Method and system for determining treatment strategy for vessel with multiple or diffuse lesions

Abstract

To provide a method and system for non-invasively determining a treatment strategy for a blood vessel.SOLUTION: The method comprises: obtaining 102 a prediction of a total pressure drop value in the blood vessel based on features generated from a plurality of images of the blood vessel, the images representing all anatomic parts of the blood vessel; and calculating 104 a contribution of one or more portions of the blood vessel to the total pressure drop value. Based on the calculated contribution, a new simulated total pressure drop value of the blood vessel is calculated 106 by neutralizing the contribution of the one or more portions to the total pressure drop value to confirm if the new simulated pressure drop improves, thereby indicating 108 portions which need to be treated to restore healthy pressure drop values.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to the automatic analysis of conduits, particularly blood vessels having multiple lesions or diffuse lesions.

Background Art

[0002] Coronary artery disease (CAD) is typically caused by lesions or stenoses, which are local constrictions in the coronary arteries that limit blood flow to the myocardium. Particularly difficult CAD can include tandem lesions, which are multiple stenoses in series along the coronary arteries, and / or diffuse lesions characterized by multiple or particularly long continuous plaques or other lesions within the coronary arteries.

[0003] The severity of a stenosis or lesion can be evaluated by measuring the pressure difference across the entire stenosis using a technique called the fractional flow reserve (FFR). FFR is defined as the post-stenosis (distal) pressure relative to the pre-stenosis pressure under hyperemia, and thus is a comparison of the maximum flow within the conduit in the presence of the stenosis to the maximum flow if the stenosis were virtually absent. Severe stenoses or lesions are treated by angioplasty and stent placement, but the diagnosis and treatment of diffuse lesions and tandem lesions are difficult.

[0004] Although pressure measurement can evaluate the physiological significance of a stenosis in detail, in the presence of multiple lesions or diffuse lesions, it can be difficult to predict the effect of stent placement on a particular stenosis, which is hampered by the flow interaction between stenoses in the hyperemic state. In the situation of a conduit having two tandem lesions, the hyperemic flow reaching the proximal lesion is affected not only by the shape of the stenosis but also by the second distal lesion. Similarly, the hyperemic flow reaching the second lesion corresponds to its particular shape and the inlet flow, which is affected by the proximal stenosis. As a result, in the case of diffuse disease, an accurate FFR cannot be determined for each individual stenosis.

[0005] Techniques referred to as the Instantaneous Wave Free Ratio or the Instant Flow Reserve isolate a specific period during diastole (when the heart is at rest), called the wave - free period, during which the coronary arteries are least affected by pulsatile blood flow and competing forces (waves) that affect coronary artery flow are at rest, meaning that pressure and flow are linearly related compared to the rest of the cardiac cycle. After obtaining a baseline iFR value at the target location, iFR measurements can be obtained at multiple points along the catheter and plotted across the entire catheter to create a pull - back curve. The pull - back curve can be useful in identifying pressure gradients along the artery, can assist in identifying focal and diffuse coronary artery disease, and can be used to evaluate the physiological significance of stenosis even in the case of tandem or diffuse disease.

[0006] iFR pull - back is typically performed using an invasive coronary pressure wire placed in the coronary artery and slowly withdrawn (pulled back) while the heart is at rest. This invasive procedure typically requires time and high levels of expertise.

[0007] Recently developed non - invasive assessment methods for coronary artery disease typically rely on flow dynamics analysis from the patient's coronary artery image data. One of the other non - invasive methods proposed involves using a trained model to calculate the target hemodynamic quantity at multiple points along the coronary artery tree. Calculation of the target hemodynamic quantity at points along healthy segments of the coronary artery tree is performed using a first trained regression model, and calculation of the target hemodynamic quantity at multiple points within each of one or more lesions is performed using a second trained regression model. Next, a pull - back curve is generated based on the target hemodynamic quantities calculated at multiple points along the artery. Summary of the Invention

[0008] Embodiments of the present invention provide a method and system for non-invasively determining a treatment strategy for a blood vessel having multiple or diffuse lesions without calculating pressure measurements at multiple points along a patient's catheter and without relying on pullback curves that require time and expertise for analysis.

[0009] A method according to an embodiment of the present invention includes obtaining a prediction of the total pressure drop value of a blood vessel based on features generated from a plurality of images of the blood vessel. The features may be related to the image data and / or may be related to a particular portion of the image, for example, a portion of the blood vessel in which the feature is shown within the image. The contribution of one or more portions of the blood vessel to the total pressure drop value is calculated, and based on the calculated contribution, a newly simulated total pressure drop value of the blood vessel is calculated by neutralizing the contribution of the one or more portions to the total pressure drop. The newly simulated total pressure drop value and typically a display of the one or more portions whose contribution to the total pressure drop has been neutralized may be presented to the user.

[0010] Typically, if a portion of the catheter has a high contribution to the total pressure drop within that catheter, it may be indicated that that portion contains severe lesion(s). Neutralizing the contribution of this portion typically results in a simulated pressure drop that is decreased compared to the initially predicted total pressure drop, indicating that it may be necessary to treat this portion (e.g., stent placement, etc.). By automatically calculating the contribution of a portion of the blood vessel to the total pressure drop, a display of the portion of the catheter that needs to be treated is provided quickly and simply to reduce the predicted total pressure drop for that catheter. Thus, the user may be able to determine an appropriate treatment strategy without performing time-consuming and specialized analysis of multiple pressure measurements and pullback curves.

[0011] The present invention will be described in connection with specific examples and embodiments with reference to the following exemplary figures so that it may be more fully understood. BRIEF DESCRIPTION OF THE DRAWINGS

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7A

Figure 7B

Figure 8

Figure 9

DETAILED DESCRIPTION OF THE INVENTION

[0013] Embodiments of the present invention provide a method and system for automatically analyzing a conduit from an image of the conduit and obtaining a prediction of the functional measurement of the conduit and / or a portion of the conduit. In one embodiment, the total pressure drop value within a blood vessel (i.e., the pressure drop across the entire length of the conduit or the pressure drop from a proximal point to a distal point) is predicted. A prediction of the contribution (e.g., numerical contribution) of a portion of the analyzed conduit to the total pressure drop value is generated, and information regarding the contribution is presented to a user (e.g., a physician or technician), thereby enabling the determination of a treatment strategy for the conduit. As described above, if a portion of the conduit has a high contribution to the total pressure drop within the conduit and neutralizing the contribution of this portion results in a decrease in the total pressure drop and possibly a physiologically healthy pressure drop, treating this portion may indicate that the pressure drop within the conduit can be restored to a physiologically healthy level. Thus, by obtaining the information automatically provided by the methods and systems of the present invention, a user may be able to determine an appropriate treatment strategy without performing a specialized analysis that requires multiple pressure measurements and the time of pullback curves.

[0014] According to embodiments of the present invention, the analysis may include, in addition to the calculations and predictions described herein, the application of algorithms for obtaining diagnostic information such as the presence of a medical condition, the identification of a medical condition, the location of a medical condition, and the like.

[0015] "Conduit" may include a tube or pipe in which a body fluid is contained and transported or circulated. Thus, the term "conduit" may include veins or arteries, coronary vessels, lymphatic vessels, a portion of the digestive tract, and the like.

[0016] Images of the conduit can be obtained using suitable imaging techniques such as, for example, X-ray images, ultrasound images, magnetic resonance images (MRI), and other suitable imaging techniques. In some embodiments of the present invention, an angiography method is used that includes injecting a radiopaque contrast agent into the patient's blood vessel and imaging the blood vessel using an X-ray based technique. According to embodiments of the present invention, the images used are typically 2D longitudinal images of the conduit, as opposed to 2D cross-sectional images used in methods that require the construction of a 3D model of the conduit, such as, for example, coronary computed tomography angiography (CTA) and other computed tomography methods.

[0017] Pathological conditions can include, for example, narrowing (stenosis or occlusion) of the conduit, lesions within the conduit, and the like.

[0018] "Functional measurement" is to measure the effect of a pathological condition on blood flow through the conduit. Functional measurements can include measurements such as an estimated fractional flow reserve (FFR), instantaneous flow reserve ratio (iFR), coronary flow reserve ratio (CFR), quantitative flow ratio (QFR), resting whole cycle ratio (RFR), quantitative coronary analysis (QCA), and the like.

[0019] In embodiments of the present invention, local pressure values are not measured or estimated, but the pressure drop between two points within the conduit is estimated (e.g., predicted). Thus, the methods and systems according to embodiments of the present invention can mimic iFR measurements without performing invasive procedures and, in many cases, without generating a pullback curve that is often a cause of physician complaints.

[0020] In the following description, various aspects of the present invention will be described. For the purposes of the description, specific configurations and details are set forth in order to provide a thorough understanding of the present invention. However, it will also be apparent to those skilled in the art that the present invention can be practiced without the specific details presented herein. Additionally, well-known features may be omitted or simplified in order not to obscure the present invention.

[0021] Unless otherwise specified, as will be apparent from the following description, throughout this specification, the use of terms such as "using", "analyzing", "processing", "computer computing", "computing", "estimating", "predicting", "generating", "determining", "detecting", "identifying", etc. refers to the operation and / or process of a computer or computer computing system, or a similar electronic computer computing device, which operates on and / or transforms data represented as physical quantities, such as electronic quantities in the registers and / or memories of a computer computing system, into other data represented as similar physical quantities in the memories, registers, or other information storage, transmission, or display devices of the computer computing system. Unless otherwise specified, these terms refer to the automatic operation of a processor that is performed without relation to the actions of a human operator and without any actions of the operator.

[0022] In one embodiment schematically shown in FIG. 1, a method for non-invasively determining a treatment strategy for a blood vessel that may have multiple lesions or diffuse lesions is illustrated. This method does not include determining the pressure at any (one or more) positions along the catheter and does not depend on creating or presenting a pullback curve or graph to the user. In one embodiment, this method includes obtaining a prediction of the total pressure drop value within the blood vessel based on features generated from a plurality of images of the blood vessel (step 102). The features can be, for example, spatial features and / or temporal features.

[0023] The entire blood vessel to be analyzed is displayed in the image. That is, the image captures all of the anatomical portions of the blood vessel where pressure drop is predicted. Typically, the plurality of images are 2D images in the longitudinal direction of the conduit and are typically captured using X-ray imaging in an angiography procedure. The angiography video can be captured from different views or angles, and the plurality of images can each include individual frames obtained from different angiography videos, such that each of the plurality of images is captured from a different angle. The term "video" as used herein means any kind of sequence of frames or images.

[0024] The images can be obtained and analyzed online. For example, the images can be obtained directly from the imaging device being used during a procedure (e.g., angiography). The images can also be analyzed offline, such as in a physician's clinic. For example, the images can be part of a DICOM (Digital Imaging and Communications in Medicine) file or any other format used for storage and exchange of medical images and related information, such as metadata like patient information and image parameters. DICOM files can be used offline to provide the images to be analyzed according to embodiments of the present invention.

[0025] In step 104, the contribution to the total pressure drop value of one or more portions of the blood vessel is calculated, and in step 106, a newly simulated total pressure drop value of the blood vessel is calculated by using the calculated contribution to neutralize the contribution to the total pressure drop value of the one or more portions. For example, the total pressure drop value can be calculated by ignoring the component representing the one or more portions whose contribution has been neutralized or by assigning a zero or null value.

[0026] One or more of the portions may be a predetermined portion. For example, each portion may be a different anatomical portion of the conduit (e.g., proximal, middle, and distal). Alternatively or additionally, a portion may be defined by the detected medical condition. For example, a portion may be defined as going from one medical condition to another, and a portion may also be defined as a portion that includes both a healthy portion and a diseased portion of the conduit (e.g., a portion including at least one medical condition).

[0027] In step 108, information regarding one or more of the portions is presented to the user. The presented information may include the newly simulated total pressure drop value and / or a display of one or more of the portions with the contributions neutralized. In some embodiments, the predicted total pressure drop value may also be presented to the user. In some embodiments, information may be presented to the user via a graphical presentation (e.g., a figure, graph, and / or chart, etc.) showing the contribution of different portions and / or medical conditions to the total pressure drop along the conduit. Other graphical presentations may be presented to the user.

[0028] In some embodiments, the method may include presenting to the user a cluster of images characterized by the same medical condition (e.g., the same lesion or other affected area), and the images within the cluster are each captured from a different angle. The display of one or more of the portions with the contributions neutralized can be overlaid on the presented cluster (e.g., one or more of the images constituting the cluster), and for example, a graphic mark may be overlaid on one or more of the images of the presented cluster.

[0029] Another embodiment of the present invention is schematically shown in FIG. 2. In the embodiment illustrated in FIG. 2, for a method of non-invasively determining a treatment strategy for a blood vessel that may have multiple or diffuse lesions, it is determined whether the newly simulated total pressure drop value is within a physiologically healthy range (e.g., above or below a threshold value), and only if the newly simulated total pressure drop value is within a physiologically healthy range, the newly simulated total pressure drop value and / or a proposed treatment strategy is presented to the user (e.g., presenting a display of one or more portions where the contribution to the total pressure drop is neutralized).

[0030] In step 202, a prediction of the total pressure drop value within the blood vessel is obtained based on features generated from a plurality of images of the blood vessel, as described below with reference to, for example, FIG. 1.

[0031] In step 203, it is determined whether the predicted total pressure drop value of the blood vessel is within a physiologically healthy range. For example, the physiologically healthy range can be determined by a healthcare institution or by a physician or technician. In one example, if the total pressure drop (e.g., the difference in pressure values between the starting point and the ending point of a portion of the blood vessel or its conduit) is less than 10% (i.e., less than 0.1), it can be considered a normal drop, but a pressure drop exceeding 10% can be a sign of a medical condition. Thus, the physiologically healthy range can be from 0 to 0.1, and any total pressure drop value exceeding 0.1 can be a sign of a medical condition.

[0032] If the predicted total pressure drop value is within a physiologically healthy range, no treatment is necessary. The predicted total pressure drop value (within the physiologically healthy range) may be presented to the user, and this method may proceed to identify another conduit or a different portion of the conduit. If the total pressure drop value is not within the physiologically healthy range, the contribution to the total pressure drop value of one or more portions of the blood vessel is calculated (step 204), and using the calculated contribution, a newly simulated total pressure drop value of the blood vessel is calculated by neutralizing the contribution to the total pressure drop value of the one or more portions (step 206). Next, it is determined whether a newly simulated total pressure drop value within the physiologically healthy range (e.g., less than 0.1) is generated when the contribution of the one or more portions is neutralized. In step 207, if a newly simulated total pressure drop value within the physiologically healthy range is generated when the contribution of the one or more portions is neutralized, possible treatment strategies may be presented to the user (step 208). This strategy typically includes a display of the one or more portions whose contribution has been neutralized, thereby providing the user with a display of the portion(s) that need to be treated to reduce the total pressure drop value to a physiologically healthy level.

[0033] In step 207, if a newly simulated total pressure drop value that is not within the physiologically healthy range (e.g., the newly simulated value may exceed 0.1 and may correspond to an iFR value of less than 0.9) is generated when the contribution of the one or more portions is neutralized, the method may proceed to identify the contribution of another portion(s) of the conduit.

[0034] If all parts of the conduit are analyzed (e.g., the contribution of each part to a newly simulated total pressure drop value (not within the physiologically healthy range) is neutralized) and a physiologically healthy total pressure drop value cannot be obtained, the initially predicted total pressure drop value (not within the physiologically healthy range) is presented without any indication of a part, i.e., the predicted total pressure drop value is presented without a possible treatment strategy (step 210). In some embodiments, the predicted total pressure drop value presented in step 210 may be marked (e.g., by color or other indication) as not being within the physiologically healthy range.

[0035] Another embodiment of the present invention, schematically shown in FIG. 3, involves determining the part of the conduit that contributes most to the total pressure drop and then neutralizing the contribution of this part to the total pressure drop. In the embodiment illustrated in FIG. 3, the contribution of each of a plurality of parts of the patient's conduit to the total pressure drop value is calculated in step 302. In step 303, it is determined whether the contribution of a particular part is the most significant, i.e., whether the contribution of that part to the total pressure drop value is the largest. This determination may be made, for example, by using percentage calculations or value comparisons. In some embodiments, this determination can be made by using a machine learning model, as will be described in more detail below with reference to FIG. 7B, for example.

[0036] If in step 303 a particular part is determined to have the most significant contribution, then in step 304, a newly simulated total pressure drop is calculated by neutralizing the contribution of this particular part. If a particular part is determined not to have the most significant contribution, the contribution of the particular part is not neutralized and is used in the calculation of the newly simulated total pressure drop in step 306.

[0037] In some embodiments of the present invention, the steps described above (e.g., referring to FIGS. 1 - 3) are performed only after or typically after the presence of a diffuse disease or multiple lesions in the blood vessel has been determined.

[0038] The presence of a diffuse disease or multiple lesions within a blood vessel can be determined by tracking the blood vessel within an image of the conduit, where each image is captured from a different view or angle. In one embodiment, the blood vessel can be marked by a user, for example, on a user interface where an image of the blood vessel is presented, by the user entering a mark on the conduit. Alternatively or additionally, a computer vision algorithm can be used to identify the blood vessel within the first image. Next, the blood vessel can be marked, for example, using a virtual mark, which enables it to be tracked, i.e., it can be easily detected in future images even if the future images are captured from an angle different from the capture angle of the first image and have different visual characteristics from the first image. Similarly, a pathological condition (such as a diffuse disease or multiple lesions) can be marked by a user or identified using computer vision techniques. The pathological condition marked or identified by the user can be marked, for example, using a virtual mark (which may be location-based, such as based on the location of the pathological condition relative to a portion or structure of the conduit), which enables it to be tracked, i.e., it can be easily identified in future images even if the future images are captured from an angle different from the capture angle of the first image and have different visual characteristics from the first image.

[0039] In the embodiment illustrated in FIG. 4, a plurality of images captured from different angles are obtained (step 402). The plurality of images may include frames respectively selected from one of the plurality of angiography videos, and each video is captured from a different angle. Blood vessels can be identified and traced from the plurality of images, and pathological conditions within the blood vessels can be identified. When a diffuse disease and / or multiple lesions are identified within the blood vessels (step 403), the method proceeds to the different calculations described above (step 404). For example, the method can obtain a prediction of the contribution of each of a plurality of portions of a blood vessel to the total pressure drop, and as described above, provide a display of the newly simulated total pressure drop value and one or more portions where the contributions are neutralized (see, for example, FIG. 1). Thus, in this embodiment, the presence of a diffuse disease or multiple lesions within the blood vessels is determined before starting any of the different calculations described above, for example, before obtaining a prediction of the total pressure drop value of the blood vessel or before obtaining a newly simulated total pressure drop.

[0040] If a diffuse disease and / or multiple lesions are not identified within the blood vessels (step 403), the method can analyze a new set of images for the presence of a pathological condition without performing calculations.

[0041] FIG. 5 schematically shows a method of using combined data from a plurality of images, each image being captured from a different angle, in accordance with an embodiment of the present invention.

[0042] In step 502, a plurality of images captured from different angles are obtained. Each of the plurality of images may be selected, for example, from one of a plurality of angiography videos in which each video is captured from a different angle. Blood vessels can be identified and traced from the plurality of images (e.g., by marking the blood vessels identified in each of the images as described above). In step 504, the complete path of the identified blood vessels is mapped. The "complete path" typically refers to the length of the conduit that may include all pathologies that can contribute to the total pressure drop. For example, the complete path of the left main coronary artery (LMCA) may include the path from the starting point of the artery to after the artery branches, and, for example, the complete path of the right coronary artery (RCA) may include the path from the starting point of the artery until the artery becomes too narrow to obtain meaningful information therefrom. By mapping the complete path of each artery, it becomes possible to analyze each artery individually.

[0043] Once the complete path of the conduit is mapped, in step 506, using data from images at different angles (images captured from different angles) that characterize the same conduit (determined, for example, by tracking the conduit as described above), a single data representation is created. In one embodiment, the single data representation may include data combined from images at different angles. For example, images at different angles may be aggregated and converted, for example, by image registration, into a single coordinate image data representation. In another embodiment, the single data representation may include non-image data. For example, a single data representation can be created based on the arrangement of one-dimensional (1D) characteristics of the conduit (information regarding a specific position along the conduit, such as the width of the conduit at a position along the conduit or the presence of a pathology at a position along the conduit), creating a data representation that is a 1D signal. In other embodiments, the single data representation need not include data (such as the combined data described above), but rather, data from images at different angles can be used and, for example, mathematically manipulated (such as averaged) to create a single data representation based on the data from images at different angles, but without including the data itself.

[0044] Next, at step 508, using a single data representation, features can be extracted therefrom. The extracted features (which may be, for example, spatial features and / or temporal features) can be used to obtain a prediction of the total pressure drop value, as will be further described below with reference to, for example, FIG. 7B.

[0045] For example, the temporal features may be extracted from a signal created from the attributes of an image of the conduit recorded over time. The attributes of the image of the conduit can be visible features of the image or other features unique to the image. For example, the attributes can include pixel intensity or pixel color or gray level. In one example, the attributes can include the number of pixels having a color or intensity above a threshold, and / or the distribution of these pixels.

[0046] In some embodiments, the temporal features include the calculation of combinations of attribute values determined from multiple images, while the spatial features can include, for example, visible attributes of the image.

[0047] FIG. 6 schematically shows one possible way of selecting an image from, for example, an angiography video, such as in steps 402 and / or 502 above.

[0048] At step 602, a first angiography video of a patient captured from a first angle is obtained, and at step 604, a second angiography video of the (same patient and same location of the patient) captured from a second angle is obtained. Angiography diagrams or images can be obtained from different angles, for example, by moving a camera between two viewpoints and / or by rotating the patient on, for example, an angiography table.

[0049] In steps 606 and 608, the first and second angiography video frames captured during the diastolic (rest) phase of the patient's cardiac cycle are determined (correspondingly). In step 610, using the frames determined to be captured during diastole, a single data representation is created, for example, as described above. In step 612, using the single data representation, features can be extracted therefrom (as described herein with reference to FIGS. 5 and 7B, for example).

[0050] Thus, the frames captured during diastole from multiple angiography videos captured from different angles may be used as multiple images for which further calculations (such as obtaining the predictions described herein) are performed.

[0051] In some embodiments, determining the frames captured during the diastolic phase of the cardiac cycle includes selecting the frame with the highest contrast agent density (from angiography videos captured from different angles), and then selecting the frame characterized by the catheter filled with the contrast agent. The selected frames are input into an ML model trained to find the key points of the cardiac cycle based on the frames characterized by the catheter filled with the contrast agent, and the frames are each captured from different angles and the frames captured during diastole are selected.

[0052] Examples of key points of the cardiac cycle may include, for example, the R peak determined by an electrocardiogram (ECG).

[0053] A system for non-invasively determining a treatment strategy for blood vessels with multiple lesions or diffuse lesions according to an embodiment of the present invention is schematically shown in FIG. 7A.

[0054] In one embodiment, the system 700 includes a machine learning (ML) unit 708 that predicts the total pressure drop value of a blood vessel from an image of the blood vessel. Typically, the image includes longitudinal 2D images captured from different angles (e.g., images selected from an angiography video captured from different angles). The total pressure drop value is predicted without obtaining pressure measurement values at positions along the catheter.

[0055] The system 700 capable of implementing the methods or steps described herein includes a processor 702 that communicates with the ML unit 708 and the user interface device 706. The processor 702 receives one or more images 703 of the patient's catheter. The images 703 may include images at different angles, typically frames of an angiography video, and frames may be selected therefrom, for example, as described herein.

[0056] A single data representation can be created from the images 703 by the processor 702 or another processing unit. Features (e.g., spatial features and / or temporal features) are extracted from the single data representation by the processor 702 or another processing unit. The extracted features can be input into the ML unit 708. The ML unit 708 can provide a prediction of the total pressure drop value of the blood vessel characterized in one or more of the images 703, and / or a contribution to the total pressure drop by one or more portions of the blood vessel, and / or a newly simulated total pressure drop value of the blood vessel taking into account the neutralization of the contribution to the total pressure drop by one or more portions. A portion of the output of the ML 708 can be calculated by the processor 702 or other processing unit.

[0057] The newly simulated total pressure drop value, and optionally, a display of one or more portions whose contribution has been neutralized to obtain the newly simulated total pressure drop value, can be presented to the user interface device 706, for example, as part of a possible treatment strategy. In some embodiments, the total pressure drop value may be presented to the user, for example, on the user interface device 706.

[0058] Processor 702 may cause a display of one or more portions to be presented on an image of a patient's conduit (e.g., image 703), which is likewise presented, via user interface device 706. The display of one or more portions presented on the display of user interface device 706 may include, for example, graphics such as text, numbers, symbols, different colors and shapes that can be overlaid on a video image(s) of the patient's conduit. Alternatively or additionally, the display of one or more portions may include, for example, a textual description of the portion with reference to the anatomical location or other anatomical features of the portion.

[0059] In some embodiments, user input may be received by processor 702 via user interface device 706. User interface device 706 may include a display, such as a monitor or screen, for presenting images, instructions, and / or notifications to the user (e.g., via graphics, images, text, or other content presented on the monitor). User interface device 706 may be designed to receive input from the user. For example, user interface device 706 may include or be communicable with mechanisms for inputting data, such as a keyboard, and / or a mouse, and / or a touch screen, so that the user can input data.

[0060] Processor 702 may include, for example, one or more processors and may be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a microprocessor, a controller, a chip, a microchip, an integrated circuit (IC), or other suitable general-purpose or specific processor or controller. Processor 702 and / or ML unit 708 may be embedded locally or may be remote, such as in the cloud.

[0061] Processor 702 typically communicates with memory unit 712. In one embodiment, when implemented by processor 702, memory unit 712 stores executable instructions that facilitate the implementation of the operations of processor 702, as described herein. Memory unit 712 can also store at least some of the image data of image 703 (such as data representing the intensity of light that has passed through body tissue and / or been reflected from a contrast agent within tissue or a conduit and received by the image sensor, including data such as pixel values, as well as partial or complete images or videos, etc.).

[0062] Memory unit 712 can include, for example, random access memory (RAM), dynamic RAM (DRAM), flash memory, volatile memory, non-volatile memory, cache memory, buffers, short-term memory units, long-term memory units, or other suitable memory or storage units.

[0063] All or some of the components of system 700 can communicate either wired or wirelessly and can include suitable ports such as USB connectors and / or network hubs.

[0064] In one embodiment, image(s) 703 include 2D longitudinal images of a patient's conduits (such as angiographic images and / or videos). Processor 702 receives one or more images and uses ML techniques and / or additional computer vision algorithms to detect and identify the conduits within image 703. In one example, a classifier such as a random forest classifier or a CNN classifier can be pre-trained with training data such as 2D longitudinal X-ray angiographic images of anatomically classified conduits to identify and anatomically classify arteries from the images. Anatomical classification of the conduits can include classifying the conduits based on anatomical location (such as left / right) and / or other anatomical features (such as the right coronary artery (RCA), left anterior descending branch (LAD), and left circumflex artery (LCX)).

[0065] The processor 702 may further obtain an indication of the presence of a pathological condition (e.g., stenosis) within the conduit and, if possible, the location of the lesion, for example, by applying a classifier to the image(s) 703. A classifier such as DenseNet, CNN (Convolutional Neural Network), EfficientNet, etc. may be used to obtain a determination of the presence of the pathological condition and to determine the location of the pathological condition. According to an embodiment of the present invention, the classifier may be pre-trained with training data including 2D longitudinal X-ray angiography images of conduits that may include a pathological condition (e.g., stenosis). The neural network constituting the classifier may be trained by a supervised learning or semi-supervised learning scheme.

[0066] Applying a classifier to the angiography image using ML techniques and / or additional or other computer vision techniques enables automatic detection of the conduit and / or the pathological condition and mapping of the entire length of the conduit, for example, without the need for user input.

[0067] In some embodiments, the longitudinal 2D image is the optimal image selected as the most detailed image from a plurality of 2D images of the patient's conduit. In the case of an angiography image, it includes a contrast agent injected into the patient to visualize the conduit (such as a blood vessel) on the X-ray image, and the optimal image can be an image of the blood vessel showing a large / maximum amount of contrast agent. Thus, the optimal image can be detected on the sequence of images by applying an image analysis algorithm (e.g., to detect the image frame with the most colored pixels). In one embodiment, the optimal image is the image captured at the time corresponding to the maximum relaxation of the heart. Thus, the optimal image can be detected based on, for example, the capture time of the image compared to the measurement of the electrical activity of the patient's heartbeat (e.g., ECG print). In other embodiments, an angiography video frame characterized by a conduit filled with contrast agent is input into an ML model trained to find important objective points (e.g., R peak) within the cardiac cycle, and the optimal image may be selected by selecting the frame captured during diastole. The trained ML model assists in the selection of the frame captured during diastole.

[0068] In one embodiment, the processor 702 receives a plurality of consecutive images of the patient's conduit and determines the presence and, in some cases, the location of diffuse disease and / or multiple lesions (such as stenosis) within the conduit in one of the images before calculating the total pressure drop value of the conduit and / or before determining the contribution of one or more portions of the conduit to the total pressure drop and / or before any of the other subsequent steps described herein (e.g., by applying a machine learning model to one of the images and applying a classifier to the image as described above).

[0069] An embodiment of the present invention is schematically shown in FIG. 7B for an ML unit (such as ML unit 708). In this embodiment, the ML unit 708 further includes a first machine learning (ML) model 721 that provides a prediction 722 of an intermediate value related to one or more portions of a blood vessel based on features 720 related to an image of the blood vessel. The features 720 may include, for example, spatial features and / or temporal features.

[0070] The ML unit 708 also includes a second ML model 723 for obtaining a prediction 726 of the total pressure drop value of the blood vessel based on the predicted intermediate value. Next, a processor 724 communicating with the second ML model 723 analyzes the calculation of the second ML model 723 to determine the contribution 728 to the total pressure drop value of each portion.

[0071] The processor 724 or another processor may calculate a newly simulated total pressure drop value 729 of the blood vessel by neutralizing the contribution 728 to the total pressure drop of one or more portions. The processor 724 communicates with a user interface device (such as the devices described herein) that may include a monitor or other display, and presents to the user possible treatment strategies that may include information regarding one or more portions (such as the display of one or more portions and, in some cases, the newly simulated total pressure drop value).

[0072] The methods and systems described herein can cooperate to provide a fully automated solution that begins with a patient's angiogram (800) and ends with possible treatment strategies (810) for a multi-focal or diffuse disease. As schematically shown in FIG. 8, a fully automated process that does not require user input at any stage includes automatically selecting the optimal image(s) (802) from the patient's angiogram (800), for example, selecting an image captured at the time corresponding to the maximum relaxation of the heart, as described with reference to FIG. 6 above. The arteries in the optimal image are automatically identified (804) and fully mapped (as described, for example, with reference to FIG. 5 above). Thereby, these arteries can be traced, and the presence of a multi-focal or diffuse disease in the arteries can be automatically determined from the optimal image(s) (806), as described with reference to FIG. 4 above. When the presence of a multi-focal or diffuse disease is determined, it can trigger the performance of predictions and calculations (808) as described above with reference to FIGS. 7A and 7B, and based on the calculations triggered, possible treatment strategies can be provided to the user (810).

[0073] In some embodiments, the above steps can use user input, for example, for image selection and / or identification of arteries and / or conditions.

[0074] The methods described herein may be performed online during angiography, stent implantation, or another procedure on a patient, or offline in a physician's or technician's office using previously collected data during the procedure, as described herein, for example, using DICOM files.

[0075] A display according to an embodiment of the present invention is schematically shown in FIG. 9.

[0076] The display 900 is part of a user interface device that can communicate with a processor such as the processor described herein. The display 900 is configured to present an image 903 of an artery having one or both of a diffuse disease and multiple lesions, for example, an angiographic image obtained via a processor. The image may be, for example, an image obtained online during a procedure and / or an image presented offline, for example, during analysis at a physician's clinic after the procedure. Additional information such as information about the patient may be attached to the image.

[0077] Also, presented on the display is a display 904 of one or more portions of an artery in which the total pressure drop value of the artery changes to a value within a predetermined range (e.g., within a physiologically healthy range) when the contribution to the total pressure drop value of the artery is neutralized. The display 904 can provide the user with information regarding the portion of the presented artery that needs to be treated in order to change the total pressure drop within the artery to be within a physiologically healthy range. That is, it is information about the portion of the artery that needs to be treated in order to obtain an artery showing a normal and healthy total pressure drop. Thus, the display 900 can provide possible treatment strategies for the artery.

[0078] In some embodiments, the display includes buttons 906 that enable user input within a desired or required range. The buttons 906 may be virtual buttons and / or may include graphical elements. Thus, the user can, for example, input new total pressure drop values and / or options for portions of the artery whose contribution to the total pressure drop value of the artery needs to be confirmed in order to calculate whether the total pressure drop value changes to a value greater than, for example, the input value when the contribution of a portion input by the user is neutralized.

[0079] In some embodiments, the total pressure drop value 907 of the artery may be presented and / or a newly simulated total pressure drop value 909 may be presented on the display 900.

[0080] In some embodiments, image 903 may include a graphical representation of an artery rather than an actual angiographic image.

[0081] In some embodiments, image 903 includes a cluster of images that characterize the same medical condition, and each of the images within the cluster is captured from a different angle. The cluster of images may be combined into a single image data representation. In other embodiments, image 903 or the cluster of images may include a sequence or group of images captured from different angles.

[0082] Display 904 may include, for example, graphics such as letters, numbers, symbols, different colors and shapes, and may be overlaid on image 903.

[0083] Using display 900 and methods and systems such as those described herein, information is automatically provided, enabling a user to simply and easily determine a treatment strategy (e.g., stent placement) for blood vessels having multiple or diffuse lesions. These calculations can be performed as quickly as possible when an image (such as an angiogram) is received, and the resulting information can be presented to the user, so that a decision regarding a treatment strategy (e.g., which catheter to insert a stent into) may be made online during a procedure (such as an angiography). Further, according to embodiments of the present invention, the system may receive, as an input, a file including an image (such as an angiogram), for example, a DICOM file, and may perform an analysis on the image from the file, thereby enabling a treatment strategy to be determined even offline.

Claims

1. A method for non-invasively determining a treatment strategy for a blood vessel, comprising: obtaining a prediction of a total pressure drop value within the blood vessel based on features generated from a plurality of images of the blood vessel, wherein the images represent all anatomical parts of the blood vessel; calculating the contribution of one or more portions of the blood vessel to the total pressure drop value; calculating a newly simulated total pressure drop value of the blood vessel by neutralizing the contribution of the one or more portions to the total pressure drop value based on the calculated contribution; presenting information about the one or more portions to a user.

2. The method of claim 1, wherein the information includes a display of the one or more portions and the newly simulated total pressure drop value, and presenting the information to the user when the newly simulated total pressure drop is determined to be within a physiologically healthy range.

3. The method of claim 1, further comprising determining a portion of the one or more portions that contributes most to the total pressure drop and neutralizing the contribution of the determined portion to the total pressure drop value.

4. The method of claim 1, wherein each of the plurality of images is captured from a different angle.

5. The method of claim 1, further comprising identifying the blood vessel in each of the plurality of images and tracking the same blood vessel from the plurality of images.

6. The method of claim 5, further comprising identifying the blood vessel in each of the plurality of images and mapping a complete path of a predetermined portion of the blood vessel in each image.

7. The method of claim 6, wherein the predetermined portion of the blood vessel includes a portion from the beginning of the left main coronary artery (LMCA) to after the bifurcation of the artery.

8. The method of claim 5, further comprising creating a single data representation using data from the images and extracting features from the single data representation, wherein the features are for obtaining the prediction of the total pressure drop value.

9. The method of claim 1, wherein the features include spatial features and / or temporal features.

10. obtaining a prediction of an intermediate value associated with the one or more portions of the blood vessel from a first machine learning (ML) model. ​ ​ ​ Inputting the intermediate value into a second ML model to obtain the prediction of the total pressure drop value of the blood vessel; Analyzing the calculation of the second ML model to determine the contribution of each part to the total pressure drop value of the blood vessel, the method according to claim 1.

11. The method according to claim 1, wherein one or more parts of the blood vessel include a healthy part of the blood vessel and a part including at least one pathological condition.

12. Tracking the blood vessel from the plurality of images, each of the images capturing the blood vessel from a different angle; Determining the presence of a diffuse disease or multiple lesions in the blood vessel before obtaining the prediction of the total pressure drop value of the blood vessel, the method according to claim 1.

13. The plurality of images include angiography images; Determining a frame of an angiography video of a patient captured during the diastolic phase of the cardiac cycle of the patient; Using the frame captured during the diastolic phase of the cardiac cycle of the patient as the plurality of images from a plurality of angiography videos captured from different angles respectively, the method according to claim 1.

14. The method according to claim 13, wherein determining the frame captured during the diastolic phase is Selecting a frame from the angiography video of the patient, the frame being characterized by a catheter filled with a contrast agent; Inputting the frame into an ML model trained to find key points of the cardiac cycle based on frames characterized by a catheter filled with a contrast agent, the frames being captured from different angles respectively, and the frame captured during the diastolic phase being selected, the method including.

15. The method according to claim 14, wherein the key point of the cardiac cycle includes an R peak determined by an ECG.

16. A system for non-invasively determining a treatment strategy for a blood vessel having multiple lesions or diffuse lesions, comprising: A first machine learning (ML) model that provides a prediction of an intermediate value associated with one or more parts of the blood vessel based on features associated with an image of the blood vessel; A second ML model for obtaining a prediction of the total pressure drop value of the blood vessel based on the intermediate measurement; A system comprising a processor that analyzes the calculation of the second ML model to determine the contribution of each part to the total pressure drop value. **Claim 17** The system according to claim 16, wherein the processor calculates a newly simulated total pressure drop value of the blood vessel by neutralizing the contribution of the one or more parts to the total pressure drop value. **Claim 18** The system according to claim 16, wherein the features include spatial features and / or temporal features. **Claim 19** The system according to claim 16, comprising a display for presenting information about the one or more parts to a user. **Claim 20** A display of a user interface device, wherein the user interface device communicates with a processor, and the display is configured to present: an image of an artery having one or both of a diffuse disease and multiple lesions; a display showing a part of the artery in which the total pressure drop value changes to a value within a physiologically healthy range when the contribution to the total pressure drop value of the artery is neutralized.

Images