Patent Application
20230326127
The present disclosure relates to medical image processing methods for analysis of coronary artery stenoses, and in particular to simulation of stents.
Coronary artery disease develops when atherosclerotic plaques build up inside the coronary artery, leading to stenosis in lumen, which reduces blood supply to the myocardium. Revascularization of coronary stenosis, including percutaneous coronary intervention (PCI), may become necessary when the stenosis results in ischemia.
Although the usage of invasive fractional flow reserve (FFR) measurement is recommended, its usage is low due to a variety of factors, including cost, time and the potential adverse effects associated with the invasive FFR measurement. Its application is also challenging for bifurcation and tandem lesions. We have adopted a system to assess non-invasive FFR, named as FFRB, by combining both computed tomography coronary angiography (CTCA) and reduced-order computational fluid dynamics (CFD) techniques. Based on the platform, we propose to model virtual stenting for assessing the functional status of the coronary lesions post-PCI so that the operator can review and compare different implantation strategies and their impact on coronary physiology post-PCI. This will enhance appropriate patient selection for invasive treatment strategy, and individualized stent selection and deployment planning, potentially transforming clinical decision-making and treatment planning in coronary artery disease.
FFR is the gold standard to select the patients who are likely to benefit from the revascularization and assess the functional improvement post-revascularization. However, FFR is measured during invasive coronary angiography (ICA) at hyperaemic state induced by the administration of adenosine, adenosine 5′-triphosphate or papaverine. To measure FFR, a pressure-sensor-tipped wire is inserted into the target diseased coronary artery via the catheter. The pressures distal and proximal to the stenosis are recorded and their ratio used to quantify FFR. For coronary artery disease patients with lesions FFR >0.8, medical therapy is recommended. Coronary revascularization is advocated for those lesions with FFR ≤0.8. Post-revascularization FFR >0.8 signifies procedural success. Although FFR-guided coronary revascularization has been shown to enhance clinical outcomes, FFR measurements are not widely used in clinical practice [5] due to the high medical cost, additional procedure time and potential complications involved in the invasive procedure. A non-invasive methodology is needed to discriminate the ischemic lesion and assess the outcome of revascularization.
In terms of revascularization, PCI has proven to be effective for restoring the blood flow to heart for coronary artery disease patients. The global coronary stents market at 5.2 billion US dollar in 2017 is projected to increase to 8.4 billion US dollars by 2025. Decisions for implantation location, stents selection (e.g., size/type/number) and stent implantation strategies remain complex and challenging, especially when treating bifurcation and tandem lesions, which are prone to in-stent restenosis (ISR), stent thrombosis and the associated adverse clinical events. Although use of FFR measurement is recommended during PCI for treating bifurcation lesions, it is challenging when re-crossing the pressure wire through the struts of the first implanted stent in main vessel. For tandem lesions, it is challenging to assess the severity of individual stenoses with FFR measurements, due to physiological interplay. Therefore, a tool to model the functional status of the coronary lesions after PCI with different implantation strategies is needed to aid PCI treatment planning.
CTCA is a valuable non-invasive test to assess coronary artery disease based on high-resolution anatomical depiction of the lesions. It has high sensitivity and low specificity, however, for discriminating ischemic coronary lesions. Combining CTCA and CFD modelling of intracoronary hemodynamics has made non-invasive assessment of the functional significance of coronary lesions feasible. An example of a non-invasive method FFRB using reduced order CFD modelling and novel outlet boundary conditions is set out in the following publication: Zhang JM, Zhong L, Luo T, Lomarda AM, Huo Y, Yap J, Lim ST, Tan RS, Wong ASL, Tan JWC, Yeo KK, Fam JM, Keng FYJ, Wan M, Su B, Zhao X, Allen JC, Kassab GS, Chua TSJ, Tan SY. Simplified models of non-invasive fractional flow reserve based on CT images. PLOS ONE 2016;11(5):e0153070 (DOI:10.1371/journal.pone.0153070).
According to a first aspect of the present disclosure, a method of simulating a stent in a coronary artery lumen structure is provided. The method comprises: reconstructing a three-dimensional coronary artery tree from segmented coronary lumen contours; replacing part of the three-dimensional artery tree with a candidate stent structure; and simulating pressure distribution through the coronary artery tree to determine a non-invasive fractional flow reserve through the candidate stent structure.
In an embodiment, the method further comprises identifying a candidate location for the candidate stent structure.
In an embodiment, identifying the candidate location for the candidate stent structure comprises: determining a mean lumen area as a function of straightened length of a vessel; identifying a proximal point of a lesion and a distal point of the lesion from the mean lumen area as a function of straightened length of the vessel; and determining the candidate stent location from the proximal point of the lesion and the distal point of the lesion.
In an embodiment, identifying the proximal point of the lesion and the distal point of the lesion from the mean lumen area as a function of straightened length of the vessel comprises determining a curvature of the mean lumen area as a function of straightened length of the vessel and identifying the proximal point of the lesion and the distal point of the lesion as maxima in the absolute values of the curvature.
In an embodiment, identifying the proximal point of the lesion and the distal point of the lesion from the mean lumen area as a function of straightened length of the vessel comprises determining a change of slope of the mean lumen area as a function of straightened length of the vessel and identifying the proximal point of the lesion and the distal point of the lesion as maxima of the change of slope.
In an embodiment, identifying the proximal point of the lesion and the distal point of the lesion from the mean lumen area as a function of straightened length of the vessel comprises identifying a location of a minimum in the mean lumen area and identifying the proximal point of the lesion and the distal point of the lesion on respective sides of the minimum in mean lumen area.
In an embodiment, the method further comprises determining a diameter for the candidate stent structure from the mean lumen diameter at the proximal point of the lesion and the distal point of the lesion.
In an embodiment, the method further comprises determining a length for the candidate stent structure from the location from the proximal point of the lesion and location of the distal point of the lesion.
In an embodiment, the method further comprises adding an extension allowance to the length for the candidate stent structure.
In an embodiment, the method further comprises segmenting a coronary artery lumen structure from a set of computed tomography coronary angiography images to obtain the segmented coronary lumen contours.
In an embodiment segmenting the coronary artery lumen structure from a set of computed tomography coronary angiography images to obtain the segmented coronary lumen contours comprises: designating points at aortic sinus as starting points of coronary artery trees; determining vessel centerlines for arteries of the artery trees; using the centerlines to create a stretched multiplanar reformatted volume for segments of the artery trees; extracting longitudinal cross sections from the stretched multiplanar reformatted volume;
detecting lumen borders in the extracted longitudinal cross sections; and detecting lumen border contours in slices of the multiplanar reformatted volume using the detected lumen borders.
In an embodiment, the cross sections are extracted from the stretched multiplanar reformatted volume at 45 degree intervals.
In an embodiment, the vessel centerlines are determined using a Hessian filter.
According to a second aspect of the present disclosure, a computer readable carrier medium carrying processor executable instructions which when executed on a processor cause the processor to carry out a method as set out above.
According to a third aspect of the present disclosure, a medical image processing system for simulating a stent in a coronary artery lumen structure is provided. The medical image processing system comprises: a processor and a data storage device storing computer program instructions operable to cause the processor to: reconstruct a three-dimensional coronary artery tree from segmented coronary lumen contours; replace part of the three-dimensional artery tree with a candidate stent structure; and simulate pressure distribution through the coronary artery tree to determine a non-invasive fractional flow reserve through the candidate stent structure.
In an embodiment, the data storage device further stores computer program instructions operable to cause the processor to: identify a candidate location for the candidate stent structure.
In an embodiment, the data storage device further stores computer program instructions operable to cause the processor to identify the candidate location for the candidate stent structure by: determining a mean lumen area as a function of straightened length of a vessel; identifying a proximal point of a lesion and a distal point of the lesion from the mean lumen area as a function of straightened length of the vessel; and determining the candidate stent location from the proximal point of the lesion and the distal point of the lesion.
In an embodiment, identifying the proximal point of the lesion and the distal point of the lesion from the mean lumen area as a function of straightened length of the vessel comprises determining a curvature of the mean lumen area as a function of straightened length of the vessel and identifying the proximal point of the lesion and the distal point of the lesion as maxima in the absolute values of the curvature.
In an embodiment, identifying the proximal point of the lesion and the distal point of the lesion from the mean lumen area as a function of straightened length of the vessel comprises determining a change of slope of the mean lumen area as a function of straightened length of the vessel and identifying the proximal point of the lesion and the distal point of the lesion as maxima of the change of slope.
In an embodiment, identifying the proximal point of the lesion and the distal point of the lesion from the mean lumen area as a function of straightened length of the vessel comprises identifying a location of a minimum in the mean lumen area and identifying the proximal point of the lesion and the distal point of the lesion on respective sides of the minimum in mean lumen area.
In an embodiment, the data storage device further stores computer program instructions operable to: determine a diameter for the candidate stent structure from the mean lumen diameter at the proximal point of the lesion and the distal point of the lesion.
In an embodiment, the data storage device further stores computer program instructions operable to: determine a length for the candidate stent structure from the location from the proximal point of the lesion and location of the distal point of the lesion.
In an embodiment, the data storage device further stores computer program instructions operable to: add an extension allowance to the length for the candidate stent structure.
In an embodiment, the data storage device further stores computer program instructions operable to: segment a coronary artery lumen structure from a set of computed tomography coronary angiography images to obtain the segmented coronary lumen contours.
In an embodiment, the data storage device further stores computer program instructions operable to: segment a coronary artery lumen structure from a set of computed tomography coronary angiography images to obtain the segmented coronary lumen contours by: designating points at aortic sinus as starting points of coronary artery trees;
In an embodiment, the cross sections are extracted from the stretched multiplanar reformatted volume at 45 degree intervals.
In an embodiment, the vessel centerlines are determined using a Hessian filter.
In the following, embodiments of the present invention will be described as nonlimiting examples with reference to the accompanying drawings in which:
The current disclosure relates to systems and methods to quantify the coronary artery stenosis, model the percutaneous coronary intervention (PCI) procedure and assess the hemodynamics before and after virtual stenting. The whole procedure includes coronary artery segmentation, centerline extraction, centerline tracking, cross sectional artery image generation, artery lumen segmentation, stenosis detection/quantification, shape-restoration to mimic the scenario with implanted stents (size, length and number), deployment, non-invasive fractional flow reserve (FFRB) assessment before and after virtual stenting.
In a CTCA imaging step 12, computed tomography coronary angiography image acquisition is carried out. The imaging is carried out according to the Society of Cardiovascular Computed Tomography guidelines. The CTCA imaging may be performed on contemporary multi-slice computed tomography scanners yielding high image spatial resolution such as the following: Toshiba Aquilion One 320 slice, Canon Aquilion ONE / Genesis 640, Siemens Somatom Force Dual Source and the Philips Brilliance iCT and others. Heart rate moderating beta-blockers or other drugs including ivabradine may be administrated to patients with resting heart rate >65 beats/minute, and sublingual glyceryl trinitrate may be administered prior to each scan. Prospective ECG-triggered scanning mode may be used for CTCA scans. All CTCA images may be saved in DICOM format.
In an image segmentation step 14, CTCA image processing is performed to segment the coronary artery lumen structure. The contour detection in the transverse slice is guided by the intersection points of each transverse slice with the longitudinal contours. The delineation of the lumen contours is supported in both the transversal and longitudinal planes. The segmented lumen contours are saved for the processing in the next step.
In a 3D patient specific model reconstruction step 16, all segmented coronary lumen contours are merged together to generate the patient-specific 3D reconstructed coronary artery model.
In a simulate stenting step 18, part of the 3D model of the patient’s coronary artery tree is replaced by a simulated stent. In some embodiments, a stenotic region of the coronary artery tree is detected and this stenotic region is replaced by the simulated stent. In other embodiments the stenotic region is manually selected by a user for replacement by the simulated stent. More than one stenotic region may be replaced by a stimulated stent.
Following the simulate stenting step 18, a model cleaning/meshing step 20 is carried out. In the model cleaning/meshing step 20, the 3D model is cleaned to have the inlet and outlet boundaries perpendicular to the respective vessel centerlines. Next, the computational domain of the cleaned-up model is discretized into tetrahedral, prisms or hexahedral shaped elements with boundary inflation layers.
Then, computational fluid dynamics (CFD) simulation 22 is carried out. The continuity and momentum-conservation (also known as the Navier-Stokes equations) equations for blood flow in the coronary arteries are solved for the computational domain using finite volume method, subject to patient-specific boundary conditions. Boundary conditions refer to the physiological conditions existing at the boundaries of the model being simulated. In our method, the inlet condition is set according to the mean brachial blood pressure of the patient. Outlet boundary conditions are specified by user defined functions which model the coronary microvasculature (parameterized using patient specific information). A no-slip boundary condition is set at the wall.
Following CFD simulation 22, a post processing step is carried out. In the post processing step 24, the pressure distribution on the coronary artery tree is used to calculate the non-invasive FFRB values.
The method 10 shown in
The medical image processing system 100 comprises a processor 110, a working memory 112, an input interface 114, a user interface 116, an output interface 118, program storage 120 and data storage 140. The processor 110 may be implemented as one or more central processing unit (CPU) chips. The program storage 120 is a non-volatile storage device such as a hard disk drive which stores computer program modules. The computer program modules are loaded into the working memory 112 for execution by the processor 110. The input interface 114 is an interface which allows data, such as patient computed tomography coronary angiography (CTCA) image data to be received by the medical image processing system 100. The input interface 114 may be a wireless network interface such as a Wi-Fi or Bluetooth interface, alternatively it may be a wired interface. The user interface 116 allows a user of the medical image processing system 100 to input selections and commands and may be implemented as a graphical user interface. The output interface 118 outputs data and may be implemented as a display or a data interface.
The program storage 120 stores a segmentation module 122, a 3D model reconstruction module 124, a stent simulation module 126, and a fluid dynamics simulation module 128. The computer program modules cause the processor 110 to execute various medical image processing which is described in more detail below. The program storage 120 may be referred to in some contexts as computer readable storage media and/or non-transitory computer readable media. As depicted in
The method 300 is carried out on medical image data such as computed tomography coronary angiography (CTCA) image data of a patient.
In step 302, the segmentation module 122 is executed by the processor 110 of the medical image processing system 100 to segment the coronary artery structure of the patient from the medical image data of the patient.
CTCA image processing is performed to segment the coronary artery lumen structure. Points at the aortic coronary sinuses may be used to designate the starting points of the left and right coronary artery trees. Next, vessel centerline may be obtained using Hessian filter and vessel detection, or artificial intelligence (Al) methods, or in combination. Based on the centerline, a stretched multi-planar-reformatted (MPR) volume may be created for each segment of interest. Subsequently, 4 longitudinal cross sections, at 45° interval, may be extracted from the multiplanar reformatted MPR volume. Then, lumen borders in these 4 longitudinal images may be detected by a model-guided minimum cost approach (MCA). Similarly, lumen border contours may be detected in the transverse slice of the multiplanar reformatted MPR volume using MCA with a circular lumen model or other lumen models. Alternatively, as well as in combination, Al approaches can be adopted to generate the contours as well.
As described above, Al may be used as alternative to Hessian filter and vessel detection, and Al can be used as alternative to or in combination with MCA approach for contour delineation. Al methods for centerline extraction may include convolutional neural network as described in: Wolterink JM, van Hamersvelt RW, Viergever MA, Leiner T, Išgum I. Coronary artery centerline extraction in cardiac CT angiography using a CNN-based orientation classifier. Med Image Analysis 2019;51:46-60 and Ziqing Wan, Weimin Huang, Su Huang, Zhongkang Lu, Liang Zhong, Zhiping Lin. Coronary Artery Extraction from CT Coronary Angiography with Augmentation on Partially Labelled Data. Annu Int Conf IEEE Eng Med Biol Soc 2021.
Al methods used for coronary artery lumen segmentation may include deep learning framework using 3D U-Net: Weimin Huang, Lu Huang, Zhiping Lin, Su Huang, Yanling Chi, Jiayin Zhou, Junmei Zhang, Ru-San Tan, Liang Zhong. Coronary Artery Segmentation by Deep Learning Neural Networks on Computed Tomographic Coronary Angiographic Images. Annu Int Conf IEEE Eng Med Biol Soc 2018: 608-611, deep learning encoder-decoder such as Residual U-Net: Cheng Zhu, Xiaoyan Wang, et al. Cascaded residual U-net for fully automatic segmentation of 3D carotid artery in high-resolution multi-contrast MR images, Phys Med Biol 2021;66(4):045003, or other methods and/or combinations.
Returning now to
In step 306, the stent simulation module 124 is executed by the processor 110 of the medical image processing system 100 to simulate stenting by replacing one or more parts of the three-dimensional coronary artery tree with a candidate stent structure. The critical parameters for deploying stent include the starting and ending points of the stent or stents, and each corresponding diameter and length.
In some embodiments step 306 comprises determining locations and dimensions of a candidate stent structure and then simulating stenting by replacing part or parts of the three-dimensional coronary artery tree with a candidate stent structure or multiple candidate stent structure.
In other embodiments, step 306 comprises receiving a user input indicating one or more candidate stent structures and locations for the candidate stent structures and then simulating stenting by replacing part or parts of the three-dimensional coronary artery tree with a candidate stent structure or multiple candidate stent structure.
As shown in
As shown in
Using either the maximal absolute value of the curvature or the maximal absolute value of the change of slope, the two points “P” and “D” can be automatically determined separately by starting from “S” and then proceeding to the left and to the right, respectively. The distance from points “P” to “D” represents the minimum length of stenotic segment that needs to be replaced by a virtual stent. A user defined extension allowance (e.g. 2 to 3 mm) may be added to both ends of the “P” and “D” to ensure adequate landing zones and will allow some degree of freedom for commercial stent length selection within a range.
As can be seen from
In the 3D model, we can search the planes corresponding to the area of 8.04 mm2 and 6.16 mm2 respectively to decide the starting and ending planes of the stent segment (this is shown in
Current generation of commercial stents have nominal diameters and recommended balloon expansion limits due to their open-cell design. The diameter of the commercial stent chosen should have a range that spans the mean diameters at proximal and distal locations. Many coronary stents have open-cell design that allows for expansion beyond the nominal diameters. Depending on the stent type, the recommended maximum stent expansion limit may exceed 50% of nominal stent diameter.
The methodology of simulating deployment of a stent will now be described with reference to
Returning now to
During the fluid dynamics simulation, the continuity and momentum-conservation (also known as the Navier-Stokes equations) equations for blood flow in the coronary arteries are solved for the computational domain using finite volume method, subject to patient-specific boundary conditions. Boundary conditions refer to the physiological conditions existing at the boundaries of the model being simulated. In our method, the inlet condition is set according to the mean brachial blood pressure of the patient. Outlet boundary conditions are specified by user defined functions which model the coronary microvasculature (parameterized using patient specific information). A no-slip boundary condition is set at the wall.
Following the same technique, additional virtual stents with different corresponding stent diameters and lengths can be “deployed” at multiple locations in the 3D coronary artery model. The operator can then interactively review the anatomical reconstruction of employing various stenting strategies that can differ in terms of stent number, location, length and diameter. Once the model with deployed stent is ready, non-invasive FFRB after virtual stenting can be calculated. With this, the operator can review the physiological effects of the planned revascularization strategy.
As described above, the proximal and distal points, stent length, and stent diameter at each coronary lesion may be determined during the execution of the method. User definable allowances proximal and distal to the proximal and distal ends of the virtual stent (e.g. 2 to 3 mm, in order to secure adequate landing zones for the stent), respectively, as well as user definable allowance for stent diameter expansion (to simulate stent inflation beyond nominal diameter in clinical practice) can be manually input by user to generate one or more user virtual stents. CFD modelling can be performed to generate pressure and FFRB maps for one or more virtual stent/s per coronary lesion.
In an embodiment of the proposed clinical application of the method, the user can select commercial stents that best fit the one or more virtual stents with favourable coronary hemodynamic properties. Commercial stents that approximate and are shorter and narrower than the selected virtual stent/s taking into account user defined virtual stent length and diameter allowances will be identified. The length and diameter of each selected commercial stent can then be fed into the model to simulate commercial stent implant, and CFD modelling and FFRB map generated automatically. The placement of the stent can be automatically performed by centring the centre of the stent with the centre of the lesion segment. The user has option to make manual adjustment of virtual stent placement, with CFD modelling and FFRB map generated automatically. Available as well as user selected commercial stent lengths, nominal diameters and diameter expansion limits with corresponding technical parameters of balloon expansion can be uploaded and updated into a library maintained by the proposed service and/or at user site.
An example to demonstrate the application an embodiment of the present invention for comparing different stenting strategies for tandem lesions along the left anterior descending artery is shown in
As shown in
A three dimensional artery tree model 820 is generated for the patient. The model includes the first lesion 812 and the second lesion 814. From this model, the non-invasive FFRB is calculated as 0.64.
Using the methods described above, three different stenting strategies are simulated. A first stenting strategy 830 involves applying a first stent 822 in the region of the first lesion 812 and a second stent 824 in the region of the second lesion 814. As shown in
Based on the simulation of stenting strategies described above with reference to
To validate the methodology, a proof-of-concept study was carried out. 18 patients with suspected or known coronary artery disease were recruited and underwent CTCA followed by invasive coronary angiogram with FFR measurements and subsequently PCI to treat hemodynamically significant lesions. FFR was measured both before and after stenting for a total of 21 vessels (22 lesions).
The calculated FFRB showed an excellent correlation (R=0.88, p<0.001) with FFR before stenting and a fair correlation (R=0.55, p< 0.001) after stenting (
Overall, FFRB was a promising index to diagnose the hemodynamic significance of coronary stenosis, as well as compute the hemodynamic outcomes of the stenting procedure.
Coronary artery disease causes myocardial ischemia and contributes to 13% of deaths globally (projected to increase to 15% by 2030, World Health Organization 2011). PCI involving the implantation of intracoronary stents is an effective revascularization therapy to reduce ischemia in coronary artery disease. Decisions for stent selection (e.g., size /type/number), implantation location and strategies are challenging, especially when treating complex bifurcation and tandem lesions, which are prone to in-stent restenosis, late thrombosis and the associated adverse clinical events. Therefore, a tool to predict the functional status of the coronary lesions before and after PCI treatment will great help in management decision and procedure planning.
| Number | Date | Country | Kind |
|---|---|---|---|
| 10202008205U | Aug 2020 | SG | national |
The present application is a national phase entry under 35 U.S.C. § 371 of International Application No. PCT/SG2021/050510, filed Aug. 26, 2021, published in English, which claims the benefit of the filing date of Singapore Patent Application No. 10202008205U, filed Aug. 26, 2020, the disclosures of which are incorporated herein by reference.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/SG2021/050510 | 8/26/2021 | WO |
