Coronary artery automatic frame selection, classification and recommendation method and device based on DSA image

Abstract

The application discloses a coronary artery automatic frame selection, classification, recommendation method and device based on DSA images, and the method comprises the following steps: acquiring a DSA image group to be tested and adopting a pre-constructed automatic frame selection model to identify each DI COM image in the DSA image group to be tested, so as to obtain key frame images; based on a pre-constructed automatic classification model, the key frame images are classified by blood vessel types, so as to obtain probability response values corresponding to the blood vessel categories of the key frame images; the probability response values are arranged in descending order, and every two key frame images from the rear to the front form an image pair; based on the mean value of the probability response values corresponding to the image pair and the imaging angle difference, the image frame image pair is scored, and the image pair with the highest score is recommended. The application can efficiently and accurately identify the image pair suitable for reconstruction in the coronary artery DSA image.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method and apparatus for automated coronary artery frame selection, classification, and recommendation based on DSA images. Background Technology

[0002] Coronary heart disease, also known as coronary atherosclerotic heart disease or ischemic heart disease, refers to heart disease caused by myocardial ischemia and hypoxia due to atherosclerosis of the coronary arteries. The coronary arteries are the only blood vessels supplying the heart; their shape resembles a coronary, hence the name. Like other blood vessels throughout the body, these vessels harden and undergo atherosclerotic changes, causing impaired blood circulation to the heart, leading to myocardial ischemia and hypoxia, which is coronary heart disease. Coronary heart disease is a common and frequently occurring disease among middle-aged and elderly people, and it can seriously endanger life.

[0003] Routine methods for diagnosing coronary artery disease include simple, non-invasive electrocardiography (ECG), coronary coronary angiography (CTA) which only provides static images, and invasive methods such as intravascular ultrasound (IVUS) and dynamic coronary angiography. Among these methods, coronary angiography is considered the "gold standard" for diagnosing coronary artery disease. As the primary imaging technique for diagnosing coronary artery disease, the morphology of the coronary arteries is obtained through real-time visualization in the catheterization lab during coronary angiography. Quantitative coronary angiography (QCA) can also provide objective quantitative measurements.

[0004] Since the reconstruction of images based on two-angle DSA images requires doctors to actively select the two best images from a series of images, and patients may have a variety of image data, it takes a lot of effort to select reasonable data from a large amount of data, and doctors also need to make judgments based on experience on whether the two images are suitable for reconstruction. This greatly increases the burden on doctors. At present, there is no function module on the market that can select the two most suitable images for reconstruction from the patient's original images and recommend them to doctors.

[0005] Therefore, a new frame selection classification recommendation method is needed in this field to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for automated frame selection, classification and recommendation of coronary arteries based on DSA images, aiming to solve the problem of how to efficiently and accurately identify suitable image pairs for reconstruction in coronary artery DSA images.

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

[0008] An automated coronary artery frame selection, classification, and recommendation method based on DSA images, the method comprising:

[0009] The DSA image group to be tested is acquired and each DICOM image in the DSA image group to be tested is identified using a pre-built automatic frame selection model to obtain key frame images;

[0010] The keyframe image is classified into blood vessel types based on a pre-built automatic classification model to obtain the probability response value of the keyframe image corresponding to the blood vessel category;

[0011] Arrange the keyframe images in descending order of the probability response values, and form an image pair from back to front;

[0012] The image frame pairs are scored based on the mean of the probability response values ​​corresponding to the image pairs and the imaging angle difference, and the image pairs with the highest scores are recommended.

[0013] In one embodiment, the automatic frame selection model is a binary classification neural network model, and the automatic classification model is a four-class classification neural network model. The blood vessel categories include LAD, LCX, RCA and OTHERS, where OTHERS are blood vessel categories other than LAD, LCX and RCA.

[0014] In one embodiment, the network training steps of the "automatic frame selection model" include:

[0015] Acquire DICOM images generated from coronary angiography;

[0016] The DICOM image frames in which the contrast agent is fully filled, clearly visible, and in the end of diastole are set as keyframe images, and a data recording table is formed.

[0017] Iterate through the data record table, select keyframe images and set their labels to 1, and select non-keyframe images and set their labels to 0.

[0018] An automatic frame selection dataset is constructed based on the keyframe images and non-keyframe images, and the automatic frame selection model is trained using machine learning algorithms.

[0019] In one embodiment, the non-keyframe image includes one frame from the first 10 frames of the randomly selected DI COM image, one frame from the previous 5 frames of the keyframe image, and one frame from the next 5 frames of the keyframe image.

[0020] In one embodiment, the network training steps of the "automatic classification model" include:

[0021] Acquire DICOM images generated from coronary angiography;

[0022] The DICOM image frames in which the contrast agent is fully filled, clearly visible, and in the end of diastole are read and set as keyframe images;

[0023] The keyframe images are divided into LAD, LCX, RCA and OTHERS according to blood vessel categories and assigned image labels of [1,0,0,0], [0,1,0,0], [0,0,1,0], [0,0,0,1] respectively to generate a blood vessel classification dataset;

[0024] The automatic classification model is trained based on the blood vessel classification dataset and using machine learning algorithms.

[0025] In one embodiment, the method further includes adjusting the contrast of the DICOM image according to the window width and window level while simultaneously adjusting the DICOM image to a fixed size.

[0026] In one embodiment, the step of "scoring the image frame pair based on the mean of the probability response values ​​corresponding to the image pair and the imaging angle difference of the corresponding image pair" includes:

[0027] The score is calculated using the recommended method shown in the following formula:

[0028] z = a × x + (1 - a) × y

[0029] Where z is the score, a is the preset scaling factor, x is the mean of the probability response values ​​of the image pair, and y is the angle difference of the image pair.

[0030] In one embodiment, the step of determining 'a' includes:

[0031] Obtain multiple image pairs for reconstruction;

[0032] The image pairs are sorted according to their reconstruction quality, and the scaling factor is adjusted based on the sorting to ensure that the rating and sorting results calculated by the recommendation algorithm are consistent with the actual sorting.

[0033] A storage device storing a plurality of programs adapted for loading and execution by a processor to implement the above-described method for automated coronary artery frame selection, classification, and recommendation based on DSA images.

[0034] A control device includes: a processor adapted to execute various programs; and a storage device adapted to store multiple programs; the programs being adapted to be loaded and executed by the processor to implement the above-described automated coronary artery frame selection and classification recommendation method based on DSA images.

[0035] In summary, the automated coronary artery frame selection, classification, and recommendation method and apparatus based on DSA images provided by this invention have the following beneficial effects:

[0036] This invention can automatically, quickly and accurately identify suitable image pairs for reconstruction in coronary DSA images, reducing the workload of doctors in selecting images and having high clinical application value. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating an automated coronary artery frame selection, classification, and recommendation method based on DSA images according to the present invention.

[0038] Figure 2 This is a schematic diagram of the network training process for the automatic frame selection mode of the present invention;

[0039] Figure 3 This is a schematic diagram of the network training process for the automatic classification model of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0041] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in this disclosure, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such use of data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein.

[0042] It should be understood that in the various embodiments of this disclosure, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this disclosure.

[0043] It should be understood that in this disclosure, "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or device.

[0044] It should be understood that in this disclosure, "multiple" refers to two or more. "And / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, "and / or B" can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "Contains A, B, and C", "Contains A, B, and C" means that all three A, B, and C are contained; "Contains A, B, or C" means that one of A, B, and C is contained; "Contains A, B, and / or C" means that any one, two, or three of A, B, and C are contained.

[0045] It should be understood that in this disclosure, "B corresponding to A", "B corresponding to A", "A corresponds to B", or "B corresponds to A" means that B is associated with A, and B can be determined based on A. Determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information. Matching A and B is defined as a similarity between A and B that is greater than or equal to a preset threshold.

[0046] Depending on the context, "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection."

[0047] The technical solutions of this disclosure will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0048] This invention provides a method and apparatus for automated frame selection, classification, and recommendation of coronary artery images based on DSA (Digital Subtraction Angiography) imaging. This method efficiently and accurately extracts keyframe images, identifies vessel types, and recommends the best image pairs suitable for reconstruction from coronary artery DSA images. Its primary purpose is to address the tedious task of physicians manually sifting through a large number of images to select specific vessel images suitable for reconstruction during coronary artery reconstruction.

[0049] See appendix Figure 1 , Figure 1 This example illustrates the main workflow of an automated coronary artery frame selection, classification, and recommendation method based on DSA imagery. For instance... Figure 1 As shown in this embodiment, an automated coronary artery frame selection and classification recommendation method based on DSA images includes:

[0050] Step S1: Obtain the DSA image group to be tested and use a pre-built automatic frame selection model to identify each DICOM image in the DSA image group to be tested in order to obtain keyframe images.

[0051] Specifically, a standard catheter can be used for catheterization via the femoral or radial catheter, and DICOM images are digitally recorded using coronary angiography. A trained automatic frame selection model is used to identify each DICOM image to obtain a keyframe image. The automatic frame selection model is a binary classification neural network model. The keyframe image is a relatively high-quality image where the contrast agent is fully filled, clearly visible, and in the end of diastole. The output of the automatic frame selection model is the probability value of each DICOM image frame belonging to the keyframe image. The probability values ​​of the frames before and after the keyframe image location are also relatively high, so the DICOM image frame with the highest probability value at the keyframe image location can be selected as the keyframe image at that location.

[0052] Step S2: Classify the blood vessel type of the keyframe image based on the pre-built automatic classification model to obtain the probability response value of the blood vessel category corresponding to the keyframe image.

[0053] Specifically, all keyframe images are input into a trained automatic classification model to classify the keyframe images. The automatic classification model is a four-class neural network model. Vessel categories include LAD (left anterior descending artery), LCX (left circumflex artery), RCA (right coronary artery), and OTHERS (vessel categories other than LAD, LCX, and RCA). When classifying keyframe images, the automatic classification model outputs the probability response value of the keyframe image belonging to each vessel category. The vessel category corresponding to the highest probability response value is the vessel category of that keyframe image, and the highest probability response value is the probability response value of that keyframe image.

[0054] Step S3: Sort the images in descending order of probability response values ​​and form an image pair from back to front for every two keyframe images.

[0055] Step S4: Score the image frame pairs based on the mean of the probability response values ​​corresponding to the image pairs and the imaging angle difference of the corresponding image pairs, and recommend the image pairs with the highest scores.

[0056] Specifically, the mean of the probability response values ​​corresponding to the image pair is the sum of the probability response values ​​of the two keyframe images divided by 2. The imaging angle difference is the difference between the imaging angles of the image pair corresponding to the two keyframe images.

[0057] The score can be calculated using the recommended method shown in formula (1):

[0058] z = a × x + (1 - a) × y (1)

[0059] Where z is the score value, a is the preset scaling factor, x is the mean of the probability response values ​​corresponding to the image pair, and y is the angle difference between the image pairs corresponding to the image pairs.

[0060] The steps for determining the scaling factor 'a' include: acquiring multiple image pairs for reconstruction; sorting the image pairs according to their reconstruction quality; and adjusting the scaling factor based on the sorting as a reference, so that the score ranking result calculated by the recommendation method is consistent with the actual ranking.

[0061] After obtaining the rating results for each image pair, the image pair with the highest rating is recommended to the doctor.

[0062] See appendix Figure 2 , Figure 2 An example is shown illustrating the network training steps for automatic frame selection. For instance... Figure 2 As shown, the network training steps for the "automatic frame selection model" include:

[0063] Step S11: Obtain DICOM images generated based on coronary angiography.

[0064] Step S12: Set the DICOM image frames that are fully filled with contrast agent, clearly visible, and in the end of diastole as keyframe images, and form a data record table.

[0065] This step can be performed by a team of clinical physicians who will read the DICOM images, extract the keyframes of each image, and record them in a data log table.

[0066] Step S13: Traverse the data record table, select keyframe images and set their labels to 1, select non-keyframe images and set their labels to 0.

[0067] Specifically, all keyframe images are extracted and labeled as 1. Non-keyframe images are selected according to certain rules. The labels for non-keyframe images are set to 0. Non-keyframe images consist of three parts: one frame randomly selected from the first 10 frames of the DICOM image (this part represents the image frame where the contrast agent is not fully filled); one frame randomly selected from the previous 5 frames of the keyframe image (this part represents the diastolic image frame before the keyframe image); and one frame randomly selected from the next 5 frames of the keyframe image (this part represents the systolic image frame after the keyframe image). These three parts do not overlap and do not exceed the normal sequence number range of the image sequence. The randomness is introduced to cover as many non-keyframe images as possible and to avoid the limitations of subjective judgment. After selecting keyframe and non-keyframe images, it is also necessary to adjust the contrast of the images according to the window width and window level, and simultaneously resize the DICOM image to a fixed size.

[0068] Step S14: Construct an automatic frame selection dataset based on keyframe images and non-keyframe images, and train an automatic frame selection model using machine learning algorithms.

[0069] See appendix Figure 3 , Figure 3An example is shown illustrating the network training steps for an automatic classification model. For example... Figure 3 As shown, the network training steps for the "automatic classification model" include:

[0070] Step S21: Obtain DICOM images generated based on coronary angiography.

[0071] Step S22: Read DICOM image frames that are fully filled with contrast agent, clearly visible contrast agent, and in the end of diastole and set them as keyframe images.

[0072] Step S23: Divide the keyframe images into LAD, LCX, RCA and OTHERS according to the blood vessel category and assign image labels to each to generate a blood vessel classification dataset.

[0073] Specifically, the image labels corresponding to the blood vessel categories LAD, LCX, RCA, and OTHERS are [1,0,0,0], [0,1,0,0], [0,0,1,0], and [0,0,0,1], respectively. Similarly, when generating the blood vessel classification dataset, a step of resizing the images according to the window width and window level can also be included to adjust the images to a fixed size.

[0074] Step S24: Train an automatic classification model based on the blood vessel classification dataset and using machine learning algorithms.

[0075] This embodiment also provides a storage device storing multiple programs adapted for loading and execution by a processor to implement the above-described method for automated coronary artery frame selection and classification recommendation based on DSA images.

[0076] This embodiment also provides a control device, including: a processor adapted to execute various programs; and a storage device adapted to store multiple programs; the programs are adapted to be loaded and executed by the processor to implement the above-described automated coronary artery frame selection and classification recommendation method based on DSA images.

[0077] The following describes in detail the automated coronary artery frame selection, classification, and recommendation method based on DSA images provided by the present invention, with reference to a specific embodiment.

[0078] 1. Acquire DSA images

[0079] Catheterization was performed using a standard catheter via the femoral or radial catheter, and coronary angiography was digitally recorded to generate DICOM images. Personal patient information in DICOM format was removed using an anonymization tool.

[0080] 2. Annotate the DICOM images

[0081] To generate keyframe images and category labels for DICOM images, a team of clinical experts interpreted the angiography sequences. DICOM image frames with complete contrast agent filling, clear contrast agent visibility, and in the end-diastolic phase were used as keyframe images and recorded in a data log table. Simultaneously, images were placed into corresponding folders according to the main vessel category suitable for reconstruction. The folder structure based on vessel category was LAD, LCX, RCA, and OTHERS. At the next level, based on the patient, all DICOM images from a patient were categorized into these four folders according to the following rules: images with clear contrast agent visibility and suitable angles and vessel overlap for reconstruction were placed in the LAD, LCX, and RCA folders, respectively. Images of low quality unsuitable for reconstruction, as well as other non-coronary images, including images from other sites, reports, or images containing only guidewires, were placed in the OTHERS folder. The final usable data included a three-level directory structured by patient, vessel category, and specific image, along with a data log table recording the keyframe images.

[0082] 3. Dataset Construction

[0083] Since the training data used by the neural network consists of frame images from DICOM image sequences, a dataset needs to be constructed based on the original folders and data record tables. First, data for training the search for keyframe images needs to be built. Images corresponding to the keyframe numbers in the data record table are extracted, and non-keyframe images are selected according to certain rules. Data from the OTHERS folder is not used here. Next, data for training classification and recommendation needs to be built. Data from all four folders is used here. If the OTHERS folder contains single images or non-DICOM image data, it can be used directly.

[0084] 4. Model Construction and Training

[0085] The model building process mainly consists of three steps.

[0086] The first step is to construct an automatic frame selection dataset for training the automatic frame selection model. The specific details of constructing the automatic frame selection dataset include: traversing the file data according to a table, reading in each frame image and adjusting it according to window width and level; then extracting the corresponding image as the keyframe image according to the keyframe image sequence number recorded in the data record table. For non-keyframe images, selection is divided into three parts: the first part is the image frame before the contrast agent is fully filled, defined as one frame randomly selected from the first 10 frames of the DICOM image; the second part is the image frame in the diastolic phase before the keyframe, defined as one frame randomly selected from the previous 5 frames; and the third part is the image frame in the systolic phase after the keyframe, defined as one frame randomly selected from the next 5 frames. After organizing the keyframe and non-keyframe images, each image is labeled as 1 for keyframe images and 0 for non-keyframe images, generating the automatic frame selection dataset. A binary classification model is trained on the automatic frame selection dataset using a classic neural network structure. The input is a uniform 224×224 image, and the output is a list of binary probability values, with the sum of the two values ​​being 1.

[0087] The second step is to construct a blood vessel classification dataset for training the automatic classification model. This dataset is built upon the automatically selected frame dataset by adding the OTHERS category. Selected images are labeled as [1,0,0,0], [0,1,0,0], [0,0,1,0], and [0,0,0,1] respectively for LAD, LCX, RCA, and OTHERS, generating the blood vessel classification dataset. A classic neural network structure is then used to train a four-class classification model on this dataset. The input is a uniform 224×224 image, and the output is a list of four-class probability values, with the sum of the four values ​​being 1.

[0088] The third step is to configure the parameter of the proportional coefficient in the image recommendation method. The data required to determine this parameter needs to be constructed on a patient-by-patient basis. All image data from the patient that can be used for reconstruction needs to be selected, and then each image is sorted according to image quality to provide the best image pair. The proportional coefficient can be adjusted subsequently based on the manual sorting as a reference. That is, after obtaining the image pair, the mean of the probability response value and the imaging angle difference of the image pair are calculated, and a score is obtained using formula (1). Finally, the image pair with the highest score is selected as the best recommended image pair. During this process, the proportional coefficient is continuously adjusted to ensure that the score ranking result calculated by the recommendation method is consistent with the actual ranking, thus completing the determination of the proportional coefficient.

[0089] 5. Testing Process

[0090] Given a patient's DICOM image, a trained automatic classification model is used to predict each frame of each DICOM image and give the probability value of each frame to obtain keyframe images. For all keyframe images, the trained automatic classification model is used to classify the images and obtain the probability response value of the blood vessel category corresponding to the keyframe image. The images are sorted according to the probability response value to form image pairs. The score is calculated using formula (1), and the image pair with the highest score is extracted and recommended to the doctor.

[0091] This completes the recommendation of the best image pair.

[0092] In summary, this invention, by combining frame selection, classification, and image pair recommendation, can automatically, quickly, and accurately identify suitable image pairs for reconstruction in coronary DSA images, reducing the workload of doctors in selecting images and possessing high clinical application value.

[0093] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0094] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0095] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0096] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0097] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0098] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processing unit of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0099] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions that execute on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0100] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0101] Note that, unless otherwise explicitly stated, all features disclosed in this specification (including any appended claims, abstract, and drawings) may be replaced by alternative features for achieving the same, equivalent, or similar purpose. Therefore, unless explicitly stated otherwise, each disclosed feature is merely one example of a set of equivalent or similar features. Where used, "further," "preferably," "even further," and "more preferably" are simple starting points for describing another embodiment based on the foregoing embodiments, the combination of which with the foregoing embodiments constitutes the complete configuration of another embodiment. Any combination of several "further," "preferably," "even further," or "more preferably" settings following the same embodiment constitutes yet another embodiment.

[0102] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and any variations or modifications may be made to the implementation of the present invention without departing from the stated principles.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit them. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this disclosure.

Claims

1. A method for automated coronary artery frame selection, classification, and recommendation based on DSA images, characterized in that, The method includes: The DSA image group to be tested is acquired and each DICOM image in the DSA image group to be tested is identified using a pre-built automatic frame selection model to obtain key frame images; The keyframe image is classified into blood vessel types based on a pre-built automatic classification model to obtain the probability response value of the keyframe image corresponding to the blood vessel type. Arrange the keyframe images in descending order of the probability response values, and form an image pair from back to front; The image pairs are scored based on the mean of the probability response values ​​corresponding to the image pairs and the imaging angle difference of the corresponding image pairs, and the image pairs with the highest scores are recommended.

2. The automated coronary artery frame selection, classification, and recommendation method based on DSA images as described in claim 1, characterized in that, The automatic frame selection model is a binary classification neural network model, and the automatic classification model is a four-class classification neural network model. The blood vessel types include LAD, LCX, RCA and OTHERS, where OTHERS are blood vessel types other than LAD, LCX and RCA.

3. The automated coronary artery frame selection, classification, and recommendation method based on DSA images as described in claim 2, characterized in that, The network training steps for the "automatic frame selection model" include: Acquire DICOM images generated from coronary angiography; The DICOM image frames in the DICOM image that are fully filled with contrast agent, clearly visible contrast agent, and in the end of diastole are set as keyframe images, and a data recording table is formed. Iterate through the data record table, select keyframe images and set their labels to 1, and select non-keyframe images and set their labels to 0. An automatic frame selection dataset is constructed based on the keyframe images and non-keyframe images, and the automatic frame selection model is trained using machine learning algorithms.

4. The automated coronary artery frame selection, classification, and recommendation method based on DSA images as described in claim 3, characterized in that, The non-keyframe image comprises three parts: one frame randomly selected from the first 10 frames of the DICOM image generated based on coronary angiography, one frame randomly selected from the previous 5 frames of the keyframe image, and one frame randomly selected from the next 5 frames of the keyframe image.

5. The automated coronary artery frame selection, classification, and recommendation method based on DSA images as described in claim 2, characterized in that, The network training steps for the "automatic classification model" include: Acquire DICOM images generated from coronary angiography; The DICOM image frames in the DICOM image that are fully filled with contrast agent, clearly visible contrast agent, and in the end of diastole are selected as keyframe images; The keyframe images are divided into LAD, LCX, RCA and OTHERS according to blood vessel type and assigned image labels of [1,0,0,0], [0,1,0,0], [0,0,1,0], [0,0,0,1] respectively to generate a blood vessel classification dataset; The automatic classification model is trained based on the blood vessel classification dataset and using machine learning algorithms.

6. A computer-readable storage medium, characterized in that, It stores multiple programs, which are loaded and executed by a processor to implement the coronary artery automated frame selection and classification recommendation method based on DSA images as described in any one of claims 1-5.

7. A control device, comprising: Memory, used to store multiple programs; Processor, used to execute the program; The program is characterized in that it is loaded and executed by a processor to implement the coronary artery automated frame selection and classification recommendation method based on DSA images as described in any one of claims 1-5.

Images