Longitudinal display of coronary artery calcium burden

Abstract

The present disclosure provides systems and methods for receiving OCT or IVUS image data frames and outputting one or more representations of a vessel segment. The image data frames may be stretched and / or aligned using various windowing, binning, or alignment features. Arterial features, such as calcium load, may be detected in each of the image data frames. The arterial features may be scored. The score may be stent malabsorption risk. The representation may include a representation of the arterial features and their respective scores. The representation may be color-coded.

Description

Technical Field

[0002] , , , [Figure 1A] , ,

[0004] , , ,

[0003] , , ,

[0001] [Cross - Reference to Related Applications] This application claims the benefit of the filing date of U.S. Provisional Patent Application No. 62 / 883,066, entitled "LONGITUDINAL DISPLAY OF CORONARY ARTERY CALCIUM BURDEN", filed on August 5, 2019, the disclosure of which is incorporated herein by reference in its entirety.

Background Art

[0002] Generally, coronary plaques are fibrous, lipidic, calcified, thrombotic, or some combination thereof. Calcified plaques can pose a particular risk to treatment success because they can prevent a balloon from fully expanding during stent implantation or angioplasty. Therefore, it is extremely important for physicians to identify calcium and appropriately treat plaques. Generally, optical coherence tomography (OCT) is particularly useful for identifying plaque compositions in each cross - sectional frame. OCT and other imaging modalities can be used to evaluate various arterial characteristics.

[0003] [[ID=S17]] The calcification of coronary plaques can prevent the stent from fully expanding. A poorly expanded stent increases the risk of in - stent restenosis and the need for future treatments. Therefore, it is very important for cardiologists involved in interventions to recognize this risk and adapt treatment strategies accordingly. However, current standard medical practices and the state of the art do not provide physicians with appropriate tools for quantifying the risk. Angiography and OCT are useful for visualizing calcium deposits, but there are manual rules of thumb for estimating the risk of poor expansion, and there is still a risk of insufficient quantification and evaluation.

Brief Description of the Drawings

[0004] [Figure 1A] An exemplary system according to aspects of the present disclosure. [Figure 1B] This is an example of multiple interface components according to the aspects of this disclosure. [Figure 2A] This is an existing example of a scoring methodology. [Figure 2B] Figure 2A is an illustrative graphic representation of the scoring methodology. [Figure 2C] This is an illustrative graphic representation of the scoring methodology shown in Figure 2B according to the embodiments of this disclosure. [Figure 3] This is an exemplary representation in accordance with the present disclosure. [Figure 4] This is an exemplary representation in accordance with the present disclosure. [Figure 5] This is an exemplary system according to the aspects of this disclosure. [Figure 6] This is an exemplary blood vessel according to the embodiments of this disclosure. [Figure 7] This is an example of multiple interface components according to the aspects of this disclosure. [Figure 8A] This is an exemplary graphic representation of cross-validation results. [Figure 8B] This is an exemplary graphic representation of cross-validation results. [Figure 9A] This is an exemplary representation in accordance with the present disclosure. [Figure 9B] These are exemplary interface components according to aspects of the present disclosure. [Figure 10] This is an exemplary representation in accordance with the present disclosure. [Figure 11] This is a flowchart illustrating a method for outputting a representation of blood vessels according to the aspects of this disclosure. [Modes for carrying out the invention]

[0005] Some parts of the detailed description relate to methods, for example, algorithms and symbolic representations of operations on data bits in computer memory. These descriptions and representations of algorithms are available to those skilled in the computer and software-related fields. In one embodiment, an algorithm is considered here, and generally, to be a self-consistent sequence of operations that produces a desired result. Operations performed as method stops or otherwise described herein require the physical manipulation of physical quantities. These quantities, though not necessarily, usually take the form of electrical or magnetic signals that can be manipulated by storage, transfer, combination, conversion, comparison, and other means.

[0006] The algorithms and representations described herein are not inherently related to any particular computer or other device. Various general-purpose systems may be used with programs following the teachings herein, or it may become apparent that it is convenient to construct more specialized devices to perform the necessary method steps. The necessary structures for these various systems will become apparent from the following description.

[0007] The aspects, embodiments, features, and examples of this disclosure should be considered illustrative in all respects and are not intended to limit the disclosure, the scope of which is defined solely by the claims. Other embodiments, variations, and uses will be apparent to those skilled in the art without departing from the essence and scope of the claimed invention.

[0008] The use of headings and sections in this application is not intended to limit the invention. Each section may apply to any aspect, embodiment, or feature of the invention.

[0009] Throughout this application, where a composition is described as having, including, or comprising certain components, or where a process is described as having, including, or comprising certain process steps, the compositions of this instruction are also intended to consist essentially of or consist of the described components, and the processes of this instruction are also intended to consist essentially of or consist of the described process steps.

[0010] Where it is stated in this application that an element or component is included in and / or selected from the list of elements or components described, it should be understood that the element or component may be any one of the elements or components described, and may be selected from a group of two or more elements or components described. Furthermore, it should be understood that the elements and / or features of the compositions, apparatus, or methods described herein can be combined in various ways, whether expressly or implicitly, without departing from the framework and scope of this teaching.

[0011] The use of the words "include," "includes," "have," or "having" should generally be understood as open-ended and non-restrictive unless otherwise specified.

[0012] In this specification, the use of the singular form includes the plural form (and vice versa) unless otherwise specified. Furthermore, the singular forms "a, an" and "the" include the plural form unless explicitly stated in the context. In addition, where the word "about" is used before a quantitative value, this instruction also includes that particular quantitative value itself unless otherwise specified. As used herein, the word "about" means a variation of ±10% from the nominal value. All numerical values ​​and numerical ranges disclosed herein are considered to include "about" before each value.

[0013] As long as the present teachings are implementable, it should be understood that the order or sequence of steps for performing a particular operation is not important. Further, two or more steps or operations may be performed simultaneously.

[0014] When a range or list of values is provided, each value intervening between the upper and lower limits of the range or list of values is considered individually and is incorporated into the invention as if each value was specifically recited herein. Further, smaller ranges between the upper and lower limits (including the upper and lower limits) of a given range are also contemplated and incorporated into the invention. Listing exemplary values or ranges as a list does not disclaim other values or ranges between the upper and lower limits (including the upper and lower limits) of a given range.

[0015] [Introduction] Systems and methods may perform feature detection and alignment of relative imaging data sets from intravascular imaging pullbacks. For example, the intravascular imaging pullback may be an OCT or intravascular ultrasound (「IVUS」) pullback. The imaging data set may be imaged at one or more time points corresponding to different arterial events or procedures. One or more representations of the artery may be displayed based on the imaging data set. The representation may include a display of the identification of calcium loading after performing calcium detection on one or more (groups or subsets) of the image data frames. One or more representations may be displayed to a user.

[0016] Calcium in the image data may be detected by image processing techniques and / or machine learning. The frames of the pullback may be stretched and aligned using various windows or bins of alignment features. The alignment may be informed or optimized using various inputs or constraints, such as flexibility to stretch the pullback using the lumen data frame or other data to maximize alignment, points for aligning calcium edges, points for aligning stents, and points for aligning side branches (SBs).

[0017] Detection of calcium of interest, or other characteristics or tissue quality, can be performed. The detected characteristics can then be identified as a cluster or group. The cluster or group of detected characteristics can be displayed for one or more pullbacks over time. For example, the pullbacks can be taken before treatment, after treatment, before stent implantation, after stent implantation, before atelectomy, after atelectomy, before angioplasty, after angioplasty, after optimization, etc. According to some examples, the pullbacks can be taken after stent implantation and / or after the physician has further ballooned the stent at various balloon diameters and pressures. The pullbacks can be aligned using common characteristics identified in each frame. The common characteristics can include, for example, collateral branches, stents, previous stents, and other characteristics using the techniques disclosed herein. [[ID=Z]]

[0018] Information related to the lumen profile, detected stents, measurements of minimum lumen area, average lumen area, minimum stent area ("MSA"), etc. can be displayed. The display can also include the risk level associated with the calcium load. In some examples, the risk level can be the risk of stent under-expansion caused by the calcium load. The risk of stent under-expansion can be automatically estimated and / or calculated based on the determined calcium load.

[0019] The display can include a longitudinal view of the blood vessel. According to some examples, the display can highlight the area of calcium load within the blood vessel. The calcium load begins as a lesion. The lesion can begin as lipid and may gradually harden into a combination of fibrous plaque and calcified plaque. When the lesion calcifies, the lesion can harden and may pose additional resistance to stent expansion. Different calcium scoring techniques may be used to improve decision-making, such as where to place a stent, or where to perform angioplasty or atelectomy.

[0020] The display may also include a view of calcium loading in the OCT pullback of blood vessels, such as coronary arteries. Various different outputs may be displayed. For example, there may be one or more horizontal charts. Horizontal charts may include keys to identify the display colors. However, in some examples, the display may be a color display and therefore may not include keys, as keys may not be necessary.

[0021] The chart may plot calcium loading using the OCT pullback frame number on the x-axis and the calcium loading level on the y-axis. The chart may show that there are eight different calcified plaques shown during the pullback. According to one example, only one of the calcified plaques may be strongly red "R". A calcified plaque identified as red "R" may indicate that it should be viewed when determining the pre- and / or post-treatment plan. According to several examples, a calcified plaque identified as red "R" may be a calcified plaque that is likely to cause problems with stent expansion. Other labeled calcified plaques, for example, labeled as orange "O", may indicate that it is worthwhile to view them when determining the pre- and / or post-treatment plan. A calcified plaque labeled as green "G" may be a calcified plaque that is unlikely to cause problems with stent expansion and therefore may not be worthwhile to view when determining the pre- and / or post-treatment plan.

[0022] [Example System] Figure 1A shows a data acquisition system 100 for use in acquiring intravascular data. The system may include a data acquisition probe 104 that can be used to image a blood vessel 102. A guidewire (not shown) may be used to introduce the probe 104 into the blood vessel 102. The probe 104 may be introduced and withdrawn along the length of the blood vessel during data acquisition. As the probe 104 is withdrawn, i.e., retracted, multiple scan or OCT and / or IVUS datasets may be acquired. The datasets, or frames of image data, may be used to identify features, such as calcium.

[0023] The probe 102 may be connected to subsystem 108 via optical fiber 106. Subsystem 108 may include a light source, such as a laser, an interferometer having a sample arm and a reference arm, various optical paths, a clock generator, a photodiode, and other OCT and / or IVUS components.

[0024] The probe 102 may be connected to the photodetector 110. According to some examples, the photodetector 110 may be a balanced photodiode-based system. The photodetector 31 may be configured to receive the light collected by the probe 102.

[0025] The subsystem may include a computing device 112. The computing device may include one or more processors 113, memory 114, instructions 115, data 116, and one or more modules 117.

[0026] One or more processors 113 may be any conventional processor, such as a commercially available microprocessor. Alternatively, one or more processors may be a dedicated device, such as an application-specific integrated circuit (ASIC) or other hardware-based processor. Although Figure 1B functionally shows the processor, memory, and other elements of device 110 as being in the same block, it will be understood by those skilled in the art that a processor, computing device, or memory may actually include multiple processors, computing devices, or memories that may or may not be housed in the same physical enclosure. Similarly, memory may be a hard drive or other recording medium located in a different enclosure than that of device 112. Thus, a reference to a processor or computing device will be understood to include a reference to a collection of processors or computing devices or memories that may or may not operate in parallel.

[0027] Memory 114 may store information accessible by the processor, including instructions 115 that can be executed by the processor 113, and data 116. Memory 114 may be a type of memory that operates to store information accessible by the processor 113, including non-temporary computer-readable media, or electronic devices, such as hard drives, memory cards, read-only memory ("ROM"), random access memory ("RAM"), optical discs, and other media for storing data that are readable by other writable and read-only memories. The subjects disclosed herein may include different combinations of the above, thereby storing different parts of instructions 101 and data 119 in different types of media.

[0028] Memory 114 may be retrieved, stored, or modified by the processor 113 in accordance with instructions 115. For example, although this disclosure is not limited by a specific data structure, data 115 may be stored in computer registers, tables with multiple different fields and records, XML documents, or relational databases as flat files. Data 115 may also be formatted in a computer-readable format, for example, but not limited to, binary values, ASCII, or Unicode. Further exemplary, data 115 may be stored as a bitmap consisting of pixels stored compressed or uncompressed, or as various image formats (e.g., JPEG), vector-based formats (e.g., SVG), or computer instructions for drawing graphics. Furthermore, data 115 may contain enough information to identify relevant information, such as numbers, descriptive text, proprietary codes, pointers, references to data stored in other memories (including other network locations), or information used by functions that compute the relevant data.

[0029] Instruction 115 may be any set of instructions executed by the processor 113 directly, for example, in machine code, or indirectly, for example, in a script. In this respect, the terms “instruction,” “application,” “step,” and “program” are interchangeable in this specification. Instructions can be stored in object code format for direct processing by the processor, or in any other computing device language, including a script or set of independent source code modules that are translated on demand or pre-compiled. The function, method, and routines of instructions are described in more detail below.

[0030] Module 117 may include a plaque detection module, such as a calcium plaque detection module, a display module, a stent detection module, or other detection and display modules. For example, computing device 112 may have access to a calcification detection module for detecting the presence of calcium plaque in blood vessels. According to some examples, the module may include an image data processing pipeline or its component modules. The image processing pipeline may be used to convert the collected OCT data into two-dimensional ("2D") and / or three-dimensional ("3D") views and / or representations of blood vessels, stents, and / or detection areas.

[0031] The computing device 112 may include a machine learning module. Image data from previous cases may be collected and stored in data 116. Each frame from previous cases may be analyzed to determine the impact on calcium loading and stent expansion failure. The analyzed information may be stored and used as input to a machine learning model. The machine learning model may predict the risk of stent expansion failure. The machine learning model is described further below with respect to Figure 5.

[0032] Subsystem 108 may include a display 118 for outputting content to the user. As shown, the display 118 is independent of the computing device 112, but according to some examples, the display 118 may be part of the computing device 112. The display 118 may output image data related to one or more features detected within a blood vessel. For example, the output may include, but is not limited to, cross-sectional scan data, longitudinal scans, diameter graphs, image masks, lumen boundaries, plaque size, plaque circumference, visual markers of plaque location, visual markers of the risk posed to stent expansion, etc. The display 118 may identify features using text, arrows, color coding, highlighting, contour lines, or other appropriate human or machine-readable markers.

[0033] According to some examples, the display 118 may be a graphical user interface ("GUI"). One or more steps may be performed automatically, i.e., without user input, for purposes such as navigating images, entering information, selecting inputs, and / or interacting with inputs. The display 118 may, alone or in combination with the computing device 112, enable switching between one or more view modes in response to user input. For example, a user may be able to switch between different side branches on the display 118, for example, by selecting a particular side branch and / or by selecting a view associated with a particular side branch.

[0034] In some examples, the display 118 may include menus, either alone or in combination with the computing device 112. These menus may allow the user to display or hide various features. Two or more menus may exist. For example, there may be a menu for selecting the features of blood vessels to display. Additionally or alternatively, there may be a menu for selecting the virtual camera angle of the display.

[0035] Figure 1B shows various interface components that may be displayed in display 118B. As shown in Figure 1B, display 118B may include four different interface components. Chart 10 may display an OCT image in Cartesian coordinates in a single frame. As shown, chart 10 may display frame 92 superimposed with the detected elliptical lumen edge. Chart 20 may display tissue characterization for the same frame (frame 92) in Cartesian coordinates centered on the lumen. The lumen may be displayed in gray, the guidewire shadow in dark gray, the medium in red, and calcium in white. In this example, frame 92 does not contain any detected visible medium. The display may show each of these features in a different color depending on the type of display. For example, if the display is black and white, the features may be displayed in grayscale. In examples where the display is color, the features may be displayed in any predetermined or user-preferred color. Charts 10 and 20 may show the angular spread and thickness of calcium in frame 92. For example, Chart 20 may show the angular spread and thickness of arcs "Ca arc 1" and "Ca arc 2".

[0036] Charts 30 and 40 may display longitudinal data with frame numbers on the x-axis and frame-by-frame measurements on the y-axis. Chart 50 may display a longitudinal view of lumen area and highlighted collateral branches (SB). Charts 30, 40, and 50 may be aligned so that the frames in Chart 50 are visually below or coincide with the same frames in Charts 30 and 40. Chart 50 may include the display of a first reference frame "RF1", a second reference frame "RF2", and a GUI reference frame. The reference frames may be selected based on a machine learning model or by the user. Reference frames RF1, RF2, and the GUI may be used for calculations and / or to display various other views on the screen when selected.

[0037] Chart 30 may show the calcification risk score developed by the Cardiovascular Research Foundation ("CRF") in "A New Optical Coherence Tomography-based Calcium Scoring System to Predict Stent Underexpansion" 13(18)EUROINTERVENTION e2182-e2189 (Apr. 6, 2018) by Fujino, A et al. Chart 30 can be determined by identifying the largest calcium deposit for each frame and measuring its total radial area (degrees·mm). Due to OCT tissue penetration and, in some cases, calcium detection accuracy may be limited to 1 mm or less, the thickness (mm) may be limited to 1 mm or less. This calcium radial area becomes the height of the bar displayed in the chart and can be associated with the risk of stent underexpansion in each frame. The color of the bars in Chart 40, identified by "R", "O", and "G", may correlate with the average radial area in a longitudinal window of 25 frames (5 mm). Chart 40 may provide an informational view of the accumulation risk posed by adjacent frames of large amounts of calcium. For example, Chart 40 may provide an additional or alternative method for indicating calcium load. The colors used in Chart 30 may correspond to the colors used in Chart 30. A bar in Chart 30 may be entirely red if it reaches a sliding window radial area of ​​180 degrees × 0.5 mm = 90 degrees·mm.

[0038] Figures 2A–2C illustrate examples of how frames may be scored, analyzed, and / or displayed based on Chart 30. Figure 2A is referenced from Fujino, A et al.

[0039] Figure 2A includes a vascular segment 228 and a cross-sectional image corresponding to frame 220 of the image data. The maximum angle 222 and / or maximum thickness 224 of the calcium within frame 220 can be determined. Additionally or alternatively, the length 226 of the calcium 221 can be determined based on the longitudinal representation of the vascular structure 228.

[0040] Each frame can be analyzed. For each frame, one or more of the following can be calculated: 1. If the largest calcium arc within the frame is greater than 180 degrees, the frame score is 2 points; otherwise, it is 2 × arc(degree) / 180 points. 2. If the largest calcium arc within the frame has a thickness greater than 0.5 mm, an additional point is added to the frame; otherwise, 1 × thickness (mm) / 0.5 is added. 3. If the length of the calcium deposit is greater than 5 mm (approximately 25 frames), an additional point is assigned to each frame in the deposit; otherwise, 1 × length (mm) / 5 is assigned. 4. If a calcium score of 4 is assigned to calcium deposits in multiple frames, the entire area of ​​those deposits will be colored red. Scores of 2 and 3 are orange, and a score of 1 is green.

[0041] OCT-based calcium scores based on one or more calculations are described with respect to Figure 2A. For example, an OCT-based calcium score may be a value between 0 and 4 points. The score may be based on the maximum calcium angle, maximum calcium thickness, and / or calcium length. In cases where the maximum calcium angle is 180 degrees or less, the score may be 0. In cases where the maximum calcium angle is greater than 180 degrees, the score may be 2. In cases where the maximum calcium thickness is 0.5 mm or less, the score may be 0. In cases where the maximum calcium thickness is greater than 0.5 mm, the score may be 1. In cases where the calcium length is 5.0 mm or less, the score may be 0. In cases where the calcium length is greater than 5.0 mm, the score may be 1. The total score may be determined based on individual scores for one or more of the maximum calcium angle, maximum calcium thickness, and calcium length. Actual measurements and / or thresholds may vary based on machine learning models. Coefficients for these factors may be determined by machine learning models, as they are also relevant to stent expansion in case study inputs. Assessment and scoring may be completed by a physician. However, the systems and methods described herein can automate evaluation and scoring. The computing device 112 may use the chart in Figure 1B, for example, chart 30, to apply evaluation criteria and provide an automated score.

[0042] In some examples, the calcium score may be calculated using a sliding window measure. A sliding window measure may include a window or range around each given point in the line. The window may slide down the line as each point is recalculated. In some examples, the risk score may be recalculated for all frames by considering the total length of the calcified plaque. In some examples, the risk score may be recalculated, additionally or alternatively, based on the calcium thickness and / or calcium angle in each particular frame. The sliding window measure may be calculated by multiplying the calcium length (mm) by the calcium thickness (mm) and the calcium angle (degrees). For example, the formula may be: Sliding window measurement = Calcium length × Calcium thickness × Calcium angle

[0043] The radial area can be determined by multiplying the calcium thickness by the calcium angle. The total radial area can be measured within a 5 mm window. However, the 5 mm window is merely an example and therefore not intended to be limiting, as the window may be larger or smaller than 5 mm.

[0044] Figure 2B shows an exemplary graphical representation of the calculated OCT-based score. A CRF score of 4 may indicate that the stent is not failing to expand. In some cases, lesions with a CRF score of 4 may not fail to expand as frequently as cases with lower scores (less than 80%) (50±16% vs. 53±12%). In some cases, lesions with a CRF score of 4 may require additional optimization with a 50% probability (compared to 13% in cases with lower scores).

[0045] Figure 2C shows an exemplary graphical representation of the calcium content 200D. A negative correlation may exist between calcium content and stent expansion. For example, approximately 67% of cases with high calcium content (greater than 200° × mm²) may have poorly expanded stents compared to cases with low calcium content (less than 80% vs. 48%). According to some examples, approximately 50% of cases with high calcium content may require additional optimization after OCT compared to 24% of cases with low calcium content.

[0046] Figures 2C and 2D may show that windowed calcium levels can better estimate the risk of stent expansion failure than the CRF score.

[0047] Figure 3 shows an exemplary output for display. OCT pullback images from before and after stent placement may be aligned to provide information correlating calcium loading with stent expansion. The correlation between calcium loading and stent expansion may allow a physician or user to assess the role of calcium loading in stent expansion. Display 318 may include charts 330, 332, 334, and 336. Chart 330 may display data or frames related to PCI calcium before percutaneous coronary intervention ("PCI"), chart 332 may display data or frames related to lumen before PCI, chart 334 may display data or frames related to lumen after PCI, and chart 336 may display data or frames related to calcium after PCI. Views of calcium and lumen before PCI (charts 330 and 332, respectively) may be displayed adjacent to views of calcium and lumen with respect to pullback after PCI (charts 334 and 336, respectively). The pullback view after PCI may include the stent placement area 337.

[0048] Display 318 includes W1, W2, W3, and W4, as shown. Each window W1, W2, W3, and W4 may have a different purpose or provide a different display of the detected features. In some examples, windows may be output for display. In other examples, windows are not output for display.

[0049] Window W1 may display regions of vessels that have flexibility to extend pullback to maximize alignment. Window W2 may display regions of vessels containing points for aligning calcium. Window W3 may display regions of vessels containing points for aligning stents. Window W4 may display regions of vessels containing points for aligning collaterals. The windows are shown in Figure 3, but this is merely to aid in explaining the output and therefore may not be output for display.

[0050] The indication may include an indication of the region of blood vessels where the extent of pretreatment, such as balloon angioplasty before stent placement, may influence the reduction of the effects of calcium loading.

[0051] The system may use the Needleman-Wunsch algorithm or a modified version of the Needleman-Wunsch algorithm for scoring to align calcium, existing stents, collateral branches, and relative lumen area. According to some examples, candidate alignments obtain points to have at least one of the following: 1. Number of similar arc angles with calcium within the frame: min(pre,post) / max(pre,post,60 degrees). 2. Similar total diameter of lateral branches within the frame: min(pre,post) / max(pre,post,0.5mm). 3. Similar relative lumen area within a frame: min(rla_pre,rla_post) / max(rla_pre,rla_post,0.1), where rla_pre and rla_post are the relative lumen areas within the frame, and are calculated as follows. rla_pre[f] = (luminal area at f) / (luminal area at F') rla_post[f] = (luminal area at f) / (luminal area at F) In the formula, F is the proximal reference frame in the post-PCI pullback, and F' is the pre-PCI frame corresponding to F in this candidate alignment.

[0052] Because both the calcium profile and the lumen profile can change due to the action of stent placement, the area of ​​calcium or lumen in the stent placement area may not match depending on the algorithm. In some cases, this discrepancy in the area of ​​calcium or lumen in the stent placement area can be compensated for by using an error margin or other statistical correction.

[0053] Figure 4 shows an exemplary representation. Representation 418 may include chart 440 showing the CFR score of the vessel, chart 430 showing the calcium of the vessel before PCI, chart 442 showing a 2D longitudinal representation of the vessel, chart 434 showing the lumen of the vessel after PCI, and chart 444 showing tapered dilation. According to some examples, tapered dilation may be determined by recalculating an appropriate reference area for each frame based on the natural taper of the vessel from proximal to distal as the side branches bypass the blood flow. Chart 430 may use the sliding window measurement described above. Additionally or alternatively, chart 430 may be based on the windowed calcium amount plotted in chart 40. Chart 430 may be similar to chart 40 shown in Figure 1B. Chart 442 may be similar to chart 50 shown in Figure 1B. Chart 434 may be similar to chart 50 in Figure 1B, but instead may show only the same region of the vessel after stent placement.

[0054] In some examples, the display may show or identify areas of detected calcium. Identified areas of detected calcium may show a negative correlation with stent expansion by limiting the amount of stent expansion, as indicated by local minima in the stent expansion threshold plot. Chart 444, which may show tapered expansion, may show the change to the lumen profile calculated using a reference frame at either the proximal and / or distal end of the pullback.

[0055] Although not shown in charts 440, 430, 442, 434, and 444 of Display 418, red areas may indicate high or increased calcium loads that may require consideration when stenting a blood vessel. Green areas may indicate decreased or low calcium loads. In some cases, an orange area "O" may indicate an intermediate range of calcium loads where the effects of calcium load may warrant further analysis or other views using cross-sectional, l-mode, and other intravascular views and analyses. This is just one example, and it should be understood that in other cases, risks may be color-coded using different color schemes.

[0056] [Stent expansion failure] The calcium loading described herein, as stated above, may be used to estimate the risk of stent expansion failure. The risk of stent expansion failure may be estimated using a trained machine learning ("ML") model. Each training example may be a case from a clinical trial and / or in-situ. The ML model may compare pre-PCI information with the post-PCI outcome for each case. The ML model estimation may be used to provide a quantitative assessment of the risk of expansion failure to a physician or end-user.

[0057] Figure 5 shows an exemplary system 500 that uses data from past PCI procedures to predict the risk of future stent expansion failure in patients. System 500 may include a development environment 552 and a catheter lab 560. Although shown as a catheter lab 560, the catheter lab 560 could be any location where a physician inserts or implants a stent in a patient. For example, a catheter lab could be a hospital, an outpatient surgery site, etc. Therefore, identifying the location as a catheter lab 560 is merely an example and not intended to be limiting.

[0058] The development center 552 may include a training database 5545, a machine learning system 556, and a trained predictive model 558A. The training database may include multiple levels of PCI information. For example, the training database 554 may include coarse statistics from published clinical studies, records and images for individual PCIs from clinical trials, and data on PCIs collected in the field. These data may be in the form of input-output pairs, in which case the input is all observable information before the target vessel is prepared and the stent is placed, and the output is the stent expansion and other outcomes (complications, readmission, TVR, etc.) that occurred. An input-output pair may be one or more image frames. According to some examples, the input may be multiple images of the target vessel before it is prepared, and the output may be multiple images of the target vessel after stent expansion, etc. Multiple images of the input may correspond to multiple images of the output such that the first frame of the multiple images of the input is from the same position as the first frame of the multiple images of the output within the target vessel.

[0059] The machine learning system 556 can learn or model the relationship between these inputs and outputs. For example, the machine learning system 556 can detect different values ​​from each of a set of input and output images. These values ​​may include, but are not limited to, calcium angle, maximum thickness, stent expansion rate, etc., for each of the set of input and output images. Each of these values ​​may later be used to predict the risk level of stent expansion failure. In some examples, the machine learning system 556 can learn the relationship by tuning its internal parameters to minimize errors in its output predictions.

[0060] As some examples show, linear models, such as logistic regression, can have their internal parameters—the multiplicative weights assigned to each predictive attribute—adjusted. For example, one model might look like this: Expansion = (w1) × (calcium) + (w0)

[0061] Expansion may be the predicted stent expansion percentage. Calcium may be the maximum windowed calcium amount. "w1" and "w0" may be values ​​that the algorithm can adjust to best fit the training data.

[0062] In some examples, a machine learning system can learn any number of decision trees such that their internal parameters can be rules governing each tree. In one example, a machine learning system might determine that good expansion can be predicted if the calcium level is less than 0.3.

[0063] In some cases, when the machine learning system 556 tunes the model parameters to minimize prediction errors, the machine learning system 556 may rerun the data (557) to create an additional model.

[0064] A model created by the machine learning system 556 may be a trained predictive model 558. The trained predictive model 558 may predict the risk of stent expansion failure. The trained predictive model 558 may be transmitted to the catheterization lab 560. According to some examples, the trained predictive model 558 may be shared over a network. In the catheterization lab 560, the trained predictive model 558 may use and / or incorporate information on new target lesions 562 to generate a stent expansion prediction 564 to assist physicians in refining interventional strategies. The stent expansion prediction 564 may include the risk of stent expansion failure.

[0065] Figure 6 shows exemplary lesions within 600 target vessels. Lesion "Y" may be a thick, eccentric calcium deposit. Lesion "O" may be a thick, circumferential calcium deposit. Lesion "G" may be a thin, eccentric calcium deposit. In some cases, lesion "Y" may be colored yellow, lesion "O" may be colored orange, and lesion "G" may be colored green in the representation.

[0066] In some cases, the calcification of a lesion may be eccentric, with the calcification being only on one side of the vessel, or it may be circumferential, encircling the inner circumference of the vessel. Circumferential calcium can hinder the omnidirectional expansion of the stent balloon, thus posing the greatest risk to stent expansion. Not all calcium resists stent expansion. Some calcium deposits are thin enough for the stent to expand through them. Other calcium deposits may not form a sufficiently large arc around the vessel, causing the stent to puncture the calcium at one or more hinge points. There may also be mixed plaque types, with calcium deposits within a lipid and / or fibrous matrix.

[0067] Figure 7 shows an example of multiple interface components that may be output to display 718. Multiple interface components may be used to determine lesions within a blood vessel. Frame 710 may be an OCT image of a single cross-sectional frame of a coronary artery with a calcified lesion. Frame 720 may be a tissue characterization of frame 710, including calcium 772, the shadow of the guidewire 774, and the identified lumen 776. Output 770 may be a frontal projection of calcium. In some examples, the horizontal position of each pixel may represent the position of the longitudinal frame. Additionally or alternatively, each vertical position may represent the angular position around the center of the lumen. Pixel intensity shown in grayscale may represent the calcium thickness in that frame and arc. Output 758 may display the diameter of the lumen across the pullback. Mark 777 may display a frame in which the estimation of the lumen diameter may have a larger error. Mark 778 may display a detected side branch. Mark 779 may display a frame shown in cross-sectional frames 710 and 720.

[0068] For the vascular region where a stent should be placed, tissue characterization can be generated for that region in the OCT pullback. Pixels within the frame can be identified as calcium. From the pixels identified as calcium, the thickness of the calcium can be measured at each radial angle around the center of the vascular lumen. Frame 720 may contain 360-degree calcium with a thickness greater than 0.0 mm. According to some examples, frame [upper right] may contain 45-degree calcium with a thickness greater than 0.8 mm, and 0-degree calcium with a thickness greater than 1.0 mm. Frame 720 can be used as an input feature vector in a machine learning model. According to some examples, six thickness thresholds can be used to measure the largest continuous arc of calcium at these thresholds. The thresholds may be, for example, 0.0 mm, 0.2 mm, 0.4 mm, 0.6 mm, 0.8 mm, and 1 mm. However, the thresholds may be any value in mm, e.g., 0.1, 0.25, etc. Therefore, the thresholds of 0.0 mm, 0.2 mm, 0.4 mm, 0.6 mm, 0.8 mm, and 1 mm are merely examples of the six thresholds and are not intended to be limiting.

[0069] With respect to frame 720, the use of exemplary thresholds of 0.0 mm, 0.2 mm, 0.4 mm, 0.6 mm, 0.8 mm, and 1 mm allows for 360-degree calcium with a thickness greater than 0.0 mm, while only the 45-degree calcium may have a thickness greater than 0.8 mm, and the 0-degree calcium may have a thickness greater than 1.0 mm. Based on the six thresholds, the arc vectors for frame 720 could be 360, 360, 210, 180, 45, and 0.

[0070] Stent expansion can be influenced not only by calcium but also by one or more other factors. For example, maximum balloon diameter, maximum balloon pressure, balloon / arterial diameter ratio, rough vessel location including LAD, LCX, and RCA, reference lumen area, minimum lumen area, stenosis percentage, stenosis divided by maximum balloon pressure, minimum lumen area divided by maximum balloon area, stent length, lumen eccentricity, and calcium depth (mm) measured as the amount of tissue between the lumen edge and the first pixel of calcium within the frame. One or more of these factors can be used as input in machine learning models.

[0071] Figures 8A and 8B show an example of cross-validation results for machine learning model predictions regarding whether a stent expanded well or poorly. In this example, a stent expanded well if it expanded to more than 90%, and poorly if it expanded to less than 70%.

[0072] As shown in Figure 8A, cases of desirable stent expansion [blue] may be assigned a lower risk estimate, while cases of poorly expanded stents [red] may be found to have a higher risk estimate.

[0073] Figure 8B correlates specificity and sensitivity at varying thresholds. A machine learning algorithm may predict either a numerical value, such as the expansion rate, or a classification marker, such as good expansion or poor expansion. As shown, Figure 8B may illustrate the performance of a classifier when the algorithm generates a numerical score rather than just predicting the marker. The classifier may be from an algorithm that predicts a classification marker. Each point on the curve may represent performance at a different decision threshold. For example, a score of 0.9 might correspond to a sensitivity of 0.9 and a specificity of 0.7.

[0074] According to some examples, risk assessments, or the risk of stent expansion failure, may be displayed as probabilities in the display or GUI. For example, a display might show that there is an 80% chance that the stent will fail to expand.

[0075] Figure 9A shows an exemplary representation including a risk assessment. Representation 918 may include a longitudinal representation of a vessel 984, a selected frame 980 within the longitudinal representation 984, and an informational representation 982. The longitudinal representation 984 may include a representation or marker 981 that identifies the selected frame 980. The informational representation may include the total angle of calcium within the selected frame 980, the maximum thickness of calcium within the selected frame 980, and a risk assessment 987.

[0076] Risk Assessment 987 may include indications or warnings about the risk of stent expansion failure. In cases where there is a high risk of stent expansion failure, Risk Assessment 987 may include a warning symbol or stop sign as shown in the illustration. In cases where there is a low risk of stent expansion failure, Risk Assessment 987 may include a green light (not shown) or an "OK" symbol. Risk Assessment 987 may be color-coded, with high risk of stent expansion failure sometimes written in red to warn the physician, and low risk of stent expansion failure sometimes written in green to give the physician an "OK". High risk may be associated with red and low risk with green, but Risk Assessment 987 may use any color system, including grayscale.

[0077] The longitudinal view 984 may, in addition or alternatively, include a risk assessment 986. The longitudinal risk assessment 986 may be displayed above or below the longitudinal view, for example, as a bar. The longitudinal risk assessment 986 may be color-coded such that one color signifies high risk and another color signifies low risk. The color coding of the longitudinal risk assessment 986 may correspond to the color coding of the risk assessment 987. For example, a red longitudinal risk assessment may be high risk, and a green longitudinal risk assessment may be low risk. In some examples, there may be additional color risks, such as orange or yellow, which may be moderate risk or risk to consider. According to some examples, individual frames within the longitudinal view 984 may be color-coded or highlighted. The color coding and / or highlighting may be similar to the color coding used for the longitudinal risk assessment 986.

[0078] According to some examples, display 918 may additionally or alternatively include a histogram, for example, histogram 988 shown in Figure 9B. Histogram 988 may graphically present information regarding stent expansion for similar cases. As shown, histogram 988 graphically displays the percentage of stent expansions and the number of cases for similar cases. As mentioned above, similar cases may be determined based on a machine learning model.

[0079] The data used to create the histograms can be based on a similarity metric. As in some examples, the similarity metric can vary depending on the selected machine learning model. For instance, if a linear model such as regression is selected, a Euclidean distance metric weighted by the coefficients of logistic regression may be used. Additionally, or alternatively, if a decision tree ensemble is selected, the histograms may be generated based on cases corresponding to the same leaf nodes as the test cases. In either the case of a linear-like model or a decision tree ensemble, each histogram can be augmented by displaying risk estimates for different balloon sizes and balloon pressures. This can provide the user with some insight into whether risk can be mitigated by balloon selection.

[0080] Figure 10 shows an exemplary representation including risk assessment and predicted stent expansion. Similar to the representation shown in Figure 9, representation 1018 may include a longitudinal representation of the vessel 1084, a selected frame 1080 within the longitudinal representation 1084, an informational representation 1082, and a representation of predicted expansion 1092. The longitudinal representation 1084 may include a representation or marker 1081 that identifies the selected frame 1080. Although not shown, representation 1018 may additionally or alternatively include one or more of the interface components described above. For example, representation 1018 may include stent expansion risk as shown in Figure 9A, or a histogram as shown in Figure 9B.

[0081] Information display 1082 may include information regarding calcium loading, such as total angle and maximum thickness. However, this is merely an example. Information display 1082 may also include information relating to any arterial characteristics, such as the presence of a visible medium, lipid plaque or thin coronary fibrous dysplasia, and measurements of lumen diameter and eccentricity. The visible medium may include an external elastic lamina.

[0082] The predicted dilation representation 1090 may be a predicted stent dilation based on detected arterial characteristics. For example, calcium loading may be determined based on image data obtained from one or more pullbacks. The predicted dilation is shown as a percentage of predicted dilation, for example, 70%. Additionally or alternatively, a box plot 1094 may be included in the predicted dilation representation 1092. The box plot 1094 may represent the range of likely dilation.

[0083] Figure 11 shows an exemplary method for outputting a representation of blood vessels. The following steps do not need to be performed in the exact order described below. Rather, the various steps may be performed in a different order or simultaneously, and steps may be added or omitted.

[0084] For example, in block 1110, one or more processors may receive one or more frames containing image data of a vascular segment. These frames may be obtained during one or more imaging pullbacks. For example, pullbacks may be imaged before, after, before, after stent placement, before, after atherectomy, before, after angioplasty, after optimization, etc. According to some examples, pullbacks may be imaged after stent placement and / or after the physician has further balloon-inflated the stent with various balloon diameters and pressures.

[0085] In block 1120, one or more processors may detect arterial features in each of one or more frames. Arterial features may include, in some examples, calcium load, visible medium, presence of lipid plaque or thin coronary fibrous dysplasia, and measurements of lumen diameter and eccentricity.

[0086] In block 1130, one or more processors may score the arterial characteristics in each of one or more frames. The score may be a stent expansion failure risk score. The score may be determined using a machine learning model. The machine learning model may compare pre-PCI information with post-PCI results for each case.

[0087] In block 1140, one or more processors may identify regions of interest based on arterial characteristic scores. In cases where the score is a stent expansion failure risk score, the region of interest may be a calcium loading region. Additionally or alternatively, in cases where the score is a stent expansion failure risk score, the region of interest may be a region where there is a high, moderate, or low risk of stent expansion failure.

[0088] In block 1150, one or more processors may output a representation of a vascular segment, including a visual display of scores for a region of interest. The representation may be two-dimensional and / or three-dimensional. The visual display of scores may be color-coded. The display may be color-coded based on the score of the arterial features. For example, if the score is a stent expansion failure risk score, the display color may be red, yellow, or green. A red display may indicate the presence of a high risk of stent expansion failure, yellow may indicate the presence of a moderate risk, and green may indicate the presence of a low risk of stent expansion failure. In some cases, such as when the display is not in color, the display may be color-coded in grayscale.

[0089] The representation may be a bar in shape, for example, a rectangle parallel to the longitudinal section of the vascular representation. The bar may extend the length of arterial features within the representation. In cases where the arterial feature is a calcium load, the bar may extend the length of the calcium load within the representation. Based on the color of the bar, the physician can quickly recognize whether or not a calcium load may pose a risk of stent expansion failure.

[0090] [User Workflow] The computing devices described above can assist end users in navigating through lesion assessment, stent sizing, placement, and post-placement evaluation. For example, the computing device may output lesion morphology. The output may be in a manner that allows end users to easily assess lesion morphology. In some examples, the output may include a color-coded representation of vascular segments. Each color in the color-coded representation may represent a level of severity. Additionally or alternatively, each color in the color-coded representation may represent a predicted, estimated, or determined risk of stent expansion failure based on arterial characteristics. Arterial characteristics may include, for example, calcium load. The color-coded risk estimates may motivate, prompt, and / or enable users to adapt vascular preparation and / or stent placement strategies. For example, if the risk posed by calcium is high, the user may see at least a portion of the vessel as red, representing high risk. In some examples, instead of or in addition to color-coded, the risk level may be shown in grayscale, hatching, etc. Therefore, color-coded is merely an example and is not intended to be limiting.

[0091] By quickly and easily assessing the high risk of stent expansion failure, indicated by specific colors, users can decide to modify their treatment plan. For example, if the risk posed by calcium loading is high and the user sees a lot of red, the user may choose to perform more aggressive ballooning or vascular preparation before stent placement. In some cases, high risk may indicate to the user that the atherectomy technique should be adapted to remove or break up calcium before stent placement using techniques such as cutting balloons, scoring balloons, orbital or rotary drills, or ultrasonic lithotripsy.

[0092] In some examples, color-coding allows users to forgo stent placement for a particular lesion and instead focus on a different lesion, such as one with a higher risk. In some examples, users may use color-coding to perform more aggressive post-dilation after stent placement.

[0093] The computing device may provide suggestions for balloon size and / or type based on color coding. For example, based on the machine learning model described above, the computing device may predict and therefore suggest balloon size and / or type based on similar cases. Additionally or alternatively, the computing device may provide suggestions for device use. Suggestions for use may be based on specific plaque morphologies in vascular segments. In some examples, suggestions for device use may be based on machine learning models that focus on device use on a case-by-case basis.

Claims

1. A method for displaying characteristics of one or more arteries related to a first group of frames obtained from a first pullback and a second group of frames obtained from a second pullback, comprising: one or more processors receiving the first group of frames and the second group of frames; the one or more processors detecting structural artery characteristics in each frame of the first group of frames and the second group of frames; the one or more processors scoring the structural artery characteristics detected in each frame of the first group of frames and the second group of frames; the one or more processors outputting a representation of the first group of frames and the second group of frames; wherein the output includes a difference between scores of the structural artery characteristics detected in at least one frame of the first group of frames and at least one frame of the second group of frames, or a visual representation of a change in the structural artery characteristics detected in the at least one frame of the first group of frames and the at least one frame of the second group of frames, and the representation of the second group of frames includes a representation of a percutaneous intervention.

2. The method of claim 1, further comprising the one or more processors aligning the at least one frame of the first group of frames with the at least one frame of the second group of frames based on the score of the structural artery characteristics detected in the at least one frame of the first group of frames and the score of the structural artery characteristics detected in the at least one frame of the second group of frames.

3. Outputting a representation further includes the one or more processors outputting at least one value, mark, or visual cue, wherein the at least one value, mark, or visual cue includes color or hatching. The method of claim 1.

4. The method of claim 3, wherein the color is a color-coding based on the score of the detected artery characteristics.

5. The method of claim 1, wherein the detected structural artery characteristic is a calcium load.

6. ​ Scoring the calcium load is the method according to claim 5, based on the determined calcium arc or the determined calcium amount. **Claim 7** The method according to claim 5, further comprising, by the one or more processors, predicting stent expansion in frame units based on the scored calcium load. **Claim 8** The method according to claim 1, wherein the structural arterial characteristics include at least one of a visible media, a lipid plaque, a thin fibrotic dysplasia, the inner lumen diameter of the lumen, the eccentricity of the lumen, the external elastic lamina, or a calcium load.

Images