WO2025087726A1

WO2025087726A1 - Methods and systems for deriving patient-specific 3d flow field parameters for a medical procedure

Info

Publication number: WO2025087726A1
Application number: PCT/EP2024/078957
Authority: WO
Inventors: Nicole VARBLE; Bernardus Hendrikus Wilhelmus Hendriks; Leili SALEHI; Vipul Shrihari Pai Raikar; Alyssa Torjesen; Rene Leonardus Jacobus Marie UBACHS
Original assignee: Koninklijke Philips NV
Current assignee: Koninklijke Philips NV
Priority date: 2023-10-23
Filing date: 2024-10-15
Publication date: 2025-05-01
Anticipated expiration: 2026-04-23

Abstract

A hemodynamic analysis system (200) configured to determine one or more patient-specific three-dimensional (3D) flow field parameters from imaging data of a patient during a medical procedure. The imaging data comprises patient-specific anatomical data of an imaged blood vessel. The system includes a processor (220) configured to derive the one or more patient-specific 3D flow field parameters for the imaged blood vessel based on analysis of the imaging data and determine a treatment prediction for the medical procedure based on the derived one or more patient-specific 3D flow field parameters.

Description

METHODS AND SYSTEMS FOR DERIVING PATIENT- SPECIFIC 3D FLOW FIELD PARAMETERS FOR A MEDICAL PROCEDURE

Field of the Disclosure

[0001] The present disclosure is directed generally to methods and systems configured to determine one or more patient-specific three-dimensional (3D) flow field parameters from imaging data of a patient during or for a medical procedure.

Background

[0002] Understanding the patient-specific hemodynamic environment of a particular blood vessel can be extremely important during certain medical procedures. However, modeling the patient-specific hemodynamic environment requires intense computation and utilization of patient-specific three-dimensional (3D) flow field parameters, a process that is non-trivial and computationally expensive.

[0003] An example of a medical procedure that benefits from modeling the patient-specific hemodynamic environment of a particular blood vessel is thoracic endovascular aortic arch repair (TEVAR). TEVAR is a challenging procedure with high incidence of post-procedure adverse events. The angulation of the proximal landing zone (PLZ), where the implant is placed, represents a well-established risk factor for endograft failure. However, current clinical practice doesn’t consider the hemodynamic environment of the aortic arch and the potential effect of the pulsatile displacement forces (DF) on the implant that may cause an insufficient proximal seal or endograft migration. However, determining this patient-specific biomechanical information intra- procedurally is both challenging and computationally expensive.

[0004] Another example of a medical procedure that benefits from modeling the patientspecific hemodynamic is during intracranial aneurysm (IA) diagnosis and repair. IAS are outpouchings in the arteries of the brain. If IAs rupture, it most often leads to death or disability. The most widely used metric for determining risk of rupture is IA size. However small aneurysms rupture at approximately the same rate as large IAs. It has been suggested that rupture risk can be determined more specifically based on 3D aneurysm shape and blood flow in and around the IA. [0005] Yet another example of a medical procedure that benefits from modeling the patientspecific hemodynamic environment of a particular blood vessel is coronary vessel stenosis diagnosis and repair. Narrowing of the coronary arteries is typically caused by atherosclerosis. The extent of narrowing determines whether or not treatment is warranted. The amount of narrowing is quantified by fractional flow reserve (FFR), which measures blood pressure after a stenosis relative to the pressure before the stenosis. Typically, FFR values below 0.8 are considered as significant enough to treat by, for example, balloon angioplasty or by stenting. FFR can be determined invasively by interventional devices such as a flow wire but can also be calculated non-invasively from a coronary computed tomography angiogram (CCTA). The hemodynamics before and after the stenosis can be derived from computational fluid dynamics (CFD) modeling. The blood flow before and after real or virtual treatment has been shown to determine treatment effectiveness.

[0006] Hemodynamics derived from CFD modelling can be useful in describing disease pathophysiology or determining treatment effectiveness. However, CFD simulations are computationally expensive, require lengthy set ups (including mesh preparation, discretization, boundary conditions), require many assumptions (such as inflow and outflow conditions, laminar and homogenous flow, etc.), and requires solving of partial differential equations (PDEs). This process can hamper the clinical application and validation of CFD as a tool. A method that can provide hemodynamic results while avoiding lengthy CFD simulations and incorporating patientspecific boundary conditions, could be a vital tool for modeling the patient-specific hemodynamic environment during a procedure, such as modeling TV AR, determining IA rupture risk, or determining FFR for coronary artery disease.

Summary of the Disclosure

[0007] There is thus a continued need for methods and systems that efficiently and accurately describe the patient-specific hemodynamic environment of a particular blood vessel using patientspecific boundary conditions, while also avoiding expensive and time-consuming CFD simulations.

[0008] Various embodiments and implementations are directed to a method and system for determining one or more patient-specific three-dimensional (3D) flow field parameters from imaging data of a patient during a medical procedure. According to an embodiment, the imaging data comprises patient-specific anatomical data of an imaged blood vessel. The system includes a processor which is configured, adapted, or programmed to analyze the received imaging data to derive the one or more patient-specific 3D flow field parameters for the imaged blood vessel. These derived patient-specific one or more 3D flow field parameters are then utilized during the medical procedure, which can be diagnosis, pre-treatment, or treatment, among other options.

[0009] According to an aspect, a hemodynamic analysis system for determination of a treatment prediction is provided. The system includes a processor configured to receive imaging data of a patient during a medical procedure, wherein the imaging data comprises patient-specific anatomical data of an imaged blood vessel, derive one or more patient-specific three-dimensional (3D) flow field parameters for the imaged blood vessel based on analysis of the imaging data, and determine the treatment prediction for the medical procedure based on the one or more patientspecific 3D flow field parameters.

[0010] According to an embodiment, the imaging data further comprises data regarding blood flow within the imaged blood vessel.

[0011] According to an embodiment, the processor comprises a trained physics-informed hemodynamic determination model configured to analyze the received imaging data to derive the one or more patient-specific 3D flow field parameters for the imaged blood vessel.

[0012] According to an embodiment, the system further includes a user interface configured to display a flow pattern within the imaged blood vessel, the flow pattern based on the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel.

[0013] According to an embodiment, displaying the flow pattern within the imaged blood vessel further comprises displaying the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel.

[0014] According to an embodiment, displaying the flow pattern comprises displaying a plurality of flow fields each comprising a flow direction and magnitude within the imaged blood vessel.

[0015] According to an embodiment, the processor is further configured to analyze one or more of patient data and/or blood vessel imaging data parameters when deriving the one or more patientspecific 3D flow field parameters for the imaged blood vessel. [0016] According to an embodiment, the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel comprise at least one of pressure, pressure gradient, and 3D velocity fields (u, v, and w).

[0017] According to an embodiment, the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel comprise one or more of wall shear stress (WSS), oscillatory shear index (OSI), and fractional flow reserve (FFR).

[0018] According to an embodiment, the processor is further configured to generate, from the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel, a risk score quantifying one or more of a predicted likelihood of success for a treatment of the imaged blood vessel and a predicted likelihood of rupture of the imaged blood vessel.

[0019] According to an embodiment, the processor is further configured to generate, from the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel, a confidence metric quantifying confidence in the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel.

[0020] According to an embodiment, the anatomical data of the imaged blood vessel is obtained via one or more of digital subtraction angiography (DSA), computerized tomography (CT) scan, positron emission tomography (PET), magnetic resonance imaging (MRI), and ultrasound.

[0021] According to another aspect is a hemodynamic analysis system for determination of an implant placement prediction. The system includes a processor configured to: receive imaging data comprising anatomical data of an imaged blood vessel of the patient, derive one or more patientspecific 3D flow field parameters for the imaged blood vessel at each of a plurality of different locations along the blood vessel based on analysis of the imaging data, determine, based on the derived one or more patient-specific 3D flow field parameters for each of the plurality of different locations, one or more implant 3D flow field parameters of a blood vessel implant within the imaged blood vessel, and determine a placement prediction of the blood vessel implant within the imagined blood vessel based on the one or more implant 3D flow field parameters.

[0022] According to an embodiment, the processor is further configured to generate, using the derived one or more patient-specific 3D flow field parameters for each of the plurality of different locations, a predicted best placement for the blood vessel implant within the imaged blood vessel. [0023] According to an embodiment, the processor is further configured to generate a risk determination quantifying a risk of placement failure for the blood vessel implant at one or more of the plurality of different locations.

[0024] According to an embodiment, the processor comprises a trained physics-informed hemodynamic determination model configured to analyze the received imaging data to derive one or more patient-specific 3D flow field parameters for the imaged blood vessel.

[0025] According to another aspect is a hemodynamic analysis system configured to analyze fluid dynamics relative to a blood vessel implant in a patient. The system includes a processor configured to: (i) analyze imaging data comprising anatomical data of an imaged blood vessel of the patient, to predict one or more patient-specific 3D flow field parameters for the imaged blood vessel at each of a plurality of different locations along the blood vessel; and (ii) determine, using the derived one or more patient-specific 3D flow field parameters for each of the plurality of different locations, one or more implant 3D flow field parameters following a predicted placement of the blood vessel implant within the imaged blood vessel.

[0026] It should be appreciated that all combinations of the foregoing concepts and additional concepts discussed in greater detail below (provided such concepts are not mutually inconsistent) are contemplated as being part of the inventive subject matter disclosed herein. In particular, all combinations of claimed subject matter appearing at the end of this disclosure are contemplated as being part of the inventive subject matter disclosed herein. It should also be appreciated that terminology explicitly employed herein that also may appear in any disclosure incorporated by reference should be accorded a meaning most consistent with the particular concepts disclosed herein.

[0027] These and other aspects of the various embodiments will be apparent from and elucidated with reference to the embodiment s) described hereinafter.

Brief Description of the Drawings

[0028] In the drawings, like reference characters generally refer to the same parts throughout the different views. The figures showing features and ways of implementing various embodiments and are not to be construed as being limiting to other possible embodiments falling within the scope of the attached claims. Also, the drawings are not necessarily to scale, emphasis instead generally being placed upon illustrating the principles of the various embodiments.

[0029] FIG. 1 is a flowchart of a method for determining patient-specific three-dimensional (3D) flow field parameters from imaging data of a patient, in accordance with an embodiment.

[0030] FIG. 2 is a schematic representation of a hemodynamic analysis system, in accordance with an embodiment.

[0031] FIG. 3 is a flowchart of a method for training a hemodynamic determination model, in accordance with an embodiment.

[0032] FIG. 4 is a schematic representation of an example network architecture for a hidden physics model, in accordance with an embodiment.

[0033] FIG. 5 depicts example flow fields as output, in accordance with an embodiment.

[0034] FIG. 6 is an example display of a secondary parameter of hemodynamics, in accordance with an embodiment.

[0035] FIG. 7 is an example display comprising detailed calculated values of FFR for various lesions, in accordance with an embodiment.

[0036] FIG. 8 is a schematic representation of a physics-informed neural network, in accordance with an embodiment.

[0037] FIG. 9 is a schematic representation of a real-time patient specific simulation and training phantom setup, in accordance with an embodiment.

Detailed Description of Embodiments

[0038] The present disclosure describes various embodiments of a system and method configured to determine patient-specific three-dimensional (3D) flow field parameters from imaging data of a patient. More generally, Applicant has recognized and appreciated that it would be beneficial to provide a method and system to efficiently and accurately describe the patientspecific hemodynamic environment of a particular blood vessel using patient-specific boundary conditions. Thus, a hemodynamic analysis system receives imaging data comprising patientspecific anatomical data of an imaged blood vessel. The system includes a processor which is configured, adapted, or programmed to analyze the received imaging data to derive one or more patient-specific 3D flow field parameters for the imaged blood vessel. These derived patientspecific one or more 3D flow field parameters are then utilized during the medical procedure, which can be diagnosis, pre-treatment, or treatment, among other options.

[0039] The embodiments and implementations disclosed or otherwise envisioned herein can be utilized with any system or process that may utilize or benefit from analysis of the patient-specific hemodynamic environment of a blood vessel. The embodiments and implementations disclosed or otherwise envisioned herein can be utilized with any system that generates the imaging data comprising patient-specific anatomical data of an imaged blood vessel, including but not limited to Philips® imaging modalities and devices (manufactured by Koninklijke Philips, N.V.), among other products. However, the disclosure is not limited to these devices or systems, and thus disclosure and embodiments disclosed herein can encompass any system that may utilize or benefit from analysis of the patient-specific hemodynamic environment of a blood vessel.

[0040] Referring to FIG. 1, in one embodiment, is a flowchart of a method 100 for determining one or more patient-specific three-dimensional (3D) flow field parameters from imaging data of a patient during a medical procedure, using a hemodynamic analysis system. The methods described in connection with the figures are provided as examples only, and shall be understood not to limit the scope of the disclosure. The hemodynamic analysis system can be any of the systems described or otherwise envisioned herein. The hemodynamic analysis system can be a single system or multiple different systems.

[0041] At step 110 of the method, a hemodynamic analysis system 200 is provided. Referring to an embodiment of a hemodynamic analysis system 200 as depicted in FIG. 2, for example, the system comprises one or more of a processor 220, memory 230, user interface 240, communications interface 250, and storage 260, interconnected via one or more system buses 212. It will be understood that FIG. 2 constitutes, in some respects, an abstraction and that the actual organization of the components of the system 200 may be different and more complex than illustrated. Additionally, hemodynamic analysis system 200 can be any of the systems described or otherwise envisioned herein. Other elements and components of the hemodynamic analysis system 200 are disclosed and/or envisioned elsewhere herein. [0042] According to an embodiment, the hemodynamic analysis system 200 comprises or is in direct or indirect communication with an imaging modality 270. The imaging modality can be any modality sufficient to obtain imagery utilized by the hemodynamic analysis system 200 to determine one or more patient-specific three-dimensional (3D) flow field parameters. The most common forms of imaging modality are X-ray, magnetic resonance imaging (MRI), ultrasound, computed tomography scan (CT scan), and nuclear imaging such as Positron Emission Tomography (PET), although many other types of health- or medicine-based imaging modalities are possible. For example, the imaging modality may be an interventional or diagnostic ultrasound system capable of generating real-time 2D (2D + time) or 3D (3D + time) ultrasound images (such as EPIQ® system, Philips Lumify®, and Philips Affinity®, among many others), and/or an interventional x-ray imaging system capable of acquiring still as well as real time fluoroscopy images (such as Philips fixed c-arm X-ray systems Azurion® and Allura®, or Philips mobile c- arm systems Zenition® and Veradius®, among many others). The images obtained using the imaging modality may be obtained from a clinical provider or other individual.

[0043] According to an embodiment, the hemodynamic analysis system 200 comprises or is in direct or indirect communication with an image database 280. The image database may be any image database, and may comprise images or videos or reports or other data obtained using any imaging modality, such as imaging modality 270. The image database 280 may be local to the hemodynamic analysis system, and may optionally be a component of the system. The image database 280 may alternatively be remote to the hemodynamic analysis system, and thus is in direct or indirection communication with the hemodynamic analysis system. The imaging database may also contain previously obtained or calculated hemodynamic parameters from other patientspecific geometries.

[0044] According to an embodiment, the hemodynamic analysis system 200 comprises or is in direct or indirect communication with an electronic medical record system and/or an electronic medical records (EMR) database from which the information about patients, including demographic, diagnosis, and/or treatment information, may be obtained or received. For example, EMR database may comprise information about an imaging or treatment procedure for a patient, including the anatomy that will be imaged during the procedure. According to an embodiment, the electronic medical record system may be a local or remote database and is in direct and/or indirect communication with system 200. Thus, according to an embodiment, the system comprises an electronic medical record database or system.

[0045] At optional step 120 of the method, the hemodynamic analysis system 200 receives patient information relevant to the patient being imaged for analysis. This patient information can be utilized by the system at any point described or otherwise envisioned herein. The information can be received from an electronic medical record database or system. Alternatively, the information can be received via a user interface or other input from a physician or technician. As another alternative, the information can be extracted from an image taken of the patient. In some cases, the received information may include more than just the identification of a subject’s anatomy to be imaged or analyzed, including demographic information, treatment information, diagnosis, and/or other information. Once received, the information may be utilized immediately, and/or it may be temporarily or permanently stored in local and/or remote memory for future use.

[0046] At step 130 of the method, imaging of the patient is obtained using the imaging modality 270 of the hemodynamic analysis system 200. The patient may be any subject that could potentially benefit from a hemodynamic analysis. For example, the patient may be suspected of a vascular issue that could require hemodynamic analysis, or may be diagnosed with a vascular issue and requires an intervention. The imaging of the patient may be obtained during a procedure, which may be only imaging or may be imaging and a medical intervention. Thus, the imaging of the patient may be obtained before, during, and/or after a medical intervention. Other situations are possible. Accordingly, one or more types of imaging may be obtained of the relevant blood vessel(s) of the patient. The imaging can be obtained by a technician, a physician, or any other healthcare professional. The imaging may be done for any vessel of the patient. According to an embodiment, the most common locations for this imaging will be cranial vessels, coronary vessels, the aorta, and many other vessels. Once obtained, the imaging may be utilized immediately, and/or it may be temporarily or permanently stored in local and/or remote memory for future use, including in image database 280.

[0047] According to an embodiment, 3D images are obtained of patient anatomy/geometry using one or more of digital subtraction angiography (DSA), computerized tomography (CT) scan, positron emission tomography (PET), magnetic resonance imaging (MRI), X-ray, and ultrasound. To include patient-specific boundary conditions into the network, angiographic images, which are routinely taken, are leveraged. For example, DSA images can be recorded and a region of interest at the inflow and outflow parent vessels can be defined. The contrast injection rate and the parent vessel diameter may need to be recorded. The time density curve of contrast through the parent vessel are used to extract the velocity field in the parent vessel, which can be used as an input the to the physics neural model. This information can be noisy when compared to a detailed CFD model but instead complements that data with real world flow information. Thus, the imaging may be X-ray angiography and/or ultrasound doppler imaging sequences (2D or 3D). Additionally, 4D MRI flow information can also be added whenever this is available. It can be a good intermediate option and a reference standard for flow between DSA and CFD. According to an embodiment, therefore, the imaging data further comprises data regarding blood flow within the imaged blood vessel.

[0048] At step 140 of the method, the hemodynamic analysis system 200 receives input necessary to determine one or more patient-specific three-dimensional (3D) flow field parameters from imaging data. Thus, the input comprises imaging data of a patient during a medical procedure, such as the imaging obtained at step 120 of the method. According to an embodiment, the imaging data includes patient-specific 3D anatomical data of an imaged blood vessel. According to another embodiment, the input comprises information about the patient such as patient health records, demographic information, smoking history, blood pressure and other vitals, and potentially any other patient information that might be relevant to a hemodynamic analysis. According to another embodiment, the input comprises blood vessel imaging data parameters received or obtained from or about the imaging modality. Once obtained, the input may be utilized immediately, and/or it may be temporarily or permanently stored in local and/or remote memory for future use.

[0049] At step 150 of the method, the hemodynamic analysis system 200 analyzes the received input to derive the one or more patient-specific 3D flow field parameters for the imaged blood vessel. These derived patient-specific one or more 3D flow field parameters can then be utilized during the medical procedure, which may be imaging or medical intervention. The hemodynamic analysis system can analyze the input, which includes the patient-specific 3D anatomical data of an imaged blood vessel for example, in a variety of different ways to predict or determine the patient-specific 3D flow field parameters. For example, a processor of the hemodynamic analysis system may be configured or programmed to receive and analyze the input to predict or determine the patient-specific 3D flow field parameters. In some embodiments, the processor applies a trained physics-informed hemodynamic determination model configured to analyze the input to predict or determine the patient-specific 3D flow field parameters. The input may comprise, for example, data regarding blood flow within the imaged blood vessel. Once generated, the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel, the optional risk score, and/or the optional confidence metric may be utilized immediately, and/or it may be temporarily or permanently stored in local and/or remote memory for future use.

[0050] According to an embodiment, the hemodynamic analysis system analyzes the received patient-specific 3D anatomical data for an imaged blood vessel, and optionally further includes within that analysis the patient information received at step 120 of the method from an EMR database or system, or other source such as via a user interface or other input from a physician or technician. For example, this information may include demographic information, treatment information, diagnosis, and/or other information. Thus, received information such as information about the procedure to be performed, the weight or age of the patient, and other information could be informative when the system is analyzing the received patient-specific 3D anatomical data.

[0051] According to an embodiment, the one or more predicted or determined patient-specific 3D flow field parameters can include one or more of pressure, pressure gradient, and 3D velocity fields (u, v, and w). Additionally or alternatively, the one or more predicted or determined patientspecific 3D flow field parameters can include one or more of wall shear stress (WSS), oscillatory shear index (OSI), and fractional flow reserve (FFR). Other patient-specific 3D flow field parameters are possible.

[0052] According to an embodiment, in addition to analyzing the received input to derive the one or more patient-specific 3D flow field parameters for the imaged blood vessel, the hemodynamic analysis system may be configured or programmed to generate, from the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel, a risk score quantifying one or more of a likelihood of rupture of the imaged blood vessel and a likelihood of success for a treatment of the imaged blood vessel. The method for generating a risk score by the hemodynamic analysis system is as described or otherwise envisioned herein.

[0053] According to an embodiment, in addition to analyzing the received input to derive the one or more patient-specific 3D flow field parameters for the imaged blood vessel, the hemodynamic analysis system may be configured or programmed to generate, from the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel, the geometric features of the patient blood vessels, or patient clinical features (such as age, sex, race, heart rate, blood pressure, etc.), a confidence metric quantifying confidence in the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel. The method for generating a confidence metric by the hemodynamic analysis system is as described or otherwise envisioned herein.

[0054] According to an embodiment, the hemodynamic analysis system uses a trained hemodynamic determination model configured to analyze the received input, including but not limited to patient-specific 3D anatomical data of an imaged blood vessel (as input to the model) to derive one or more patient-specific 3D flow field parameters for the imaged blood vessel (as output of the model). In some embodiments, the model is a trained physics-informed hemodynamic determination model. The trained hemodynamic determination model can be any model that can be trained to utilize the input to generate the output, as described or otherwise envisioned herein. For example, the hemodynamic determination model can be a neural network or other trained machine learning model, for example a convolutional neural network (CNN) or a transformer network or other neural network. Thus, according to an embodiment, the hemodynamic analysis system comprises a trained hemodynamic determination model that receives the input data and outputs at least patient-specific 3D flow field parameters, as well as optionally a risk score and/or a confidence metric. The risk score can be calculated from flow parameters as well as other parameters such as geometric features or clinical features. A high risk score indicates a high similarity to adverse outcomes (such as rupture or treatment failure).

[0055] The hemodynamic determination model can be trained in a variety of different ways. According to one embodiment, the hemodynamic determination model is trained in an unsupervised manner by designing loss functions that capture some or all of the user needs. Training of the hemodynamic determination model is further described elsewhere herein.

[0056] Referring to FIG. 3, in one embodiment, is a flowchart of a method 300 for training the hemodynamic determination model of the hemodynamic analysis system 200. This method may be performed by the hemodynamic analysis system, or may be performed by another system such as a specialized machine learning model training system. [0057] At step 310 of the method, the training system receives training data which will be used to train the model. The training data can be any data sufficient to train the model to utilize the described input data to generate the described output. For example, the training data may comprise imaging for each of a plurality of patients and procedures. This training data, which could be utilized in a supervised or unsupervised manner, can comprise imaging for 100s or 1000s of patients and/or procedures, and can be updated with new imaging. The training data may also comprise one or more patient-specific 3D flow field parameters which are generated by other mechanisms and are associated with the training images. The training data may also comprise other information, such as patient information. This training data may be curated by an expert such as a clinician, or it may be obtained and utilized without curation. The training data may be received from any source. For example, the training data may be received from an electronic medical record database or system, or any other component of the system or a training system. According to an embodiment, system 200 comprises or is in direct or indirect communication with an imaging database which comprises some or all of the training data set.

[0058] According to an embodiment, the training system may comprise a data pre-processor or similar component or algorithm configured to process the received training data. For example, the data pre-processor analyzes the training data to remove noise, bias, errors, and other potential issues. The data pre-processor may also analyze the input data to remove low quality data. Many other forms of data pre-processing or data point identification and/or extraction are possible.

[0059] At step 320 of the method, the training system trains the hemodynamic determination model, using the training data, to determine from input data the one or more patient-specific 3D flow field parameters, a risk score, and/or a confidence metric. The hemodynamic determination model is trained using any method for training such a model. The trained hemodynamic determination model is a unique model based on the training data used to train the model. Following training, the system comprises a trained hemodynamic determination model.

[0060] Thus, following training, the hemodynamic determination model is a specialized model configured to receive the specialized input (namely, imaging of a blood vessel) and generate the very specific output, namely the one or more patient-specific 3D flow field parameters, a risk score, and/or a confidence metric. [0061] At step 430 of the method, the trained hemodynamic determination model is stored for future use. According to an embodiment, the trained hemodynamic determination model may be stored in local or remote storage.

[0062] Returning to method 100 in FIG. 1, at step 160 of the method, the hemodynamic analysis system provides, via a user interface, one or more of the derived patient-specific 3D flow field parameters for the imaged blood vessel, the optional risk score, and/or the optional confidence metric. For example, according to an embodiment, the system may provide a flow pattern within the imaged blood vessel, where the flow pattern is based on the derived one or more patientspecific 3D flow field parameters for the imaged blood vessel. As another example, displaying the flow pattern within the imaged blood vessel further comprises displaying the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel. According to yet another example, displaying the flow pattern comprises displaying a plurality of flow fields each comprising a flow direction and magnitude within the imaged blood vessel. The user interface may be any display or interface that enables display of the generated output, such as via text, a visual representation, and/or other formats.

[0063] Referring to FIG. 2 is a schematic representation of a hemodynamic analysis system 200. System 200 may be any of the systems described or otherwise envisioned herein, and may comprise any of the components described or otherwise envisioned herein. It will be understood that FIG. 2 constitutes, in some respects, an abstraction and that the actual organization of the components of the system 200 may be different and more complex than illustrated.

[0064] According to an embodiment, system 200 comprises a processor 220 capable of executing instructions stored in memory 230 or storage 260 or otherwise processing data to, for example, perform one or more steps of the method. Processor 220 may be formed of one or multiple modules. Processor 220 may take any suitable form, including but not limited to a microprocessor, microcontroller, multiple microcontrollers, circuitry, field programmable gate array (FPGA), application-specific integrated circuit (ASIC), a single processor, or plural processors.

[0065] Memory 230 can take any suitable form, including a non-volatile memory and/or RAM. The memory 230 may include various memories such as, for example LI, L2, or L3 cache or system memory. As such, the memory 230 may include static random access memory (SRAM), dynamic RAM (DRAM), flash memory, read only memory (ROM), or other similar memory devices. The memory can store, among other things, an operating system. The RAM is used by the processor for the temporary storage of data. According to an embodiment, an operating system may contain code which, when executed by the processor, controls operation of one or more components of system 200. It will be apparent that, in embodiments where the processor implements one or more of the functions described herein in hardware, the software described as corresponding to such functionality in other embodiments may be omitted.

[0066] User interface 240 may include one or more devices for enabling communication with a user. The user interface can be any device or system that allows information to be conveyed and/or received, and may include a display, a mouse, and/or a keyboard for receiving user commands. In some embodiments, user interface 240 may include a command line interface or graphical user interface that may be presented to a remote terminal via communication interface 250. The user interface may be located with one or more other components of the system, or may located remote from the system and in communication via a wired and/or wireless communications network.

[0067] Communication interface 250 may include one or more devices for enabling communication with other hardware devices. For example, communication interface 250 may include a network interface card (NIC) configured to communicate according to the Ethernet protocol. Additionally, communication interface 250 may implement a TCP/IP stack for communication according to the TCP/IP protocols. Various alternative or additional hardware or configurations for communication interface 250 will be apparent.

[0068] Storage 260 may include one or more machine-readable storage media such as readonly memory (ROM), random-access memory (RAM), magnetic disk storage media, optical storage media, flash-memory devices, or similar storage media. In various embodiments, storage 260 may store instructions for execution by processor 220 or data upon which processor 220 may operate. For example, storage 260 may store an operating system 261 for controlling various operations of system 200.

[0069] It will be apparent that various information described as stored in storage 260 may be additionally or alternatively stored in memory 230. In this respect, memory 230 may also be considered to constitute a storage device and storage 260 may be considered a memory. Various other arrangements will be apparent. Further, memory 230 and storage 260 may both be considered to be non-transitory machine-readable media. As used herein, the term non-transitory will be understood to exclude transitory signals but to include all forms of storage, including both volatile and non-volatile memories.

[0070] While system 200 is shown as including one of each described component, the various components may be duplicated in various embodiments. For example, processor 220 may include multiple microprocessors that are configured to independently execute the methods described herein or are configured to perform steps or subroutines of the methods described herein such that the multiple processors cooperate to achieve the functionality described herein. Further, where one or more components of system 200 is implemented in a cloud computing system, the various hardware components may belong to separate physical systems. For example, processor 220 may include a first processor in a first server and a second processor in a second server. Many other variations and configurations are possible.

[0071] According to an embodiment, system 200 comprises or is in direct or indirect communication with an imaging modality 270. The imaging modality can be any modality sufficient to obtain imagery utilized by the hemodynamic analysis system 200 to determine one or more patient-specific three-dimensional (3D) flow field parameters. The most common forms of imaging modality are X-ray, magnetic resonance imaging (MRI), ultrasound, computed tomography scan (CT scan), and nuclear imaging such as Positron Emission Tomography (PET), although many other types of health- or medicine-based imaging modalities are possible. For example, the imaging modality may be an interventional or diagnostic ultrasound system capable of generating real-time 2D (2D + time) or 3D (3D + time) ultrasound images (such as EPIQ® system, Philips Lumify®, and Philips Affinity®, among many others), and/or an interventional x- ray imaging system capable of acquiring still as well as real time fluoroscopy images (such as Philips fixed c-arm X-ray systems Azurion® and Allura®, or Philips mobile c-arm systems Zenition® and Veradius®, among many others).

[0072] According to an embodiment, system 200 comprises or is in direct or indirect communication with an image database 280. The image database may be any image database, and may comprise images or videos or reports or other data obtained using any imaging modality, such as imaging modality 270. The image database 280 may be local to the hemodynamic analysis system, and may optionally be a component of the system. The image database 280 may alternatively be remote to the hemodynamic analysis system.

[0073] According to an embodiment, the system 200 may also comprise or be in direct or indirect communication with an electronic medical record system and/or an electronic medical records (EMR) database from which the information about patients, including demographic, diagnosis, and/or treatment information, may be obtained or received. For example, EMR database may comprise information about an imaging or treatment procedure for a patient, including the anatomy that will be imaged during the procedure. According to an embodiment, the electronic medical record system may be a local or remote database and is in direct and/or indirect communication with system 200. Thus, according to an embodiment, the system comprises an electronic medical record database or system.

[0074] According to an embodiment, storage 260 of system 200 may store one or more algorithms, modules, and/or instructions to carry out one or more functions or steps of the methods described or otherwise envisioned herein. For example, storage 260 may comprise, among other instructions or data, a trained hemodynamic determination model 262, training instructions 263, and/or reporting instructions 264.

[0075] According to an embodiment, the trained hemodynamic determination model 262 of the hemodynamic analysis system 200 is trained to analyze the received input, including but not limited to patient-specific 3D anatomical data of an imaged blood vessel (as input to the model) to derive one or more patient-specific 3D flow field parameters for the imaged blood vessel (as output of the model). The trained hemodynamic determination model can be any model that can be trained to utilize the input to generate the output, as described or otherwise envisioned herein. For example, the hemodynamic determination model can be a neural network or other trained machine learning model, for example a convolutional neural network (CNN) or a transformer network or other neural network. Thus, according to an embodiment, the hemodynamic analysis system comprises a trained hemodynamic determination model that receives the input data and outputs at least patient-specific 3D flow field parameters, as well as optionally a risk score and/or a confidence metric.

[0076] According to an embodiment, training instructions 263 direct the system to train a hemodynamic determination model 262 of the hemodynamic analysis system 200. The instructions direct the system to retrieve, obtain, or receive training data. The training data can be any data sufficient to train the model to utilize the described input data to generate the described output. For example, the training data may comprise imaging for each of a plurality of patients and procedures, as well as one or more patient-specific 3D flow field parameters which are generated by other mechanisms and are associated with the training images. The training data may also comprise other information, such as patient information. This training data may be curated by an expert such as a clinician, or it may be obtained and utilized without curation. The training data may be received from any source. For example, the training data may be received from an electronic medical record database or system, or any other component of the system or a training system. According to an embodiment, system 200 comprises or is in direct or indirect communication with an imaging database which comprises some or all of the training data set. The training instructions 263 further direct the system to train the hemodynamic determination model using the obtained training data. The hemodynamic determination model can be trained using a variety of different training methods. The training instructions 263 further direct the system to store the trained hemodynamic determination model for future use.

[0077] According to an embodiment, the hemodynamic analysis system 200 is configured to process many thousands or millions of datapoints in the input data used to train the hemodynamic determination model 262, such as via the training instructions 263. For example, generating a functional and skilled trained hemodynamic determination model from a corpus of training data requires processing of millions of datapoints from input data and generated features. This can require millions or billions of calculations to generate a novel trained hemodynamic determination model from those millions of datapoints and millions or billions of calculations. As a result, each trained hemodynamic determination model is novel and distinct based on the input data and parameters of the model, and thus improves the functioning of the system. Generating a functional and skilled trained hemodynamic determination model comprises a process with a volume of calculation and analysis that a human brain cannot accomplish in a lifetime, or multiple lifetimes.

[0078] According to an embodiment, reporting instructions 264 direct the system to provide the output of the system to a user, such as a clinician, via a user interface. The provided output can be any of the information as described or otherwise envisioned herein. The system may provide the information to a user via any mechanism, including but not limited to a visual display, an audible notification, a page, or any other method of notification. The information may be communicated by wired and/or wireless communication to another device. For example, the system may communicate the information to a mobile phone, computer, laptop, wearable device, and/or any other device configured to allow display and/or other communication of the information.

[0079] EXAMPLES

[0080] The following are provided as examples of the methods and systems configured to determine one or more patient-specific three-dimensional (3D) flow field parameters from imaging data of a patient during or for a medical procedure, as described or otherwise envisioned herein. Accordingly, it is understood that these are non-limiting examples, and that the methods and systems described or otherwise envisioned herein can be utilized for a variety of purposes, applications, and situations.

[0081] EXAMPLE 1 - Intracranial Aneurysm Rupture and Intervention

[0082] According to an embodiment, the methods and systems described or otherwise envisioned herein can be utilized for intracranial aneurysm (IA) analysis. By leveraging a physics- informed neural network model and patient-specific flow parameters from quantified digital subtraction angiography (DSA), this method can produce quick and accurate IA flow results. Quantifying IA flow using DSA alone has been shown to be a potential method for determining treatment effectiveness, but has been shown to be more accurate within the more uniform parent vessel rather than in the complex aneurysm sac. This makes the methods and systems described or otherwise envisioned herein appealing for defining patient-specific boundary conditions using computational fluid dynamic (CFD) modeling versus determining IA flow from DSA alone.

[0083] The IA flow results produced according to the methods and systems described or otherwise envisioned herein can streamline visualizations, which inform clinicians of the flow patterns and jet locations to plan for treatment strategies. The process also provides quantified parameters such as WSS and OSI that can be interpreted to an overall risk score regarding rupture risk or treatment effectiveness. Such an efficient and accurate tool can be easily incorporated into the clinical workflow.

[0084] Armed with flow parameters derived according to the methods and systems described or otherwise envisioned herein, data-driven decision making can be done. Flow quantification provides additional information that is previously difficult to access or is inaccessible. [0085] According to an embodiment, intracranial aneurysm analysis utilizes a trained hemodynamic determination model uses or is a trained physics-inspired neural network, as well as patient-specific flow parameters derived from DSA. The model produces flow parameters (pressure, pressure gradient, velocity) and derived flow parameters (WSS, OSI, FFR, etc.) that can be visualized and quantified within a patient-specific aneurysm geometry. A risk score can be produced from the quantified flow parameters that indicates similarity of an aneurysm to previously ruptured aneurysms or the likelihood of treatment success. A confidence metric may be output by the model which indicates the certainty of the model in its estimations. The risk score can be provided by a machine learning model developed based on assessment of IA cases as described or otherwise envisioned herein. Boundary conditions (inflow and outflow) could be determined from DSA that is routinely collected as a part of diagnosis. Risk scores may also include geometry information (size, morphology) and patient clinical information (age, sex, race, etc.).

[0086] Accordingly, for intracranial aneurysm analysis, the hemodynamic analysis system receives input that will be utilized for analysis. The input includes, but is not limited to, 3D patient anatomy/geometry typically taken from 3D DSA, CT, or MR images, among other possible sources. The input further includes 3D angiographic flow from imaging or detailed CFD simulations. This information can also be generated by setting up high detail CFD simulations based on patient anatomy from a collection of pre-procedure scans. High quality CFD simulations are difficult to setup but provide a dense sampling of data points which are accurate within the setup of the computational model but may not be patient or pathology specific. To include patientspecific boundary conditions into the network, angiographic images, which are routinely taken, can be leveraged. DSA images can be recorded and a region of interest at the inflow and outflow parent vessels can be defined. The contrast injection rate and the parent vessel diameter may need to be recorded. The time density curve of contrast through the parent vessel are used to extract the velocity field in the parent vessel, which can be used as an input the to the physics neural model. This information can be noisy when compared to a detailed CFD model but instead complements that data with real world flow information. Additionally, 4D MRI Flow information can also be added whenever this is available. Optionally, the input may include information about the patient such as patient health records, demographic information, smoking history, blood pressure and other vitals, and potentially any other patient information that might be relevant to a hemodynamic analysis. According to another embodiment, the input comprises blood vessel imaging data parameters received or obtained from or about the imaging modality.

[0087] According to an embodiment, the input is analyzed by the hemodynamic analysis system to generate the outputs. The outputs can include, for example, patient-specific 3D flow fields including 3D velocity fields (u, v, w) and pressure, as well as secondary flow parameters including WSS and OSI. The outputs may be displayed as an average (time or spatial average) or may be displayed dynamically over a cardiac cycle. The output may also include a risk stratification score, and/or a confidence metric.

[0088] According to an embodiment, the input is analyzed by a trained hemodynamic determination model of the hemodynamic analysis system to generate the outputs. As further described herein, the network can be trained to optimize a loss function that minimizes the residuals based on Navier-stokes equations. The architecture can be a combination of fully connected neural layers or any other architecture that optimizes for regression.

[0089] According to an embodiment, the system comprises an application or run-time network and visualization controller that utilizes a network architecture with a hidden physics model. Referring to FIG. 4, in one embodiment, is an example network architecture for a hidden physics model. Symbols x, y, z, and t represent input such as 3D patient anatomy, 2D angiographic flow, a pre-trained CFD model with corresponding flow fields, among other possible input. Symbols w, v, w, and P represent output such as patient-specific 3D flow field parameters (e.g., u, v, w and pressure), and secondary parameters WSS and OSI, among other possible output. Notably, other imaging modalities such as 4D DS A or 4D MRI may be used as input instead of or in addition to the quantitative DS A in order to increase network accuracy.

[0090] In the embodiment of FIG. 4, the Navier-stokes loss equations are as follows:

[0091] In the embodiment of FIG. 4, the experimental data loss is as follows:

[0092] Referring to FIG. 5, in one embodiment, is an example flow fields as output of the model, where the flow direction and magnitude are shown by streamlines. Visualizations like these can help decide treatment strategy. Thus, the system can display detailed real-time 3D flow patterns, such as velocity streamlines or vectors. This can be informative to a clinician as they can utilize it to identify locations of jets or flow stagnation points. Although color is removed from FIG. 5, velocity streamlines and vectors can be colored based on velocity magnitude.

[0093] Referring to FIG. 6, in one embodiment, is an example of a secondary parameter of hemodynamics derived by the system, namely normalized WSS in the aneurysm (top, although color is removed) and a risk score (bottom). Showing and displaying parameters calculated from velocity and pressure such as WSS and OSI can also be informative as clinicians can identify points of potential vessel wall vulnerability. Additionally, these parameters can be quantified as “aneurysm averaged” or “normalized based on the parent vessel.” The flow parameters may also be shown dynamically over a cardiac cycle. The quantified parameters may be incorporated into a risk score as illustrated in FIG. 6.

[0094] According to an embodiment, the risk score can be the output of a machine learning or Al model that has been previously developed or the model can be continuously improved based on new case information. Rupture risk can be on a continuous scale (such as 0 = not similar to ruptured IAS, 100% = very similar to ruptured IAS). Treatment success can also be on a continuous scale (such as 100% = likely to fully occlude, 0% = not likely to occlude). Treatment can be coiling, flow diversion, clipping etc. and a model for each treatment modality can be developed or the treatment modality can be incorporated into the treatment success score. [0095] According to an embodiment, the system may additionally output a map demonstrating the voxel-wise confidence of the network in the 3D reconstruction quality. Parameters such as 2D image quality (e.g., image contrast, framerate, pixel size, image resolution, etc.) might affect the reconstruction confidence.

[0096] According to an embodiment, a model trained with drop-out layers may be run multiple times on the same input during inference to generate slightly different outputs (since dropout drops the outputs from a specified number of nodes at random). These different outputs can be used to compute the mean and variance in the reconstructed 3D flow patterns (e.g., flow velocities, flow direction, etc.). The variance can be used to indicate a level of confidence in the output (i.e., high variance indicates that network output is not consistent and, therefore, confidence is low, while low variance indicates consistent output and high confidence).

[0097] According to an embodiment, physics-informed neural networks trained for solving 3D fluid dynamics PDEs have shown to be capable of generating 3D information based on information derived from 2D cross sections of flow. X-ray DSA provides these snapshots along with the estimation of flow parameters as input. In another embodiment, the network can also be trained to predict parameters to generate optimal c-arm angles and contrast injection parameters to get the best DSA images.

[0098] EXAMPLE 2 - Coronary Artery Analysis

[0099] According to an embodiment, the methods and systems described or otherwise envisioned herein can be utilized for analyzing coronary arteries. For example, determination of extent of the stenosis of a coronary artery or treatment effectiveness is addressed by calculating the FFR based on coronary angiograms using computer calculations. The methods significantly reduce calculation time, such that the method and system can be employed while a patient is in a catheterization laboratory for a coronary angiogram measurement, and eliminates the need for offline evaluation. Diagnosis and treatment of coronary artery disease can potentially be done in the same session and definitive diagnostic tests can be done more rapidly in emergency situations.

[00100] According to an embodiment, coronary artery analysis utilizes a trained hemodynamic determination model uses or is a trained physics-inspired neural network, as well as patientspecific flow parameters derived from DSA. The model produces flow parameters (pressure, pressure gradient, velocity) and derived flow parameters (WSS, OSI, FFR, etc.) that can be visualized and quantified within a patient-specific aneurysm geometry. A risk score can be produced from the quantified flow parameters that indicates the risk of rupture or the likelihood of treatment success. A confidence metric may be outputted by the model which indicates the certainty of the model in its estimations. The risk score can be provided by a machine learning model developed based on assessment of IA cases as described or otherwise envisioned herein. Boundary conditions (inflow and outflow) could be determined from DSA that is routinely collected as a part of diagnosis.

[00101] Accordingly, for coronary artery analysis, the hemodynamic analysis system receives input that will be utilized for analysis. The input includes, but is not limited to, 3D patient anatomy/geometry typically taken from CTA, 3D DSA, PET, or MR images, among other possible sources. The input further includes 3D angiographic flow from imaging or detailed CFD simulations. This information can also be generated by setting up high detail CFD simulations based on patient anatomy from a collection of pre-procedure scans. High quality CFD simulations are difficult to setup but provide a dense sampling of data points which are accurate within the setup of the computational model but may not be patient or pathology specific.

[00102] According to an embodiment, the hemodynamic analysis system receives as input pressure P from aortic pressure measurements or angiographic images. CTA or DSA images can be recorded and a region of interest at the inflow and outflow parent vessels can be defined. The contrast injection rate and the parent vessel diameter may need to be recorded. The time density curve of contrast through the parent vessel are used to extract the velocity field in the parent vessel, which can be used as an input the to the physics neural model. This information can be noisy when compared to a detailed CFD model but instead complements that data with real world flow information.

[00103] Optionally, the input may include information about the patient such as patient health records, demographic information, smoking history, blood pressure and other vitals, and potentially any other patient information that might be relevant to a hemodynamic analysis. According to another embodiment, the input comprises blood vessel imaging data parameters received or obtained from or about the imaging modality.

[00104] According to an embodiment, the input is analyzed by a trained hemodynamic determination model of the hemodynamic analysis system to generate the outputs. As further described herein, the network can be trained to optimize a loss function that minimizes the residuals based on Navier-stokes equations. The architecture can be a combination of fully connected neural layers or any other architecture that optimizes for regression.

[00105] According to an embodiment, the system comprises an application or run-time network and visualization controller that utilizes a network architecture with a hidden physics model. Referring again to FIG. 4, in one embodiment, is an example network architecture for a hidden physics model. Symbols x, y, z, and t represent input such as 3D patient anatomy, 2D angiographic flow, a pre-trained CFD model with corresponding flow fields, among other possible input. Symbols w, v, w, and P represent output such as patient-specific 3D flow field parameters (e.g., u, v, w and pressure), and secondary parameters WSS and OSI, as well as an FFR score for one or more lesions, and a risk stratification score, among other possible output.

[00106] Referring to FIG. 7, in one embodiment is a display comprising detailed calculated values of FFR for various lesions. These example FFR calculations are for various lesions before the proposed treatment (left) as well as a prediction of how it would look after the proposed treatment (right).

[00107] According to an embodiment, the hemodynamic analysis system can further generate a risk score. For example, for each detected coronary artery narrowing a risk score can be provided, which may indicate whether the lesion should be treated and what the effected heart mass is. Hence, apart from the calculated FFR values of the lesion also a recommendation for treatment can also be. The effect of this treatment on the FFR values can also be provided. According to an embodiment, the hemodynamic analysis system can further generate a map demonstrating the voxel-wise confidence of the network in the 3D reconstruction quality. Parameters such as 2D image quality (e.g., image contrast, framerate, etc.) might affect the reconstruction confidence.

[00108] According to an embodiment, a network trained with drop-out layers may be run multiple times on the same input during inference to generate slightly different outputs (since dropout drops the outputs from a specified number of nodes at random). These different outputs can be used to compute the mean and variance in the reconstructed 3D flow patterns (e.g., flow velocities, flow direction, etc.). The variance can be used to indicate a level of confidence in the output (i.e., high variance indicates that network output is not consistent and, therefore, confidence is low, while low variance indicates consistent output and high confidence). [00109] According to an embodiment, physics-informed neural networks trained for solving 3D fluid dynamics PDEs have shown to be capable of generating 3D information based on information derived from 2D cross sections of flow. X-ray DSA provides these snapshots along with the estimation of flow parameters as input. In another embodiment, the network can also be trained to predict parameters to generate optimal c-arm angles and contrast injection parameters to get the best DSA images.

[00110] The methods and systems described or otherwise envisioned herein provide a significant improvement over prior hemodynamic analysis systems. Indeed, the methods and systems described or otherwise envisioned herein vastly increase the speed at which FFRCT results are produced. To produce FFRCT results using prior art methods, images are obtained from a CTA and sent to a third-party company or institution outside of the cath lab. Results would be produced in approximately half a working day due to the use of traditional mesh generation and PDE solving methods. However, when implemented, the methods and systems described or otherwise envisioned herein not only obviate the need for analysis by a third party, they vastly increase speed of the analysis itself.

[00111] According to an embodiment, in addition to intracranial aneurysm rupture or restenosis, abdominal aneurysm growth and rupture, atherosclerotic plaque stability, and cardiovascular disease severity (stenosis), the methods and systems described or otherwise envisioned herein can be used for a variety of other applications. Additional applications where the methods and systems described or otherwise envisioned herein could be utilized include but are not limited to: 1. structural modeling or fluid structure interactions, such as with pulsatile flow, elastic walls, device deployment or device fatigue, 2. heat equations, such as ablation treatment in liver, kidney, and prostate, and 3. airway flow dynamics such as particle deposition.

[00112] EXAMPLE 3 - Implant Placement

[00113] According to an embodiment, the methods and systems described or otherwise envisioned herein can be utilized for facilitating implant placement within or near a blood vessel.

[00114] For example, aortic arch and descending aorta repair is a challenging procedure with surgical intervention being the gold standard. In patients who would tolerate such procedures, open surgery shows low procedural morbidity and good mid-term outcomes. In recent years, with the development of new endovascular therapies, Thoracic Endovascular Aortic Repair (TEVAR) has also emerged as a viable option for patients who would be at high risk of complications in an open procedure. Various clinical reports have highlighted it as a safe and effective alternative to a conventional approach.

[00115] However, TEVAR of the aortic arch and the descending aorta requires careful planning of the type, size, and placement location or landing zone of the implant with respect to the anatomy and associated pathology in addition to creating an interventional navigation plan. This is typically done based on pre-procedural CTA imaging. The size of the implant, location of the landing zone as well as the interventional approach all have a strong impact on post procedural complications. TEVAR requires proximal and distal landing zones of diameter (<40 mm) and length (>20 mm) and a viable iliofemoral or infrarenal aortic access route. Improper planning, particularly of the proximal landing zone, can lead to an insufficient seal and potential migration of the graft thereby failing to exclude the aneurysmal sac or pathology. This can have severe post-procedural complications ranging from endoleaks to aneurysm rupture. Studies have shown that the risk of these failure cases is dependent on the angulation and tortuosity of the arch. Steeper angulation of the arch necessitates a longer minimum length of the landing zone for minimizing complications. However, there isn’t a clear guideline for pre-operative planning that takes into account and provides a risk stratification based on the geometric variability of the anatomy seen in the population. So far, the majority of pre-operative planning and decision making is based on Ishimaru’ s aortic map classification which doesn’t factor in the tortuosity or angulation of the arch.

[00116] While careful planning based on geometric information helps the clinician deploy the implants with the potential to minimize post procedure complications, the local biomechanics in the arch is another important factor that can have a critical impact on efficacy. The tissue-device interface along with hemodynamics are central to proper implantation of the scaffolds in a turbulent environment. The local fluid dynamics contribute to the implant experiencing various displacement forces due to the elastic and pulsatile nature of the aorta. In addition to tissue motion and the complexity of the geometry, these forces can contribute to implant failure (improper sealing) and migration. The primary contributors to the displacement forces in the arch are the blood pressure and the geometry of the arch. Understanding these local mechanics at the device tissue interface may help in clinical decision making and improve the efficacy of TEVAR implants in the aortic arch and the descending aorta. [00117] Thus, TEVAR of the aortic arch can be a challenging procedure with high post procedural complications and implant failure. The properties of the landing zones for implants and the geometry of the aortic arch are important factors leading to failure. In addition, the local biomechanics of the tissue-implant interface also plays a major role due to the displacement forces generated in this dynamic environment. While creating accurate and patient specific biomechanical models for decision support are an open area of research, in-silico modelling methods such as CFD provide a good approximation. However, effectively developing such patient specific models is quite challenging as well as computationally expensive. This renders them ineffective for widespread use as such methods require handcrafting the simulations. Towards improving this, the methods and systems described or otherwise envisioned here can be used to create data-driven models leveraging the advances in AI/ML techniques to provide patient specific decision support and risk stratification during pre-procedure and intra-procedure steps in TEVAR of the aortic arch.

[00118] According to an embodiment, the methods and systems described or otherwise envisioned herein may be utilized for TEVAR decision support by providing data driven patientspecific risk classification for TEVAR procedures based calculating biomechanical information using imaging and pre-procedure information. According to an embodiment, the system provides a risk classification of stent migration or failure based on image derived hemodynamic environment/biomechanical properties of the tissue-device interface.

[00119] According to an embodiment, the methods and systems described or otherwise envisioned herein model biomechanical relationships between aortic arch anatomy and an implant based on hybrid and partial clinical patient data using deep neural networks, by providing stratification of risk as well as real-time navigation guidance based on intra-procedural data.

[00120] According to an embodiment, creating patient specific realistic biomechanical scenarios of the localized conditions in the aortic arch requires development of fluid dynamics models that capture the behavior of the tissue-blood-implant system by solving the Navier-Stokes equations. These equations are setup to characterize the motion of blood as an incomprehensible fluid, and the impact of such motion in a tissue domain, i.e., aortic arch. These methods, when performed using inefficient conventional methods, also require the explicit definition of the 3D geometry of the domain as well as the specific boundary conditions to solve for. This can involve handcrafting a set of equations for every data point (dataset or patient) and solving complex partial differential equation (PDEs) which can often only be approximated using computationally expensive algorithms. While a rich source of data when defined well, creating these computational models is cumbersome when applied to patient-specific scenarios and hence, can be hard to generalize and scale over a population. This can make development of patient specific models to use in interventional scenarios a challenging problem.

[00121] The advent of new machine learning methods has seen the development of a new class of neural networks that take a hybrid data driven approach to computational modelling. These new architectures, called physics-informed neural nets (PINN), utilize a hybrid approach of learning from data and maintaining the physical relationship in the natural process that the data represents by solving the mathematical equations (PDEs) implicitly via the process of auto-differentiation. The PDE is incorporated into the loss function and thereby, is solved implicitly during the optimization of the neural network. These networks have shown to be efficient and effective for solving ill-posed and inverse problems by modelling the physical relationships from noisy data.

[00122] Referring to FIG. 8, in one embodiment, is a schematic representation of a physics- informed neural network that utilizes high quality data from computational modelling done in an ideal setting with limited patient geometries along with sparse data coming from patient specific pre-procedural and intra-procedural imaging and measurements. The network is designed to implicitly solve the PDEs that describe biomechanical properties by incorporating them into the overall loss function that is governed by the available training data. Once optimized, the PDE solving is now learned and as an advantage, the network is capable of calculating parameters outside of what they were explicitly trained on but can be mathematically described by the underlying PDE’s that the network models. In the embodiment of FIG. 8, the Navier-stokes loss equations are as shown in Eq. 1 - Eq. 4 and the experimental data loss is as shown in Eq. 5.

[00123] Thus, the hemodynamic analysis system leverages partial sources of data generated based on CFD simulations, 2D fluoroscopy, as well as ultrasound imaging to create a decision support model that provides risk stratification during the planning and deployment of TEVAR implants in patient specific scenarios.

[00124] Accordingly, the hemodynamic analysis system receives input data comprising one or more of: (i) imaging from an interventional or diagnostic ultrasound system capable of generating real-time 2D (2D + time) or 3D (3D + time) ultrasound images (such as Philips EPIQ, Philips Lumify, and Philips Affinity, among other systems): (ii) images from an interventional x-ray imaging system capable of acquiring still as well as real time fluoroscopy images (such as a Philips fixed c-arm X-ray system like Azurion and Allura, or a Philips mobile c-arm -ray system like Zenition and Veradius, among other systems).

[00125] The hemodynamic analysis system is configured to process one or more of: (i) preprocedure Computed Tomography Angiography (CTA) imaging or Magnetic resonance imaging for flow (4D MR Flow); (ii) intra-procedural x-ray fluoroscopy or angiography images; (iii) intraprocedural ultrasound imaging; (iv) endovascular pressure data from pressure wires, and/or (v) computational modelling solvers and neural network models.

[00126] According to an embodiment, the hemodynamic analysis system comprises a computational modelling controller configured to perform and generate fluid dynamics data by: (i) receiving retrospective 2D/3D angiography/ 4D MR Flow data from a set of patient population, where the data comprises (1) the anatomical information of the patient and disease state of the aortic arch and (2) flow information of blood; (ii) generating high dimensional tetrahedral mesh of the anatomy from the CTA images; (iii) solve incompressible Navier-Stokes equations assuming constant viscosity for blood; (iv) iteratively generate and process data generated by solvers; and (v) iteratively solve for (1) various boundary conditions as defined based on the pathology and (2) conditions that are defined by the presence of various interventional tools and implants within the arch at typical deployment states.

[00127] According to an embodiment, the hemodynamic analysis system comprises a hemodynamic determination model that is trained by receiving and processing retrospective ground truth data from patient population of healthy and diseased aortic arch aneurysms. The data is: (1) data that consists of blood flow vectors (velocity and direction), concentration of blood in the boundary conditions defined within the fluid model, pressure, wall shear stress; (2) paired data from intervention or diagnostic ultrasound that provides doppler flow information (velocity and direction) as well as pressure measurement information from an endo-vascular pressure sensor at various points in the arch (this data can be sparse); and/or (3) paired data from interventional fluoro-angiography that provides flow information in the form of contrast flow patterns (direction and velocity) as well as pressure measurements from an endo-vascular pressure sensor at various points in the arch (this data can be sparse).

[00128] The hemodynamic determination model is trained based on the inputs as being flow information of location of the blood flow vectors in time and output being velocities of blood (pressure if available). An example architecture can comprise of a set of fully connected layers of varying width and depth. The activation function used can be sine and the network can be optimized using the ADAM optimizer.

[00129] According to an embodiment, the hemodynamic determination model is optimized based on minimizing the following loss function:

LOSS Ldata d” L_PE)E (Eq. 6) where L_data minimizes the velocities based given the positional information of the flow vectors based on dense observations from fluid dynamics models and sparse observations from ultrasound and fluoroscopy imaging.

[00130] According to an embodiment, the L_PDE loss term is of the type:

where,

, described by the Navier — Stokes equations.

[00131] During optimization, the L_PDE loss term forces the network to implicitly optimize the Navier-stokes equations for incompressible fluids thereby learning to predict derived fluid dynamic properties governed by physics-based constraints. During this optimization Pe (Peclet) and Re (Reynolds) constants can be provided as hyperparameters but can also be tunable as learned parameters. As the network is designed to optimize for a data loss and simultaneously learn the solutions for the Navier-stokes equations, other parameters such as wall shear stress (WSS) can be derived while purely optimizing for the velocity term.

[00132] According to an embodiment, the hemodynamic determination model can be trained on retrospective and longitudinal data, and can also be fine-tuned for patient specific pre-procedural CTA data, often taken days before the procedure, which can add to its predictive power.

[00133] According to an embodiment, an inference phase controller of the system is configured to receive flow information from x-ray angiography and/or ultrasound doppler imaging sequences (2D or 3D) intra-procedurally, and then the trained hemodynamic determination model predicts the 3D velocity and pressure at any sampled location within the anatomy from 2D/3D inputs. More input samples of 2D planes (XY, YZ) provide better estimates of velocity and pressure at the 3D locations. [00134] According to an embodiment, the hemodynamic determination model can also be trained and configured to provide biomechanical and fluid dynamics information in the presence of various types of implant devices as well as different navigation tools. The network when fully trained is capable of providing inference at interactive frame rates.

[00135] According to an embodiment, the parameters that can be generated as the output from the neural PDE solver (PINN) can then be passed into a second stage classifier which can be trained to create a decision surface for procedure specific tasks. An example task for such a classifier would be to provide a visual indicator of a safe landing zone during the intervention given a particular type of device and the patient geometry. Typically, even with the methods described here, the information generated about biomechanical properties is still limited to pre-procedure setting. However, the information can be utilized in real-time by using the parameters generated from the network output as inputs to a risk classifier. Such classifiers can either be trained on prior data in a supervised fashion (e.g., support vector machines, SVM) or in an unsupervised fashion using dynamic clustering (e.g., gaussian mixture model (GMM)). These classifiers can be made task specific to operate on focused steps of the procedure (e.g., navigation of the ascending aorta, expanding and ballooning of the implant etc.).

[00136] Referring to FIG. 9, in one embodiment, is a schematic representation of a real-time patient specific simulation and training phantom setup that utilizes trained PINNs to provide realtime biomechanical feedback. The system receives as input 3D imaging of the patient anatomy, and utilizes the trained PINN, with live data (e.g., ultrasound) and information about the device being used, to provide real-time biomechanical feedback.

[00137] According to an embodiment, the hemodynamic analysis system can be used to create a hybrid simulation environment that can be used to train clinicians in challenging cases. With the advent of rapid prototyping and new closed loop phantom setups, a patient-specific anatomy can be created virtually as well as printed using materials that are a close approximation to tissue. In the simulation interface the phantom has a virtual model that is powered by pretrained PINN which provides all the necessary biomechanical information on screen. This can then be used to mock an experiment and try various devices as well as understand the potential risks during the actual procedure.

[00138] According to an embodiment, the hemodynamic analysis system can be used to generate biomechanical data retrospectively on patient cases where pre-procedural 3D data, fluoroscopy runs with contrast flow information, implant information, implant deployment information and maneuvers are recorded. This can form the basis to either generating biomechanical data of the procedure retrospectively or can be used to update the model. This can also then be used to create a biomechanical timeline of the intervention and then tag implant success or outcomes at timepoints post procedure. Furthermore, the model can be continually updated based on data from post procedural scans.

[00139] All definitions, as defined and used herein, should be understood to control over dictionary definitions, definitions in documents incorporated by reference, and/or ordinary meanings of the defined terms.

[00140] The indefinite articles “a” and “an,” as used herein in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean “at least one.”

[00141] The phrase “and/or,” as used herein in the specification and in the claims, should be understood to mean “either or both” of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with “and/or” should be construed in the same fashion, i.e., “one or more” of the elements so conjoined. Other elements may optionally be present other than the elements specifically identified by the “and/or” clause, whether related or unrelated to those elements specifically identified.

[00142] As used herein in the specification and in the claims, “or” should be understood to have the same meaning as “and/or” as defined above. For example, when separating items in a list, “or” or “and/or” shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly indicated to the contrary, such as “only one of’ or “exactly one of,” or, when used in the claims, “consisting of,” will refer to the inclusion of exactly one element of a number or list of elements. In general, the term “or” as used herein shall only be interpreted as indicating exclusive alternatives (i.e. “one or the other but not both”) when preceded by terms of exclusivity, such as “either,” “one of,” “only one of,” or “exactly one of.”

[00143] As used herein in the specification and in the claims, the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified.

[00144] It should also be understood that, unless clearly indicated to the contrary, in any methods claimed herein that include more than one step or act, the order of the steps or acts of the method is not necessarily limited to the order in which the steps or acts of the method are recited.

[00145] In the claims, as well as in the specification above, all transitional phrases such as “comprising,” “including,” “carrying,” “having,” “containing,” “involving,” “holding,” “composed of,” and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases “consisting of’ and “consisting essentially of’ shall be closed or semi-closed transitional phrases, respectively.

[00146] While several inventive embodiments have been described and illustrated herein, those of ordinary skill in the art will readily envision a variety of other means and/or structures for performing the function and/or obtaining the results and/or one or more of the advantages described herein, and each of such variations and/or modifications is deemed to be within the scope of the inventive embodiments described herein. More generally, those skilled in the art will readily appreciate that all parameters, dimensions, materials, and configurations described herein are meant to be exemplary and that the actual parameters, dimensions, materials, and/or configurations will depend upon the specific application or applications for which the inventive teachings is/are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific inventive embodiments described herein. It is, therefore, to be understood that the foregoing embodiments are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, inventive embodiments may be practiced otherwise than as specifically described and claimed. Inventive embodiments of the present disclosure are directed to each individual feature, system, article, material, kit, and/or method described herein. In addition, any combination of two or more such features, systems, articles, materials, kits, and/or methods, if such features, systems, articles, materials, kits, and/or methods are not mutually inconsistent, is included within the inventive scope of the present disclosure.

Claims

Claims What is claimed is:

1. A hemodynamic analysis system (200) for determination of a treatment prediction, the system comprising: a processor (220) configured to: receive imaging data of a patient during a medical procedure, wherein the imaging data comprises patient-specific anatomical data of an imaged blood vessel, derive one or more patient-specific three-dimensional (3D) flow field parameters based on analysis of the imaging data, and determine the treatment prediction for the medical procedure based on the one or more patient-specific 3D flow field parameters.

2. The system of claim 1, wherein the imaging data further comprises data regarding blood flow within the imaged blood vessel.

3. The system of claim 1, wherein the processor comprises a trained physics-informed hemodynamic determination model configured to analyze the received imaging data to derive the one or more patient-specific 3D flow field parameters for the imaged blood vessel.

4. The system of claim 1, further comprising a user interface (240) configured to display a flow pattern within the imaged blood vessel, the flow pattern determined based on the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel.

5. The system of claim 4, wherein displaying the flow pattern within the imaged blood vessel further comprises displaying the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel.

6. The system of claim 4, wherein displaying the flow pattern comprises displaying a plurality of flow fields each comprising a flow direction and magnitude within the imaged blood vessel.

7. The system of claim 1, wherein the processor is further configured to analyze one or more of patient data and/or blood vessel imaging data parameters when deriving the one or more patient-specific 3D flow field parameters for the imaged blood vessel.

8. The system of claim 1, wherein the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel comprise at least one of pressure, pressure gradient, and 3D velocity fields (u, v, and w).

9. The system of claim 1, wherein the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel comprise one or more of wall shear stress (WSS), oscillatory shear index (OSI), and fractional flow reserve (FFR).

10. The system of claim 1 , wherein the processor is further configured to generate, from the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel, a risk score quantifying one or more of a predicted likelihood of success for a treatment of the imaged blood vessel and a predicted likelihood of rupture of the imaged blood vessel.

11. The system of claim 1 , wherein the processor is further configured to generate, from the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel, a confidence metric quantifying confidence in the derived one or more patient-specific 3D flow field parameters for the imaged blood vessel.

12. The system of claim 1, wherein the anatomical data of the imaged blood vessel is obtained via one or more of digital subtraction angiography (DSA), computerized tomography (CT) scan, positron emission tomography (PET), magnetic resonance imaging (MRI), and ultrasound.

13. A hemodynamic analysis system (300) for determination of an implant placement prediction, the system comprising: processor (320) configured to: receive imaging data comprising anatomical data of an imaged blood vessel of the patient, derive one or more patient-specific 3D flow field parameters for the imaged blood vessel at each of a plurality of different locations along the blood vessel based on analysis of the imaging data, determine, based on the derived one or more patient-specific 3D flow field parameters for each of the plurality of different locations, one or more implant 3D flow field parameters of a blood vessel implant within the imaged blood vessel, and determine a placement prediction of the blood vessel implant within the imagined blood vessel based on the one or more implant 3D flow field parameters.

14. The system of claim 13, wherein the processor is further configured to generate, based on the derived one or more patient-specific 3D flow field parameters for each of the plurality of different locations, a predicted best placement for the blood vessel implant within the imaged blood vessel.

15. The system of claim 13, wherein the processor is further configured to generate a risk determination quantifying a risk of placement failure for the blood vessel implant at one or more of the plurality of different locations.

16. The system of claim 13, wherein the processor comprises a trained physics- informed hemodynamic determination model configured to analyze the received imaging data to derive one or more patient-specific 3D flow field parameters for the imaged blood vessel.