A system, method, and computer-accessible medium for providing plaque load assessment.

Abstract

For example, exemplary embodiments of methods, systems, and computer-accessible media for measuring certain exemplary clinical features that can guide clinical decision-making during percutaneous coronary intervention (PCI), such as plaque load, which is important for assessing vascular health and patient or plaque vulnerability, and external elastic lamina diameter, which is important for sizing and positioning stents, balloons, and other coronary reshaping tools during PCI.

Description

Technical Field

[0001] (Cross - Reference to Related Applications) This application claims the benefit of priority from U.S. Provisional Patent Application No. 63 / 466,773, filed May 16, 2023, the entire disclosure of which is incorporated herein by reference in its entirety.

[0002] (Field of Disclosure) This disclosure generally relates to the analysis of plaque, and more specifically, to systems, methods, and computer - accessible media for providing plaque burden assessment.

Background Art

[0003] (Background Information) The coronary artery has a layered structure. Layers such as the external elastic lamina (EEL) or external elastic media (EEM) in most clinical cases can represent the true size of the coronary vessels for treatment planning and are thus important tissue structures to be assessed during the planning of percutaneous coronary intervention (PCI) procedures. Stents, balloons, and other coronary vessel modification methods can benefit from planning, sizing, and positioning using the measured EEL diameter compared to sizing using a stenosed lumen that no longer represents the true vessel size in atherosclerotic lesions. The EEL - to - lumen ratio can also be an important metric for assessing vascular health status and vulnerability to future adverse events. However, the EEL in the presence of atherosclerosis is a subsurface tissue structure and is conventionally difficult to visualize and measure using optical methods and has been plagued by reduced image resolution, and intravascular ultrasound has been the preferred methodology for vascular sizing.

[0004] Intravascular ultrasound (IVUS) is effective for imaging deep tissue structures due to the nature of the ultrasound imaging modality. However, IVUS offers poor spatial resolution compared to some optical modalities and therefore provides incomplete data when assessing and sizing arterial structures important for clinical decision-making, such as thrombosis and dissection. The resolution of optically sensing modalities (e.g., intravascular optical coherence tomography (IVOCT)) and their ability to measure relatively deep structures such as EELs would be significantly beneficial when making important clinical decisions during PCI.

[0005] Therefore, there is a need to provide devices, systems, computer-accessible media, and / or methods to address and / or overcome at least some of these shortcomings. Furthermore, providing techniques for inferring the location of layered structures, for example, when they are not directly visible in images, can provide beneficial clinical utility. [Overview of the project] [Means for solving the problem]

[0006] (Summary of exemplary embodiments) According to exemplary embodiments of this disclosure, it is possible to overcome the challenges of using optical imaging to detect and assess the location and size of layered structures (e.g., intima, media, EEL). To this end, exemplary embodiments of methods, systems, and computer-accessible media for measuring certain exemplary clinical features that can guide clinical decision-making during percutaneous coronary intervention (PCI), such as plaque load, which is important for assessing vascular health and patient or plaque vulnerability, and EEL diameter, which is important for sizing and positioning stents, balloons, and other coronary artery repair tools during PCI.

[0007] Methods, systems, and computer-accessible media according to exemplary embodiments can be used to visualize and estimate the location and size of EELs, for example, in both two and three dimensions, along with downstream clinical features, using intravascular imaging (e.g., ultrasound imaging, e.g., optical imaging). In one exemplary embodiment, improved visualization of EELs is achieved via a highly sensitive optical coherence tomography (OCT) imaging system with improved depth imaging. In some exemplary embodiments, there are image processing procedures for automatically detecting visible portions of EELs or any other laminar coronary structures (e.g., segmentation using machine learning / deep learning procedures), and optionally pre-processing (e.g., aligning, enhancing, denoising, etc.) visually obscured or invisible portions and inferring (e.g., interpolation, extrapolation, estimation, etc.). In some exemplary embodiments of this disclosure, improved depth imaging can be combined with exemplary image processing methods to provide further EEL localization accuracy. In some exemplary embodiments of the present disclosure, such improved visualization, detection, preprocessing, and inference can yield highly accurate EEL diameter measurements, which are important for inducing PCI. In some exemplary embodiments of the present disclosure, such improved visualization, detection, preprocessing, and inference can yield highly accurate plaque load estimates, which are important for assessing the risk of adverse events (e.g., at the patient or plaque level).

[0008] Accordingly, exemplary methods, systems, and computer-accessible media according to exemplary embodiments of the present disclosure can be provided, which enable the acquisition of images of at least one segment of the arterial wall, the automatic detection of one or more visible portions of the arterial external elastic lamina of the segment, the automatic determination or estimation of one or more invisible portions of the arterial external elastic lamina based on the automatically detected visible portions, the automatic combination of visible and invisible portions, and the provision of measurements of the external elastic lamina.

[0009] For example, it is possible to automatically estimate plaque load from at least one of one or more first parts or one or more second parts. For example, using a display device, it is possible to display an external elastic plate or a representation of plaque load based on a viewport in the user interface, for example, in response to reaching a predetermined threshold of plaque load. The intravascular image may be an optical coherence tomography (OCT) image received from an OCT system with a sensitivity of approximately 100 dB or higher. One or more further (e.g., invisible) parts can be determined or inferred using machine learning procedures. The machine learning procedure can determine or infer a second (e.g., invisible) part by interpolation from neighboring frames.

[0010] According to another exemplary embodiment of the present disclosure, a method, system, and computer-accessible medium for determining or providing one or more measurements of an arterial wall can be provided. Thus, for example, it is possible to determine or estimate plaque load by acquiring an image of at least one segment of the arterial wall, automatically detecting one or more first segments of the extraarterial elastic lamina of the segment of the arterial wall, automatically detecting one or more second segments of the arterial wall, and processing the detected first and second segments. It is also possible, for example, to display a representation of the plaque load on a viewport of a user interface when a predetermined threshold of plaque load is reached.

[0011] According to yet another exemplary embodiment of the present disclosure, a method, system, and computer-accessible medium for determining or providing one or more measurements of a coronary artery can be provided. Thus, for example, it is possible to take an image of at least one portion of a coronary artery and, using a machine learning-based procedure, directly determine or estimate at least one measurement of plaque load based on one or more features or information relating to the portion of the image.

[0012] According to further exemplary embodiments of the present disclosure, methods, systems, and computer-accessible media for determining or providing one or more measurements of a coronary artery can be provided. Thus, for example, it is possible to acquire an image of at least one portion of a coronary artery and use a machine-trained procedure to determine or estimate the location of the external elastic lamina within at least one portion of the image. For example, the image can be generated using a first imaging modality. It is possible to train a machine-trained procedure to determine or estimate the location of the external elastic lamina using alignment data from a second imaging modality different from the first imaging modality.

[0013] In one exemplary embodiment, the alignment data may have a lower (e.g., IVUS) or higher (e.g., histology, e.g., confocal imaging) resolution than that of the image from the first imaging modality (e.g., IVOCT). The machine-learned procedure can determine or estimate the position of the external elastic plate within an area where the external elastic plate is invisible in the image. The machine-learned procedure can generate further images adjacent to the image.

[0014] These and other purposes, features, and advantages of the exemplary embodiments of this disclosure will become apparent upon careful reading of the following detailed description of the exemplary embodiments of this disclosure, when considered in conjunction with the appended claims. [Brief explanation of the drawing]

[0015] Further objectives, features, and advantages of this disclosure will become apparent from the following detailed description, in conjunction with the accompanying diagrams illustrating illustrative embodiments of this disclosure.

[0016] [Figure 1] Figure 1 is a cross-sectional diagram of a coronary artery to illustrate plaque loading.

[0017] [Figure 2] Figure 2 is a block diagram of a workflow and method according to an exemplary embodiment of the present disclosure.

[0018] [Figure 3] Figure 3 is an exemplary illustration set of an atherosclerotic plaque, and the risk associated with plaque load, where the higher the plaque load, the higher the risk of atherosclerotic / vulnerable plaque.

[0019] [Figure 4] Figure 4 is an exemplary illustration of an exemplary interpolation of an EEL contour using automatically detected / visible EEL segments.

[0020] [Figure 5] Figure 5 is an exemplary illustration of the alignment of the EEL prior to interpolation.

[0021] [Figure 6] Figure 6 is an exemplary illustration of the ND interpolation of the EEL.

[0022] [Figure 7] Figure 7 is an exemplary illustration of the ND interpolation of the EEL using an ellipticity constraint, optimized / trained using IVUS or ground truth data.

[0023] [Figure 8] Figure 8 is an exemplary illustration of the ND interpolation within a continuous region defined by automatically detected branches.

[0024] [Figure 9] Figure 9 is an exemplary illustration of an exemplary UI. It is a multi-panel and includes a threshold processed plaque load measurement overlay that exceeds a predefined value, along with an EEL overlay.

[0025] [Figure 10] Figure 10 is a schematic diagram of an exemplary intravascular imaging system according to an exemplary embodiment of the present disclosure.

[0026] [Figure 11]Figure 11 illustrates an exemplary block diagram of an exemplary system according to one exemplary embodiment of the present disclosure. [Modes for carrying out the invention]

[0027] Throughout the drawings, unless otherwise noted, the same reference numerals and letters are used to represent similar features, elements, components, or parts of the illustrated embodiments. Furthermore, this disclosure will be described in detail herein with reference to the drawings, in relation to illustrative embodiments and not limited to the specific embodiments illustrated in the drawings and accompanying paragraphs.

[0028] (Detailed description of exemplary embodiments) In some exemplary embodiments of this disclosure, the location and size of EELs are inferred directly via frame-based (e.g., image-based) or volume-based (e.g., multi-frame, multi-image, etc.) machine learning procedures, which can take frames or volumes of data or pre-processed versions of such data (e.g., alignment, registration, enhancement, denoising, etc.) as input, in addition to physical parameters associated with image acquisition and / or clinical parameters associated with the patient. In some exemplary embodiments of this disclosure, the location and size of EELs can be inferred indirectly via frame-based or volume-based machine learning procedures, which first output the area in which the EEL is clearly visible. A custom shape-fitting model can then be used in each frame, and when it is not directly visible in the circumferential direction, the location and size of the EEL can be interpolated and / or extrapolated.

[0029] According to some exemplary embodiments of this disclosure, a shape-fitting model can utilize visible EEL data from multiple frames to interpolate and / or extrapolate the position and size of the EEL in a single target frame. In some exemplary embodiments of this disclosure, prior information such as a normative atlas or patient-specific atlas (e.g., generated by another imaging modality, such as coronary computed tomography angiography (CCTA)) relating to a given vessel can be used to inform preprocessing and / or interpolation and / or inference procedures. In some exemplary embodiments of this disclosure, additional physiological information relating to the vessel (e.g., ischemia index, e.g., local FFR measurements, e.g., longitudinal FFR measurements, e.g., location of side branches, vessel of interest such as RCA, LCX) can be used to further specialize the machine learning model, and / or any of the associated processing methods (such as shape-fitting) can be used in conjunction with the exemplary embodiments of this disclosure.

[0030] According to various exemplary embodiments of this disclosure, exemplary EEL detection can be inferred / interpolated using a shape-fitting model (e.g., circular or elliptic fit) with defined constraints (e.g., ellipticity constraints). In some exemplary embodiments of this disclosure, for example, the elliptic fitting model may use optimized fit constraints (e.g., maximum ellipticity, e.g., maximum diameter), and these optimized constraints may be automatically updated based on some automatically detected features (e.g., EEL diameter). Furthermore, plaques of various compositions have varying levels of hardness. For example, calcium plaques are hard and crystalline structures, while lipid pools can be soft. In some exemplary embodiments, the optimized fit constraints (e.g., ellipticity constraints, e.g., eccentricity constraints) are automatically updated with respect to the automatically detected plaque type (e.g., calcium or lipid-rich). Also, luminal stenosis of the coronary artery and other morphological characteristics (e.g., tortuosity) can provide information about potential deformation of the EEL. For example, a highly eccentric plaque causing eccentric lumen stenosis may cause the EEL to expand non-concentrically on the side of the plaque (which is potentially narrowing). In some exemplary embodiments of this disclosure, lumen morphology and / or other vascular morphological characteristics may be used to inform optimized fit constraints (e.g., ellipticity constraints, eccentricity constraints, elliptic center, etc.) and / or any other processing techniques (e.g., detection, preprocessing, inference, etc.) that do not include / require constraints.

[0031] Patient and arterial categories can influence the diameter of a given artery and the rate of diameter reduction along the length of the vessel. In some exemplary embodiments, the optimized fit constraints are automatically updated with respect to known or detected categories of the patient (e.g., history of CAD) or artery (e.g., left anterior descending artery (LAD), right coronary artery (RCA), etc.).

[0032] The lateral aspect of the ductus arteriosus lumen, for example, the automatically detected lumen eccentricity, can also assist in the automatically detected inference of the EEL diameter and eccentricity. In some exemplary embodiments of this disclosure, the optimized fit constraints can be automatically updated based on the lumen profile (e.g., lumen diameter or, for example, lumen eccentricity, lumen depression, lumen center, etc.).

[0033] Interpolating automatically detected portions of EELs can be achieved using machine learning-prepared procedures or machine learning procedures (e.g., machine learning-prepared elliptic fitting procedures, machine learning-prepared semantic segmentation, machine learning-prepared detection, etc.). In some exemplary embodiments of this disclosure, optimized fitting constraints may be machine learning-prepared constraints (e.g., through training on aligned IVUS training datasets, histology training datasets, etc.). In some exemplary embodiments of this disclosure, machine learning procedures may be trained on expert annotations. In some exemplary embodiments of this disclosure, expert annotations may only include annotations if the EEL is directly visible within the intravascular imaging modality. In some exemplary embodiments of this disclosure, expert annotations may include annotations if the EEL is not directly visible and can be inferred (e.g., interpolated by a human) based on the expert technical knowledge of the annotator of the EEL form. In some exemplary embodiments of this disclosure, a machine learning procedure can be trained with registered images from different imaging modalities from an intravascular imaging modality (e.g., registered higher resolution images, registered lower resolution images, registered histological images, registered IVUS images, registered OCT images, registered CCTA images, etc.).

[0034] Information from neighboring frames (e.g., automatically detected objects) can also be used to assist in the inference of fit constraints for a particular frame. For example, in some exemplary embodiments of this disclosure, the inference method or procedure and / or optimized fit constraints can be automatically updated, at least in part, based on the proximity of automatically detected vascular features (e.g., collaterals) within a given frame or neighboring frames.

[0035] For example, any optimization of the fit constraints can be performed to optimize the accuracy of the plaque load measurement.

[0036] According to further additional or alternative exemplary embodiments of the present disclosure, the user interface may summarize a set of EEL measurements to the user, potentially along with other metrics (e.g., lumen size, lipids, etc.), to inform clinical decisions. Exemplary measurements and indicators may include when any of the EEL diameter, area, plaque load, etc., and any of the measurements exceed or fall below a computer- or user-generated threshold. For example, a plaque load greater than approximately 50%, e.g., almost or all of it, may trigger a visual indicator to the user, and / or, for example, a plaque load less than approximately 50%, may be indicated as a recommended location to terminate a stent or balloon (e.g., shown on an angiography viewport, shown on a longitudinal viewport, shown in a transverse intravascular viewport, etc.). According to further additional or alternative exemplary embodiments of the present disclosure, the user interface may display a 3D rendering of any of such measurements.

[0037] In additional or alternative exemplary embodiments of this disclosure, exemplary systems, methods, and computer-accessible media may be used to summarize the integrity of treatment using information characterized herein, for example, that the residual plaque load outside the stent is, for example, less than about 50%. For example, EEL measurements, including vessel size and plaque load, may be used to directly guide a PCI plan, such as one recommended by this system, including the locations where the stent should be started and terminated, the diameter and length of the stent, and the diameter and length of the expansion balloon.

[0038] According to further or alternative exemplary embodiments of this disclosure, exemplary systems, methods, and computer-accessible media include and / or may be available procedures and / or user interfaces mounted on a mobile or integrated set of hardware designed to acquire, process, and provide or otherwise display to a physician during a PCI procedure, findings from an intravascular imaging system (e.g., an optical coherence tomography intracoronary imaging system). For example, EEL measurements may be coposition-correlated with several other intravascular sensing parameters, including lumen size, calcium size and location, lipid size and location, stent location, and side branches, and summarized for a clinician user to guide PCI decision-making and confirm the completeness of treatment. In addition, or alternatively, EEL measurements, including plaque load, may be detected along with other intravascular sensing parameters (e.g., capsule thickness, lipids), and may be used to identify the need to treat an area within a vessel. The exemplary systems, methods, and computer-accessible media may, when using these exemplary measurements, recommend progressive treatment with stents, scaffolds, or other methodologies to proactively treat plaques at high risk of future adverse events. Recommendations may be based on machine learning procedures or score-based threshold systems. In some exemplary embodiments of this disclosure, any recommendations provided by the exemplary systems and / or methods may be interpreted and communicated to a user through an existing or custom-trained large-scale language model (LLM), which may also access intermediate features (e.g., hidden features) of any other exemplary procedure pipeline, similarly deployed on the exemplary systems and / or methods.

[0039] In additional or alternative exemplary embodiments of this disclosure, a user-determined clinical guidance threshold may be used to filter EEL measurements, including plaque load. When an EEL measurement exceeds or falls below a threshold, the user may be notified by the user interface, for example, via a color bar, shaded area, text, or other visual indicator. For example, EEL measurements may be presented to the user in a viewport (e.g., multiple viewports) to guide clinical decision-making. Exemplary viewports may include cross-sections, longitudinal views, or other planes of the intravascular modality. Exemplary measurements may also be visually overlaid on other views of the co-aligned modality, including angiography, summary measurement viewports (e.g., risk viewports), computed tomography, etc. Furthermore, or alternatively, EEL measurements may be co-correlated with other intravascular sensing parameters in a display that summarizes insights of importance to the physician. Combined insights may include co-positional correlations of large or small vessel size, high or low plaque load, large or small lumen size, presence or absence and amount of lipids, thin or thick fibrous cap, presence or absence and amount of calcium, and other clinically significant combinations.

[0040] In some exemplary embodiments of this disclosure, a preprocessing step / procedure aimed at spatially aligning neighboring frames may be used first (e.g., using the central position of a partially detected EEL) prior to using multiple neighboring frames to interpolate / extrapolate the position and size of the EEL within the target frame. In some exemplary embodiments of this disclosure, the interpolation or extrapolation step includes 3D topology interpolation. In some exemplary embodiments of this disclosure, the interpolation or extrapolation step includes multiple frames of intravascular imaging data. In some exemplary embodiments of this disclosure, the interpolation or extrapolation step may include multiple frames of intravascular imaging data, while each frame may include only a partial segment of the detected EEL.

[0041] According to further exemplary embodiments of the present disclosure, prior knowledge of the physical and optical properties (e.g., light attenuation, refractive index, etc.) of tissues, blood, and wash media can be used to preprocess input data or to postprocess intermediate or final results.

[0042] In some exemplary embodiments of this disclosure, multimodality data (e.g., OCT and IVUS, OCT, NIRS, IVUS, and NIRS, and / or angiography and / or FFR) can be used to preprocess input data or to postprocess intermediate or final results of a plaque loading estimation procedure, and / or, in some cases, can be used directly for EEL location detection / inference (e.g., 2D detection / inference, e.g., 3D detection / inference). In some exemplary embodiments of this disclosure, plaque composition can be determined from one imaging modality (e.g., angiography), and this information can be used to refine the EEL estimation within an intravascular imaging modality (e.g., OCT). In some exemplary embodiments of this disclosure, plaque composition (e.g., lipid concentration, lipid subtyping, etc.) can be determined from a spectroscopic modality, e.g., NIRS images, and this information can be used to refine the estimation. In another exemplary embodiment of the present disclosure, prior knowledge or inferred knowledge of tissue (e.g., automatically detected tissue) and properties (e.g., mechanical properties, elasticity, shear modulus, stress, strain) may be used to preprocess input data for estimation procedures and / or postprocess intermediate or final results.

[0043] A transfer learning procedure can be used to take EEL measurements from an imaging modality with supplemental EEL visibility and infer the location of the EEL in a different imaging modality. This can be done by training a neural network to learn the relationship between EEL shape and structure using two different imaging modalities, one of which may have lower visibility than the other. The Trans-UNet architecture, with its ability to capture high-level semantic context, is well-suited for this task. Partially visible EELs can be inferred in 360 degrees using only the visible portion of the EEL in the image. This can be done using a highly contextualized neural network trained by learning the shape of the EEL in the presence of other AI-detectable structural features such as lumen and calcium. This technique can be further improved by combining it with image information using multi-channel inputs.

[0044] According to additional exemplary embodiments of the present disclosure, exemplary procedures designed to detect other plaque features (e.g., calcium, lipids, TCFAs, cholesterol) can be used to further refine the local nature of the plaque and to guide the inference of the location and size of the EEL. In some exemplary embodiments of the present disclosure, templates of similar cross-sections from prior data can be used to aid in refinement (e.g., templates from histology that match, e.g., similar cross-sections). In some exemplary embodiments of the present disclosure, plaques are automatically detected and typed based on imaging data (e.g., NIRS data, OCT data, angiography data, etc.).

[0045] In some exemplary embodiments of this disclosure, high-sensitivity OCT (e.g., >100 dB) can be deployed to image arterial walls to greater depths, and the extended depth of imaging allows for improved visualization of the EEL. For example, image processing procedures (which may utilize generative adversarial networks (GANs)) can be applied to OCT images to improve EEL visibility and automated EEL detection (e.g., to improve plaque load measurements). In another embodiment, the image processing procedure may include image denoising (e.g., using custom machine learning-based denoising procedures, machine learning-based diffusion models, etc.) to improve the visualization of images and / or specific features. In another embodiment, in some exemplary variations, preprocessed images for improved image depth visualization are provided to machine learning procedures (e.g., deep learning procedures, e.g., U-Net, e.g., Res-Net, transformer networks, diffusion models) to improve EEL detection (e.g., EEL segmentation). In some exemplary embodiments of this disclosure, a machine-learned or machine-learned procedure (e.g., a regression model) detects (e.g., measures) plaque load (e.g., plaque load percentage, plaque load index, 0-100%, 0-1, etc.) directly from an image (e.g., IVUS image, OCT image, etc.) or a set of images (e.g., a set of angiography images, a set of CCTA images, etc.).

[0046] According to further exemplary embodiments of this disclosure, automated detection procedures (e.g., automated segmentation of EELs, e.g., automated detection and typing of plaques), preprocessing procedures (e.g., inter-image alignment, image denoising, image enhancement, edge filtering, etc.), 3D interpolation methods (e.g., measured from intravascular imaging or angiography data, e.g., automated pre-alignment of frames based on EEL portions, automated pre-alignment of frames based on lumen portions, automated pre-alignment of frames based on EEL portions and lumen morphology, etc.), automated measurement procedures (e.g., automated plaque load measurement), display procedures (e.g., thresholded display, color-encoded display, aligned overlay, etc.), user input procedures (e.g., user-selectable thresholds), and any other novel and beneficial procedures themselves disclosed herein can be applied to any imaging modality, including intravascular ultrasound, computed tomography angiography (CCTA), magnetic resonance imaging (MRI), etc. For example, in some embodiments of the present disclosure, automatic detection and measurement of EEL diameter and / or plaque load estimates derived from OCT can be co-aligned with angiographic images and overlaid on a UI display port (e.g., angiography UI display port). For example, in some embodiments of the present invention, automatic detection and measurement of EEL diameter and / or plaque load estimates derived from IVUS can be co-aligned with angiographic images and overlaid on a UI display port (e.g., angiography UI display port). For example, in some exemplary embodiments of the present disclosure, automatic detection and measurement of EEL diameter and / or plaque load estimates derived from multimodal OCT-NIRS can be co-aligned with lumen diameter or vascular characteristics (e.g., calcium characteristics derived from CCTA) derived from angiography (e.g., CCTA) and overlaid on a UI display port (e.g., also a longitudinal representation of intravascular and / or extravascular data).

[0047] In further exemplary embodiments of this disclosure, automatically detected (e.g., measured) EEL shape (e.g., angle) and size (e.g., diameter) can inform and improve image-derived physiological function measurements (e.g., fractional flow reserve (FFR) measurements, virtual fractional flow reserve (vFFR) measurements). For example, angiography-derived FFR measurements can be improved by incorporating information about EEL diameter and / or plaque load. For example, OCT-derived FFR measurements can be improved by incorporating information about EEL diameter and / or plaque load. In some exemplary embodiments of this disclosure, angiography-derived coronary assessment (ADCA) can be improved using intravascular detected EEL measurements (e.g., angiography-derived measurements such as structural measurements (e.g., stenosis measurements), absolute or relative coronary flow (CF), index of microcirculatory resistance (IMR), fractional flow reserve (CFR), coronary microvascular resistance (CMR), etc.). For example, automatically detected EEL measurements from intravascular OCT and / or IVUS data can be used to recalibrate automatically detected lumen measurements from extravascular data (e.g., lumen measurements calculated by radiographic angiography) based, for example, on the difference or ratio between the automatically detected lumen and the automatically detected EEL from intravascular data.

[0048] (Technical explanation of an example of optimized plaque load measurement) According to additional exemplary embodiments of the present disclosure, exemplary optimized plaque load measurements can be facilitated by exemplary procedures for determining EEL location when boundaries are obscured by lipids, calcium, and other highly optically scattering media. For example, EEL boundaries can be enhanced by OCT interframe co-registration, which uses visible EEL from neighboring frames to infer its location posterior to the plaque (e.g., based on multiple pre-processing steps). Exemplary OCT co-registration can be performed by exemplary procedures for correcting, for example, catheter eccentricity, precession, vascular tortuosity, and NURD. In some exemplary embodiments of the present disclosure, OCT co-registration can be assisted by angiography-detectable lumen morphological characteristics (e.g., tortuosity) and angiography-detectable intravascular catheter location (e.g., radiopaque marker location, imaging location, imaging path, etc.). In some exemplary embodiments of this disclosure, OCT images can be eccentric and / or concentric through a constrained least-squares elliptic fit of arterial boundaries in all frames, based either on OCT data alone or on OCT data referencing angiography-derived information (e.g., imaging pathway).

[0049] In some exemplary embodiments of this disclosure, angular corrections for NURD and precession can be performed using exemplary speckle analysis and polar phase uncorrelated analysis, respectively. For example, EEL boundaries from neighboring frames can be merged with detected EEL boundaries within a given frame using interpolation on smooth low eccentricity curves (e.g., optimization-based interpolation, machine learning-based interpolation, etc.). Furthermore, EEL boundary estimation can be further enhanced, for example, through context-based inference using automatically detected lipids, calcium, and side branches.

[0050] According to various exemplary embodiments of this disclosure, exemplary EEL detection can be trained / informed based on co-aligned IVUS and histological images that show EEL boundary behavior posterior to plaques that would obscure the EEL in optical imaging. In some exemplary embodiments of this disclosure, side branches (e.g., side branches exceeding a certain diameter) are used to modulate the EEL boundary (e.g., based on prior information in a stepwise manner). For example, such automatically detected information can be used to constrain the frames used to infer the EEL location from neighboring frames so that it can be understood that vessel size decreases after significant side branches (e.g., based on Murray's Law). The exemplary procedure can infer the EEL in frames with small side branches, which would give inaccurate measurements if single-frame analysis were used.

[0051] Figure 1 shows an illustration to aid in explaining the plaque load (i.e., plaque load) shown therein. As illustrated in Figure 1, the ratio of the area of ​​the lumen 30 to the area of ​​the EEL 20 can explain the plaque load (e.g., in a specific longitudinal section along the coronary artery). In some exemplary embodiments of this disclosure, the plaque load can be determined or calculated as any ratio or relationship between the lumen area and / or plaque area and the EEL area. In some exemplary embodiments of this disclosure, the plaque load measurement may be a volumetric measurement, using a relationship in volume (3D) instead of an area (2D) relationship. In some exemplary embodiments of this disclosure, a specific plaque risk and / or patient risk may be provided based on a volumetric assessment of the plaque along the length of the affected portion (e.g., total plaque volume).

[0052] Figure 2 illustrates a flowchart or schematic diagram 200 that may be implemented by an exemplary system according to an exemplary embodiment of the present disclosure, which is configured to automatically calculate plaque load measurements based on automatically detected portions of the lumen and EEL in intravascular images and to further display some representation of the automatically detected portions and / or measurements. For example, as shown in Figure 2, in step 202, at least one processor (e.g., a system containing such a processor) can receive synchronized extravascular and intravascular images (e.g., IVOCT images) of the coronary vascular system. In step 204, the processor automatically detects portions of the lumen and EEL. In step 206, the processor may optionally perform processing on the automatically detected portions of the lumen and / or EEL (e.g., interpolation across frames, e.g., constrained interpolation, e.g., alignment) to improve the performance of the automatic detection (e.g., accuracy) (e.g., improve the automatic calculation of the EEL area). In step 208, the processor may calculate or otherwise determine a plaque load measurement based on automatically detected portions that are processed at will. In step 210, the processor may control the display to output or otherwise show a representation of the automatically detected portions and / or a representation of the measurement (for example, to guide treatment).

[0053] Figure 3 shows a set of exemplary illustrations of automatically detected portions of intravascular images associated with plaque load, with and without the presence of plaque (e.g., atherosclerotic plaque with high plaque load and high risk). For comparison purposes, a first vessel section 302 and a second vessel section 304 with automatically detected features (e.g., a lumen portion, e.g., an EEL portion) are shown. In the first vessel section 302, an automatically measured lumen area 306, defined by an automatically detected lumen circumference 308, may be contained within a relatively healthy vessel wall 310. The automatically measured lumen area 306 may further be contained within an automatically detected EEL circumference 312. The automatically detected circumference can be used to automatically measure areas relating to various vessel regions (e.g., the EEL circumference can be used to automatically measure the EEL area). Vessels with sections such as those shown in item 302 may have a low plaque load and therefore a low corresponding risk. In the second vessel cross-section 304, large plaques 314 can be automatically detected and / or measured within the lumen wall 320. The automatically measured plaque load can be calculated based on any combination of the automatically measured or detected plaque 314 (e.g., plaque circumference, plaque area, etc.), the automatically measured lumen area 316, the automatically detected lumen circumference 318, and / or the automatically detected EEL circumference 322.

[0054] Figure 4 illustrates exemplary automated processing (e.g., interpolation) of automatically measured EEL area using automatically detected portions (e.g., contours, e.g., perimeters) of EEL in intravascular images, according to exemplary embodiments of the present disclosure. As shown in Figure 4, an intravascular section 402 is illustrated with automatically detected portions 404 of the EEL (e.g., visible portions of the EEL in the image) and automatically inferred portions 406 of the EEL (e.g., invisible portions of the EEL in the image). In some exemplary embodiments of the present disclosure, visible portions of EEL in intravascular images (e.g., IVOCT images) can be automatically detected (e.g., segmented) and automatically interpolated laterally (e.g., interpolated along the image axes, e.g., in 2D) using a machine learning-trained detection network (e.g., a convolutional neural network) to further infer the EEL contour (e.g., to improve the accuracy of the EEL area measurement). In some exemplary embodiments of this disclosure, interpolation (e.g., smoothing, averaging, etc.) may be performed longitudinally and / or transversely (e.g., interpolated in 3D) to further infer the EEL contour and / or to provide volumetric plaque load calculations (e.g., to improve the accuracy of EEL area measurement values). In some exemplary embodiments of this disclosure, plaque load measurement may be performed only when the visible and automatically detected portions of the EEL are adequate (e.g., above a threshold, detected in sections covering above 30 degrees, detected in sections covering above 60 degrees, detected in sections covering above 90 degrees, detected in sections covering above 180 degrees, detected in sections covering above 220 degrees, etc.). For example, if a small portion of the EEL is detected, for example, on the opposite side of a cross section, the threshold may be applied based on an interpolation confidence index (e.g., constrained interpolation, eccentricity-constrained interpolation, etc.). Figure 4 shows that three portions of the EEL are detected, and interpolation may be performed based on these three portions. In one exemplary case, interpolation may be performed on only one or two of the automatically detected parts based on an automatically assessed quality index (e.g., confidence index) for each specific part.

[0055] Figure 5 illustrates exemplary graphs 502 and 504, which provide exemplary results obtained when automatically detected EEL portions are aligned based on automatically detected characteristics in intravascular images, according to exemplary embodiments of the present disclosure. As shown in Figure 5, the first graph 502 displays eccentric EEL portions automatically detected from an intravascular imaging dataset, according to exemplary embodiments of the present disclosure. The second graph 504 displays the aligned EEL portions based on a processing and aligning method (e.g., an automated processing and aligning method), according to exemplary embodiments of the present disclosure. For example, in each intravascular image, the automatically detected portions 506 of the vascular lumen (not shown in Figure 5) and / or EEL, provided on the first graph 502, can be processed and grouped into segments. Such exemplary segments 510 are depicted in grayscale in graphs 502 and 504, respectively. Each portion or segment can be stratified based on an automatically measured vascular center 508, which may be derived based on automatically detected lumen portions and / or EEL portions and / or plaque portions. In some exemplary embodiments of the present disclosure, lumen and / or EEL inference may include interpolation after vascular stratification is applied (e.g., in 2D) to improve stratification performance, for example. In some exemplary embodiments of the present disclosure, lumen and / or EEL inference may include interpolation after vascular stratification is applied (e.g., interpolation in 3D) to improve interpolation performance and / or downstream measurement performance (e.g., plaque load measurement performance).

[0056] Figure 6 illustrates ND processing and interpolation of EELs based on automatically detected characteristics in intravascular images, according to an exemplary embodiment of the present disclosure. For example, an automatically detected portion of an EEL spanning several images, starting in a first image 602 and ending in a second image 604, can utilize composite processing (e.g., symmetry) and interpolation to improve automatic EEL inference in other frames, such as frames with poor EEL visibility 606. Figure 6 illustrates a dark solid line showing an automatically detected portion 610 of an EEL in several intravascular images, a dashed line showing an automatically inferred portion 614 of an EEL, and a light dotted line showing an automatically inferred vascular center axis 612 based on an image-based automatically inferred vascular center 608. As shown in Figure 6, an EEL may be visible in one frame at a single circumferential position but invisible at the same circumferential position in subsequent frames. For example, an EEL may be completely invisible overall within a single frame.

[0057] The systems, methods, and computer-accessible media according to exemplary embodiments of this disclosure can be used to address such invisible portions by automatic 3D alignment (e.g., weighted alignment) and interpolation (e.g., machine learning-based interpolation) that take into account vascular morphology (e.g., plaque locations, e.g., side branch locations, e.g., stent locations) in relation to the position of the imaging probe. Figure 6 shows that automatically detected portions may be offset longitudinally in the lateral direction, and this offset may be caused by the probe position within the vessel, or by vascular tortuosity, or by a combination of such factors and many others (e.g., luminal protrusions caused by plaque, e.g., myocardial bridges). As described herein, automatically processing and alignment of automatically detected features in a longitudinal context (e.g., 3D context) can facilitate improved vascular feature detection (e.g., EEL location detection) and thus improve the accuracy of vascular measurements (e.g., plaque load measurements).

[0058] Figure 7 illustrates multidimensional (ND) processing (e.g., 3D, 4D, where one dimension may be time) and interpolation of an EEL using optimized and / or machine-learned ellipticity constraints. In exemplary embodiments of this disclosure, as shown in Figure 7, an automatically detected portion 702 of an EEL can be interpolated using a high ellipticity constraint 706, a low ellipticity constraint 710, and an optimized interpolation constraint 708. The optimized interpolation constraint may be optimal based on vascular specific morphology, patient specific based on patient history, or coronary artery specific based on imaging from other modalities (e.g., radiographic angiography, CTA, etc.). In some exemplary embodiments of this disclosure, the optimized interpolation constraint is or may include a machine-learned constraint. In some embodiments of this disclosure, the machine-learned interpolation constraint may be trained using images and / or annotations from a secondary imaging modality (e.g., another intravascular imaging modality, histological imaging modality, etc.). For example, during training, at least partially aligned data from a secondary imaging modality (e.g., annotations) can be used as a ground truth target for the intravascular imaging modality (e.g., the second imaging modality may not be required for prediction).

[0059] According to some embodiments of the present disclosure, the optimized interpolation constraint may be a constraint based on physical information (e.g., based on the mechanical properties of a human blood vessel). In further exemplary embodiments of the present disclosure, the optimized interpolation constraint may be provided with information based on automatically detected vascular objects such as adjacent images in the longitudinal direction or plaque (e.g., plaque with known hardness). For example, in some exemplary embodiments of the present disclosure, an automated plaque detection and / or characterization procedure may be used to provide the type of plaque in order to adjust the constraint based on whether the plaque is lipid-based or calcium-based. In some additional exemplary embodiments of the present disclosure, the shape of the lumen that is automatically detected may also be used to provide information to the interpolation constraint. For example, when the lumen wall protrudes into the lumen (e.g., eccentrically), the constraint may be adjusted to promote a higher ellipticity resulting from increased pressure on the EEL in its circumferential direction. In additional exemplary embodiments of the present disclosure, temporal data (e.g., 2D image data with a temporal dimension, 3D image data with a temporal dimension, etc.) may be used to improve automated EEL detection and interpolation. For example, multiple longitudinal image pullbacks from different points in time (e.g., sequential, pre-PCI and post-PCI, etc.) can be used to improve EEL detection and inference, and / or to measure temporal characteristics of EEL (e.g., biomechanical properties) and / or plaque load (e.g., after alignment).

[0060] Figure 8 shows exemplary graphs and illustrations of ND processing and interpolation of EEL diameter in a continuous vascular region defined by automatically detected vascular objects, according to exemplary embodiments of the present disclosure. As illustrated in Figure 8, vessel 802 may comprise automatically detected side branch vessels 804, 806. Interpolation of automatically detected and / or inferred EEL and / or measured EEL diameter can only be interpolated between side branches in a segment. In exemplary graph 818 of EEL diameter versus pullback distance (e.g., from intravascular imaging pullback), the EEL diameter may decrease monotonically along the length of the artery and also decrease in a stepwise manner at the location of major side branches, and this effect may make interpolation more inaccurate if not considered. In some exemplary embodiments of the present invention, a first interpolation segment 808 may extend upward to a first lateral branch 804, a second segment 810 may extend from the first lateral branch to a second lateral branch 806, and a third segment 812 may extend beyond the second lateral branch. These segments 810, 812 can each be interpolated over them. Exemplary interpolation in these embodiments may refer to interpolating an automatically detected portion of a vascular structure (e.g., lumen, EEL, etc.), which may also refer to interpolating over an automatically measured value such as an automatically measured EEL diameter 616 (e.g., if the segments are interpolated differently over it). According to some exemplary embodiments of the present disclosure, the designated segments may all be ignored together during segmented (e.g., stepwise) interpolation, such as the location of the detected lateral branch 614.

[0061] Figure 9 shows a set of exemplary illustrations of an exemplary multi-panel user interface (UI) for PCI planning and post-PCI assessment according to exemplary embodiments of the present invention. For example, a multi-panel UI for displaying automated output during PCI planning and post-PCI assessment is illustrated. The exemplary UI may include an angiography display window 902, an intravascular window 904, and a longitudinal representation window 906. Coronary arteries 908 are provided within the angiography display window 902 and displayed together with a longitudinal EEL diameter overlay 910 alignment representation and a alignment threshold representation overlay 912 (e.g., depicting locations of high plaque load above a threshold). The overlays 910, 912 can be automatically generated from intravascular imaging data and aligned with each other according to exemplary embodiments of the present disclosure.

[0062] As shown in Figure 9, a cross-section of a coronary vessel 920 provided within the intravascular imaging window 904 can be displayed along with exemplary representations of an automatically detected and optionally interpolated lumen contour 914, an automatically detected and optionally interpolated EEL contour 916, and an automatically detected and optionally interpolated plaque contour 918. A longitudinal representation of the EEL diameter 922 is provided within the longitudinal representation window 906 and may also include illustrations of the lumen diameter 924 and pressure gradients (e.g., image-derived pressure gradients, angiography-derived FFR, etc.) 928. A locally thresholded plaque load representation 926 (also included within window 906) can be aligned along the longitudinal axis with a thresholded plaque load representation 912 in the angiography window 902. Although not shown in Figure 9, other automatically detected objects from either angiographic or intravascular data, such as stent locations, lipid locations, calcium locations, macrophage locations, calcium crystal locations, neointima locations, thin capsule locations, weak capsule locations, necrotic core locations, fibrous atheroma locations, and malabposition locations, may also be displayed in any of the UI windows described. Any of the described objects may also be displayed only when a thresholded measurement of their presence is reached.

[0063] Figure 9 is illustrative only, and it should be understood that the overlays displayed on any of the UI windows may take on any shape or form with respect to any automatically detected, processed, or measured structure according to the exemplary embodiments of this disclosure. Any of the windows described may also be occupied by a 3D visualization of a longitudinal structure, including an EEL overlay and a thresholded plaque load measurement overlay exceeding a predefined value.

[0064] Figure 10 illustrates a system diagram of an intravascular imaging system 1000 according to an exemplary embodiment of the present disclosure. For example, the exemplary imaging system 1000 may include a display configured to acquire intravascular data and to align the intravascular data with extravascular data. In some exemplary cases, the intravascular data may include one or more intravascular images of a blood vessel. In one exemplary case, the extravascular data may include one or more extravascular images (e.g., radiographic angiography, MRI, etc.) of vascular shape, physiological function, anatomical structure, or any combination thereof. In one exemplary case, the imaging system may include a computer system 1006 for processing intravascular, extravascular, user interaction, or any combination thereof.

[0065] User interaction data may include user input data to the imaging system 1000, which may include patient information, landmark designation, system operating mode selection, image processing functions, or any combination thereof. In one exemplary case, the user may input data to the imaging system 1000 using a mouse and / or keyboard electrically coupled to the computer system 1006. The user may visualize a view (i.e., a user interface) configured for inputting data to the system via a first monitor 1002 and / or a second monitor 1004. In one exemplary case, the first monitor 1002 and / or the second monitor 1004 may include a touchscreen interface and keyboard for actions performed on intravascular and / or extravascular data, such as interaction, acquisition, or any combination thereof. In one exemplary case, user interaction data may include data resulting from the user interacting with extravascular and / or intravascular data (e.g., rotation, zoom in, contrast adjustment, brightness adjustment, distance measurement, etc.). The computer system 1006 includes, or can communicate with, an electronic display (e.g., a first monitor 1002 and / or a second monitor 1004) having one or more view configurations (i.e., user interfaces (UIs)) as described elsewhere herein, for viewing intravascular data, extravascular data, aligned intravascular and extravascular data and / or combinations thereof, or any combination thereof.

[0066] In an exemplary case, the computer system (1006) may include an input interface 1005, which may include one or more input points and / or ports electrically coupled to the computer system 1006. The input interface 1005 may receive one or more data and / or streams of data from one or more imaging systems. For example, the input interface 1005 may receive X-ray angiography data, which the computer system 1006 can then align with intravascular data. In an exemplary case, the input interface 1005 may receive data from one or more medical devices, such as angiography-derived physiological functions, MRI, computed tomography, spatial position, intravascular sensors (e.g., intravascular physiological functions), or any combination thereof, to be displayed and / or aligned with intravascular data. In some exemplary cases, the input interface 1005 may receive data to be aligned to extravascular data wirelessly via a wireless communication platform such as ad-hoc Wi-Fi, Bluetooth®, radio frequency, or any combination thereof, as described elsewhere in this specification.

[0067] In some exemplary embodiments of this disclosure, the computer system 1006 may process data using one or more processors. In some exemplary cases, the one or more processors may comprise one or more graphics processing units, integrated circuit processors, or any combination thereof. Graphics processing units facilitate the processing of complex, large datasets due to their highly parallelized processor architecture. For example, processing data using one or more graphics processing units provides the system with the ability to align intravascular data with real-time streams of extravascular data, which would otherwise not be achievable using conventional multi-core processors.

[0068] In some additional exemplary embodiments of this disclosure, the computer system 1006 may be configured to process intravascular and extravascular data and / or images. The computer system 1006 may comprise a central processing unit and / or graphics processing unit (CPU and / or GPU, also herein referred to as “processor” and “computer processor”), which may be a single-core or multi-core processor or multiple processors for parallel processing. The computer system 1006 may further comprise memory or memory locations (e.g., random-access memory, read-only memory, flash memory), electronic storage units (e.g., hard disks), communication interfaces for communicating with one or more other devices (e.g., network adapters), and peripheral devices such as caches, other memory, data storage devices, and / or electronic display adapters. The memory, storage units, communication interfaces, and peripheral devices (e.g., mouse, keyboard, etc.) may communicate with the CPU and / or GPU through a communication bus (solid line), such as a motherboard. The storage units may be data storage units (or data repositories) for storing data. Computer system 1006 can be operationally coupled to a computer network ("Network") with the help of a communication interface 1408. The Network may be the Internet, the Internet and / or an extranet, or an intranet and / or extranet communicating with the Internet. In some exemplary cases, the Network may be a telecommunications and / or data network. The Network may include one or more computer servers that can facilitate distributed computing, such as cloud computing. In some exemplary cases, the Network may implement a peer-to-peer network with the help of computer system 1006 that can facilitate devices coupled to computer system 1006 acting as clients or servers.

[0069] The CPU and / or GPU can execute a sequence of machine-readable instructions, which may be embodied in a program or software. Instructions can be directed to the CPU and / or GPU, which may then be programmed or otherwise configured to acquire and / or process data produced by an imaging system, as described elsewhere herein.

[0070] In some exemplary embodiments of this disclosure, a central processing unit and / or graphics processing unit of a computer system (1006) may execute machine-executable or machine-readable code, which may be provided in software form, to transfer data generated by the imaging system to a network and / or cloud for further processing, classification, data clustering, or any combination thereof. In some exemplary cases, the data may comprise intravascular and / or extravascular data, as described elsewhere in this specification. In some exemplary cases, the data may comprise image pixel data. In some exemplary cases, the pixel data may comprise image pixel data from optical coherence tomography, radiographic angiography, computed tomography, intravascular ultrasound, spectroscopy, MRI, or any combination thereof.

[0071] In some further exemplary embodiments of this disclosure, the CPU and / or GPU may be part of a circuit such as an integrated circuit. One or more other components of the exemplary system may be included in the circuit. In yet further exemplary embodiments of this disclosure, the circuit may comprise an application-specific integrated circuit (ASIC).

[0072] The storage unit may store files such as drivers, libraries, and saved programs. The storage unit may store data and / or images obtained from radiographic angiography, optical coherence tomography, intravascular ultrasound, near-infrared spectroscopy, photoacoustics, or any combination thereof. In an exemplary case, intravascular and / or extravascular data and / or images may be stored in a cloud, a medical system electronic medical record (e.g., EPIC), or any combination thereof. In an exemplary case, the computer system 1006 may include one or more additional data storage units located outside the computer system 1006, such as on a remote server that communicates with the computer system 1006 via an intranet or the internet.

[0073] In one exemplary case, the imaging system 1000 communicates electrically and / or optically with the imaging probe actuator 1010, and an imaging probe 1012, as seen in the imaging system 1000, can communicate electrically and / or optically with the imaging probe actuator 1010 through one or more electrical and / or optical communication wires 1008. In one exemplary case, the imaging probe 1012 can be removably coupled to the imaging probe actuator 1010 such that a first imaging probe can be removed from the imaging probe actuator and replaced with a second imaging probe.

[0074] According to further exemplary embodiments of the present disclosure, the imaging probe may comprise an intravascular imaging probe. The intravascular imaging probe may comprise an imaging probe for optical coherence tomography, intravascular ultrasound, reflectivity, photoacoustics, near-infrared spectroscopy, fluorescence, or any combination thereof. In some exemplary cases, the imaging probe may acquire, collect, and / or detect intravascular data from the inner lumen and / or body of a blood vessel. In some exemplary cases, the intravascular data may comprise two-dimensional (e.g., circular section data) and / or volumetric intravascular data (i.e., one or more two-dimensional circular section data as a function of the length of the optical axis of the imaging probe). In some exemplary cases, the imaging probe may comprise one or more radiopaque markers and / or marks that can be visualized on an extravascular imaging modality, e.g., radiographic angiography, computed tomography, MRI, or any combination thereof.

[0075] In additional exemplary embodiments of this disclosure, the imaging probe actuator 1010 can rotate and / or translate the imaging probe 1012 to acquire two- and / or three-dimensional intravascular data sets. In some exemplary cases, the probe can be rotated by a stepper motor, a DC brushless motor, or any combination thereof, coupled to an optical rotary joint. In some exemplary embodiments, the imaging probe actuator 1010 can translate the imaging probe 1012 using a stage, the stage may comprise a linear and / or planar translation stage. The stage translation and rotation of the imaging probe actuator 1010 can be set and / or adjusted by the user via one or more interfaces of the imaging system 1000, as described elsewhere in this specification. In some exemplary embodiments, the stage translation and rotation of the imaging probe actuator 1010 can be determined and / or set by the system based on pre-set standard values ​​relating to a particular type of imaging procedure or frequently used settings.

[0076] Exemplary embodiments of the systems, computer-accessible media, and methods provided herein, such as computer system 1006, can be embodied in programming. Various exemplary aspects of the Art can typically be considered “products” or “manufactured goods” in the form of machine (or processor) executable code and / or associated data, carried on or embodied on a certain type of machine-readable medium. Machine-executable code can be stored on electronic storage units such as memory (e.g., read-only memory, random-access memory, flash memory) or hard disks. The “storage” type medium can include any tangible memory of a computer, processor, equivalent, or associated module thereof, such as various semiconductor memories, tape drives, disk drives, and equivalents, which can provide non-transient storage for software programs from time to time. All or part of the software may be communicated from time to time through the Internet or various other telecommunication networks. Such communication can facilitate, for example, the loading of software from one computer or processor to another, such as from a management server or host computer to an application server computer platform. Therefore, other types of media that can carry software elements include optical, electrical, and electromagnetic waves, such as those used across physical interfaces between local devices, through wired and optical terrestrial networks, and via various wireless links. Physical elements that carry such waves, such as wired or wireless links, optical links, or equivalents, can also be considered media that carry software. Unless limited to non-transient tangible “storage” media as used herein, the terms computer or machine “readable medium,” etc., refer to any medium involved in providing instructions to a processor for execution.

[0077] Therefore, machine-readable media such as computer executable code may take many forms, but are not limited to tangible storage media, carrier media, or physical transmission media. Non-volatile storage media may include optical or magnetic disks, such as any storage device in any computer or equivalent, which may be used to implement databases, etc. Volatile storage media may include dynamic memory, such as the main memory of such a computer platform. Tangible transmission media include coaxial cables, copper wires, and optical fibers, including wires that form buses within a computer system. Carrier transmission media may take the form of electrical or electromagnetic signals or acoustic or optical waves, such as those generated during radio frequency (RF) and infrared (IR) data communications. Common forms of computer-readable media can therefore include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, DVDs or DVD-ROMs, any other optical media, punched card paper tapes, any other physical storage media with perforation patterns, RAM, ROMs, PROMs and EPROMs, flash EPROMs, any other memory chips or cartridges, carriers for transporting data or instructions, cables or links for transporting such carriers, or any other media from which a computer can read programming code and / or data. Many of these forms of computer-readable media can be involved in transporting one or more sequences of one or more instructions to a processor for execution.

[0078] In some further exemplary embodiments of the Disclosure, the exemplary system 1100 may comprise a computer system (1006) suitable for implementing machine learning procedures and / or predictive models configured to analyze, process, segment, and / or label extravascular and / or intravascular data collected by an imaging system 1000, an imaging probe 1012, and an imaging probe actuator 1010, as described elsewhere in the Spec. In some exemplary cases, one or more intravascular and / or extravascular images may be generated from the intravascular and / or extravascular data. In additional exemplary embodiments of the Disclosure, the predictive model, e.g., a machine learning model and / or machine learning procedure, may perform operations such as analysis, extraction, summarization, reduction, prediction, processing, classification, segmentation, or any combination thereof, on the intravascular and / or extravascular data. In some exemplary embodiments of the Disclosure, the systems disclosed herein may implement one or more machine learning procedures and / or models to identify, classify, process, and / or segment regions of interest in the intravascular and / or extravascular data.

[0079] In some exemplary cases, one or more categories and / or features of data may be provided to one or more therapeutic parameter machine learning models and / or procedures to determine a proposed treatment and / or therapeutic parameters (e.g., the type of stent to be placed and the spatially best location for stent placement to achieve the clinical efficacy of the treatment). One or more therapeutic parameter machine learning models may be trained with prior features and corresponding therapeutic efficacy (i.e., whether any complications persisted after a clinical intervention using the system) to generate one or more trained therapeutic parameter machine learning models to predict effective treatment. The spatial orientation of labeled features and their relationships to each other may be other features determined and considered by the therapeutic parameter machine learning models. In some exemplary cases, one or more categories and / or features of data relating to extravascular data may comprise background data, healthy vascular morphology, stenotic vascular morphology, or occluded vessels. In some exemplary cases, one or more categories of data relating to intravascular data may comprise epithelial vascular tissue, intimal vascular tissue, adventitia vascular tissue, plaque within vascular tissue, EEL, plaque load, fragile plaque within vascular tissue, or any combination thereof. In some exemplary cases, one or more categories and / or features of intravascular data may have a spectroscopic (e.g., near-infrared) signature of intravascular vascular tissue. For example, one or more categories and / or features may classify the composition of vascular plaque based on their spectroscopic signature. In some exemplary cases, one or more categories may have a spectroscopic signature of calcium or lipids. In some exemplary cases, machine learning models and / or procedures may preprocess intravascular and / or extravascular data prior to classifying the features of the data. In some exemplary cases, preprocessing intravascular and / or extravascular data may involve mathematical manipulation of the data, such as denoising, smoothing, averaging, sharpening, brightness and / or contrast adjustment, or any combination thereof.In some exemplary cases, features and / or categories of intravascular and / or extravascular data can be extracted without preprocessing steps / procedures.

[0080] In some exemplary cases, machine learning procedures can extract and / or derive relationships between features where conventional statistical techniques may be insufficient. In some exemplary cases, machine learning procedures can be used in conjunction with conventional statistical techniques. In some exemplary cases, conventional statistical techniques can provide pre-processed features to machine learning procedures.

[0081] In some exemplary embodiments of this disclosure, any number of features can be classified by a machine learning procedure. The machine learning procedure may classify at least one feature. In some exemplary cases, the feature set may include about 1 to 5 features. In some exemplary cases, the feature set may include about 5 to 10 features. In some exemplary cases, the feature set may include about 10 to 50 features.

[0082] In some exemplary embodiments of this disclosure, the machine learning procedure may be, for example, an unsupervised learning procedure, a supervised learning procedure, or a combination thereof. Unsupervised learning procedures may include, for example, clustering, hierarchical clustering, k-means clustering, mixed models, DBSCAN, OPTICS procedure, VoxelMorph procedure, anomaly detection, local outlier factorization, neural networks, autoencoders, deep belief networks, Hebb learning, generative adversarial networks, self-organized maps, expectation maximization (EM) procedures, methods of moments, blind source separation techniques, principal component analysis, independent component analysis, non-negative matrix factorization, singular value decomposition, or a combination thereof. Supervised learning procedures may include, for example, support vector machines, linear regression, logistic regression, linear discriminant analysis, decision trees, k-nearest neighbor procedures, neural networks, similarity learning, or a combination thereof. In some exemplary embodiments of this disclosure, the machine learning procedure may comprise a deep neural network (DNN). The deep neural network may comprise a convolutional neural network (CNN). CNNs can be, for example, U-Net, ImageNet, LeNet-5, AlexNet, ZFNet, GoogleNet, VGGNet, ResNet18, or ResNet. Other neural networks can be, for example, deep feedforward neural networks, recurrent neural networks, LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), autoencoders, variational autoencoders, adversarial autoencoders, denoising autoencoders, sparse autoencoders, Boltzmann machines, RBMs (Restricted Beam Networks), deep belief networks, generative adversarial networks (GANs), deep residual networks, capsule networks, or attention / transformer networks.

[0083] In some illustrative examples, machine learning models may include clustering, scalar vector machines, kernel SVMs, linear discriminant analysis, quadratic discriminant analysis, neighboring component analysis, manifold learning, convolutional neural networks, reinforcement learning, random forests, naive Bayes, Gaussian mixtures, hidden Markov models, Monte Carlo, restricted Boltzmann machines, linear regression, or any combination thereof.

[0084] In some exemplary cases, machine learning procedures can include ensemble learning procedures such as bagging, boosting, and stacking. These machine learning procedures can be applied individually to multiple features being extracted.

[0085] In some exemplary embodiments of this disclosure, the exemplary system and / or method may apply one or more machine learning procedures and / or ensembles of machine learning procedures.

[0086] In some exemplary embodiments of this disclosure, the machine learning procedure may include and / or utilize a variety of parameters. These parameters may include, for example, the learning rate, small batch size, the number of reference time points to be trained, momentum, learning weight decay rate, or neural network layers.

[0087] Figure 11 shows a block diagram of another exemplary embodiment of the system according to the present disclosure. For example, the exemplary procedures according to the present disclosure described herein can be carried out by a processing array and / or computing array (e.g., a computer hardware array) 705. Such a processing / computing array 1105 is, for example, a computer / processor 1110, which, in whole or in part, includes, for example, one or more microprocessors and can use instructions stored on a computer-accessible medium (e.g., RAM, ROM, hard drive, or other storage device).

[0088] As shown in Figure 11, for example, a computer-accessible medium 1115 (for example, a storage device such as a hard disk, floppy disk, memory stick, CD-ROM, RAM, ROM, etc., or a collection thereof, as described above herein) may be provided (for example, to communicate with the processing array 1105). The computer-accessible medium 1115 may contain executable instructions 1120 thereon. In addition, or alternatively, a storage array 1125 may be provided separately from the computer-accessible medium 1115, which may provide instructions to the processing array 1105 to configure the processing array to perform certain exemplary procedures, processes, and methods, as described above herein.

[0089] Furthermore, the exemplary processing array 1105 comprises, or may include, input / output ports 1135, which may include, for example, wired networks, wireless networks, the internet, intranets, data acquisition probes, sensors, etc. As shown in Figure 11, the exemplary processing array 1105 may communicate with an exemplary display array 1130, which may be a touchscreen, configured for inputting information into the processing array, in addition to outputting information from the processing array, according to an exemplary embodiment of the present disclosure. Furthermore, the exemplary display array 1130 and / or storage array 1125 may be used to display and / or store data in a user-accessible format and / or user-readable format.

[0090] The foregoing merely illustrates the principles of the present disclosure. Various modifications and alterations of the embodiments described will be obvious to those skilled in the art, taking into account the teachings herein. Therefore, it should be understood that it will be possible for those skilled in the art to devise numerous systems, arrangements, and procedures, not expressly shown or described herein, that embody the principles of the present disclosure and thus may be in the spirit and scope of the present disclosure. Various different exemplary embodiments may be used together and interchangeably with each other, so as should be understood by those skilled in the art. In addition, certain terms used in the present disclosure, including, for example, data and information, in its specification, drawings, and claims, may be used synonymously in certain cases. It should be understood that there may be cases where these words and / or other words that may be synonymous with each other may be used synonymously herein, but where such words may not be intended to be used synonymously. Furthermore, to the extent that prior art knowledge is not expressly incorporated herein by reference above, this is expressly incorporated herein as a whole. All referenced publications are incorporated herein as a whole by reference.

Claims

1. It is a device, An imaging catheter configured to acquire one or more intravascular images, At least one computer processing configuration and Equipped with, The aforementioned at least one computer processing configuration is a) Automatically detect one or more first parts of the external elastic plate, b) Automatically inferring or determining one or more second parts of the external elastic plate, c) Automatically determining at least one measurement of the diameter of the outer elastic plate based on the first part and the one or more second parts. A device configured to perform the following actions.

2. The apparatus according to claim 1, wherein the at least one computer processing configuration is configured to automatically estimate the plaque load from at least one of the one or more first parts or the one or more second parts.

3. The apparatus according to claim 2, further comprising a display device configured to display a representation of the external elastic plate or the plaque load based on a viewport in the user interface.

4. The apparatus according to claim 3, wherein the display device displays the expression in response to the plaque load reaching a predetermined threshold.

5. The apparatus according to claim 1, wherein the one or more intravascular images are optical coherence tomography (OCT) images, and the OCT images are received from an OCT system having a sensitivity of approximately 100 dB or more.

6. The apparatus according to claim 5, wherein the one or more second parts are determined or inferred using a machine learning procedure.

7. The apparatus according to claim 6, wherein the machine learning procedure determines or infers one or more invisible portions by interpolation from neighboring frames.

8. A method for determining or providing one or more measurements of an arterial wall, To obtain an image of at least one segment of the arterial wall, Automatically detect one or more visible portions of the extraarterial elastic lamina of at least one of the aforementioned sections, Based on the one or more automatically detected visible portions, one or more invisible portions of the arterial external elastic lamina are automatically determined or estimated. The visible and invisible portions are automatically combined to provide one or more measurement values ​​of the external elastic plate. Methods that include...

9. The method according to claim 8, wherein the one or more invisible portions are determined or estimated using a machine learning procedure.

10. The method according to claim 9, wherein the machine learning procedure determines or estimates one or more invisible portions by interpolation from neighboring frames.

11. The method according to claim 8, further comprising providing one or more further measurements of plaque load.

12. The method according to claim 11, further comprising displaying a representation of at least one of the external elastic plate or the plaque load on a viewport of the user interface.

13. The method according to claim 12, wherein the expression is displayed in response to reaching a predetermined threshold of the plaque load.

14. The method according to claim 1, wherein the intravascular imaging is performed using an optical coherence tomography (OCT) system having a sensitivity of more than 100 dB.

15. It is a device, - An imaging catheter configured to acquire an image of at least one segment of the arterial wall, - At least one computer processing configuration and Equipped with, The aforementioned at least one computer processing configuration is i. Automatically detecting one or more first portions of the extraarterial elastic lamina of at least one section, ii. Automatically detect one or more second portions of the ductus arteriosus lumen, iii. Determining or estimating the plaque load by processing the one or more first portions and the one or more second portions that have been detected. A device configured to perform the following actions.

16. The apparatus according to claim 15, further comprising a display device configured to display a representation of the plaque load on a viewport of the user interface.

17. The apparatus according to claim 16, wherein the expression is displayed when a predetermined threshold of the plaque load is reached.

18. A method for determining or providing one or more measurements of an arterial wall, - To obtain an image of at least one segment of the arterial wall, - Automatically detect one or more first portions of the extraarterial elastic lamina of at least one section, - Automatically detect one or more second portions of the arterial wall, - Determining or estimating the plaque load by processing the one or more first portions and the one or more second portions detected. Methods that include...

19. The method according to claim 18, further comprising displaying a representation of the plaque load on a viewport of the user interface.

20. The method according to claim 19, wherein the expression is displayed when a predetermined threshold of the plaque load is reached.

21. A method for determining or providing one or more measurements of a coronary artery, - To obtain an image of at least one portion of the coronary artery, - Using a machine learning-based procedure, directly determine or estimate at least one measurement of plaque load based on one or more features or information relating to at least one portion of the image. Methods that include...

22. It is a device, - An imaging catheter configured to acquire an image of at least one segment of the coronary artery, - At least one computer processing configuration configured to directly determine or estimate at least one measurement of plaque load based on one or more features or information relating to at least one portion of the image, using a machine learning-based procedure, and A device equipped with the following features.

23. A method for determining or providing one or more measurements of a coronary artery, - To obtain an image of at least one portion of the coronary artery, - Using a machine learning-based procedure, determine or estimate the position of the external elastic plate within at least one portion of the image. Methods that include...

24. The method according to claim 23, wherein the image is generated using a first imaging modality, and the method further comprises training the machine-learned procedure to determine or estimate the position of the external elastic plate using alignment data from a second imaging modality different from the first imaging modality.

25. The method according to claim 24, wherein the alignment data has a lower resolution than that of the image from the first imaging modality.

26. The method according to claim 23, wherein the alignment data had a lower resolution than that of the first imaging modality.

27. The method according to claim 23, wherein the first resolution is histology or IVUS.

28. The method according to claim 23, wherein the machine learning procedure determines or estimates the position of the external elastic plate in an area where the external elastic plate is invisible in the image.

29. The method according to claim 20, wherein the machine learning procedure generates further images adjacent to the image.

30. It is a device, - An imaging catheter configured to acquire an image of at least one segment of the coronary artery, - A computer processing configuration configured to directly determine or estimate the position of an external elastic plate within at least one portion of the image using a machine learning-based procedure, and A device equipped with the following features.

31. The apparatus according to claim 30, wherein the image is generated using a first imaging modality, and the machine-trained procedure is trained to determine or estimate the position of the external elastic plate using alignment data from a second imaging modality different from the first imaging modality.

32. The apparatus according to claim 31, wherein the alignment data has a lower resolution than that of the image from the first imaging modality.

33. The apparatus according to claim 30, wherein the alignment data had a lower resolution than that of the first imaging modality.

34. The apparatus according to claim 30, wherein the first resolution is histology or IVUS.

35. The apparatus according to claim 30, wherein the machine learning procedure determines or estimates the position of the external elastic plate in an area where the external elastic plate is invisible in the image.

Images