域
本发明涉及数据处理技术领域，特别涉及一种基于血管造影视频进行冠脉造影定量分析的方法和装置。
背景技术
冠状动脉狭窄会导致心脏供血不足，从而引起心肌机能出现障碍和(或)病变。血管造影技术是基于X光无法穿透显影剂的原理，将显影剂注入检测对象血管中，并对X光下显影剂通过血管的过程进行影像拍摄从而输出血管造影视频。对冠状动脉的狭窄情况进行检测时，常规情况下先基于血管造影技术获得冠脉血管的血管造影视频，再由医生根据个人经验从血管造影视频中筛选出带有狭窄段血管的视频图像进行血管狭窄率的定量分析也称为定性比较分析(Qualitative Comparative Analysis，QCA)从而计算出对应的血管狭窄率。这种操作模式过于依赖人为因素，诸如人员从业经验、人眼的识别能力等，极容易出现视频图像提取不准确、狭窄率计算精度不够等问题。
发明内容
本发明的目的，就是针对现有技术的缺陷，提供一种基于血管造影视频进行冠脉造影定量分析的方法、装置、电子设备及计算机可读存储介质，对冠脉 血管的血管造影视频进行视频截取和视频帧图像提取处理，基于图像目标检测与语义分割模型对提取出的图像序列进行狭窄段血管的目标检测和语义分割处理，基于目标识别的置信度对提取图像进行优选，在优选图像上对每段狭窄段血管进行冠脉造影定量分析生成对应的血管狭窄率。通过本发明可摆脱传统做法中对人工经验的过度依赖，提高图像提取准确度和狭窄率计算精度。
为实现上述目的，本发明实施例第一方面提供了一种基于血管造影视频进行冠脉造影定量分析的方法，所述方法包括：
获取冠状动脉造影的血管造影视频；
对所述血管造影视频进行视频帧图像提取生成对应的第一图像序列；
基于预设的图像目标检测与语义分割模型，对所述第一图像序列中各个第一图像进行狭窄段血管的目标检测和语义分割处理，从而在每个所述第一图像上得到一个或多个用于标记狭窄段血管的目标检测框以及在每个所述目标检测框中的一段狭窄段血管掩膜图像；每个所述目标检测框对应一个检测框置信度；
根据所述检测框置信度，对所述第一图像序列进行图像优选处理得到指定数量的优选图像；
在各个所述优选图像上对每个所述目标检测框内的所述狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率。
优选的，所述对所述血管造影视频进行视频帧图像提取生成对应的第一图像序列，具体包括：
对所述血管造影视频的进行视频截取，保留造影剂充盈冠状动脉阶段的视频内容生成对应的截取造影视频；
按时间先后顺序，对所述截取造影视频进行视频帧图像提取处理生成对应的视频帧图像序列，并对所述视频帧图像序列的视频帧图像数量进行统计生成对应的第一总数；所述视频帧图像序列包括多个视频帧图像；
当所述第一总数未超过预设的图像总数阈值时，将各个所述视频帧图像 作为对应的所述第一图像，并按时间先后顺序对得到的所有所述第一图像进行排序生成所述第一图像序列；
当所述第一总数超过所述图像总数阈值时，从所述视频帧图像序列中提取排序索引全为奇数索引的所述视频帧图像作为对应的所述第一图像，或者提取排序索引全为偶数索引的所述视频帧图像作为对应的所述第一图像；并按时间先后顺序对得到的所有所述第一图像进行排序生成所述第一图像序列。
优选的，所述图像目标检测与语义分割模型包括Mask R-CNN模型；所述图像目标检测与语义分割模型具体为Mask R-CNN模型时，采用残差网络ResNet50作为其特征提取骨干网络。
优选的，所述根据所述检测框置信度，对所述第一图像序列进行图像优选处理得到指定数量的优选图像，具体包括：
在所述第一图像序列中，对各个所述第一图像上所有所述目标检测框的所述检测框置信度进行均值计算，生成对应的第一图像平均置信度；
按对应的所述第一图像平均置信度从大到小的顺序，对所有所述第一图像进行排序，并将排序靠前的指定数量的所述第一图像作为所述优选图像。
优选的，所述在各个所述优选图像上对每个所述目标检测框内的所述狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率，具体包括：
在当前优选图像上对各个所述目标检测框进行遍历，并将当前遍历的所述目标检测框记为当前目标检测框；
对所述当前目标检测框内的所述狭窄段血管掩膜图像进行血管边缘和血管中心线识别生成对应的第一血管边缘和第一中心线；所述第一中心线包括多个中心线像素点P i，其中第一个中心线像素点P 1为按血流方向距离冠状动脉入口最近的点，最后一个中心线像素点P N为按血流方向距离冠状动脉入口最远的点，1≤i≤N，N为所述第一中心线的中心线像素点总数；
根据所述第一血管边缘，对所述第一中心线上各个所述中心线像素点P i对应的血管直径长度进行分析生成对应的第一血管直径d i；
根据所述中心线像素点P 1到所述中心线像素点P N的血管线性变化关系，以及各个所述中心线像素点P i的所述第一血管直径d i，对各个所述中心线像素点P i对应的狭窄率进行分析生成对应的第一狭窄率r i；
从得到的所有所述第一狭窄率r i中，选择最大值作为与所述当前目标检测框内的所述狭窄段血管掩膜图像对应的所述血管狭窄率。
进一步的，所述根据所述第一血管边缘，对所述第一中心线上各个所述中心线像素点P i对应的血管直径长度进行分析生成对应的第一血管直径d i，具体包括：
按所述中心线像素点P i与其邻八域像素点的方向关系，过所述中心线像素点P i做四条直线分别记为第一、第二、第三和第四直线；所述第一直线过所述中心线像素点P i的左上相邻像素点、所述中心线像素点P i和所述中心线像素点P i的右下相邻像素点；所述第二直线过所述中心线像素点P i的上方相邻像素点、所述中心线像素点P i和所述中心线像素点P i的下方相邻像素点；所述第三直线过所述中心线像素点P i的右上相邻像素点、所述中心线像素点P i和所述中心线像素点P i的左下相邻像素点；所述第三直线过所述中心线像素点P i的右方相邻像素点、所述中心线像素点P i和所述中心线像素点P i的左方相邻像素点；
将所述第一、第二、第三和第四直线与所述第一血管边缘相交的线段分别记为对应的第一、第二、第三和第四线段；并对所述第一、第二、第三和第四线段的线段长度进行计算生成对应的第一、第二、第三和第四线段长度；并从所述第一、第二、第三和第四线段长度中，选择最小值作为与所述中心线像素点P i对应的所述第一血管直径d i。
进一步的，所述根据所述中心线像素点P 1到所述中心线像素点P N的血管线性变化关系，以及各个所述中心线像素点P i的所述第一血管直径d i，对各个所述中心线像素点P i对应的狭窄率进行分析生成对应的第一狭窄率r i，具体包括：
根据所述第一血管直径d 1和所述第一血管直径d N，构建反映所述中心线像素点P 1到所述中心线像素点P N血管线性变化关系的线性函数f(i)，f(i)＝d 1+k*(i-1)，k＝(d N-d 1)/(N-1)；
根据所述线性变化关系函数f(i)，对各个所述中心线像素点P i对应的线性变化直径长度进行计算生成对应的第一参考直径d’ i；
根据所述第一血管直径d i和所述第一参考直径d’ i，计算各个所述中心线像素点P i对应的所述第一狭窄率r i，r i＝1-d i/d’ i。
本发明实施例第二方面提供了一种实现上述第一方面所述的方法的装置，包括：获取模块、图像预处理模块、狭窄段血管处理模块、图像优选模块和定量分析模块；
所述获取模块用于获取冠状动脉造影的血管造影视频；
所述图像预处理模块用于对所述血管造影视频进行视频帧图像提取生成对应的第一图像序列；
所述狭窄段血管处理模块用于基于预设的图像目标检测与语义分割模型，对所述第一图像序列中各个第一图像进行狭窄段血管的目标检测和语义分割处理，从而在每个所述第一图像上得到一个或多个用于标记狭窄段血管的目标检测框以及在每个所述目标检测框中的一段狭窄段血管掩膜图像；每个所述目标检测框对应一个检测框置信度；
所述图像优选模块用于根据所述检测框置信度，对所述第一图像序列进行图像优选处理得到指定数量的优选图像；
所述定量分析模块用于在各个所述优选图像上对每个所述目标检测框内的所述狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率。
本发明实施例第三方面提供了一种电子设备，包括：存储器、处理器和收发器；
所述处理器用于与所述存储器耦合，读取并执行所述存储器中的指令，以实现上述第一方面所述的方法步骤；
所述收发器与所述处理器耦合，由所述处理器控制所述收发器进行消息收发。
本发明实施例第四方面提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机指令，当所述计算机指令被计算机执行时，使得所述计算机执行上述第一方面所述的方法的指令。
本发明实施例提供了一种基于血管造影视频进行冠脉造影定量分析的方法、装置、电子设备及计算机可读存储介质，对冠脉血管的血管造影视频进行视频截取和视频帧图像提取处理，基于图像目标检测与语义分割模型对提取出的图像序列进行狭窄段血管的目标检测和语义分割处理，基于目标识别的置信度对提取图像进行优选，在优选图像上对每段狭窄段血管进行冠脉造影定量分析生成对应的血管狭窄率。通过本发明摆脱了传统做法中对人工经验的过度依赖，提高了图像提取准确度和狭窄率计算精度。
附图说明
图1为本发明实施例一提供的一种基于血管造影视频进行冠脉造影定量分析的方法示意图；
图2为本发明实施例二提供的一种基于血管造影视频进行冠脉造影定量分析的装置的模块结构图；
图3为本发明实施例三提供的一种电子设备的结构示意图。
具体实施方式
为了使本发明的目的、技术方案和优点更加清楚，下面将结合附图对本发明作进一步地详细描述，显然，所描述的实施例仅仅是本发明一部份实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其它实施例，都属于本发明保护的范围。
本发明实施例一提供的一种基于血管造影视频进行冠脉造影定量分析的 方法，如图1为本发明实施例一提供的一种基于血管造影视频进行冠脉造影定量分析的方法示意图所示，本方法主要包括如下步骤：
步骤1，获取冠状动脉造影的血管造影视频。
这里，冠状动脉造影是将显影剂注入检测对象血管中并对X光下显影剂通过冠状动脉血管的过程进行影像拍摄，血管造影视频为该由影像拍摄得到的视频数据。
步骤2，对血管造影视频进行视频帧图像提取生成对应的第一图像序列；
具体包括：步骤21，对血管造影视频的进行视频截取，保留造影剂充盈冠状动脉阶段的视频内容生成对应的截取造影视频；
这里，因为血管造影视频包含了从造影剂注入血管到逐渐充盈整个冠状动脉再到逐渐消散的全过程视频内容，而本发明实施例关注的应是造影剂到达冠状动脉之后的视频内容，为提高数据分析效率所以要预先要对血管造影视频进行视频截取，截取的方式有多种；其中一种是根据冠状动脉造影的实施经验可设定一个相对时间阈值并将血管造影视频中处于该相对时间阈值之前的视频数据截除，保留该相对时间阈值之后的视频数据作为造影剂充盈冠状动脉阶段的视频内容生成对应的截取造影视频；
步骤22，按时间先后顺序，对截取造影视频进行视频帧图像提取处理生成对应的视频帧图像序列，并对视频帧图像序列的视频帧图像数量进行统计生成对应的第一总数；其中，视频帧图像序列包括多个视频帧图像；
这里，每类视频数据都默认带有一个对应的视频采样帧率参数，按血管造影视频对应的视频采样帧率参数对截取造影视频进视频帧图像提取，提取出的每帧图像就是视频帧图像，按时间先后顺序对视频帧图像进行排序就能得到视频帧图像序列；
步骤23，当第一总数未超过预设的图像总数阈值时，将各个视频帧图像作为对应的第一图像，并按时间先后顺序对得到的所有第一图像进行排序生成第一图像序列；当第一总数超过图像总数阈值时，从视频帧图像序列中提取 排序索引全为奇数索引的视频帧图像作为对应的第一图像，或者提取排序索引全为偶数索引的视频帧图像作为对应的第一图像；并按时间先后顺序对得到的所有第一图像进行排序生成第一图像序列。
这里，若视频帧图像序列中的图像数量过多会影响后续步骤中的模型运算效率，为提高模型运算效率则预先需要对视频帧图像序列进行图像数量控制；控制的方法就是预先设定一个图像总数阈值，并对视频帧图像序列中的图像总数也就是第一总数是否超过该阈值进行识别；若未超过该阈值，说明无需对视频帧图像序列进行图像数量缩减，直接将各个视频帧图像作为第一图像，并将由第一图像组成的第一图像序列送入后续步骤进行处理；若超过该阈值，说明需要对视频帧图像序列进行图像数量缩减，为保证不因图像缩减造成有效数据丢失，在缩减时采用对相邻图像进行抽帧缩减的方式来达到不丢失有效数据的目的，当前步骤从视频帧图像序列中提取所有奇数帧或者提取所有偶数帧构成第一图像序列就是在对相邻图像进行抽帧缩减处理。
步骤3，基于预设的图像目标检测与语义分割模型，对第一图像序列中各个第一图像进行狭窄段血管的目标检测和语义分割处理，从而在每个第一图像上得到一个或多个用于标记狭窄段血管的目标检测框以及在每个目标检测框中的一段狭窄段血管掩膜图像；
其中，每个目标检测框对应一个检测框置信度；图像目标检测与语义分割模型包括Mask R-CNN模型；图像目标检测与语义分割模型具体为Mask R-CNN模型时，采用残差网络ResNet50作为其特征提取骨干网络。
这里，图像目标检测与语义分割模型用于对输入的第一图像进行狭窄段血管目标检测，从而得到一个或多个用于标记狭窄段血管的目标检测框，每个目标检测框的检测框置信度用于标识框内图像为狭窄段血管图像的可信度；图像目标检测与语义分割模型还用于在每个目标检测框中对已识别目标也就是已识别的狭窄段血管进行图像语义分割，并将分割出的掩膜图像的作为狭窄段血管掩膜图像；
图像目标检测与语义分割模型的实现方式有多种，其中一种就是基于Mask R-CNN模型的神经网络架构进行实现；当图像目标检测与语义分割模型具体为Mask R-CNN模型时，其神经网络结构可参考由作者Kaiming He、Georgia Gkioxari、Piotr Doll′ar和Ross Girshick发表的文章《Mask R-CNN》，包括：特征提取网络层、区域候选网络(Region Proposal Network，RPN)层、区域对齐(Region Of Interest Align，ROI Align)网络层和区域头部(ROI HEAD)网络层；特征提取网络层与区域候选网络层连接，区域候选网络层与区域对齐网络层连接；区域对齐网络层与区域头部网络层连接；区域头部网络层包括；两个子网络分别为目标检测分支网络和目标分割分支网络；目标检测分支网络用于输出狭窄段血管的目标检测框和检测框置信度，目标分割分支网络用于输出狭窄段血管掩膜图像；
本发明实施例的特征提取网络层具体由五级残差网络(Residual Network,ResNet)和对应的五级特征金字塔网络(Feature Pyramid Networks，FPN)构成；区域候选网络层包括五级区域候选网络，与五级特征金字塔网络对应；在实现五级残差网络时，本发明实施例使用ResNet-50网络结构进行实现并以此作为特征提取的骨干网络。
步骤4，根据检测框置信度，对第一图像序列进行图像优选处理得到指定数量的优选图像；
具体包括：在第一图像序列中，对各个第一图像上所有目标检测框的检测框置信度进行均值计算，生成对应的第一图像平均置信度；按对应的第一图像平均置信度从大到小的顺序，对所有第一图像进行排序，并将排序靠前的指定数量的第一图像作为优选图像。
这里，第一图像平均置信度越高说明对应的第一图像上狭窄段血管特征越明显；指定数量根据具体要求进行设定，例如指定数量为3则当前步骤会从第一图像序列中提取狭窄段血管特征最明显的3个第一图像作为优选图像。
步骤5，在各个优选图像上对每个目标检测框内的狭窄段血管掩膜图像进 行冠脉造影定量分析生成对应的血管狭窄率；
具体包括：步骤51，在当前优选图像上对各个目标检测框进行遍历，并将当前遍历的目标检测框记为当前目标检测框；
步骤52，对当前目标检测框内的狭窄段血管掩膜图像进行血管边缘和血管中心线识别生成对应的第一血管边缘和第一中心线；
其中，第一中心线包括多个中心线像素点P i，其中第一个中心线像素点P 1为按血流方向距离冠状动脉入口最近的点，最后一个中心线像素点P N为按血流方向距离冠状动脉入口最远的点，1≤i≤N，N为第一中心线的中心线像素点总数；
这里，对狭窄段血管掩膜图像进行血管边缘识别的实现方法有多种；其中一种是，对当前目标检测框的图像内容进行二值化处理得到第一二值图,第一二值图上原有的狭窄段血管掩膜图像的所有像素点的像素值会被转换为预设的前景像素值A，狭窄段血管掩膜图像之外的所有像素点的像素值会被转换为预设的背景像素值B；然后，对第一二值图上的像素值为前景像素值A的各个像素点进行逐点遍历，遍历时若当前遍历像素点的邻八域像素点中有一个的像素值为背景像素值B则将当前遍历像素点视为边缘点；完成遍历后，按顺时针或逆时针方式对各个边缘点进行依次连接得到的封闭曲线就是血管边缘；此处，所谓邻八域像素点实际就是像素点的左上相邻像素点、上方相邻像素点、右上相邻像素点、右方相邻像素点、右下相邻像素点、下方相邻像素点、左下相邻像素点和左方相邻像素点这八个像素点；
这里，对狭窄段血管掩膜图像进行血管中心线识别的实现方法有多种；其中一种是，对当前目标检测框的图像内容进行二值化处理得到第二二值图,第二二值图上原有的狭窄段血管掩膜图像的所有像素点的像素值会被转换为预设的前景像素值A，狭窄段血管掩膜图像之外的所有像素点的像素值会被转换为预设的背景像素值B；在不改变血管图像拓扑性质的前提下，基于拓扑细化方法对第一二值图的狭窄段血管掩膜图像进行中心线提取处理生成第一中 心线；此处，血管图像拓扑性质主要指的是血管的连通性；
在得到第一中心线时为给该中心线标识出方向，特意将距离冠状动脉入口最近的点也就是狭窄段入口点作为第一中心线的第一个中心线像素点P 1，将距离冠状动脉入口最远的点也就是狭窄段出口点作为第一中心线的最后一个中心线像素点P N；
步骤53，根据第一血管边缘，对第一中心线上各个中心线像素点P i对应的血管直径长度进行分析生成对应的第一血管直径d i；
具体包括：步骤531，按中心线像素点P i与其邻八域像素点的方向关系，过中心线像素点P i做四条直线分别记为第一、第二、第三和第四直线；
其中，第一直线过中心线像素点P i的左上相邻像素点、中心线像素点P i和中心线像素点P i的右下相邻像素点；第二直线过中心线像素点P i的上方相邻像素点、中心线像素点P i和中心线像素点P i的下方相邻像素点；第三直线过中心线像素点P i的右上相邻像素点、中心线像素点P i和中心线像素点P i的左下相邻像素点；第三直线过中心线像素点P i的右方相邻像素点、中心线像素点P i和中心线像素点P i的左方相邻像素点；
步骤532，将第一、第二、第三和第四直线与第一血管边缘相交的线段分别记为对应的第一、第二、第三和第四线段；并对第一、第二、第三和第四线段的线段长度进行计算生成对应的第一、第二、第三和第四线段长度；并从第一、第二、第三和第四线段长度中，选择最小值作为与中心线像素点P i对应的第一血管直径d i；
这里，首先已知过中心线像素点P i到第一血管边缘的血管直径应在所有过中心线像素点P i的直线范围内；而在图像上过任一像素点作直线实际只有4条可做，也就上述第一、第二、第三和第四直线，也就是说过中心线像素点P i到第一血管边缘的血管直径只能是第一、第二、第三和第四线段中的一条；在确定了血管直径的选择范围之后，将其中最短线段的长度作为中心线像素点P i对应的第一血管直径d i；
步骤54，根据中心线像素点P 1到中心线像素点P N的血管线性变化关系，以及各个中心线像素点P i的第一血管直径d i，对各个中心线像素点P i对应的狭窄率进行分析生成对应的第一狭窄率r i；
具体包括：步骤541，根据第一血管直径d 1和第一血管直径d N，构建反映中心线像素点P 1到中心线像素点P N血管线性变化关系的线性函数f(i)，f(i)＝d 1+k*(i-1)，k＝(d N-d 1)/(N-1)；
步骤542，根据线性变化关系函数f(i)，对各个中心线像素点P i对应的线性变化直径长度进行计算生成对应的第一参考直径d’ i；
这里，在没有发生血管狭窄突变的情况下，血管管径应与其位置距离冠状动脉入口的远近存在一定的线性关系，对于一段分支血管而言其入口与出口位置的管径也存在一定的线性关系；通过确认一段血管入口、出口的线性关系就可以得到这段血管上任一点的正常管径也就是第一参考直径d’ i；
步骤543，根据第一血管直径d i和第一参考直径d’ i，计算各个中心线像素点P i对应的第一狭窄率r i，r i＝1-d i/d’ i；
这里，若某段血管中某个位置发生了狭窄突变，那么在得到该位置的狭窄管径也就是第一血管直径d i之后，再基于其对应的第一参考直径d’ i就可以算出该位置的血管狭窄率也就是第一狭窄率r i；
步骤55，从得到的所有第一狭窄率r i中，选择最大值作为与当前目标检测框内的狭窄段血管掩膜图像对应的血管狭窄率；
这里，本发明实施例将一段血管中的最大狭窄率作为这段血管也就是当前目标检测框内的狭窄段血管掩膜图像的血管狭窄率；
步骤56，将下一个尚未处理的目标检测框作为当前目标检测框，并转至步骤52继续进行处理直到当前优选图像上所有目标检测框内狭窄段血管掩膜图像的血管狭窄率都确认了为止。
通过上述步骤1-5，可以将从一段血管造影视频中提取出多张优选图像，并可通过冠脉造影定量分析获得每张优选图像上的一个或多个狭窄段血管的 血管狭窄率。在得到多张优选图像的分析结果之后，可将多张带有狭窄段血管目标检测框、血管狭窄率的优选图像同时提供给医生作为参数数据；也可对多张优选图像进行图像融合，并将相同位置的狭窄段血管的血管狭窄率进行均值计算，最后将带有一张带有狭窄段血管目标检测框、血管狭窄率均值的融合图像提供给医生作为参数数据。
图2为本发明实施例二提供的一种基于血管造影视频进行冠脉造影定量分析的装置的模块结构图，该装置可以为实现本发明实施例方法的终端设备或者服务器，也可以为与上述终端设备或者服务器连接的实现本发明实施例方法的装置，例如该装置可以是上述终端设备或者服务器的装置或芯片系统。如图2所示，该装置包括：获取模块201、图像预处理模块202、狭窄段血管处理模块203、图像优选模块204和定量分析模块205。
获取模块201用于获取冠状动脉造影的血管造影视频。
图像预处理模块202用于对血管造影视频进行视频帧图像提取生成对应的第一图像序列。
狭窄段血管处理模块203用于基于预设的图像目标检测与语义分割模型，对第一图像序列中各个第一图像进行狭窄段血管的目标检测和语义分割处理，从而在每个第一图像上得到一个或多个用于标记狭窄段血管的目标检测框以及在每个目标检测框中的一段狭窄段血管掩膜图像；每个目标检测框对应一个检测框置信度。
图像优选模块204用于根据检测框置信度，对第一图像序列进行图像优选处理得到指定数量的优选图像。
定量分析模块205用于在各个优选图像上对每个目标检测框内的狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率。
本发明实施例提供的一种基于血管造影视频进行冠脉造影定量分析的装置，可以执行上述方法实施例中的方法步骤，其实现原理和技术效果类似，在此不再赘述。
需要说明的是，应理解以上装置的各个模块的划分仅仅是一种逻辑功能的划分，实际实现时可以全部或部分集成到一个物理实体上，也可以物理上分开。且这些模块可以全部以软件通过处理元件调用的形式实现；也可以全部以硬件的形式实现；还可以部分模块通过处理元件调用软件的形式实现，部分模块通过硬件的形式实现。例如，获取模块可以为单独设立的处理元件，也可以集成在上述装置的某一个芯片中实现，此外，也可以以程序代码的形式存储于上述装置的存储器中，由上述装置的某一个处理元件调用并执行以上确定模块的功能。其它模块的实现与之类似。此外这些模块全部或部分可以集成在一起，也可以独立实现。这里所描述的处理元件可以是一种集成电路，具有信号的处理能力。在实现过程中，上述方法的各步骤或以上各个模块可以通过处理器元件中的硬件的集成逻辑电路或者软件形式的指令完成。
例如，以上这些模块可以是被配置成实施以上方法的一个或多个集成电路，例如：一个或多个特定集成电路(Application Specific Integrated Circuit，ASIC)，或，一个或多个数字信号处理器(Digital Signal Processor，DSP)，或，一个或者多个现场可编程门阵列(Field Programmable Gate Array，FPGA)等。再如，当以上某个模块通过处理元件调度程序代码的形式实现时，该处理元件可以是通用处理器，例如中央处理器(Central Processing Unit，CPU)或其它可以调用程序代码的处理器。再如，这些模块可以集成在一起，以片上系统(System-on-a-chip，SOC)的形式实现。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。该计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行该计算机程序指令时，全部或部分地产生按照本发明实施例所描述的流程或功能。上述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。上述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，上述计算机指令可 以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线路(Digital Subscriber Line，DSL))或无线(例如红外、无线、蓝牙、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。上述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。上述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，DVD)、或者半导体介质(例如固态硬盘(solid state disk，SSD))等。
图3为本发明实施例三提供的一种电子设备的结构示意图。该电子设备可以为前述的终端设备或者服务器，也可以为与前述终端设备或者服务器连接的实现本发明实施例方法的终端设备或服务器。如图3所示，该电子设备可以包括：处理器301(例如CPU)、存储器302、收发器303；收发器303耦合至处理器301，处理器301控制收发器303的收发动作。存储器302中可以存储各种指令，以用于完成各种处理功能以及实现本发明上述实施例中提供的方法和处理过程。优选的，本发明实施例涉及的电子设备还包括：电源304、系统总线305以及通信端口306。系统总线305用于实现元件之间的通信连接。上述通信端口306用于电子设备与其他外设之间进行连接通信。
在图3中提到的系统总线可以是外设部件互连标准(Peripheral Component Interconnect，PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture，EISA)总线等。该系统总线可以分为地址总线、数据总线、控制总线等。为便于表示，图中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。通信接口用于实现数据库访问装置与其他设备(例如客户端、读写库和只读库)之间的通信。存储器可能包含随机存取存储器(Random Access Memory，RAM)，也可能还包括非易失性存储器(Non-Volatile Memory)，例如至少一个磁盘存储器。
上述的处理器可以是通用处理器，包括中央处理器CPU、网络处理器(Network Processor，NP)等；还可以是数字信号处理器DSP、专用集成电路 ASIC、现场可编程门阵列FPGA或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
需要说明的是，本发明实施例还提供一种计算机可读存储介质，该存储介质中存储有指令，当其在计算机上运行时，使得计算机执行上述实施例中提供的方法和处理过程。
本发明实施例还提供一种运行指令的芯片，该芯片用于执行上述实施例中提供的方法和处理过程。
本发明实施例提供了一种基于血管造影视频进行冠脉造影定量分析的方法、装置、电子设备及计算机可读存储介质，对冠脉血管的血管造影视频进行视频截取和视频帧图像提取处理，基于图像目标检测与语义分割模型对提取出的图像序列进行狭窄段血管的目标检测和语义分割处理，基于目标识别的置信度对提取图像进行优选，在优选图像上对每段狭窄段血管进行冠脉造影定量分析生成对应的血管狭窄率。通过本发明摆脱了传统做法中对人工经验的过度依赖，提高了图像提取准确度和狭窄率计算精度。
专业人员应该还可以进一步意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本发明的范围。
结合本文中所公开的实施例描述的方法或算法的步骤可以用硬件、处理器执行的软件模块，或者二者的结合来实施。软件模块可以置于随机存储器(RAM)、内存、只读存储器(ROM)、电可编程ROM、电可擦除可编程ROM、寄存器、硬盘、可移动磁盘、CD-ROM、或技术领域内所公知的任意其它形式的存储介质中。
以上所述的具体实施方式，对本发明的目的、技术方案和有益效果进行了进一步详细说明，所应理解的是，以上所述仅为本发明的具体实施方式而已，并不用于限定本发明的保护范围，凡在本发明的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (10)

  1. 一种基于血管造影视频进行冠脉造影定量分析的方法，其特征在于，所述方法包括：
    获取冠状动脉造影的血管造影视频；
    对所述血管造影视频进行视频帧图像提取生成对应的第一图像序列；
    基于预设的图像目标检测与语义分割模型，对所述第一图像序列中各个第一图像进行狭窄段血管的目标检测和语义分割处理，从而在每个所述第一图像上得到一个或多个用于标记狭窄段血管的目标检测框以及在每个所述目标检测框中的一段狭窄段血管掩膜图像；每个所述目标检测框对应一个检测框置信度；
    根据所述检测框置信度，对所述第一图像序列进行图像优选处理得到指定数量的优选图像；
    在各个所述优选图像上对每个所述目标检测框内的所述狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率。
  2. 根据权利要求1所述的基于血管造影视频进行冠脉造影定量分析的方法，其特征在于，所述对所述血管造影视频进行视频帧图像提取生成对应的第一图像序列，具体包括：
    对所述血管造影视频的进行视频截取，保留造影剂充盈冠状动脉阶段的视频内容生成对应的截取造影视频；
    按时间先后顺序，对所述截取造影视频进行视频帧图像提取处理生成对应的视频帧图像序列，并对所述视频帧图像序列的视频帧图像数量进行统计生成对应的第一总数；所述视频帧图像序列包括多个视频帧图像；
    当所述第一总数未超过预设的图像总数阈值时，将各个所述视频帧图像作为对应的所述第一图像，并按时间先后顺序对得到的所有所述第一图像进行排序生成所述第一图像序列；
    当所述第一总数超过所述图像总数阈值时，从所述视频帧图像序列中提 取排序索引全为奇数索引的所述视频帧图像作为对应的所述第一图像，或者提取排序索引全为偶数索引的所述视频帧图像作为对应的所述第一图像；并按时间先后顺序对得到的所有所述第一图像进行排序生成所述第一图像序列。
  3. 根据权利要求1所述的基于血管造影视频进行冠脉造影定量分析的方法，其特征在于，
    所述图像目标检测与语义分割模型包括Mask R-CNN模型；所述图像目标检测与语义分割模型具体为Mask R-CNN模型时，采用残差网络ResNet50作为其特征提取骨干网络。
  4. 根据权利要求1所述的基于血管造影视频进行冠脉造影定量分析的方法，其特征在于，所述根据所述检测框置信度，对所述第一图像序列进行图像优选处理得到指定数量的优选图像，具体包括：
    在所述第一图像序列中，对各个所述第一图像上所有所述目标检测框的所述检测框置信度进行均值计算，生成对应的第一图像平均置信度；
    按对应的所述第一图像平均置信度从大到小的顺序，对所有所述第一图像进行排序，并将排序靠前的指定数量的所述第一图像作为所述优选图像。
  5. 根据权利要求1所述的基于血管造影视频进行冠脉造影定量分析的方法，其特征在于，所述在各个所述优选图像上对每个所述目标检测框内的所述狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率，具体包括：
    在当前优选图像上对各个所述目标检测框进行遍历，并将当前遍历的所述目标检测框记为当前目标检测框；
    对所述当前目标检测框内的所述狭窄段血管掩膜图像进行血管边缘和血管中心线识别生成对应的第一血管边缘和第一中心线；所述第一中心线包括多个中心线像素点P i，其中第一个中心线像素点P 1为按血流方向距离冠状动脉入口最近的点，最后一个中心线像素点P N为按血流方向距离冠状动脉入口最远的点，1≤i≤N，N为所述第一中心线的中心线像素点总数；
    根据所述第一血管边缘，对所述第一中心线上各个所述中心线像素点P i对应的血管直径长度进行分析生成对应的第一血管直径d i；
    根据所述中心线像素点P 1到所述中心线像素点P N的血管线性变化关系，以及各个所述中心线像素点P i的所述第一血管直径d i，对各个所述中心线像素点P i对应的狭窄率进行分析生成对应的第一狭窄率r i；
    从得到的所有所述第一狭窄率r i中，选择最大值作为与所述当前目标检测框内的所述狭窄段血管掩膜图像对应的所述血管狭窄率。
  6. 根据权利要求5所述的基于血管造影视频进行冠脉造影定量分析的方法，其特征在于，所述根据所述第一血管边缘，对所述第一中心线上各个所述中心线像素点P i对应的血管直径长度进行分析生成对应的第一血管直径d i，具体包括：
    按所述中心线像素点P i与其邻八域像素点的方向关系，过所述中心线像素点P i做四条直线分别记为第一、第二、第三和第四直线；所述第一直线过所述中心线像素点P i的左上相邻像素点、所述中心线像素点P i和所述中心线像素点P i的右下相邻像素点；所述第二直线过所述中心线像素点P i的上方相邻像素点、所述中心线像素点P i和所述中心线像素点P i的下方相邻像素点；所述第三直线过所述中心线像素点P i的右上相邻像素点、所述中心线像素点P i和所述中心线像素点P i的左下相邻像素点；所述第三直线过所述中心线像素点P i的右方相邻像素点、所述中心线像素点P i和所述中心线像素点P i的左方相邻像素点；
    将所述第一、第二、第三和第四直线与所述第一血管边缘相交的线段分别记为对应的第一、第二、第三和第四线段；并对所述第一、第二、第三和第四线段的线段长度进行计算生成对应的第一、第二、第三和第四线段长度；并从所述第一、第二、第三和第四线段长度中，选择最小值作为与所述中心线像素点P i对应的所述第一血管直径d i。
  7. 根据权利要求5所述的基于血管造影视频进行冠脉造影定量分析的方 法，其特征在于，所述根据所述中心线像素点P 1到所述中心线像素点P N的血管线性变化关系，以及各个所述中心线像素点P i的所述第一血管直径d i，对各个所述中心线像素点P i对应的狭窄率进行分析生成对应的第一狭窄率r i，具体包括：
    根据所述第一血管直径d 1和所述第一血管直径d N，构建反映所述中心线像素点P 1到所述中心线像素点P N血管线性变化关系的线性函数f(i)，f(i)＝d 1+k*(i-1)，k＝(d N-d 1)/(N-1)；
    根据所述线性变化关系函数f(i)，对各个所述中心线像素点P i对应的线性变化直径长度进行计算生成对应的第一参考直径d’ i；
    根据所述第一血管直径d i和所述第一参考直径d’ i，计算各个所述中心线像素点P i对应的所述第一狭窄率r i，r i＝1-d i/d’ i。
  8. 一种用于实现权利要求1-7任一项所述的基于血管造影视频进行冠脉造影定量分析的方法步骤的装置，其特征在于，所述装置包括：获取模块、图像预处理模块、狭窄段血管处理模块、图像优选模块和定量分析模块；
    所述获取模块用于获取冠状动脉造影的血管造影视频；
    所述图像预处理模块用于对所述血管造影视频进行视频帧图像提取生成对应的第一图像序列；
    所述狭窄段血管处理模块用于基于预设的图像目标检测与语义分割模型，对所述第一图像序列中各个第一图像进行狭窄段血管的目标检测和语义分割处理，从而在每个所述第一图像上得到一个或多个用于标记狭窄段血管的目标检测框以及在每个所述目标检测框中的一段狭窄段血管掩膜图像；每个所述目标检测框对应一个检测框置信度；
    所述图像优选模块用于根据所述检测框置信度，对所述第一图像序列进行图像优选处理得到指定数量的优选图像；
    所述定量分析模块用于在各个所述优选图像上对每个所述目标检测框内的所述狭窄段血管掩膜图像进行冠脉造影定量分析生成对应的血管狭窄率。
  9. 一种电子设备，其特征在于，包括：存储器、处理器和收发器；
    所述处理器用于与所述存储器耦合，读取并执行所述存储器中的指令，以实现权利要求1-7任一项所述的方法步骤；
    所述收发器与所述处理器耦合，由所述处理器控制所述收发器进行消息收发。
  10. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机指令，当所述计算机指令被计算机执行时，使得所述计算机执行权利要求1-7任一项所述的方法的指令。
PCT/CN2022/097245 2022-01-07 2022-06-07 基于血管造影视频进行冠脉造影定量分析的方法和装置 WO2023130663A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210018415.6A CN114418977A (zh) 2022-01-07 2022-01-07 基于血管造影视频进行冠脉造影定量分析的方法和装置
CN202210018415.6 2022-01-07

Publications (1)

Publication Number Publication Date
WO2023130663A1 true WO2023130663A1 (zh) 2023-07-13

Family

ID=81271742

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/097245 WO2023130663A1 (zh) 2022-01-07 2022-06-07 基于血管造影视频进行冠脉造影定量分析的方法和装置

Country Status (2)

Country Link
CN (1) CN114418977A (zh)
WO (1) WO2023130663A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116681716A (zh) * 2023-08-04 2023-09-01 杭州脉流科技有限公司 颅内血管感兴趣区域分割方法、装置、设备和存储介质
CN116703948A (zh) * 2023-08-03 2023-09-05 杭州脉流科技有限公司 基于深度神经网络的颅内血管树分割方法和分割装置

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114418977A (zh) * 2022-01-07 2022-04-29 乐普（北京）医疗器械股份有限公司 基于血管造影视频进行冠脉造影定量分析的方法和装置
CN114708259B (zh) * 2022-05-24 2022-08-16 深圳市铱硙医疗科技有限公司 基于cta的头颈血管狭窄检测方法、装置、设备及介质
CN116612102A (zh) * 2023-05-31 2023-08-18 上海博动医疗科技股份有限公司 血管图像处理系统、装置及存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109222980A (zh) * 2018-06-19 2019-01-18 北京红云智胜科技有限公司 基于深度学习的测量冠状动脉造影图像血管直径的方法
US20190159683A1 (en) * 2017-11-28 2019-05-30 Beijing Curacloud Technology Co., Ltd. Methods and systems for determining lumen volume and coronary blood flow
CN111657883A (zh) * 2020-06-03 2020-09-15 北京理工大学 基于序列造影的冠脉syntax分数自动计算方法与系统
CN114418977A (zh) * 2022-01-07 2022-04-29 乐普（北京）医疗器械股份有限公司 基于血管造影视频进行冠脉造影定量分析的方法和装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190159683A1 (en) * 2017-11-28 2019-05-30 Beijing Curacloud Technology Co., Ltd. Methods and systems for determining lumen volume and coronary blood flow
CN109222980A (zh) * 2018-06-19 2019-01-18 北京红云智胜科技有限公司 基于深度学习的测量冠状动脉造影图像血管直径的方法
CN111657883A (zh) * 2020-06-03 2020-09-15 北京理工大学 基于序列造影的冠脉syntax分数自动计算方法与系统
CN114418977A (zh) * 2022-01-07 2022-04-29 乐普（北京）医疗器械股份有限公司 基于血管造影视频进行冠脉造影定量分析的方法和装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
KAI SHAO, ZHANG YUNFENG; BAO FANGXUN, ZHENG YONG, QIN CHAO: "A Coronary Artery Segmentation Method Based on Mask RCNN Integrated with Geometric Features", JOURNAL OF TAIYUAN UNIVERSITY OF TECHNOLOGY, vol. 52, no. 1, 22 December 2020 (2020-12-22), pages 83 - 90, XP093077190 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116703948A (zh) * 2023-08-03 2023-09-05 杭州脉流科技有限公司 基于深度神经网络的颅内血管树分割方法和分割装置
CN116703948B (zh) * 2023-08-03 2023-11-14 杭州脉流科技有限公司 基于深度神经网络的颅内血管树分割方法和分割装置
CN116681716A (zh) * 2023-08-04 2023-09-01 杭州脉流科技有限公司 颅内血管感兴趣区域分割方法、装置、设备和存储介质
CN116681716B (zh) * 2023-08-04 2023-10-10 杭州脉流科技有限公司 颅内血管感兴趣区域分割方法、装置、设备和存储介质

Also Published As

Publication number Publication date
CN114418977A (zh) 2022-04-29

Similar Documents

Publication Publication Date Title
WO2023130663A1 (zh) 基于血管造影视频进行冠脉造影定量分析的方法和装置
Hsu et al. Ratio-and-scale-aware YOLO for pedestrian detection
WO2021169128A1 (zh) 眼底视网膜血管识别及量化方法、装置、设备及存储介质
Li et al. A composite model of wound segmentation based on traditional methods and deep neural networks
Gunasekara et al. A systematic approach for MRI brain tumor localization and segmentation using deep learning and active contouring
WO2021196873A1 (zh) 车牌字符识别方法、装置、电子设备和存储介质
WO2023137948A1 (zh) 基于血管造影图像分析血流储备分数的处理方法和装置
CN110751636B (zh) 一种基于改进型编解码网络的眼底图像视网膜动脉硬化检测方法
WO2022087853A1 (zh) 一种图像分割方法、装置及计算机可读存储介质
US20220036561A1 (en) Method for image segmentation, method for training image segmentation model
CN110348345B (zh) 一种基于动作连贯性的弱监督时序动作定位方法
Abdulla Efficient computer‐aided diagnosis technique for leukaemia cancer detection
WO2023130664A1 (zh) 一种血管造影图像的分析方法和装置
WO2023130661A1 (zh) 一种二维频谱多普勒超声心动图像的处理方法和装置
Chen et al. Segmentation of overlapping cervical cells with mask region convolutional neural network
García-Aguilar et al. Optimized instance segmentation by super-resolution and maximal clique generation
Yang et al. Retinal vessel segmentation based on an improved deep forest
CN117351307B (zh) 一种模型训练方法、装置、设备及存储介质
CN112288718B (zh) 图像处理方法及装置，电子设备及计算机可读存储介质
CN110717916B (zh) 一种基于卷积神经网络的肺栓塞检测系统
CN110796129A (zh) 一种文本行区域检测方法及装置
Adu et al. MLAF-CapsNet: Multi-lane atrous feature fusion capsule network with contrast limited adaptive histogram equalization for brain tumor classification from MRI images
CN115690143B (zh) 图像分割方法、装置、电子设备及存储介质
CN111292298A (zh) 乳腺癌病理分型确定方法、装置和存储介质
Sharma et al. Detection of white blood cells using optimized qgwo

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22918119

Country of ref document: EP

Kind code of ref document: A1