Apparatus, methods, and articles for 4D flow magnetic resonance imaging.

Abstract

To provide a method for operation to be used together with a medical image system of a magnetic resonance imaging (MRI) base.SOLUTION: A method includes: a step which serves a plurality of requests by a first proxy server, the requests coming from a clinic facility network and from an external clinic facility network: a step which responds to at least some of the requests and generates plain text information from hashed patient health information by the proxy server; a step for replacing the hashed patient health information with the plain text information by the proxy server; and a step for transmitting given information to a client device.SELECTED DRAWING: None

Description

【Technical Field】 【0001】 The present invention generally relates to magnetic resonance imaging (MRI), and more particularly to four-dimensional (4D) flow MRI. 【Background Art】 【0002】 MRI is most commonly used in medical imaging, but can also be used in other fields. An MRI machine typically includes a main magnet that is an annular array of coils having a central or longitudinal bore. The main magnet can create a strong, stable magnetic field (e.g., from 0.5 tesla to 3.0 tesla). The bore is sized to receive at least a portion of the object to be imaged, such as a human body. When used in a medical imaging application, the MRI machine may include a patient table that allows a prone patient to easily slide or roll in and out of the bore. 【0003】 The MRI machine also includes gradient magnets. The gradient magnets create a relatively small, varying magnetic field (e.g., from 180 gauss (0.018 tesla) to 270 gauss (0.027 tesla)) that is smaller than the magnetic field created by the main magnet, and allows a selected portion of the object (e.g., a patient) to be imaged. The MRI machine also includes a radio frequency (RF) coil that is operated to apply high-frequency energy to a selected portion of the object (e.g., a patient) to be imaged. Different RF coils can be used to image different structures (e.g., anatomical structures). For example, one set of RF coils may be appropriate for imaging a patient's neck, while another set of RF coils may be appropriate for imaging a patient's chest or heart. The MRI machine typically includes additional magnets, such as resistive magnets and / or permanent magnets. 【0004】 MRI machines typically include or are communicatively coupled to a computer system, which is used to control the image and / or coils, and / or perform image processing to create images of the subject being imaged. Traditionally, MRI machines create magnitude data sets representing physical structures, such as anatomical structures. These data sets often conform to the Digital Imaging and Communications in Medicine (DICOM) standard. DICOM files typically contain pixel data and metadata in a predetermined format. [Prior art documents] [Patent Documents] 【0005】 [Patent Document 1] U.S. Provisional Patent Application No. 61 / 571,908 [Patent Document 2] International Publication No. 2013 / 006709 [Patent Document 3] U.S. Provisional Patent Application No. 61 / 928,702 [Overview of the project] 【0006】 In recent years, proposals have been made to create 4D flow datasets, which include anatomical data as well as velocities in three orthogonal directions, which may be named x-velocity, y-velocity, and z-velocity. 【0007】 When used, an MRI study may be defined for a single-patient session. Each MRI study typically includes several series of perfusion and 4D flow acquisitions, with multiple sets of MRI data (e.g., 100 images) acquired per series. The series may be divided into magnitude acquisition and phase acquisition portions. The resulting images may be annotated with information. 【0008】 4D flow pulsed sequence MRI is highly anticipated, particularly for cardiac MRI procedures, due to its potential to provide low-cost, high-speed, and accurate medical imaging. Several obstacles to adoption include high costs. For example, there are high monetary and opportunity costs associated with the need for clinicians (e.g., physicians) to participate during the MRI procedure (e.g., acquisition) to evaluate anatomical structures during imaging. There are also high costs associated with the powerful computers used for computation at the clinical facility where MRI medical imaging is performed, and the personnel required to operate and maintain such equipment. Techniques requiring breath-holding may not be feasible for certain patients (e.g., very young children or infants). Also, synchronizing respiration (i.e., lung or respiratory cycle) and / or cardiac cycle with image acquisition makes for a fairly long procedure. Long procedures increase costs and reduce throughput, as expensive equipment and costly personnel are assigned to a single patient during the period. It also tends to increase patient anxiety. These techniques not only tend to require the participation of highly trained technicians and clinicians, but the annotation of imaging results tends to be difficult. Furthermore, due to the subjectivity involved in interpreting images, such as anatomical structures, reproducibility tends to be low not only between sessions but also between series. 【0009】 This specification describes various apparatuses, methods, and articles that address at least one or more of these problems in part. 【0010】 Instead of acquiring only specific planes during MRI acquisition, a 4D pulse sequence is used to acquire a complete 3D volume set including phase contrast; therefore, it is named 4D phase contrast MRI. This gives clinicians the freedom to view any plane they desire, even if the patient is no longer present after acquisition. Error detection and / or correction, segmentation (e.g., boundary demarcation), quantification, verification, and visualization (e.g., merging visual representations of flow information with visual representations of anatomical structures) can be performed on the resulting MRI dataset. Much of the image processing and analysis can be performed autonomously (e.g., without human intervention). 【0011】 The MRI image processing and analysis system can be remotely located from one or more MRI acquisition systems and can perform error detection and / or correction (e.g., for phase error correction, phase aliasing, signal unwrapping, and / or other artifacts) on MRI datasets, segmentation, visualization, quantification, and verification of flow overlaid on anatomical structures (e.g., velocity, arterial vs. venous flow, shunts), and / or generation of patient-specific 4D flow protocols. Asynchronous commands and imaging pipelines enable timely and secure remote image processing and analysis, even for complex or large-scale 4D flow MRI datasets. 【0012】 Remote MRI image processing and analysis systems may offer cloud-based web services or software as services (SAS). MRI image processing and analysis systems utilize powerful computing resources, such as large sets or arrays of GPUs. The MRI image processing and analysis system can serve multiple MRI acquisition systems, which are operated by one or more diagnostic entities and / or located in one or more diagnostic facilities. This can significantly reduce costs to diagnostic facilities and the burden associated with acquiring and maintaining expensive computing equipment. 【0013】 The techniques described herein can simplify MRI procedures. These techniques provide a turnkey system for capturing complete 4D flow images and eliminate the need for clinician involvement during MRI procedures. These techniques can also significantly shorten the length of MRI procedures. This reduces the costs associated with personnel and equipment. This can also increase throughput, allowing capital-intensive MRI systems to be amortized over a large number of patients during the equipment's effective lifespan. 【0014】 The methods described herein also provide automated or even autonomous verification of the results. Relying on mass conversation principles, at least one method can identify shunts or other anatomical abnormalities. 【0015】 Improved reproducibility across patients or populations may enable the identification of new indicators and the development of new therapies for the methods described herein. 【0016】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as including the steps of: receiving a set of MRI data by at least one processor-based device, the set of MRI data including anatomical structure and blood flow information for each of a plurality of voxels; identifying one or more instances of structure in the set of MRI flow data by at least one processor-based device; and deriving contours in the set of MRI flow data by at least one processor-based device based on the identified one or more instances of structure in the set of MRI flow data. The step of identifying one or more instances of structure in the set of MRI flow data may include the step of identifying one or more instances of coherence in the set of MRI flow data. The step of identifying one or more instances of coherence in the set of MRI flow data may include the step of identifying one or more instances of directed coherence in the set of MRI flow data. The step of identifying one or more instances of coherence in the set of MRI flow data may include the step of identifying one or more instances of directional pathline or structural coherence in the set of MRI flow data. The step of identifying one or more instances of coherence in a set of MRI flow data may include the step of identifying one or more instances of discrete Fourier transform (DFT) component coherence in a set of MRI flow data. The step of identifying one or more instances of coherence in a set of MRI flow data may include the step of identifying one or more instances of acceleration coherence in a set of MRI flow data. 【0017】 The process may further include, by at least one processor, identifying one or more instances of clinical markers in the set of MRI flow data based on one or more identified instances in the set of MRI flow data. The step of identifying one or more instances of clinical markers in the set of MRI flow data may include identifying one or more instances of anatomical markers and / or temporal markers in the set of MRI flow data. The step of identifying one or more instances of clinical markers in the set of MRI flow data may include identifying one or more instances of aneurysms, stenosis, or plaques in the set of MRI flow data. The step of identifying one or more instances of clinical markers in the set of MRI flow data may include identifying one or more pressure gradients in the set of MRI flow data. The step of identifying one or more instances of clinical markers in the set of MRI flow data may include identifying one or more instances of cardiac anatomical landmarks in the set of MRI flow data. The step of deriving contours in the set of MRI flow data based on one or more identified instances of structures in the set of MRI flow data may include deriving contours in the set of MRI flow data representing various body tissues. 【0018】 The method may further include the step of autonomously segmenting blood body tissue from non-blood body tissue using at least one processor-based device. 【0019】 The method may further include the step of autonomously segmenting air from body tissue using at least one processor-based device. 【0020】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as including the steps of: receiving input in a processor-based device in a first pass, prior to a first MRI acquisition sequence, for a first investigation of a subject, wherein the input is specific to the first MRI acquisition; and generating a 4D flow localizer by the processor-based device, at least in part, based on the received input. The step of receiving input may include receiving at least one of the following: clinical indication, type or identity of contrast agent, amount of contrast agent, subject weight, subject height, and subject heart rate, amount of time elapsed since the bolus was delivered to the subject, identification of the MRI hardware manufacturer, type of coil used, and identification of at least one characteristic of the MRI machine used. 【0021】 The method may further include the steps of receiving information about a first MRI acquisition sequence in a processor-based device in a second pass, prior to a second MRI acquisition sequence, for a first investigation of a subject, and generating a high-fidelity 4D flow localizer by the processor-based device, at least in part, based on the received information. The step of receiving information about a first MRI acquisition sequence in a processor-based device may include the step of receiving output from the first MRI acquisition sequence. The step of receiving information about a first MRI acquisition sequence in a processor-based device may include the step of receiving metrics indicating the quality of the first MRI acquisition sequence. The step of receiving metrics indicating the quality of the first MRI acquisition sequence may include at least one evaluation indicating the quality of the first MRI acquisition sequence, which is evaluated by at least one person. The step of generating a high-fidelity 4D flow localizer by a processor-based device, at least in part, based on the received information, may include specifying one or more of the following: acquisition period, VENC, field of view, repetition time (TR), echo time (TE), row resolution, column resolution, slice resolution, temporal resolution, and flip angle. The step of generating a high-fidelity 4D flow localizer may include determining a value for the velocity encoding (VENC) parameter. The step of determining a value for the velocity encoding (VENC) parameter may include determining at least an approximation of the blood flow velocity in the lumen and selecting a value for VENC using a lookup table. The step of determining a value for the velocity encoding (VENC) parameter may include determining the number of channels in the coil used for acquisition and selecting a value for VENC, at least in part, based on the number of channels in the coil. 【0022】 A processor-based device can have at least one processor and at least one non-transitory processor-readable medium communicatively coupled to the at least one processor, and can be operable to perform any one of the above methods. 【0023】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system includes, for a plurality of voxels of an MRI image dataset, grouping the voxels into a number of bins by at least one processor-based device; reducing the number of bins via the at least one processor-based device; for a first bin and a second bin of the bins, determining which of the first bin or the second bin of the bins includes voxels representing arterial blood flow and which of the first bin or the second bin of the bins includes voxels representing venous blood flow; assigning a first set of colors to the voxels representing arterial blood flow; and assigning a second set of colors to the voxels representing venous blood flow, the second set of colors being different from the first set of colors. The step of assigning a first set of colors to the voxels representing arterial blood flow can include assigning a single blue color to the voxels representing arterial blood flow, and the step of assigning a second set of colors to the voxels representing venous blood flow can include assigning a single red color to the voxels representing venous blood flow. 【0024】 The step of determining which of the first bin or the second bin among the bins can contain voxels representing arterial blood flow and which of the first bin or the second bin among the bins can contain voxels representing venous blood flow can include the step of determining, by at least one processor-based device, based on the proximity of at least one of the voxels to several anatomical landmarks, which of the first bin or the second bin among the bins can contain voxels representing arterial blood flow and which of the first bin or the second bin among the bins can contain voxels representing venous blood flow. The step of determining which of the first bin or the second bin among the bins can contain voxels representing arterial blood flow and which of the first bin or the second bin among the bins can contain voxels representing venous blood flow can include the step of determining, by at least one processor-based device, based on the closest one of the voxels within each bin to several anatomical landmarks, which of the first bin or the second bin among the bins can contain voxels representing arterial blood flow and which of the first bin or the second bin among the bins can contain voxels representing venous blood flow. 【0025】 The method can further include the step of autonomously identifying anatomical landmarks by a processor-based device. 【0026】 The method can further include the step of receiving user input for identifying anatomical landmarks by a processor-based device. 【0027】 For each of at least some of the multiple voxels, the process may further include the steps of determining whether each voxel represents blood flow and logically marking the voxels determined to represent blood flow, before grouping the voxels into multiple bins. The step of determining whether each voxel can represent blood flow may include the step of autonomously determining whether each voxel can represent blood flow by at least one processor-based device. The step of determining whether each voxel can represent blood flow may include the step of receiving user input indicating whether each voxel can represent blood flow by at least one processor-based device. The step of reducing the number of bins via at least one processor-based device may include the step of merging voxels in one bin into one bin if voxels in one bin have been in contact with voxels in another bin over time. 【0028】 The method may further include the steps of determining whether there are three or more bins, and, in response to determining whether there are three or more bins, determining by at least one processor-based device whether some of the voxels represent a potential shunt. The step of determining by at least one processor-based device whether some of the voxels represent a potential shunt may include identifying regions where adjacent bins are connected by fewer voxels than a defined threshold number or by regions smaller than a defined threshold region. 【0029】 The method may further include a step of providing visual highlighting to at least one of the voxels representing a potential shunt or an area at least adjacent to a potential shunt. The step of providing visual highlighting to at least one of the voxels representing a potential shunt or an area at least adjacent to a potential shunt may include a step of assigning a third set of colors to the voxel representing the potential shunt or an area at least adjacent to a potential shunt, wherein the third set of colors is different from both the first set of colors and the second set of colors. 【0030】 The method may further include the steps of receiving input by at least one processor-based device, the input indicating a human assessment of whether a voxel represents an actual shunt, and updating the color of voxels representing potential shunts based on the received input indicating a human assessment of whether a voxel represents an actual shunt. The step of determining whether a first or second bin of the bins can contain voxels representing arterial blood flow, and whether a first or second bin of the bins can contain voxels representing venous blood flow, may include, for each of at least several bins, identifying any voxels in the bin that have a coherence value above a threshold coherence value, for any voxels in the bin that have a coherence value above a threshold coherence value, calculating the average velocity over all of a multiple time point and the angle between the centroid of each voxel and a vector connecting the centroid of the anatomical structure, and calculating the average angle between all voxels in the bin and the centroid of the anatomical structure. 【0031】 The step of determining whether the first or second bin of the bins can contain voxels representing arterial blood flow, and whether the first or second bin of the bins can contain voxels representing venous blood flow, may further include the step of assigning the bin with the largest average angle to represent arterial blood flow and the step of assigning the bin with the smallest average angle to represent venous blood flow. The step of calculating the angle between the average velocity over all of the multiple time points and a vector connecting the center of gravity of each voxel to the center of gravity of the anatomical structure may include the step of calculating the angle between the average velocity over all of the multiple time points and a vector connecting the center of gravity of each voxel to the center of gravity of the heart. 【0032】 A processor-based device may have at least one processor and at least one non-transient processor-readable medium communicatively coupled to at least one processor, and may be operable to perform any one of the above methods. 【0033】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as comprising the steps of: receiving a set of MRI data by at least one processor-based device, the set of MRI data including anatomical and blood flow information for each of a plurality of voxels; and applying a first filter for isolating blood flow based on directed coherence to at least a portion of the received set of MRI data by at least one processor-based device. The step of applying a first filter for isolating blood flow based on directed coherence to at least a portion of the received set of MRI data may include the step of calculating directed coherence for each of a plurality of voxels. The step of calculating directed coherence for each voxel may include the steps of summing a set of weighted directional coherence scores between each voxel and a plurality of neighboring voxels which are neighboring elements of each voxel, and dividing the result of the sum by the sum of all applied weights. 【0034】 The method may further include the step of determining a weighted directed coherence score between each voxel and several neighboring voxels. The step of determining a weighted directed coherence score between each voxel and several neighboring voxels may include the step of determining the dot product of normalized velocity vectors; the step of applying the trigonometric function ACOS to the result of the dot product to determine the angled difference; the step of scaling the angle difference between 0 and π to obtain a result between 0 and 1; and the step of multiplying the scaling result by the respective weights that represent the distance between each voxel and each of its neighboring voxels. 【0035】 The method may further include a step of determining each weight. The step of determining each weight may include a step of finding the minimum interval across all three dimensions and a step of dividing that minimum interval by the distance between voxels. The first filter may be applied with one volume per time point. The first filter may be applied with one volume averaged over all time points, per time point. 【0036】 The method may further include the step of applying a second filter to remove random noise from at least a portion of the received set of MRI data using at least one processor-based device. 【0037】 A processor-based device may have at least one processor and at least one non-transient processor-readable medium communicatively coupled to at least one processor, and may be operable to perform any one of the above methods. 【0038】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as including the steps of: receiving a set of MRI data by at least one processor-based device, the set of MRI data including anatomical structure and blood flow information for each of a plurality of voxels; identifying an anatomical volume in the set of MRI data; determining the flow of blood into the identified anatomical volume; determining the flow of blood out of the identified anatomical volume; and comparing the flow of blood into the identified anatomical volume with the flow of blood out of the identified anatomical volume by at least one processor-based device. 【0039】 The method may further include a step of verifying prior actions based on the results of comparing the flow of blood into an identified anatomical volume with the flow of blood outside an identified anatomical volume. 【0040】 The method may further include a step of verifying prior segmentation behavior based on the results of comparing the flow of blood into the identified anatomical volume with the flow of blood outside the identified anatomical volume. 【0041】 The method may further include a step of providing notification based on the results of comparing the flow of blood into an identified anatomical volume with the flow of blood outside an identified anatomical volume. 【0042】 The method may further include a step of providing notification of detected shunts based on the result of comparing the flow of blood into an identified anatomical volume with the flow of blood outside an identified anatomical volume. The step of identifying an anatomical volume in a set of MRI data may include a step of identifying at least one of a lumen, a portion of a lumen, a vessel, a portion of a vessel, a chamber, or a cavity or portion of a cavity within an anatomical structure. The step of identifying an anatomical volume in a set of MRI data may include a step of autonomously identifying the anatomical volume by at least one processor-based device. The step of identifying an anatomical volume in a set of MRI data may include a step of autonomously identifying the anatomical volume based on user input received by at least one processor-based device. 【0043】 The method may further include the steps of identifying natural entry points into an identified anatomical volume and identifying natural exit points from the identified anatomical volume. 【0044】 The method may further include the steps of naming a first location as an inlet to an identified anatomical volume, and naming a second location as an outlet from the identified anatomical volume, wherein the second location is spaced apart from the first location. The step of determining the flow of blood into the identified anatomical volume may include determining the dot product of the normal vector of a first plane slicing the identified anatomical volume at the first location and the velocity vector at each voxel, and the step of determining the flow of blood out of the identified anatomical volume may include determining the dot product of the normal vector of a second plane slicing the identified anatomical volume at the second location and the velocity vector at each voxel, wherein the second location is different from the first location. 【0045】 The method may further include the steps of determining the net flow of blood into an identified anatomical volume and determining the net flow of blood out of the identified anatomical volume. The step of determining the net flow of blood into an identified anatomical volume may include integrating over time the result of the dot product of the normal vector of a first plane slicing the identified anatomical volume at a first location and the velocity vector at each voxel, and the step of determining the net flow of blood out of the identified anatomical volume may include integrating over time the result of the dot product of the normal vector of a second plane slicing the identified anatomical volume at a second location and the velocity vector at each voxel, wherein the second location is different from the first location. 【0046】 The step of comparing the flow of blood into an identified anatomical volume with the flow of blood outside the identified anatomical volume may include a step of determining whether the flow values ​​at the inlets and outlets of the identified anatomical volume match within at least a defined threshold, and may further include a step of providing a mismatch indicator in response to determining that the flow values ​​at the inlets and outlets of the identified anatomical volume do not match within at least a defined threshold. The step of determining whether the flow values ​​at the inlets and outlets of the identified anatomical volume match within at least a defined threshold may include a step of autonomously determining whether the flow values ​​at the inlets and outlets match within a defined threshold by at least one processor-based device. The step of comparing the flow of blood into an identified anatomical volume with the flow of blood outside the identified anatomical volume by at least one processor-based device may include a step of comparing the flow of blood through the ascending thoracic aorta with the combined flow of blood through the superior vena cava and the descending thoracic aorta. The step of comparing the flow of blood into an identified anatomical volume with the flow of blood outside the identified anatomical volume by at least one processor-based device may include the step of comparing the combined flow of blood through the superior and inferior vena cava with the flow of blood through a set of pulmonary vascular structures. The step of comparing the flow of blood into an identified anatomical volume with the flow of blood outside the identified anatomical volume by at least one processor-based device may include the step of comparing the flow of blood through a set of pulmonary vascular structures with the combined flow of blood through a set of right pulmonary vascular structures and a set of left pulmonary vascular structures.The step of comparing the flow of blood into an identified anatomical volume with the flow of blood out of an identified anatomical volume by at least one processor-based device may include comparing the flow of blood through the left pulmonary vascular structure with the sum of the blood flows through all left pulmonary veins. The step of comparing the flow of blood into an identified anatomical volume with the flow of blood out of an identified anatomical volume by at least one processor-based device may include comparing the flow of blood through the right pulmonary vascular structure with the sum of the blood flows through all right pulmonary veins. The step of comparing the flow of blood into an identified anatomical volume with the flow of blood out of an identified anatomical volume by at least one processor-based device may include comparing the flow of blood leaving the cardiac chambers with the changes in systolic and diastolic volume for each cardiac chamber. 【0047】 A processor-based device may have at least one processor and at least one non-transient processor-readable medium communicatively coupled to at least one processor, and may be operable to perform any one of the above methods. 【0048】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as comprising the steps of: receiving a set of MRI data by at least one processor-based device, the set of MRI data containing anatomical structure and blood flow information for each of a plurality of voxels; identifying an anatomical volume in the set of MRI data; identifying a plurality of planes, each intersecting a common point, each traversing an anatomical volume in a different orientation, and having different orientations; and autonomously comparing the blood flow through each of the planes over the entire cardiac cycle by at least one processor-based device. 【0049】 The step of autonomously comparing the blood flow through each plane over the entire cardiac cycle includes the step of autonomously determining whether the blood flow through each plane matches at least within a defined threshold over the entire cardiac cycle. 【0050】 The method may further include a step of verifying prior actions based on the results of comparing blood flow through each plane over the entire cardiac cycle. 【0051】 The method may further include a step of verifying the preceding segmentation behavior based on the results of comparing the blood flow through each plane over the entire cardiac cycle. 【0052】 The method may further include a step of providing notification based on the results of comparing blood flow through each plane over the entire cardiac cycle. 【0053】 A processor-based device may have at least one processor and at least one non-transient processor-readable medium communicatively coupled to at least one processor, and may be operable to perform any one of the above methods. 【0054】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as comprising the steps of: receiving a set of MRI data by at least one processor-based device, the set of MRI data including anatomical structure and blood flow velocity information for each of a plurality of voxels; identifying a seed point in a vascular represented in the set of MRI data; identifying a point in time having the maximum flow magnitude at the seed point; determining a cross-sectional plane perpendicular to the direction of flow based on both the anatomical structure and blood flow velocity information; and determining the luminal boundary of the vascular based at least partially on the determined cross-sectional plane. 【0055】 The step of determining a cross-sectional plane perpendicular to the direction of flow, based on both anatomical structure and blood flow velocity information, may include: generating a coarse hemisphere of vectors at a seed point, wherein any vector closest to the direction of flow has the greatest weight; projecting multiple rays perpendicular to the vectors on the coarse hemisphere for each of the vectors constituting the coarse hemisphere; and terminating each ray when the magnitude of change in both anatomical pixel intensity and velocity magnitude reaches or exceeds a change threshold. 【0056】 The step of determining a cross-sectional plane perpendicular to the direction of flow, based on both anatomical structure and blood flow velocity information, may further include the step of calculating the sum of all regions of several resulting triangles for each of several rays, wherein a seed point, the end of each ray, and the end of another ray define the resulting triangle, and the step of selecting a vector on a coarse hemisphere having the smallest calculated sum of regions as the normal vector to the plane most perpendicular to the direction of blood flow. 【0057】 The method may further include the step of resampling a multi-plane reconstruction that combines both anatomical structure and blood flow velocity information at the initial seed point, along with the strongest flow time point and normal vector. 【0058】 The method may further include the step of determining the contour that depicts the luminal boundary of the blood vessel using a dynamic contour model. 【0059】 The method may further include the step of applying an energy minimization function using gradient descent to find a smooth contour. 【0060】 The method may further include the steps of determining whether a contour diverges above a threshold in terms of region or curvature, and replacing the contour with a replacement contour in response to the determination that the contour diverges above a threshold in terms of region or curvature, wherein the replacement contour is a linear blend of the contour at multiple time points adjacent to the time point of the contour being replaced. 【0061】 A processor-based device may have at least one processor and at least one non-transient processor-readable medium communicatively coupled to at least one processor, and may be operable to perform any one of the above methods. 【0062】 The method of operation for use with magnetic resonance imaging (MRI)-based medical imaging systems can be summarized as including the steps of: receiving multiple user events from multiple client devices via a first channel of asynchronous commands and image pipelines; capturing at least some of the user events in a persistence layer; determining which of the captured events to squelch; performing image processing and analysis; and providing the respective responses to the user events to the clients via a second channel of asynchronous commands and image pipelines. 【0063】 The step of receiving multiple user events from multiple client devices via a first channel of asynchronous commands and image pipelines may include the step of a server receiving user events, and the step of the server providing the user events to the MRI imaging and analysis system for image processing and analysis of the MRI dataset associated with the user events. The step of capturing at least some of the user events in a persistence layer and the step of deciding which of the captured events to suppress may include the step of receiving the user events as messages, the step of determining whether the imaging and analysis system is busy, and the step of immediately forwarding the messages to the computing server in response to the computing server determining that it is not busy. The step of capturing at least some of the user events in a persistence layer and the step of deciding which of the captured events to suppress may include the step of receiving the user events as messages, the step of determining whether the imaging and analysis system is busy, and the step of placing the messages in a slot in response to the computing server determining that it is busy. 【0064】 The method may further include the step of placing messages that are more recent than the messages already placed in the slot into the slot. 【0065】 The method may further include the steps of detecting a completion event and forwarding the respective response to the user event to the respective client. For responses that may contain image data, the response is sent as an HTTPS image request and via an HTTPS AJAX request or a WebSocket with binary support. For responses that may not contain image data, the response is sent directly via a WebSocket. 【0066】 The method may further include the steps of determining whether there is a message in the slot, waiting for a completion event to be detected in response to the determination that there is a message in the slot, sending the message from the slot to the computing server in response to the detection of the completion event, and clearing the slot. 【0067】 The method may further include the steps of rendering multiple images linked by properties as one larger image, and sending the one larger image as a single response. 【0068】 The method may further include the step of overlaying at least one of lines, markers, or planes within the image before sending one large image as a single response. 【0069】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as including the steps of: remotely providing access to information from a secure clinical facility network via a first firewall, wherein all secure patient health information is held on the secure clinical facility network; and protecting the MRI image processing and analysis system from the Internet by a second firewall. 【0070】 The method of operation for use with magnetic resonance imaging (MRI)-based medical imaging systems can be summarized as including the steps of: receiving a raw DICOM file via an anonymization service running on a clinical facility network; generating a hash of arbitrary plaintext patient health information within the raw DICOM file; and replacing any arbitrary plaintext patient health information within the raw DICOM file with its respective hash. 【0071】 The method may further include the step of identifying all plain text patient health information fields in the raw DICOM file, and the step of generating a hash of any plain text patient health information in the raw DICOM file, which may include the step of generating a salted hash. 【0072】 The method may further include the step of accessing a key maintained within a network of clinical facilities that converts plaintext patient health information into hashes using an anonymization service. 【0073】 The method may further include the step of enabling access from outside the clinical site network only via the clinical site's VPN to anonymization services running on the clinical site network. 【0074】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as comprising the steps of: a first proxy server serving a plurality of requests, the requests coming from within and outside the clinical facility network; the proxy server generating plaintext information from hashed patient health information in response to at least some of the requests; the proxy server replacing the hashed patient health information with the plaintext information; and transmitting the resulting information to a client device. 【0075】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as including the steps of: a client connecting to a server of an MRI image processing and analysis system, the server being located outside the clinical facility network; checking a flag indicating whether the client is missing information required to render patient health information in plain text; in response to the indicator that the client is missing information required to render patient health information (PHI) in plain text, the client connecting to an anonymization service; and the client requesting patient health information in plain text from the anonymization service. The step of the client requesting patient health information in plain text from the anonymization service may include providing a hash or identifier. 【0076】 The method may further include a step by the client to locally cache the received plain text patient health information. 【0077】 The method may further include a step in which the client purges cached plaintext patient health information in response to the user logging out. 【0078】 A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system can be summarized as comprising the steps of: receiving a set of MRI data by at least one processor-based device, the set of MRI data including anatomical and blood flow information for each of a plurality of voxels; applying a Discrete Fourier Transform (DFT) to each of the plurality of voxels in the set of MRI data over a plurality of available time points by at least one processor-based device; and examining several components of the DFT for each of the voxels over the available time points by at least one processor-based device. 【0079】 The step of applying DFT to each of several voxels in a set of MRI data across multiple available time points may include the step of applying DFT to three velocity components (x, y, z) of the blood flow information. The step of applying DFT to each of several voxels in a set of MRI data across multiple available time points may include the step of applying DFT to the velocity magnitude from which the blood flow information was derived. The method may further include the step of segmenting the blood pool from static tissue based at least partially on DFT components by at least one processor. The step of segmenting the blood pool from static tissue based at least partially on DFT components may include the step of segmenting the blood pool from static tissue autonomously by at least one processor based at least partially on a set of lower-order non-DC DFT components. The method may further include the step of identifying lung tissue in the MRI data. The method may further include the step of combining several DFT elements together, regardless of relative magnitude or phase, by at least one processor to create a general-purpose mask for pinpointing all blood flow within a chest scan. The method may further include the step of combining several DFT elements, taking into account the relative magnitude or phase of the DFT elements, by at least one processor to create a refined mask for identifying a specific region of the blood pool in the body. The method may further include the step of comparing the phase of the DFT components to the time of maximal systole by at least one processor and assigning a probability to each voxel based on the amount of phase deviation from the expected value. The method may further include the step of distinguishing between blood flow in the aorta and blood flow in the pulmonary artery, at least partially based on the refined mask, by at least one processor. The method may further include the step of autonomously identifying a probability cutoff value, at least partially based on a histogram of the resulting probability values, by at least one processor.The method may further include the steps of: autonomously by at least one processor identifying a probability cutoff value for an artery-specific mask based at least partially on a histogram of the resulting probability values; and determining a probability cutoff value for at least one other mask based at least partially on the artery-specific mask. The step of determining a probability cutoff value for at least one other mask based at least partially on the artery-specific mask may include performing a flood fill on the general-purpose blood mask to remove heterogeneous, unconnected elements. The method may further include the step of separating the artery-specific mask into two main parts based at least partially on some flow directions, some gradients, and / or some trajectories along the resulting probability values ​​by at least one processor. The method may further include the step of distinguishing the aorta and the pulmonary artery from each other based at least partially on at least one of the mean direction of flow or relative position of the two main parts in space by at least one processor. The method may further include the step of creating a probability mask about the walls of the heart autonomously by at least one processor. The method may further include the step of combining a probabilistic mask for the cardiac wall with a blood flow mask using at least one processor. The method may further include the step of utilizing the probabilistic mask for the cardiac wall when performing eddy current correction using at least one processor. The method may further include the step of utilizing the probabilistic mask for the cardiac wall to provide at least one of the location and / or size of the heart in the image using at least one processor. [Brief explanation of the drawing] 【0080】 In drawings, the same reference number identifies similar elements or actions. Sizes and relative positions within drawings are not necessarily depicted to actual size. For example, the shapes of various elements and angles are not necessarily to actual size, and some of these elements may be arbitrarily enlarged and repositioned to improve the readability of the drawing. Furthermore, the specific shapes of illustrated elements may not necessarily convey any information about the actual shape of those elements, but are simply chosen to facilitate recognition in the drawing. [Figure 1] This is a schematic diagram of a networked environment according to one exemplary embodiment, which includes at least one MRI acquisition system located in a clinical setting and at least one image processing system located remotely from the MRI acquisition system and connected to the MRI acquisition system via one or more networks in a communicative manner. [Figure 2] This is a functional block diagram of an MRI acquisition system and an MRI image processing and analysis system providing MRI image processing and analysis services, according to one exemplary embodiment. [Figure 3A] This is a schematic diagram of the data flow in an MRI image processing and analysis or rendering system according to one exemplary embodiment. [Figure 3B] This is a schematic diagram of the data flow in an MRI imaging and processing / analysis environment according to one exemplary embodiment. [Figure 4A] This figure shows a lookup table of exemplary velocity encoding (VENC) values ​​for a range of flow velocities in a blood vessel or lumen of the body, according to one exemplary embodiment. [Figure 4B] This figure shows a lookup table of exemplary values ​​for scan length or duration against the number of channels in a cardiac coil, according to one exemplary embodiment. [Figure 5] This flowchart illustrates how to operate using the 4D Flow Localizer and an optional refinement algorithm according to one exemplary embodiment. [Figure 6]This flowchart illustrates a high-level method of generating a blood flow color map according to one exemplary embodiment. [Figure 7] This is a flowchart illustrating how to determine whether a bin corresponds to venous or arterial flow, according to one exemplary embodiment. [Figure 8] This is a flowchart illustrating how to determine whether a bin corresponds to venous or arterial flow, according to one exemplary embodiment. [Figure 9A] This is a schematic diagram illustrating the operation of an asynchronous command and imaging pipeline in one exemplary embodiment. [Figure 9B] Figure 9A is a schematic diagram illustrating the operation of the asynchronous command and imaging pipeline at different time points. [Modes for carrying out the invention] 【0081】 In the following description, several specific details are presented to provide a complete understanding of the various disclosed embodiments. However, it will be understood by those skilled in the art that embodiments may be carried out without one or more of these specific details, or with other methods, components, materials, etc. In other examples, well-known structures associated with MRI machines, computer systems, server computers, and / or communication networks are not illustrated or described to avoid unnecessarily obscuring the description of the embodiments. 【0082】 Unless otherwise required by context, throughout the specification and subsequent claims, variations of the word “comprise” and “comprises” and “comprising” are synonymous with “including” and are inclusive or open-ended (i.e., do not exclude any additional undescribed elements or method acts). 【0083】 Any reference in this specification to “one embodiment” or “embodiment” means that a particular feature, structure, or characteristic described in relation to an embodiment is included in at least one embodiment. Therefore, the appearance of the phrase “in one embodiment” or “in an embodiment” in various places throughout this specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any suitable manner in one or more embodiments. 【0084】 As used herein and in the appended claims, the singular forms "a," "an," and "the" refer to multiple objects unless explicitly indicated otherwise in the context. It should also be noted that the term "or" is generally used to mean "and / or" unless explicitly indicated otherwise in the context. 【0085】 The title and abstract provided in this application are for convenience only and do not constitute an interpretation of the scope or meaning of the embodiments. 【0086】 Many of the implementations described herein utilize 4D flow MRI datasets that essentially capture MRI magnitude and phase information over time for a three-dimensional volume. This technique can enable the capture or acquisition of MRI datasets without requiring breath-holding or synchronization or gating to the patient's cardiac or pulmonary cycle. Instead, the MRI dataset is captured or acquired, and imaging and analysis are used to derive the desired information by re-binning the acquired information based, for example, on the cardiac or pulmonary cycle. This essentially pushes what is normally a time-intensive acquisition operation into the imaging and analysis stage. As a simplified analogy, in some respects it can be thought of as capturing a movie of an anatomical structure (e.g., chest, heart) without concern for the patient's pulmonary or cardiac cycle, i.e., processing the captured movie while considering the relative motion introduced by the pulmonary or cardiac cycle. The captured information includes both magnitude information indicating the anatomical structure and phase information indicating velocity. Phase information enables the distinction between stationary and non-stationary tissues, for example, allowing non-stationary tissues (e.g., blood, air) to be distinguished from stationary tissues (e.g., fat, bone). Phase information also allows non-stationary tissues (e.g., air) to be distinguished from other non-stationary tissues (e.g., blood). This can advantageously enable automated or even more autonomous segmentation between tissues and / or the distinction of atrial blood flow from venous blood flow. It can enable the automated or even more autonomous generation of flow visualization information that can be superimposed on anatomical information. This can also advantageously enable automated or even more autonomous flow quantification for the identification of anomalies and / or the verification of results. 【0087】 A workflow can generally be divided into three parts: 1) image acquisition, 2) image reconstruction, and 3) image processing or post-processing and analysis. Alternatively, a workflow can be divided into 1) manipulation, 2) pre-processing, and 3) visualization and quantification. 【0088】 Image acquisition can include determining, defining, generating, or otherwise configuring one or more pulse sequences, which are used to operate the MRI machine (e.g., control magnet) and acquire raw MRI. The use of 4D flow pulses enables the capture of anatomical structures, expressed in magnitude, as well as velocity, expressed in phase. At least one of the methods or techniques described herein, the generation of patient-specific 4D pulse sequences, occurs during or as part of the image acquisition portion. Image reconstruction can, for example, utilize the Fast Fourier Transform to yield an MRI dataset, which is often in a form conforming to the DICOM standard. Image reconstruction is more computationally intensive than conventional methods, and this often relies on supercomputers. Their requirements place a considerable burden on many clinical facilities. Many of the methods and techniques described herein occur during or as part of imaging processing or post-processing and analysis. They can include error detection and / or correction, segmentation, visualization including fusion of flow-related information with anatomical structures, quantification, identification of anomalies including shunts, and verification including identification of false data. Alternatively, error detection and / or correction may occur during the preprocessing phase. 【0089】 Figure 1 shows a networked environment 100 according to one exemplary embodiment, in which one or more MRI acquisition systems (one is shown) 102 are communicably coupled to at least one image processing and analysis system 104 via one or more networks 106a, 106b (two of which are shown together as 106). 【0090】 The MRI acquisition system 102 is typically located in a clinical facility, such as a hospital or a dedicated medical imaging center. Various techniques and structures, such as those described herein, can advantageously allow the image processing and analysis system 104 to be located remotely from the MRI acquisition system 102. The image processing and analysis system 104 may be located, for example, in a different building, city, state, region, or even country. 【0091】 The MRI acquisition system 102 may include, for example, an MRI machine 108, a computer system 110, and an MRI operator system 112. The MRI machine 108 may include a main magnet 114, which is typically an annular array of coils having a central or longitudinal bore 116. The main magnet 108 can create a strong, stable magnetic field (e.g., 0.5 Tesla to 2.0 Tesla). The bore 116 is sized to accommodate at least a portion of the object to be imaged, such as a human body 118. When used in medical imaging applications, the MRI machine 108 typically includes a patient table 120, which allows a prone patient 118 to easily slide or roll in and out of the bore 116. 【0092】 The MRI machine also includes a set of gradient magnets 122 (only one is shown). The gradient magnets 122 create a relatively smaller fluctuating magnetic field (for example, from 180 gauss (0.018 tesla) to 270 gauss (0.027 tesla) than the magnetic field created by the main magnet 114, allowing a selected portion of the subject (e.g., a patient) to be imaged. 【0093】 The MRI machine 108 also includes a radio frequency (RF) coil 124 (only one is shown) which is operated to deliver high-frequency energy to a selected portion of the object to be imaged (e.g., a patient 118). Different RF coils 124 may be used to image different structures (e.g., anatomical structures). For example, one set of RF coils 124 may be suitable for imaging the patient's neck, while another set of RF coils 124 may be suitable for imaging the patient's chest or heart. The MRI machine 108 typically includes additional magnets, such as normal conducting magnets and / or permanent magnets. 【0094】 The MRI machine 108 typically includes, or is communicably coupled to, a processor-based MRI control system 126, which is used to control images and / or coils 114, 122, 124. The processor-based control system 126 may include one or more processors, non-temporary computer-readable or processor-readable memory, drive circuits, and / or instance components that interface with the MRI machine 108. In some implementations, the processor-based control system 126 may also perform certain preprocessing on data resulting from MRI operation. 【0095】 The MRI operator's system 128 may include a computer system 130, a monitor or display 132, a keypad and / or keyboard 134, and / or a cursor control device such as a mouse 136, joystick, trackpad, or trackball. The MRI operator's system 128 may include or read computer or processor-executable instructions from one or more non-temporary computer-readable or processor-readable media, such as a rotating medium 138, such as a magnetic or optical disk. The operator's system 128 may enable a technician to operate the MRI machine 108 to capture MRI data from a patient 118. The various techniques, structures, and features described herein may enable the operation of the MRI machine 108 by a technician without requiring the participation of a clinician or physician. This may, advantageously, significantly reduce the cost of the MRI procedure. Also, as described herein, the various techniques, structures, and features may enable the MRI procedure to be performed much faster than using conventional techniques. This may, advantageously, increase the throughput of each MRI deployment, allowing the cost of the capital-intensive equipment to be amortized over a much larger number of procedures. For example, a high-performance computer can be remotely deployed from the clinical setting and used to serve multiple clinical facilities. Furthermore, the various techniques, structures, and features described herein can, additionally or alternatively, advantageously reduce the time each patient is exposed to the MRI procedure and reduce or alleviate the anxiety often associated with it. For example, by using the image processing and analysis techniques described herein, the time to acquisition can be significantly reduced, for example, from 8 to 10 minutes, by eliminating the need for breath-holding and / or synchronization with the patient's lung and / or cardiac cycles. 【0096】 The image processing and analysis system 104 may include one or more servers 139 for handling incoming requests and responses, and one or more rendering or image processing and analysis computers 140. The servers 139 may take the form of, for example, one or more server computers, workstation computers, supercomputers, or personal computers, and may execute server software or instructions. The one or more rendering or image processing and analysis computers 140 may take the form of one or more computers, workstation computers, supercomputers, or personal computers, and may execute image processing and / or analysis software or instructions. The one or more rendering or image processing and analysis computers 140 typically utilize one, preferably more, graphics processing units (GPUs) or GPU cores. 【0097】 The image processing and analysis system 104 may include one or more non-temporary computer-readable media 142 (e.g., magnetic or optical hard drives, RAID, RAM, Flash) for storing processor-executable instructions and / or data or other information. The image processing and analysis system 104 may include one or more image processing and analysis operator systems 144. The image processing and analysis operator system 144 may include a computer system 146, a monitor or display 148, a keypad and / or keyboard 150, and / or a cursor control device such as a mouse 152, joystick, trackpad, or trackball. The image processing and analysis operator system 144 may be communicatively coupled to a rendering or image processing and analysis computer 140 via one or more networks, such as a LAN 154. While many image processing techniques and analyses can be fully automated, the image processing and analysis operator system may allow technicians to perform some image processing and / or analysis operations on MRI data captured from patients. 【0098】 In many implementations, it is shown as a single non-temporary computer-readable or processor-readable storage medium 142, but the non-temporary computer-readable or processor-readable storage medium 142 may constitute multiple non-temporary storage media. Multiple non-temporary storage media may generally be located in a common location or distributed in various remote locations. Thus, databases of raw MRI data, pre-processed MRI data, and / or processed MRI data may be implemented on one or more non-temporary computer-readable or processor-readable storage media. Such databases may be stored separately on separate computer-readable or processor-readable storage media 142, or they may be stored on the same computer-readable or processor-readable storage media 142. The computer-readable or processor-readable storage media 142 may be located in the same location as the image processing and analysis system 104, for example, in the same room, building, or facility. Alternatively, the computer-readable or processor-readable storage media 142 may be located remotely from the image processing and analysis system 104, for example, in different facilities, cities, states, or countries. Electronic or digital information, files or records, or other collections of information may be stored in specific locations within a non-temporary computer-readable or processor-readable medium 142, i.e., in logically addressable portions of such medium that may or may not be connected. 【0099】 As described above, the image processing and analysis system 104 may be located remotely from the MRI acquisition system 102. The MRI acquisition system 102 and the image processing and analysis system 104 can communicate, for example, via one or more communication channels, such as a local area network (LAN) 106a and a wide area network (WAN) 106b. Network 106 may include packet-switched communication networks, such as the Internet, the World Wide Web portion of the Internet, extranets, and / or intranets. Network 106 may also take the form of various other types of telecommunication networks, such as cellular telephone and data networks, as well as basic telephone systems (POTS) networks. The type of communication infrastructure should not be considered limiting. 【0100】 As shown in Figure 1, the MRI acquisition system 102 is communicatively connected to a first LAN 106a. The first LAN 106a can be a network operated by or for a clinical facility and can provide local area communication for the clinical facility. The first LAN 106a is communicatively connected to a WAN (e.g., the Internet) 106b. A first firewall 156a can provide security for the first LAN. 【0101】 Furthermore, as shown in Figure 1, the image processing and analysis system 104 is communicatively connected to a second LAN 154. The second LAN 154 can be a network operated by or for an image processing facility or entity, and can provide local area communication for the image processing facility or entity. The second LAN 154 is communicatively connected to a WAN 106b (e.g., the Internet). A second firewall 156b can provide security for the second LAN 154. 【0102】 The image processing facility or entity may be independent of the clinical facilities and may, for example, be an independent entity providing services to one, two, or more clinical facilities. 【0103】 Although not shown in the diagram, the communication network may include one or more additional networking devices. Networking devices can take any variety of forms, including servers, routers, network switches, bridges, and / or modems (e.g., DSL modems, cable modems). 【0104】 Figure 1 shows a typical networked environment 100, but a typical networked environment may include many additional MRI acquisition systems, image processing and analysis systems 104, computer systems, and / or entities. The concepts taught herein may also be used in networked environments that incorporate more than those shown. For example, a single entity may provide image processing and analysis services to multiple diagnostic entities. One or more of the diagnostic entities may operate two or more MRI acquisition systems 102. For example, a large hospital or a dedicated medical imaging center may operate two, three, or even more MRI acquisition systems in a single facility. Typically, an entity providing image processing and analysis services may operate multiple entities, thereby providing an image processing and analysis system 104 that may include two, three, or even more rendering or image processing and analysis computers 140. 【0105】 Figure 2 shows a networked environment 200 comprising one or more image processing and analysis systems 104 (only one is shown) and one or more associated non-temporary computer-readable or processor-readable storage media 204 (only one is shown). The associated non-temporary computer-readable or processor-readable storage media 204 are communicably coupled to the image processing and analysis systems 104 via one or more communication channels, for example, one or more parallel cables, serial cables, or wireless channels capable of high-speed communication, such as FireWire®, Universal Serial Bus® (USB) 2 or 3, and / or Thunderbolt®, Gigabyte Ethernet®. 【0106】 The networked environment 200 also includes one or more end MRI acquisition systems 102 (only one is shown). The MRI acquisition system 102 is communicably coupled to an image processing and analysis system 104 via one or more communication channels, for example, one or more wide area networks (WANs) 210, such as the Internet or a portion of its World Wide Web. 【0107】 In operation, the MRI acquisition system 102 typically functions as a client to the image processing and analysis system 104. In operation, the image processing and analysis system 104 typically functions as a server to receive requests or information (e.g., MRI datasets) from the MRI acquisition system 102. This specification describes an overall process that utilizes asynchronous commands and imaging pipelines to enable remote image processing and analysis from the MRI acquisition system 102 (e.g., via a WAN). This approach offers several distinctive advantages, for example, enabling the MRI acquisition system 102 to be operated by technicians without requiring the participation of clinicians (e.g., physicians). Various techniques or methods for enhancing security while enabling access to medical images and patient-specific health information are also described. 【0108】 In some implementations, the image processing and analysis system 104 is shown to be located remotely from the MRI acquisition system 102, but it may also be located in the same location as the MRI acquisition system 102. In other implementations, one or more of the operations or functions described herein may be performed by the MRI acquisition system 102 or via a processor-based device located in the same location as the MRI acquisition system 102. 【0109】 The image processing and analysis system 104 receives an MRI dataset, performs image processing on the MRI dataset, and provides the processed MRI dataset to a clinician, for example, for review. The image processing and analysis system 104 can, for example, perform error detection and / or correction on the MRI dataset, such as phase error correction, phase aliasing detection, signal unwrapping, and / or detection and / or correction of other artifacts. Phase errors are related to phase, and so are phase aliasing. Signal unwrapping is related to magnitude. Various other artifacts may be related to phase and / or magnitude. 【0110】 The image processing and analysis system 104 can, for example, perform segmentation to distinguish between different tissue types. The image processing and analysis system 104 can, for example, perform quantification, such as comparing blood flow into and out of a closed anatomical structure or blood flow through two or more anatomical structures. Advantageously, the image processing and analysis system 104 can use quantification to verify results, for example, to confirm the identification of a specific tissue and / or to provide an indicator of the certainty of the results. In addition, advantageously, the image processing and analysis system 104 can use quantification to identify the presence of shunts. 【0111】 In some implementations, the image processing and analysis system 104 can generate images that reflect blood flow, including, for example, distinguishing between arterial and venous blood flow. For example, the image processing and analysis system 104 can utilize a first color map (e.g., blue) to show arterial blood flow and a second color map (e.g., red) to show venous blood flow. The image processing and analysis system 104 can use other characteristic colors or visual highlights to indicate abnormalities (e.g., shunts). Numerous different techniques for distinguishing between different tissues and between arterial and venous blood flow are described. Flow visualization can be superimposed, for example, as one or more layers on or over a visual representation of anatomical structures or magnitude data. 【0112】 In some implementations, the image processing and analysis system 104 can generate a patient-specific 4D flow protocol for use when operating the MRI acquisition system 102 for a particular patient. This may include setting the appropriate velocity encoding (VENC) for operating the MRI machine. 【0113】 The image processing and analysis system 104 can autonomously perform one or more of these actions or functions without human input. Alternatively, the image processing and analysis system 104 can perform one or more of these actions or functions based on human input, such as identifying a point, location, or plane, or identifying anatomical tissue characteristics in other forms. Several planes and views can be predefined, thereby enabling an operator, user, or clinician to quickly and easily obtain the desired view by simply selecting a plane (e.g., valve plane) or named view (e.g., two-chamber section, three-chamber section, four-chamber section). 【0114】 The networked environment 200 may utilize other computer systems and network equipment, such as additional servers, proxy servers, firewalls, routers, and / or bridges. While the image processing and analysis system 104 may be referred to in the singular form herein, this is not intended to limit embodiments to a single device, as a typical embodiment may include multiple image processing and analysis systems 104. Unless otherwise specified, the configuration and operation of the various blocks shown in Figure 2 are of conventional design. Consequently, such blocks do not need to be described in further detail herein, as they will be understood by those skilled in the art. 【0115】 The image processing and analysis system 104 may include one or more processing units 212a, 212b (collectively 212), system memory 214, and system bus 216, the system bus 216 connecting various system components, including the system memory 214, to the processing units 212. The processing units 212 may be any logic processing units, such as one or more central processing units (CPUs) 212a, digital signal processors (DSPs) 212b, application-specific integrated circuits (ASICs), or field-programmable gate arrays (FPGAs). The system bus 216 may utilize any known bus structure or architecture, including a memory bus with a memory controller, a peripheral bus, and / or a local bus. The system memory 214 includes read-only memory ("ROM") 218 ​​and random access memory ("RAM") 220. A basic input / output system ("BIOS") 222 may form part of the ROM 218 and includes basic routines that help transfer information between elements within the image processing and analysis system 104, such as during startup. 【0116】 The image processing and analysis system 104 may include a hard disk drive 224 for reading and writing to a hard disk 226, an optical disk drive 228 for reading and writing to a removable optical disk 232, and / or a magnetic disk drive 230 for reading and writing to a magnetic disk 234. The optical disk 232 may be a CD-ROM, and the magnetic disk 234 may be a magnetic floppy disk or diskette. The hard disk drive 224, the optical disk drive 228, and the magnetic disk drive 230 can communicate with the processing unit 212 via a system bus 216. The hard disk drive 224, the optical disk drive 228, and the magnetic disk drive 230 may include interfaces or controllers (not shown) coupled between such drives and the system bus 216, as is known to those skilled in the art. The drives 224, 228, and 230, and the computer-readable media 226, 232, and 234 associated with them, provide non-volatile storage for computer-readable instructions, data structures, program modules, and other data for the image processing and analysis system 104. The illustrated image processing and analysis system 104 is shown using a hard disk 224, an optical disk 228, and a magnetic disk 230, but it will be understood by those skilled in the art that other types of computer-readable media capable of storing computer-accessible data may also be used, such as WORM drives, RAID drives, magnetic cassettes, flash memory cards, digital video discs ("DVDs"), Bernoulli cartridges, RAM, ROM, smart cards, etc. 【0117】 An operating system 236, one or more application programs 238, other programs or modules 240, and program modules such as program data 242 may be stored in system memory 214. An application program 238 may include instructions that cause the processor 212 to perform image processing and analysis on an MRI image dataset. For example, an application program 238 may include instructions that cause the processor 212 to perform phase error correction on phase or velocity-related data. For example, an application program 238 may include instructions that cause the processor 212 to perform correction for phase aliasing. Alternatively, for example, an application program 238 may include instructions that cause the processor 212 to perform signal unwrapping. Additionally or alternatively, an application program 238 may include instructions that cause the processor 212 to identify and / or correct for artifacts. 【0118】 The application program 238 may include instructions to cause the processor 212 to perform segmentation to distinguish between different tissue types. The application program 238 may also include instructions to cause the processor 212 to perform quantification, for example, comparing blood flow into and out of a closed anatomical structure or blood flow through two or more anatomical structures. The application program 238 may also include instructions to cause the processor 212 to use quantification to verify the results, for example, to confirm the identification of a particular tissue and / or to provide an indicator of the certainty of the results. The application program 238 may also include instructions to cause the processor 212 to use quantification to identify the presence of a shunt. 【0119】 The application program 238 may include instructions to the processor 212 to generate images that reflect blood flow, for example, images that distinguish between arterial and venous blood flow. For example, a first color map (e.g., blue) may be used to show arterial blood flow, and a second color map (e.g., red) may be used to show venous blood flow. Other characteristic colors or visual highlights may be used to indicate abnormalities (e.g., shunts). A color transformation function may be applied to generate the color maps. The application program 238 may include instructions to the processor 212 to overlay a flow visualization (e.g., MRI phase data showing blood flow velocity and / or volume) onto a visualization or rendered image of an anatomical structure (e.g., MRI magnitude data). These instructions can cause flow visualizations to be rendered as one or more layers on an image of an anatomical structure, for example, as a color heat map and / or as vectors (e.g., arrow icons) with direction and magnitude (e.g., represented by length, line thickness), providing a fusion of anatomical structure (i.e., magnitude) and flow (i.e., phase) information. These instructions can additionally or alternatively induce the generation of spatial mappings or visualizations of signal distortion, turbulence, and / or pressure, which can be overlaid or superimposed on the spatial mappings or visualizations of the anatomical structure. Fusing phase- or velocity-related information with visualizations of anatomical information or visual representations of anatomical structures can facilitate the identification of anatomical landmarks. These instructions can leverage a set or array of graphics processing units, i.e., GPUs, to rapidly render the visualizations. 【0120】 Furthermore, a transformation function may be applied to determine which visual effect (e.g., color) to apply to which tissue. For example, arterial blood flow may be colored in shades of blue, venous blood flow in shades of red, and adipose tissue in shades of yellow. Anatomical structures represented as magnitude in an MRI image dataset may be visualized using, for example, grayscale. The depth of the view may be adjustable by the operator or user, for example, via a slider control on a graphical user interface. Thus, the visualization can take the form of a fused view that favorably merges the visual representation of velocity information with the visual representation of anatomical information or structures. 【0121】 The application program 238 may include instructions that cause the processor 212 to generate a patient-specific 4D flow protocol for use when operating the MRI acquisition system 102 for a particular patient. This protocol may be based, for example, on patient-specific input provided by a technician, and also on the specific MRI machine used to capture the MRI dataset. 【0122】 The application program 238 may include instructions that cause the processor 212 to receive an image dataset from the MRI acquisition system, process and / or analyze the image dataset, and provide the processed and / or analyzed images and other information to a remotely located user in a time-constrained and secure manner. This is described in detail herein with reference to various figures. 【0123】 Furthermore, the system memory 214 may include a server 244 that causes the image processing and analysis system 104 to supply electronic information or files via the Internet, intranet, extranet, telecommunications network, or other network, as described later. The server 244 in the illustrated embodiment may be based on a markup language such as Hypertext Markup Language (HTML), Extended Markup Language (XML), or Wireless Markup Language (WML), and may operate with a markup language that uses syntactically delimited characters added to the document data to represent the structure of the document. Several suitable servers may be commercially available from companies such as Mozilla, Google, Microsoft, and Apple Computer. 【0124】 Although shown in Figure 2 as being stored in system memory 214, the operating system 236, application programs 238, other programs / modules 240, program data 242, and server 244 can be stored on the hard disk 226 of the hard disk drive 224, the optical disk 232 of the optical disk drive 228, and / or the magnetic disk 234 of the magnetic disk drive 230. 【0125】 The operator can input commands and information to the image processing and analysis system 104 via input devices such as a touchscreen or keyboard 246, and / or a pointing device such as a mouse 248, and / or a graphical user interface. Other input devices may include microphones, joysticks, gamepads, tablets, scanners, etc. These and other input devices are connected to one or more of the processing units 212 via an interface 250, such as a serial port interface coupled to the system bus 216, although other interfaces may be used, such as a parallel port, game port, wireless interface, or Universal Serial Bus ("USB"). A monitor 252 or other display device is coupled to the system bus 216 via a video interface 254, such as a video adapter. The image processing and analysis system 104 may include other output devices such as speakers and printers. 【0126】 The image processing and analysis system 104 can operate in a networked environment 200 using logical connections to one or more remote computers and / or devices. For example, the image processing and analysis system 104 can operate in a networked environment 200 using logical connections to one or more MRI acquisition systems 102. Communication can be via wired and / or wireless network architectures, such as wired and wireless enterprise-scale computer networks, intranets, extranets, and / or the Internet. Other embodiments may include other types of communication networks, including telecommunications networks, cellular networks, paging networks, and other mobile networks. Between the image processing and analysis system 104 and the MRI acquisition system 102, there may be various computers, switching devices, routers, bridges, firewalls, and other devices in the communication path. 【0127】 The MRI acquisition system 102 typically takes the form of an MRI machine 108, as well as one or more associated processor-based devices, such as an MRI control system 126 and / or an MRI operator system 128. The MRI acquisition system 102 captures MRI information or datasets from the patient. Therefore, in some examples, the MRI acquisition system 102 may be named a front-end MRI acquisition system or MRI acquisition system to distinguish it from the MRI image processing and analysis system 104, while the MRI image processing and analysis system 104 may, in some examples, be named an MRI back-end system. Although the MRI acquisition system 102 may be referred to in the singular form herein, this is not intended to limit embodiments to a single MRI acquisition system 102. In typical embodiments, there may be two or more MRI acquisition systems 102, and often there are numerous MRI acquisition systems 102 in a networked environment 200. 【0128】 The MRI acquisition system 102 can be communicatively coupled to one or more server computers (not shown). For example, the MRI acquisition system 102 can be communicatively coupled via one or more diagnostic facility server computers (not shown), routers (not shown), bridges (not shown), LAN 106a (Figure 1), etc., which may include or implement firewalls 156a (Figure 1). The server computer (not shown) can execute a set of server commands to function as a server for several MRI acquisition systems 102 (i.e., clients) communicatively coupled via LAN 106a at a clinical facility or site, and thus can act as an intermediary between the MRI acquisition system 102 and the MRI image processing and analysis system 104. The MRI acquisition system 102 can execute a set of client commands to function as a client of a server computer communicatively coupled via a WAN. 【0129】 The MRI control system 126 typically includes one or more processors (e.g., microprocessors, central processing units, digital signal processors, graphics processing units) and non-temporary processor-readable memory (e.g., ROM, RAM, Flash, magnetic and / or optical disks). The MRI operator's system 128 can take the form of a computer that executes appropriate instructions, such as a personal computer (e.g., a desktop or laptop computer), a netbook computer, a tablet computer, a smartphone, a personal digital assistant, a workstation computer, and / or a mainframe computer. 【0130】 The MRI operator's system 128 may include one or more processing units 268, a system memory 269, and a system bus (not shown), the system bus connecting various system components, including the system memory 269, to the processing units 268. 【0131】 The processing unit 268 can be any logic processing unit, such as one or more central processing units (CPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or graphics processing units (GPUs). Non-limiting examples of commercially available computer systems include, but are not limited to, Intel Corporation's 80x86 or Pentium series microprocessors, IBM's PowerPC microprocessors, Sun Microsystems, Inc.'s Sparc microprocessors, Hewlett-Packard Company's PA-RISC series microprocessors, Motorola Corporation's 68xxx series microprocessors, ATOM processors, or A4 or A5 processors. Unless otherwise specified, the configuration and operation of the various blocks of the MRI acquisition system 102 shown in Figure 2 are of conventional design. Consequently, such blocks do not need to be described in further detail herein, as they are understood by those skilled in the art. 【0132】 The system bus can utilize any known bus structure or architecture, including a memory bus with a memory controller, a peripheral bus, and a local bus. The system memory 269 includes read-only memory ("ROM") 270 and random access memory ("RAM") 272. The basic input / output system ("BIOS") 271 may form part of the ROM 270 and includes basic routines that help transfer information between elements in the MRI acquisition system 102, such as during startup. 【0133】 The MRI operator's system 128 may also include one or more media drives 273, e.g., hard disk drives, magnetic disk drives, WORM drives, and / or optical disk drives, for reading and writing computer-readable storage media 274, e.g., hard disks, optical disks, and / or magnetic disks. The non-temporary computer-readable storage media 274 may take the form of, for example, removable media. For example, a hard disk may take the form of a Winchester drive, an optical disk may take the form of a CD-ROM, and a magnetic disk may take the form of a magnetic floppy disk or diskette. The media drives 273 communicate with the processing unit 268 via one or more system buses. The media drives 273 may include interfaces or controllers (not shown) coupled between such drives and the system buses, as is known to those skilled in the art. The media drives 273 and the non-temporary computer-readable storage media 274 associated therewith provide non-volatile storage of computer-readable instructions, data structures, program modules, and other data for the MRI acquisition system 102. Although the system is shown to utilize computer-readable storage media 274 such as hard disks, optical disks, and magnetic disks, it will be understood by those skilled in the art that the MRI operator's system 128 may also utilize other types of non-temporary computer-readable media capable of storing computer-accessible data, such as magnetic cassettes, flash memory cards, digital video discs ("DVDs"), Bernoulli cartridges, RAM, ROM, smart cards, and the like. Data or information, such as electronic or digital files or data or metadata associated therewith, may be stored on the non-temporary computer-readable storage media 274. 【0134】 The system memory 269 may store an operating system, one or more application programs, other programs or modules, and program modules such as program data. The program modules may include instructions for accessing websites, extranet sites, or other sites or services (e.g., web services) hosted or provided by the MRI processing and analysis system 104, and associated web pages, other pages, screens, or services. 【0135】 In particular, the system memory 269 may include a communication program that enables the MRI acquisition system 102 to exchange electronic or digital information, files, data, or metadata with MRI image processing and / or analysis services provided by the MRI processing and analysis system 104. The communication program may be, for example, a web client or browser that enables the MRI acquisition system 102 to access information, files, data, and / or metadata and exchange them with sources such as websites on the Internet, a corporate intranet, an extranet, or other network. Such actions may require the end-user client to have sufficient rights, permissions, privileges, or authority to access a given website, for example, one hosted by the MRI processing and analysis system 104. As discussed herein, patient-identifying data may reside on systems operated by or for clinical facilities, but may not be accessible by or through systems operated by or for imaging facilities or imaging facility personnel. Browsers can be based on markup languages ​​such as Hypertext Markup Language (HTML), Extended Markup Language (XML), or Wireless Markup Language (WML), and can work with markup languages ​​that use syntactically delimited characters added to the document's data to represent the document's structure. 【0136】 Although described as being stored in system memory 269, the operating system, application programs, other programs / modules, program data, and / or browser can be stored in computer-readable storage media 274 of media drive 273. The operator can input commands and information to the MRI operator's system 128 via input devices such as a touchscreen or keyboard 276, and / or a pointing device 277 such as a mouse. Other input devices may include microphones, joysticks, gamepads, tablets, scanners, etc. These and other input devices are connected to the processing unit 269 via an interface such as a serial port interface coupled to the system bus, although other interfaces may be used, such as a parallel port, game port, wireless interface, or Universal Serial Bus ("USB"). A display or monitor 278 may be coupled to the system bus via a video interface such as a video adapter. The MRI operator system 128 may include other output devices such as speakers and printers. 【0137】 MRI image processing and analysis systems can construct static interfaces that allow various tissue types to be subtracted from or added to MRI4D flow datasets. For example, static tissues such as fat or bone can be distinguished from non-static tissues such as air or flowing blood. MRI image processing and analysis systems can also autonomously distinguish various non-static tissues, for example, distinguishing between air (e.g., lungs) and flowing blood. Furthermore, MRI image processing and analysis systems can distinguish between arterial and venous blood flow. 【0138】 For example, MRI imaging and analysis systems can utilize the Fast Fourier Transform to identify blood tissues that are expected to have pulsating patterns or waveforms. Air or lungs tend to have patterns that appear random across a defined volume when the velocities of neighboring voxels are compared. For example, voxels with strong or fast velocities typically indicate air. MRI datasets can be quite large, for example, 256 × 256 × 256 × 20 time points. MRI imaging and analysis systems can rely on gradients (e.g., gradient descent) to detect different tissue types, and advantageously, numerical methods can be used more readily than analytical methods to handle relatively large MRI datasets quickly. By controlling the number of significant digits of the numerical method (e.g., 2), MRI imaging and analysis systems can achieve very fast results (e.g., 1 second for 30 minutes) while still obtaining results that are sufficiently accurate for specific applications. 【0139】 In some implementations, different tissue types can be subtracted one by one from a patient MRI dataset. For example, air or lungs can be subtracted, blood can be subtracted, atrial flow can be separated from venous flow, bone can be subtracted, and fat can be left. In particular, since fat is static, each voxel representing fat should have a zero velocity associated with it. MRI image processing and analysis systems can advantageously use such ground truth to correct the MRI dataset for all tissue types. 【0140】 If a non-zero velocity is found for adipose-type tissue, this can be used to adjust the entire set of data (e.g., for all tissues). For example, an MRI imaging and analysis system can generate or create a polynomial model based on an identified region or volume (e.g., adipose or soft tissue). It can be a simple polynomial (e.g., ax). 2This can be a polynomial (+bx+c), or a much more complex polynomial (e.g., a non-uniform rational B-spline). MRI imaging and analysis systems can find the coefficients of the polynomial fit to the image, for example, using linear regression or linear algebra techniques. This results in a model that the MRI imaging and analysis system can apply (e.g., subtract) to the entire field, not just fat or soft tissue. 【0141】 In one implementation, a replica body is imaged to create a reference set of data or a "phantom" model that can be subtracted from actual patient data. The replica body may be formed from a material that mimics the MRI response of a real body but does not have blood flow. The phase gradient in the reference set of data or the "phantom" model can represent noise (e.g., random noise) and can be used to correct for phase shifts. This technique advantageously avoids the need to generate polynomials to fit the 3D data. The generated reference set or phantom model may be valid for several months of MRI machine operation, but if the MRI machine is repaired or moved, a new set of reference data or a phantom model should be generated. 【0142】 MRI image processing and analysis systems can define various filters or masks to remove different tissue types or to remove venous or atrial blood flow. Filters or masks can remove abnormal blood flow, such as blood flow outside a reasonable range (e.g., too high or too fast, too slow or too low), or blood flow that appears to be present in anatomical structures where there should be no blood flow (e.g., bone). Filters or masks can also be defined to display only voxels with magnitudes having an absolute value greater than a certain threshold. Filters or masks can also be defined to display only voxels whose absolute value of the cross product of magnitude and velocity is greater than a defined threshold. Furthermore, filters or masks can be defined to show only voxels whose vectors are in the same direction as the vectors of neighboring voxels, for example, to identify or view high-velocity jets. In particular, the fact that the velocity vectors of neighboring voxels are in various directions can be an indicator of noise. 【0143】 Figure 3A shows a typical data flow 300 in an MRI image processing and analysis or rendering system according to one exemplary embodiment. 【0144】 The MRI image processing and analysis system 104 (Figure 1) receives the "raw" MRI image data 302, for example, in the form of a DICOM-compliant electronic file. In some implementations, the "raw" MRI image data 302 may have undergone some minimal image preprocessing by a processor-based system associated with the acquisition unit (MRI acquisition system 102, Figure 1). 【0145】 In step 304, the MRI image processing and analysis system performs data preprocessing on the "raw" data. 【0146】 Preprocessing may include, for example, phase error correction 306, which can be performed autonomously by an MRI image processing and analysis system. Various suitable methods for phase error correction 306 are discussed elsewhere in this specification. 【0147】 As part of preprocessing 304, the MRI image processing and analysis system may perform tissue segmentation, for example, 308. The MRI image processing and analysis system may perform tissue segmentation 306 autonomously, as indicated by the circle surrounding the letter A. Additionally or alternatively, the MRI image processing and analysis system may perform tissue segmentation 308 manually, for example, by interaction with the MRI image processing and analysis system (e.g., receiving input from a human), as indicated by the circle surrounding the letter M. Various appropriate methods for tissue segmentation 306 are discussed elsewhere in this specification. 【0148】 As part of preprocessing 304, the MRI image processing and analysis system may perform landmark identification in, for example, 310. The MRI image processing and analysis system may perform landmark identification 310 autonomously, as indicated by the circle surrounding the letter A. Alternatively or additionally, the MRI image processing and analysis system may perform landmark identification 310 manually, for example, by interaction with the MRI image processing and analysis system (e.g., receiving input from a human), as indicated by the circle surrounding the letter M. Various appropriate methods for landmark identification 310 are discussed elsewhere in this specification. 【0149】 In step 312, the MRI image processing and analysis system visualizes the preprocessed data. 【0150】 As part of the visualization 312, the MRI image processing and analysis system can perform velocity-weighted volume rendering, for example, 314. Various appropriate methods for velocity-weighted volume rendering 314 are discussed elsewhere in this specification. 【0151】 As part of visualization 312, the MRI image processing and analysis system can perform simultaneous vector field and anatomical structure visualization or rendering, for example, 316. Various appropriate methods for simultaneous vector field and anatomical structure visualization or rendering 316 are discussed elsewhere in this specification. 【0152】 As part of the visualization 312, the MRI image processing and analysis system can perform a natural volume rendering transition, for example, 318. Various appropriate methods for the natural volume rendering transition 318 are discussed elsewhere in this specification. 【0153】 As part of visualization 312, the MRI image processing and analysis system can, for example, provide or create an interactive landmarked-based view 320. Various suitable methods for providing the interactive landmarked-based view 320 are discussed elsewhere in this specification. 【0154】 As part of the visualization 312, the MRI image processing and analysis system may perform or provide interactive 3D volumetric segmentation, for example, 322. Various suitable methods for providing interactive 3D volumetric segmentation 322 are discussed elsewhere in this specification. 【0155】 As part of visualization 312, the MRI image processing and analysis system can perform visualization-guided cardiac and / or vessel interrogation, for example, 324. Various appropriate methods for performing or providing visualization-guided cardiac and / or vessel interrogation 324 are discussed elsewhere in this specification. 【0156】 As part of visualization 312, an MRI image processing and analysis system can perform or provide, for example, a three-dimensional vector field and anatomical visualization 326. Various suitable methods for providing three-dimensional vector field and anatomical visualization 326 are discussed elsewhere in this specification. 【0157】 In step 328, the MRI image processing and analysis system visualizes the preprocessed data. 【0158】 As part of performing quantification 328, the MRI image processing and analysis system can perform vascular measurements 330, for example. The MRI image processing and analysis system can perform vascular measurements 330 autonomously, as indicated by the circle surrounding the letter A. Alternatively or additionally, the MRI image processing and analysis system can perform vascular measurements 330 manually, for example, by interaction with the MRI image processing and analysis system (e.g., receiving input from a human), as indicated by the circle surrounding the letter M. Various appropriate methods for vascular measurements 330 are discussed elsewhere in this specification. 【0159】 As part of performing quantification 328, the MRI image processing and analysis system can perform blood flow quantification 332, for example. The MRI image processing and analysis system can perform blood flow quantification 332 autonomously, as indicated by the circle surrounding the letter A. Alternatively or additionally, the MRI image processing and analysis system can perform blood flow quantification 332 manually, for example, by interaction with the MRI image processing and analysis system (e.g., receiving input from a human), as indicated by the circle surrounding the letter M. Various appropriate methods for blood flow quantification 332 are discussed elsewhere in this specification. 【0160】 As part of performing quantification 328, the MRI image processing and analysis system can, for example, perform cardiac chamber volume quantification 334. The MRI image processing and analysis system can perform cardiac chamber volume quantification 334 autonomously, as indicated by the circle surrounding the letter A. Alternatively or additionally, the MRI image processing and analysis system can perform cardiac chamber volume quantification 334 manually, for example by interaction with the MRI image processing and analysis system (e.g., receiving input from a human), as indicated by the circle surrounding the letter M. Various appropriate methods for cardiac chamber volume quantification 334 are discussed elsewhere in this specification. 【0161】 As part of performing quantification 328, the MRI image processing and analysis system may perform pressure quantification 336, for example. The MRI image processing and analysis system may perform pressure quantification 336 autonomously, as indicated by the circle surrounding the letter A. Alternatively or additionally, the MRI image processing and analysis system may perform pressure quantification 336 manually, for example, by interaction with the MRI image processing and analysis system (e.g., receiving input from a human), as indicated by the circle surrounding the letter M. Various appropriate methods for pressure quantification 336 are discussed elsewhere in this specification. 【0162】 In 338, the MRI image processing and analysis system performs automated result verification. Various appropriate methods for performing automated result verification 338 are discussed elsewhere in this specification. 【0163】 In 340, the MRI image processing and analysis system performs automated structural reporting. Various appropriate methods for performing automated structural reporting 340 are discussed elsewhere in this specification. 【0164】 Figure 3B shows a data flow 350 in an MRI imaging and processing / analysis environment according to one exemplary embodiment. 【0165】 The MRI imaging and processing / analysis data flow 350 can be broken down into three segments or stages, namely, the image acquisition stage 352, the image reconstruction stage 354, and the post-processing stage 356. 【0166】 During the image acquisition phase 352, various pulse sequences (PS1, PS2, PS) are used to drive the magnets of the MRI machine. n A 4D flow PS is used to collect raw or k-space MRI data or information. The raw or k-space MRI data or information represents multiple voxels and may include magnitude and phase values ​​for each voxel. Magnitude values ​​represent anatomical structures, and phase values ​​represent blood flow or velocity. 【0167】 Raw MRI data is processed during image reconstruction stage 354. Image reconstruction can result in, for example, a standardized file, such as an MRI dataset in the DICOM standard format. Image reconstruction may be specific to the MRI machine manufacturer, such as GE, Siemens, or Philips, and may utilize software or firmware provided by the MRI machine manufacturer. Typically, image reconstruction is performed locally at the site where the MRI machine is installed. 【0168】 The MRI dataset (e.g., a DICOM file) is further processed in post-processing stage 356. In some particularly advantageous implementations, post-processing stage 356 is performed remotely from image acquisition stage 352 and image reconstruction stage 354. For example, post-processing stage 356 can be performed in a dedicated post-processing facility, which may be operated by an entity that is further separated and independent from one or more entities (e.g., a clinical facility or organization) that perform image acquisition stage 352 and image reconstruction stage 354. 【0169】 The post-processing stage 356 may be divided into the operation stage 360, the error detection and / or correction stage 362, the visualization stage 364, and the quantification stage 366. Although shown in order, these stages 360-366 may be performed in a different order or not in order. For example, many of the actions grouped in each stage may be performed simultaneously, or some actions in one stage may be performed before actions in another stage, while other actions in one stage may be performed after actions in another stage. 【0170】 Operational phase 360 ​​may include receiving requests (e.g., web requests regarding MRI datasets), queuing requests, and providing responses. Many techniques for implementing operational phases, for example, through asynchronous command and imaging pipeline architectures, are described below. 【0171】 Error detection and / or correction step 362 includes error detection and, where possible, error correction on the MRI dataset. Error detection and / or correction step 362 may include, for example, eddy current detection and / or correction, phase detection and / or correction, image unwrapping, magnitude aliasing detection and / or correction, phase aliasing detection and / or correction, and / or artifact detection and / or correction. Many error detection and / or correction techniques are described in detail elsewhere in this specification. Many error detection and / or correction activities can be performed autonomously by rendering or post-processing systems, while others may utilize human intervention or input to a computer. 【0172】 Visualization stage 364 may include various actions, such as defining or setting landmarks (e.g., anatomical landmarks or structures), and generating two-dimensional contours and / or three-dimensional volumes through segmentation. In addition, visualization stage 364 may include distinguishing information representing arterial and venous blood flow and identifying them, for example, through appropriate colors (e.g., blue, red). Furthermore, visualization stage 364 may include identifying suspected shunts representing blood flow that should not occur in a healthy subject, and identifying them, for example, through appropriate colors. Many visualization actions may be performed autonomously by rendering or post-processing systems, while others may utilize human intervention or input to the computer. 【0173】 During the quantification stage 366, various properties can be quantified and compared. For example, the quantification stage 366 can utilize the concepts of conservation of mass or conservation of flow. Thereafter, blood flow can be quantified and compared. For example, the flow of blood into a volume typically coincides with the flow of blood out of a volume. A volume can constitute a body organ, such as a vascular structure (e.g., arteries, veins, sinuses) or the heart, cardiac chambers, etc. A volume can be an enclosed structure with natural inlets and outlets. Alternatively, a volume may have defined inlets and outlets that do not correspond to any naturally occurring inlets or outlets. For example, any segment of an artery may be delineated or identified as a volume. Also, for example, two or more anatomical structures may be combined to form or define a volume. 【0174】 By comparing flow into and out of a volume, shunts (e.g., defects that allow blood flow that would not occur in normal, healthy tissue) can be identified. By comparing flow into and out of a volume, other post-processing actions can be verified or confirmed, for example, by verifying arterial flow versus venous flow segmentation or assignment to one or more voxels. 【0175】 Many quantification techniques are described in detail elsewhere in this specification. Many quantification activities can be performed autonomously by rendering or post-processing systems, while others may utilize human intervention or input to a computer. 【0176】 Pre-acquisition 4D Flow Localizer A 4D flow localizer is a unique pulse sequence (i.e., commands or drive signals that control the magnets of an MRI machine during an MRI scan or acquisition) that is tailored to collect information for a high-fidelity 4D flow scan or acquisition. Essentially, the 4D flow localizer is a first-pass scan or acquisition that helps to define the appropriate or optimized settings for the second pass (i.e., the high-fidelity 4D flow scan or acquisition). Before the patient enters the scanner, the type of 4D flow localizer is selected based on several variables. Clinical indication, type and amount of contrast agent, patient weight, patient height and heart rate, hardware manufacturer, type of coil used, and hardware type (1.5T or 3T) are used to predict the optimal localizer to use. The localizer data is then sent to a processor-based system (e.g., a server in the image processing and analysis system) to construct a patient-specific 4D flow protocol. 【0177】 For example, a localizer can help select the VENC (velocity encoding) and the required temporal resolution. For instance, the VENC may be selected based on the maximum velocity observed within the blood vessel of interest, either overall or partially. The VENC may be selected using a lookup table method and / or interpolation. Figure 4A shows a lookup table 400 with exemplary values ​​of the VENC for various ranges of velocity. 【0178】 Furthermore, for example, the scan length or duration may be selected based on the number of channels in the coil used for the MRI procedure (e.g., cardiac coils), either entirely or partially. Typically, a smaller number of channels results in a lower signal-to-noise ratio (SNR), i.e., requires a longer imaging time. Figure 4B is Table 402, showing some exemplary scan lengths or durations against the total number of channels in cardiac coils. 【0179】 In addition, as more data is collected and sent to the image processing and analysis system server, a learning or refinement algorithm can detect the optimal settings to use based on a set of input parameters. This algorithm may be trained to know or recognize whether the selected output variables were appropriate for a given scan (i.e., acquisition) by leveraging feedback from clinicians (e.g., from a quality range of 1 to 10 for high-fidelity 4D flow scans). The algorithm may utilize one or more artificial intelligence (AI) techniques, including but not limited to training neural networks. In the absence of a training set, the default values ​​for the input variables of a high-fidelity 4D flow scan will be selected. In addition to quality metrics, when images are returned to the server, this data is used to determine what the next scan settings should be. The goal is to improve scan quality over time or number of steps. Input parameters are incremented or decremented to determine whether scan quality improves or decreases. If a sufficient number of training sets are available from a particular institution, a multivariate optimization algorithm may be used to select the optimal input variables for any new patient that needs to be imaged. 【0180】 When in use, for each patient, the MRI imaging and analysis system can generate one or more patient-specific pulse sequences, which are used to operate the MRI machine to acquire MRI datasets. For example, the MRI imaging and analysis system can generate 4 to 6 pulse sequences for a patient. The MRI machine can execute pulse sequences in response to signals input by, for example, an MRI machine operator or technician. In particular, this allows for the capture or acquisition of desired MRI datasets without requiring the participation or time of a clinician (e.g., a physician). Inputs may include patient characteristics, gradient coil identification, and / or contrast. 【0181】 Figure 5 shows a method of operation 500 using a 4D flow localizer and an optional refinement algorithm according to one exemplary embodiment. 【0182】 Initial input variables 502 are generated or collected to create the 4D flow localizer 504. The input variables 502 may be collected, for example, via the MRI operator's system 112, for example, by a human, and / or automatically or autonomously. The input variables 502 may include one or more of the following: clinical indication, name or identifier of contrast agent, volume or dose of contrast agent, amount of time elapsed since the bolus was injected or delivered to the patient or subject, patient or subject weight, patient or subject height, patient or subject age, patient or subject heart rate, hardware manufacturer, maker, model, serial number or other hardware identifier, coil type (e.g., cardiac coil, number of channels), hardware magnet strength (e.g., 1.5T, 3T), and an indication of whether the patient or subject is sedated. 【0183】 The 4D flow localizer 504 may be generated via a processor-based device. In many implementations, the processor-based device is part of an image processing and analysis system, but the teachings herein are not limited to this. For example, in some implementations, the 4D flow localizer may be generated by a component or part of an MRI acquisition system. 【0184】 The operation of an MRI machine using a 4D flow localizer 504 generates or creates a set of output variables 506. In at least some examples, the output variables 506 can take the form of a 4D flow localizer image. 【0185】 At step 508, the output variables (e.g., 4D flow localizer images) are refined. For example, a processor-based device can run an algorithm, such as a machine learning algorithm. Alternatively, a neural network can be used to perform the refinement. 【0186】 As described above, machine learning or other techniques can be used to refine and generate a set of input variables 510 for the high-fidelity 4D flow. The input variables 510 for the high-fidelity 4D flow may include, for example, one or more of the following: scan or acquisition length or duration, VENC, field of view, repetition time (TR) (mm), echo time (TE) (mm), row resolution (mm), column resolution (mm), slice resolution (mm), temporal resolution (ms), and flip angle (degrees). 【0187】 A processor-based device uses the input variable 510 of the high-fidelity 4D flow to generate a high-fidelity 4D flow scan or acquisition 512. The high-fidelity 4D flow scan or acquisition 512 yields the output variable 514 of the high-fidelity 4D flow. The output of the high-fidelity 4D flow may include one or more of the following: a 4D flow image, scan quality (e.g., on a scale such as 1-10, evaluated by one or more experts), and events occurring during the scan (e.g., the patient is moved). 【0188】 visualization Arterial and venous blood flow isolation To highlight the differences between left heart (i.e., arterial) and right heart (i.e., venous) blood flow, MRI imaging and analysis systems can be configured to generate a unique color map (i.e., color gradient map) for each side. The MRI imaging and analysis system uses 2D or 3D rendering (i.e., streamlines and trajectories) to generate image data with voxels colored according to their respective color maps. Typically, arterial blood flow has red based on the color map, and venous blood flow has blue based on the color map. In a healthy system, these two blood pools never touch each other, except at the light microscopic level in the lungs, for example. To determine which voxels in the body belong to which side, the MRI imaging and analysis system runs an algorithm that, for each voxel, determines which bin the voxel belongs to and assigns the appropriate color to that voxel. 【0189】 Figure 6 shows a method 600 for generating a blood flow color map according to one exemplary embodiment. 【0190】 Method 600 starts at 602, for example, in response to being called by a calling routine. 【0191】 In 604, for each voxel in a plurality of voxels, the MRI imaging and analysis system determines whether the given voxel is within a blood pool. To make this determination, the MRI imaging and analysis system may utilize one of several segmentation algorithms or methods, for example, the various segmentation algorithms or methods described herein. The MRI imaging and analysis system may autonomously perform segmentation and / or determination to identify the enclosed volume or region or pathway. Alternatively, the MRI imaging and analysis system may receive input from a human operator or user to identify whether a given voxel is within a blood pool, or otherwise facilitate segmentation between hematopoietic and non-hematopoietic tissues. The determination may be repeated, for example, until all voxels have been analyzed. 【0192】 If a given voxel is determined to be in the blood pool, in 606, the MRI imaging and analysis system marks, identifies, or otherwise logically assigns the given voxel as a blood type tissue (e.g., logically). 【0193】 In 608, the MRI imaging and analysis system places all neighboring voxels marked or identified as blood into the same bin. If these bins come into spatial contact with each other over time, the MRI imaging and analysis system merges the bins of those contacting voxels into the same bin. Optionally, the MRI imaging and analysis system can calculate streamlines and / or trajectories in the blood pool, which can highlight the connected bins. 【0194】 In step 610, the MRI image processing and analysis system counts the number of bins. Ideally, there should be only two bins, one corresponding to arterial blood flow and the other to venous blood flow. 【0195】 At step 612, the MRI image processing and analysis system determines whether there are three or more bins. If the MRI image processing and analysis system determines that there are three or more bins, control proceeds to step 614. Otherwise, control proceeds to step 626. 【0196】 Regions containing a shunt (i.e., a connection between right and left ventricular flow) typically indicate a medical problem or abnormal condition. As shown in 614, MRI imaging and analysis systems can perform shunt detection methods and algorithms for these regions. Appropriate shunt detection methods or algorithms are described elsewhere in this specification. For example, an MRI imaging and analysis system can determine or identify regions where adjacent bins are connected by a defined threshold number of voxels or fewer voxels than a defined threshold region. For example, if the threshold is 10 voxels, an MRI imaging and analysis system can tentatively mark or identify as a shunt any region where two large volumes of blood flow are connected together by fewer than 10 voxels, as shown in 616. Care should be taken not to assume that such regions are necessarily shunts. These regions could be, for example, stenotic valves or stenoses. Optionally, in 618, an MRI imaging and analysis system visually highlights (e.g., highlights) these regions for the operator or user. The operator or user can visually confirm the presence or absence of a shunt very quickly and provide the corresponding input received by the MRI imaging and analysis system, as shown in 620. In 622, the MRI imaging and analysis system marks or identifies the corresponding voxel accordingly. In 624, the MRI imaging and analysis system updates or assigns the color associated with the region based on the user input confirming the presence or absence of a shunt. For example, if the presence of a shunt is confirmed, the MRI imaging and analysis system does not color the corresponding region with either the red or blue color map. Instead, the MRI imaging and analysis system associates a specific color with the voxel in this region to visually highlight (e.g., highlight) this region for the operator or user. 【0197】 In 626, the MRI imaging and analysis system determines whether a bin corresponds to venous or arterial flow. Various methods for determining whether a bin corresponds to venous or arterial flow are discussed herein with reference to Figures 4 and 5. In 628, the MRI imaging and analysis system associates each voxel of the two bins with a respective color or set of colors, for example, one or more blues for arterial flow and one or more reds for venous flow. Method 600 concludes in 630. 【0198】 Figure 7 shows a method 700 for determining whether a bin corresponds to venous or arterial flow, according to one illustrated embodiment. Method 700 is particularly suitable when anatomical landmarks (i.e., points in space, e.g., the apex of the left ventricle) are already defined, allowing for a simple selection of only the landmarks in the blood pool. Techniques for identifying and / or selecting landmarks are described elsewhere in this specification. 【0199】 To define arterial blood flow, in 702, the MRI image processing and analysis system can autonomously identify or select one or more anatomical landmarks associated with arterial blood flow, or it can receive user input through manual identification or selection. The anatomical landmarks may include, for example, one or more of the aortic valve point, mitral valve point, and / or the aorta (e.g., points within the aortic sinus, points within the ascending aorta, points at the top of the aortic arch, points within the descending aorta). 【0200】 To define venous blood flow, in 704, the MRI image processing and analysis system may autonomously identify or select one or more anatomical landmarks associated with venous blood flow, or it may receive user input through manual identification or selection. The anatomical landmarks may include, for example, one or more of the pulmonary valve point, tricuspid valve point, points within the left pulmonary artery, points within the right pulmonary artery, points within the inferior vena cava, and / or points within the superior vena cava. 【0201】 In 706, the MRI imaging and analysis system can autonomously determine the distance to the nearest point in each bin for all anatomical landmarks. Ideally, this distance should be zero, since each landmark is in one of each blood pool. In 708, the MRI imaging and analysis system assigns each bin closest to each anatomical landmark to either arterial or venous blood flow, depending on whether the anatomical landmark is in the arterial or venous blood pool. In 710, the MRI imaging and analysis system determines whether there is a conflict. Conflict indicates the presence of a shunt or that connected groups of voxels are in the wrong bin. In response to the identification of a conflict, in 712, the MRI imaging and analysis system provides a label or notification. 【0202】 MRI image processing and analysis systems implement image navigation, for example, autonomously or in response to operator or user input. For example, an MRI image processing and analysis system implements distance scrolling, enabling movement through a defined number of spatial slices in a defined spatial direction (e.g., along the x, y, or z axis) and comparison of data between spatial slices. For example, an MRI image processing and analysis system implements time or temporal scrolling, enabling movement through a defined number of temporally consecutive slices in a defined temporal direction (e.g., forward time, backward time) and comparison of data between temporal slices. As described elsewhere in this specification, an MRI image processing and analysis system can advantageously use comparisons between slices to identify artifacts. For example, an MRI image processing and analysis system implements zoom and / or pan operations, enabling an operator or user to zoom in on or pan the image to a portion of the image. 【0203】 Figure 8 shows a method 800 for determining whether a bin corresponds to venous or arterial flow, according to one exemplary embodiment. Method 800 is particularly suitable when anatomical landmarks are not identified. 【0204】 Method 800 starts at 802, for example, in response to a call from a calling routine. 【0205】 Method 800 is based on vector angles. In 804, the MRI image processing and analysis system determines the center of gravity of the heart. This may be done autonomously or based on operator or user input. For example, the center of gravity can be extracted from DICOM tags, since there is a patient coordinate system in which 0,0,0 in these fields corresponds to a specific point within the patient's anatomical structure. 【0206】 In 806, the MRI imaging and analysis system determines or identifies all voxels in a bin that have coherence above a threshold. In 808, for all voxels in a bin that have coherence above a threshold, the MRI imaging and analysis system calculates the average velocity over all time points and the angle between the voxel's center of mass and a vector connecting the center of mass of the heart. The MRI imaging and analysis system then calculates the average angle between all voxels in a bin and the center of mass. In 810, the MRI imaging and analysis system determines if there are additional bins to the processor and increments the bins in 812 within loop 814 to repeat calculations 806, 808 for all bins. As arterial blood flow moves away from the heart, in 816, the MRI imaging and analysis system assigns the bin with the largest average angle to be arterial blood flow. In 818, the MRI imaging and analysis system assigns the bin with the smallest average angle to be venous blood flow. 【0207】 Method 800 terminates at 820. 【0208】 Additionally or alternatively, MRI imaging and analysis systems can utilize probability. Based on several other datasets where blood pools have already been identified, the MRI imaging and analysis system can generate an atlas and compare the new dataset to the atlas. By co-registering the atlas and the new dataset together, the MRI imaging and analysis system can assign the bins to the correct blood pools. 【0209】 Additionally or alternatively, MRI image processing and analysis systems may utilize velocity signals. In particular, venous blood velocity waveforms differ from arterial blood velocity waveforms (e.g., slower flow, different waveform shapes). 【0210】 Directed coherence-based blood flow filter A filter may be used to isolate blood flow voxels by measuring directed coherence. 【0211】 A filter is a 4D feature extraction operator that determines how similar the 3D flow vector in a particular voxel is to the flow vectors of its neighboring voxels. A filter can determine how similar the 3D flow vector in a particular voxel at a given time is to the 3D flow vectors of the same voxel at time points before and after that given time. This can help filter out random noise, resulting in vascular isolation. 【0212】 The filter can be generated with one volume per time point, allowing movement through the vessels. Alternatively, the filter can be generated with one volume averaged over all time points, which typically achieves better noise reduction on its own. The averaging period may be changed from the length of the cine (i.e., the duration of the resulting acquisition) to the cardiac or respiratory phase. 【0213】 The filter can be further supplemented by another filter, such as a Boolean function or a weighted function (anatomical structure_magnitude * velocity_magnitude), to remove random noise. 【0214】 Since both filters rely on only a single input (e.g., a percentage threshold), the filters can be combined into a single 2D input gesture. For example, moving from left to right increases the directed coherence percentage threshold of the second filter, and moving from bottom to top increases the percentage threshold of the second filter. 【0215】 Since the quality / magnitude of the velocity volume varies depending on the anatomical structure, location, scanner, etc., different default values ​​for these filters may be used for each metadata combination. 【0216】 The directed coherence for a particular voxel is calculated by summing the weighted directed coherence scores between the voxel and all applicable neighboring voxels, and dividing by the sum of all applied weights. For example, if a voxel is located at the left edge of a volume, the applicable neighboring voxels would consist of all voxels within a given radius for all dimensions except the left side, and the same central voxel in the adjacent time point volume. 【0217】 The directed coherence score between two voxels is calculated by taking the dot product of the normalized velocity vectors, passing them through the ACOS function to find the angled difference, and then scaling that angle between 0 and π to obtain a result between 0 and 1. This result is then multiplied by a weight (representing the distance between the voxels). If one of the voxels has velocity at magnitude 0, a default score of 0.5 is multiplied by the weight. 【0218】 The weights applied to the directed coherence score are calculated by finding the minimum interval for all three dimensions and dividing that number by the distance between voxels (defined by the interval). For example, the "weight" for voxels on adjacent slices, columns, and rows is min_interval / sqrt(column_interval). 2 +Row_Spacing 2 +Slice_Spacing 2 ) 【0219】 After all directed coherence scores have been calculated, the minimum and maximum values ​​are determined and used to scale the scores within a given range for more effective thresholding. 【0220】 Quantification Result Verification There are physical principles that provide insight into whether flow data is correct. For example, the flow entering an enclosed volume must be equal to the flow leaving the enclosed volume. A discrepancy between the flow values ​​entering and leaving an enclosed volume may indicate a shunt or other anatomical problem. MRI imaging and analysis systems can utilize this general principle in various specific ways. It can be determined, for example, as the dot product of the normal vector of the plane slicing the volume and the velocity vector at each point or voxel. With respect to net flow, the dot product can be integrated over time. Flow can be expressed, for example, as pressure per unit time (e.g., millibars per second) or volume per unit time (e.g., liters per second). 【0221】 i) Flow entering a closed vessel or a portion thereof must match flow exiting the closed vessel. An MRI imaging and analysis system can, autonomously or by use by a human operator, identify vessels or lumens or other demarcated volumes or cavities of various organs, and then identify inlets and outlets of vessels or lumens or other demarcated volumes. Inlets and / or outlets may be anatomically significant beginnings or endings of lumens or physical holes or openings. Alternatively, inlets and / or outlets may be, for example, natural beginnings, endings, or defined, logical constructs spaced inward from openings or holes. An MRI imaging and analysis system can autonomously determine whether flow values ​​at an inlet match flow values ​​at an outlet, for example, within a defined threshold (e.g., 1%). An MRI imaging and analysis system may provide indicators of the result of the determination. For example, an MRI imaging and analysis system may provide indicators for flow mismatch, indicators for the amount or percentage of mismatch, and / or indicators for flow match. 【0222】 ii) The flow through the ascending thoracic aorta should match the combined flow through the superior vena cava (SVC) and the descending thoracic aorta. The MRI imaging and analysis system can identify the ascending thoracic aorta, SVC, and descending thoracic aorta autonomously or by being used by a human operator. The MRI imaging and analysis system can autonomously determine whether the flow value through the ascending thoracic aorta matches the combined flow through the SVC and the descending thoracic aorta within a defined threshold (e.g., 1%). The MRI imaging and analysis system can provide a label for the result of the determination. For example, the MRI imaging and analysis system can provide a label indicating that the flows do not match, a label indicating the amount or percentage of mismatch, and / or a label indicating that the flows match. 【0223】 iii) The combined flow through the splenic ventricle (SVC) and inferior vena cava (IVC) should match the flow through the pulmonary vascular structure (PA). An MRI imaging and analysis system can identify the SVC, IVC, and PA autonomously or by being used by a human operator. The MRI imaging and analysis system can autonomously determine whether the combined flow values ​​through the SVC and IVC match the flow through the PA within a defined threshold (e.g., 1%). The MRI imaging and analysis system can provide indicators of the result of the determination. For example, the MRI imaging and analysis system can provide indicators indicating that the flows do not match, indicators indicating the amount or percentage of mismatch, and / or indicators indicating that the flows match. 【0224】 iv) The flow through the PA should match the combined flow through the right pulmonary vascular structure (RPA) and the left pulmonary vascular structure (LPA). The MRI imaging and analysis system can identify the PA, RPA, and LPA autonomously or by being used by a human operator. The MRI imaging and analysis system can autonomously determine whether the flow value through the PA matches the combined flow value through the RPA and LPA within a defined threshold (e.g., 1%). The MRI imaging and analysis system can provide indicators of the result of the determination. For example, the MRI imaging and analysis system can provide indicators that the flows do not match, indicators of the amount or percentage of mismatch, and / or indicators that the flows match. 【0225】 v) The flow through the LPA should match the sum of the flows through all left pulmonary veins. The MRI imaging and analysis system can identify the LPA and the left pulmonary veins autonomously or by being used by a human operator. The MRI imaging and analysis system can autonomously determine whether the flow value through the LPA matches the flow through all left pulmonary veins within a defined threshold (e.g., 1%). The MRI imaging and analysis system can provide indicators of the result of the determination. For example, the MRI imaging and analysis system can provide indicators that the flows do not match, indicators of the amount or percentage of mismatch, and / or indicators that the flows match. 【0226】 vi) The flow through the RPA should match the sum of the flows through all right pulmonary veins. The MRI imaging and analysis system can identify the RPA and the right pulmonary veins autonomously or by being used by a human operator. The MRI imaging and analysis system can autonomously determine whether the flow value through the RPA matches the flow through all right pulmonary veins within a defined threshold (e.g., 1%). The MRI imaging and analysis system can provide indicators of the result of the determination. For example, the MRI imaging and analysis system can provide indicators that the flows do not match, indicators of the amount or percentage of mismatch, and / or indicators that the flows match. 【0227】 vii) Flow through a closed volume should be the same regardless of the orientation of the planes traversing the closed volume. An MRI imaging and analysis system can, autonomously or by use by a human operator, identify two or more planes that intersect a given point, each traversing the blood vessel in its respective orientation (i.e., two or more planes, each having different normal vectors that intersect the same point). The MRI imaging and analysis system can autonomously determine whether the flow over the entire cardiac cycle matches for all planes within at least some defined threshold. The MRI imaging and analysis system can provide indicators of the result of the determination. For example, the MRI imaging and analysis system can provide indicators that the flows do not match, indicators of the amount or percentage of mismatch, and / or indicators that the flows match. 【0228】 viii) The flow leaving the cardiac chambers should correspond to the changes in systolic and diastolic volume for each cardiac chamber. An MRI imaging and analysis system can identify one or more chambers of the heart, either autonomously or by use by a human operator. The MRI imaging and analysis system can make autonomous decisions. The MRI imaging and analysis system can provide indicators of the results of the decisions. For example, the MRI imaging and analysis system can provide indicators for flow mismatch, indicators for the amount or percentage of mismatch, and / or indicators for flow match. 【0229】 MRI image processing and analysis systems can perform any one or more of the above actions as result validation, either manually or autonomously. This can provide a confidence measure indicating the accuracy of the flow value. For example, instead of simply providing (e.g., displaying) a flow value (e.g., 5.4 L / min), result validation allows the MRI image processing and analysis system to provide a result with an indicator of accuracy (e.g., 5.4 L / min ± 0.3 L / min with a 95% confidence interval). Error estimation can be determined by evaluating the difference between the left and right sides of the relationships listed above. In addition, other error estimations can be derived by perturbing the measurements several times (e.g., in position, orientation, contour expansion / erosion), thereby generating several measurements (e.g., mean, 95% confidence interval, etc.) that can be evaluated statistically. Also, knowing the possible level of error in the data or results can assist other preprocessing algorithms (i.e., eddy current correction). For example, this can be used by the implemented algorithm to know in which direction the data should be distorted to make the data or results more accurate. 【0230】 Segmentation Flow-driven segmentation The various implementations described herein advantageously utilize blood flow datasets (i.e., x-velocity, y-velocity, and z-velocity) as input for any type of quantification. Traditionally, only anatomical (i.e., magnitude) datasets have been used for magnetic resonance (MR) quantification. As an example, blood flow datasets can serve as a substitute for blood flow quantification, but can also be used as input for determining the luminal boundaries of blood vessels or lumens (i.e., segmenting the anatomical dataset). Blood flow datasets can also be used to calculate pressure, wall shear stress, and to determine how to render 2D and 3D images. As an example of how flow can modify image rendering, MRI image processing and analysis systems can render only anatomical voxels that match specific flow conditions (i.e., when the velocity magnitude within that voxel exceeds a specific defined threshold, e.g., a user-defined threshold). 【0231】 An example is flow-driven segmentation (e.g., volume, region, pathway) to isolate blood from other tissue types. A specific example is volumetric segmentation of cardiac cavities. At some point in time during the scan, the flow in the cardiac cavities may have a very high magnitude. MRI imaging and analysis systems can use these points in time to derive a geometric model of the cavities. Once the geometric model is derived from the high-flow-magnitude points in time, the MRI imaging and analysis systems can align the model with the magnitude scan. Subsequently, the MRI imaging and analysis systems can align the model with other points in time to obtain a nonlinear transformation that aligns the model with the other points in time. 【0232】 In an exemplary algorithm for segmenting blood vessels and determining the luminal boundaries of the provided vascular cross-sections, given a seed point in the vessel, the time point with the maximum flow magnitude at the seed point is identified. This time point indicates when the strongest jet of blood is moving through the vessel. The system uses a combination of anatomical structure and velocity data to determine a cross-section perpendicular to the direction of flow. The direction of flow provides a strong indicator of the orthogonal cross-section to be used. In some cases, the flow may have turbulence, which will lead to unsatisfactory results. In addition, the system creates a coarse hemisphere of vectors at the seed point, where the vector closest to the direction of flow has the greatest weight. For each of the vectors constituting the coarse hemisphere, several rays are projected perpendicular to the vectors on the hemisphere. These rays are terminated when the magnitude of the change in both anatomical structure pixel intensity and velocity magnitude has changed sufficiently. This provides an approximation of the vessel boundary on that plane. The seed point, the termination of ray n, and the termination of ray (n+1) form a triangle. The system calculates the sum of the regions of all resulting triangles. From this calculation, the vector on the hemisphere with the smallest result region is selected as the normal to the plane that is most perpendicular to the direction of flow. 【0233】 Next, a multi-section reconstruction (MPR) is resampled, which combines (i.e. multiplies) both anatomical structure and velocity data with normal vectors from the previous step at the initial seed point and the point of strongest flow. A technique called dynamic contour modeling or snakes may be used to find the contour that depicts the luminal boundary of the blood vessel. The seed contour is created based on the endpoint of the ray projected to obtain the region measurements. The dynamic contour then attempts to find a smooth contour that depicts the blood vessel wall. This smooth contour is obtained by applying an energy minimization function using gradient descent. The forces associated with the energy function include 1) internal forces of first and second derivatives using finite differences of points on the contour, 2) external forces (balloon forces) that are always pushing outward from the center of mass, and 3) external forces that pull the point toward a region with a large gradient magnitude. 【0234】 Internal forces ensure the contour is smooth. Outward pushing forces are used to extend the contour through the MPR. Pulling forces toward the gradient magnitude should overcome the balloon and pull the vascular wall contour. Since the algorithm determining the vascular lumen boundary is heavily influenced by the gradient in the MPR, the MPR is first smoothed, for example, using Gaussian convolution. Once the contour for the strongest flow time point is determined, the remaining time points are evaluated in the same way. Because the dynamic contour model is very sensitive, if a contour diverges significantly in region or curvature, it is treated as abandoned. Any abandoned contour is replaced with a contour that is a linear blend of contours from adjacent time points. 【0235】 Overall process Asynchronous command and imaging pipeline. Commands are sent from the client to the server using a persistent connection via the WebSockets protocol. Images are returned via the same WebSockets connection through the standard. Since user events can occur much more frequently than the application can generate images and return them to the client, some events must be suppressed. Using an asynchronous command and image pipeline allows all critical user events to be sent to or processed by the server. All critical events are captured in the persistence layer, determining the time to allow them, deciding which events to suppress if necessary, and creating the best possible user experience, if any. Images are simply processed and returned to the client at a speed the network can handle. 【0236】 The architectural design for a remotely located rendering system (i.e., an image processing and analysis system) involves commands sent from a client to a web server, which interprets these commands, and one or more computing servers generate images or computation results that are sent back to the client, for example, via the web server. If the work involves tasks where messages can be distributed, multiple computing nodes may be used; otherwise, messages are routed to the appropriate computing node. The problem with a simple design where every user event is an image or computation is that user events can occur much more frequently than the rendering system generates results and the servers return them to the client over the network. Also, simply suppressing events can result in the loss of important user interaction. 【0237】 The network can utilize asynchronous command and imaging pipelines, which use two communication channels: one for the client to send commands to the server, and the other for the server to send responses to the client. This approach allows all events to be captured in a persistence layer, intelligently determining which events to suppress and which to serve, creating the best possible user experience. The results are simply processed and returned to the client at a speed at which image processing and analysis calculations can be performed and transmitted. 【0238】 Examples of the operation of asynchronous commands and the imaging pipeline 900 are shown in Figures 9A and 9B. As shown in Figure 9A, client 902 initiates message 904. The middle layer 906 receives message 904 and determines which computing server will handle the message. At 908, the middle layer 906 checks and determines whether computing server 910 (for example, the server in the image processing and analysis system 104 in Figure 1) is busy executing an equivalent command. If computing server 910 is not busy, message 904 is immediately forwarded to computing server 910, as indicated by arrow 912. Otherwise, as indicated by arrow 916, message 910 is placed in command slot 914 to be executed when a completion event is received from computing server 910. If another message 904 arrives before computing server 910 is ready, the message in command slot 914 is replaced with the newer message. Alternatively, the newer message may be added to the queue. 【0239】 As shown in Figure 9B, when the computing server 910 completes the execution of command 918, the computing server 910 generates a completion event 920. The middle layer 906 receives the completion event 920 and forwards a response 922 to the client 902, as indicated by arrow 924. At 926, the middle layer 906 determines whether there is a message in command slot 914. If there is a message in command slot 914, the completion event 920 triggers the middle layer 906 to send message 904' to the computing server 910, as indicated by arrow 928, and command slot 914 is cleared. 【0240】 Another example of a proper imaging pipeline is a workflow that requires the simultaneous rendering of multiple property-linked images. In some cases, the current workflow action might require the simultaneous display of, for example, 20 images. If each of these images were retrieved by a separate request, performance would be significantly degraded due to the considerable overhead in creating and sending the requests. Instead, all images could be rendered on or as one large image, with only a single response sent to the resulting "sprite sheet." The client then displays the images by using pixel offsets. For example, if a view has four images, each 256x256, the sprite sheet could be 256x1024, with each image placed or arranged touching each other. The client then displays the four images at 256x256 by using offsets of 0, 256, 512, and 768. 【0241】 In addition, any lines, markers, or planes within the image are drawn on the client as overlays, and information on how to render the overlays comes from the server via text-based messages. This provides higher quality rendering of the overlay data than if the overlays were rendered on the server, encoded as JPEG, and then sent. 【0242】 Workflow Architecture Description: One of the main advantages of the 4D flow is that a radiologist does not need to be on-site during acquisition. The technician only needs to position the chest or region of interest and press a single button to begin acquisition of this region of interest. This significantly alters the typical cardiac MR workflow. In addition, due to offline processing, the patient can leave the scanner immediately after acquisition. If the technician is concerned about image quality because the patient may have moved during acquisition, accelerated reconstruction can be used to give the technician one cardiac phase image in less than 5 minutes. Once the technician has confirmed the quality of that image, they can be confident in letting the patient leave. Once the full 4D flow scan is complete, the raw data is sent for automated image reconstruction (this can be done online or offline at the scanner). If this is done offline, it can be done within the hospital network or the Morpheus cloud. After image reconstruction, the workflow takes in the DICOM images and imports them into the Morpheus software. At this point, the Morpheus technician or hospital technician goes through the workflow steps highlighted above (image preprocessing, segmentation, visualization, and quantification) to process the offline case. The goal is to set up the case for the clinician so that all they need to do when they sit in front of the Morpheus software is validate and annotate the report. The goal is to complete this process in under 10 minutes. This workflow differs significantly from many other software applications on the market today, primarily because very few software applications are connected to a central database (i.e., in the cloud) and have the flexibility to allow multiple users to quickly access data on any device (e.g., a home laptop, tablet, etc.). 【0243】 Anonymization service Numerous regulatory requirements exist for a class of information typically named Protected Health Information (PHI). Many clinical facilities (e.g., hospitals) or entities operating clinical facilities do not want any PHI to leave their network or control. Similarly, entities providing services to clinical facilities may want to be isolated from PHI to avoid numerous handling and security requirements and to eliminate or reduce potential liabilities. Various implementations of networked environments can implement services that prevent PHI from leaving a clinical facility's own network. 【0244】 Several methods are described herein, and the preferred method varies depending on the location of the database holding the PHI. 【0245】 In cases where the database containing the PHI, and the computer nodes that require access to the PHI, are all within the clinical facility's own network, there are two typical different implementations that allow users to access MRI imaging and analysis services provided by a remotely located MRI imaging and analysis system outside the clinical facility's own network. 【0246】 1) The first implementation form utilizes multiple firewalls. A clinical facility (e.g., hospital) deployment should consist of several security layers. A secure clinical facility network should contain all PHI information (e.g., MONGO® database, data files). MRI image processing and analysis systems (e.g., their server components) should run in a demilitarized zone (DMZ) or perimeter network configuration with appropriate access permissions to the secure clinical facility network to access data via the firewall as needed. Another firewall should isolate the DMZ from the internet, with only ports 80 and 443 exposed to the public via the internet. In this way, users can log in via a normal URL from a remote location or network (e.g., home) and access (e.g., view) the PHI from a remote location or network. 【0247】 2) The second implementation method utilizes a virtual private network (VPN). Specific URLs for accessing services provided by the MRI image processing and analysis system are only accessible if the client device is first connected via the clinical facility's VPN. 【0248】 If a clinical site does not require PHI in a database outside of its own network, and the computer nodes that need access to PHI are hosted outside of the clinical site's network (e.g., Amazon Web Services, i.e., AWS), the anonymization service can run within the clinical site's network. The anonymization service receives raw DICOM files, anonymizes the raw DICOM files by removing any plaintext PHI, and replaces the plaintext PHI with a hash of the PHI (e.g., all PHI fields in a DICOM image are identified and converted to salted hashes). The key used to convert plaintext PHI to hashes is maintained within the clinical site's own network and is accessible by the anonymization service. 【0249】 This anonymization service can operate in at least two ways. 【0250】 1) The first implementation uses a proxy server. When a request comes from a client, either within or outside the clinical facility's network, the request goes through the proxy server, where an anonymization service replaces the hash with plaintext PHI. This allows the client to always see the PHI. In this implementation, all HTTP / S requests are routed through the proxy server. 【0251】 2) The second implementation uses a one-time lookup. When a client connects to a server of an MRI image processing and analysis system located outside the clinical facility's network, a flag indicates to the client whether it lacks the information necessary to render the PHI. If the client does not have a plaintext PHI, the client requests a PHI by connecting to an anonymization service and providing a hash or identifier (ID). Based on the provided hash or ID, the correct plaintext PHI is then provided to the client. The received PHI is then stored (cached) locally by the client as long as the session is active. When the user logs out, the client purges the cache. 【0252】 In all of the implementations described above, PHI is transmitted from the client to the anonymization service and from the encrypted server. In addition, any stored data may also be in an encrypted form. 【0253】 For customers who want their PHI hosted in the cloud, but who, in some situations, want to hide their PHI on their client (for example, for a conference), the inventors have created an anonymization filter (which is part of the anonymization service). The anonymization filter is a filter placed on the client data stream, and the traffic passing through the filter Fi It inspects the data and removes the PHI (which is a streaming data filter) before the data can proceed through the communication pipeline. 【0254】 Phase aliasing detection and / or correction Phase aliasing is typically the result of a 2π offset of certain information within an MRI dataset. Phase aliasing can manifest, for example, as voxels of the wrong color; for instance, a black voxel might represent blood flow in a first direction, while an immediately adjacent voxel is white, representing, for example, a second direction opposite to the first. Phase aliasing can also result from inaccurately (e.g., excessively large) settings of the velocity encoding (VENC) parameter. Notably, while generally desirable to set the VENC parameter high to achieve better resolution, setting the VENC parameter excessively high tends to cause voxels to appear as "negative" images. 【0255】 To detect them, the MRI imaging and analysis system can detect anatomical structures, such as the bodies of blood vessels. The MRI imaging and analysis system can then analyze each voxel within the detected boundary and determine that all flows within the bounded volume are in the same direction or within a defined angular threshold in the same direction. In particular, the MRI imaging and analysis system can compare the respective vectors for each voxel, where the vectors indicate the direction of tissue movement represented by a particular voxel. The MRI imaging and analysis system can mark voxels that appear abnormal. The MRI imaging and analysis system can check neighboring voxels, such as the closest neighbor element and / or the next closest neighbor element outward within a defined range or degree. For example, the MRI imaging and analysis system can first check the closest neighbor element to a given voxel in two dimensions or more typically three dimensions. In response to identifying one or more abnormal-looking neighboring voxels, the MRI imaging and analysis system can iterate outward for the next level of voxels. This may include all the next closest neighboring voxels to the first abnormal-looking voxel, or the closest neighboring voxels to the second or subsequently identified abnormal-looking voxel. The MRI image processing and analysis system defines or forms a region of interest (ROI) from the set of abnormal-looking voxels. 【0256】 Additionally or alternatively, MRI image processing and analysis systems can utilize continuity or mass conservation. For example, an MRI image processing and analysis system can determine whether a flow into a bounded volume coincides with a flow outside the bounded volume. 【0257】 In some implementations, MRI image processing and analysis systems can attempt to correct phase aliasing by adding, for example, a correction value (e.g., 2π). Other methods include multi-VENC techniques, in which multiple passes with various VENC parameters (i.e., different frequencies) are tried to determine which works best. 【0258】 Signal Unwrapping In some cases, certain parts of an image may appear in the wrong position. For example, the bottom or base of an image may appear at the top of the image or frame. This results from an incorrect order of data or information, and because the periodic nature of the signal often leads to inaccurate rendering at both ends of the image (e.g., top and bottom). This error is sometimes called image wrapping or magnitude aliasing. 【0259】 In such cases, the MRI image processing and analysis system can correct the image by implementing correction actions, such as cropping out incorrectly positioned (e.g., anomalous) portions from the image, or further, moving the incorrectly positioned portions to the correct position within the image or image frame. In some implementations, the MRI image processing and analysis system can identify incorrectly positioned portions depending on user input. In other implementations, the MRI image processing and analysis system can autonomously identify incorrectly positioned portions. For example, the MRI image processing and analysis system can analyze image data in terms of discontinuities. For example, the MRI image processing and analysis system can compare images across time slices by time slice consistency. The MRI image processing and analysis system can, for example, identify gradients or steps in voxel intensity values ​​(magnitude). For example, large changes in magnitude (e.g., doubling) should not occur in resting tissue, and therefore, this is detected as an error. 【0260】 Other artifacts In some cases, artifacts can appear in MRI datasets. For example, patient or body part movement can cause image blurring. Also, for example, the presence of metal can cause streaks or other visual artifacts in the resulting images. Contrast agents can also cause artifacts. 【0261】 Typically, these artifacts are difficult to correct. However, MRI image processing and analysis systems can detect the appearance of artifacts and alert healthcare providers (e.g., physicians) to their presence in the dataset. Flow-driven segmentation 【0262】 The system can, for example, automatically or autonomously identify structures in 4D flow information (e.g., 1D, 2D, 3D flow structures) and identify clinical markers. For example, the system can automatically or autonomously identify clinical markers such as anatomical and / or temporal clinical markers or clinical landmarks. Clinical markers may include one or more of, for example, aneurysms, stenosis, and plaques. For example, clinical markers may include anatomical markers or landmarks such as various structures of the heart, such as the apex of the heart, the aorta, the vena cava, and valves such as the mitral valve or tricuspid valve. Clinical markers may include, for example, other known structures and / or other traces. In some examples, the system can use structures to identify various clinical markers. The system can, for example, automatically or autonomously segment or delineate various structures or clinical markers. For example, the system can segment or delineate the lungs from other body tissues, for example, at least partially based on the dynamic properties of air in the lungs being larger compared to the relatively smaller dynamic properties of other body tissues (e.g., cardiac tissue). Furthermore, for example, the system can segment or delineate blood tissue from non-blood tissue, at least partially based on the relatively dynamic and / or periodic nature of blood tissue compared to non-blood tissue. Also, for example, the system can segment or delineate fluid matter (e.g., air in the lungs, blood in the vascular structure or heart) from non-fluid tissue (e.g., vascular structures, heart, fat). 【0263】 Blood flow at specific locations within the body can often be characterized by the velocity of blood flow over time. This is particularly true for blood flow within the heart and in the aorta and vena cava of the body. Due to the rhythmic nature of the cardiac cycle, one method for characterizing blood flow patterns in a time-resolved velocity dataset, where time points are evenly distributed over a single cardiac cycle, is to examine the components of the Discrete Fourier Transform (DFT) calculated at each voxel or pixel over the available time points. The analysis can be performed by one or more processors using transformations of the individual velocity components (x, y, and z) or the resulting velocity magnitudes. 【0264】 DFT components are useful for segmenting the blood pool from stationary tissue. For this purpose, lower-order non-DC components are most useful because they are strongly represented in both the blood flow within the heart and the aorta and vena cava connected to the heart, and the pseudovelocity signals in stationary tissue caused by eddy currents and other sources tend to be largely static in time. Air in the lungs and outside the body has its typical high-velocity signals and is also strongly represented in this transformation, but these regions are easily removed using other techniques, such as simple thresholding in images of anatomical structures. 【0265】 Because different regions of blood flow (e.g., the aorta compared to the inferior vena cava and superior aorta) have different shapes or flow patterns, a general-purpose mask designed to pinpoint all blood flow in a chest scan can be obtained by combining the desired DFT components together, regardless of their relative magnitude or phase. The application of such a mask generally works well for separating the blood pool from resting tissue. More sophisticated masks designed to identify specific regions of the blood pool can be created through a system by taking into account the phase and magnitude of the DFT components. The phase of the DFT components can be compared to the time of maximal systole, and a probability is assigned to each voxel based on the phase deviation from the expected value. The time of maximal systole can be determined through various techniques described elsewhere. Similarly, the system can examine the relative magnitudes of different DFT components and create probability values ​​based on the underlying characteristics of the flow. The system can combine these probability values ​​across the entire blood mask to create a highly accurate blood mask for a specific blood pool region of interest. In particular, the highly distinctive temporal properties of blood flow in the aorta and pulmonary artery can be easily distinguished and segmented from the rest of the blood pool using this technique. 【0266】 Blood probability masks created in this manner with respect to arterial flow from the ventricles have the property that, due to their strength in identifying arterial flow, the resulting histogram of probability values ​​can be used by the system to automatically identify appropriate probability cutoff values ​​without user assistance (i.e., autonomously). Once an appropriate probability cutoff is determined for the arterial mask, the system can use the resulting mask to help determine probability cutoff values ​​for other masks, such as the generic blood mask described above. Due to the accuracy provided by the arterial mask, it is also useful to provide points that can be guaranteed to be in the blood pool with a high probability. Such points are useful to the system for various purposes, such as performing flood-filling of less accurate generic blood masks and removing them by eliminating heterogeneous, unconnected elements. 【0267】 Once an arterial flow mask is created, the marked areas of ventricular, blood flow, or blood flow from the blood pool can be automatically divided into aortic and pulmonary artery flows. Since there is no direct connection between the arterial flow and the pulmonary artery flow, the system can separate the arterial mask into two main elements using flow direction and gradient, as well as / or trajectory lines, along with probability values. Once separated, the system can easily distinguish the aorta and pulmonary artery from each other by anatomical cues such as the mean direction of flow or the relative positions of the two segments in space. 【0268】 Because cardiac tissue moves rhythmically with the cardiac cycle, calculating the DFT per pixel of a time-resolved anatomical structure dataset can be used to generate a probabilistic mask about the cardiac wall in a similar manner to how it is used to identify blood pools in a time-resolved velocity dataset. Similar to this use above, non-zero low-frequency components are most useful for identifying motions that have the same overall period as the heartbeat. Creating a mask that accurately identifies the myocardium is useful for several purposes. The system can combine this mask with a blood flow mask to create a mask that covers the entire heart. This is useful, for example, in the aforementioned eddy current correction where it is important to ignore areas of non-resting tissue, or simply for the system to segment the heart for better visualization. The cardiac wall mask can also be used by the system in combination with other metrics to provide useful information for segmenting blood pools and cardiac cavities in different regions, and to assist the system in the automatic or autonomous identification of landmarks such as the cardiac wall. 【0269】 Furthermore, masks that accurately identify the myocardium may be used to provide metrics such as the location and size of the heart in a scan or image. The location of the heart in a scan or image has many applications. For example, its location may be used by blood flow segmentation routines performed by the system to determine whether blood is moving toward or away from the heart, for example, to separate the right blood flow from the left blood flow in the lungs. 【0270】 The automated methods described above eliminate the subjectivity inherent in conventional methods for identifying anatomical structures and flows, providing a high level of accuracy or reproducibility. This reproducibility enables new uses of MRI data. For example, MRI data from a single patient can be reviewed reliably as a trend across different sessions. Even more surprisingly, MRI data from multiple patients can be reviewed reliably as a trend across populations or demographic groups. 【0271】 The various embodiments described above may be combined to provide further embodiments. All U.S. patents, U.S. patent application publications, U.S. patent applications, foreign patents, foreign patent applications, and non-patent publications referenced herein and / or enumerated in the application datasheets, including but not limited to Patent Document 1 filed July 7, 2011, Patent Document 2 filed July 5, 2012, and Patent Document 3 filed January 17, 2014, are incorporated herein by reference in their entirety, insofar as they do not conflict with the specific teachings and definitions herein. The aspects of the embodiments may be modified as necessary to provide further embodiments by utilizing the systems, circuits, and concepts of the various patents, applications, and publications. 【0272】 In light of the detailed description above, these and other modifications may be made to the embodiments. In general, in the following claims, the terms should not be construed as limiting the claims to any specific embodiments disclosed in the specification and claims, but rather as including all possible embodiments that are equivalent in scope to the claims. Thus, the claims are not limited by this disclosure.

Claims

[Claim 1] A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system, A step of a proxy server serving multiple requests, wherein the requests originate from and outside the clinical facility network. In response to at least some of the aforementioned requests, the steps include generating plain text information based on hashed patient health information through the proxy server, The anonymization service, through the proxy server, replaces the hashed patient health information with the plain text information. The steps include sending the received information to the client device and Includes, and further The steps include filtering the client data stream via an anonymization filter within the anonymization service, The steps include: In order to determine whether or not patient health information is present in the client data of the client data stream, inspect the traffic of the anonymization filter installed in the client data stream; including A method characterized by the following: [Claim 2] A method of operation for use with a magnetic resonance imaging (MRI)-based medical imaging system, The steps include: connecting a client to a server of an MRI image processing and analysis system, wherein the server is located outside the clinical facility network; A step of checking a flag indicating whether the client is missing information necessary to render patient health information (PHI) in plain text, The steps include: connecting the client to an anonymization service in response to an indicator that the client is missing information necessary to render the patient health information in plain text; The client requests the anonymized service to provide the patient health information in plain text. Includes, and further The steps include filtering the client data stream via an anonymization filter within the anonymization service, The steps include: In order to determine whether or not patient health information is present in the client data of the client data stream, inspect the traffic of the anonymization filter installed in the client data stream; including A method characterized by the following: [Claim 3] The method according to 2, characterized in that the step of the client requesting the anonymization service for the patient health information in plain text includes the step of providing a hash or identifier. [Claim 4] The method according to the previous version, further comprising the step of locally caching the patient health information in plain text received by the client. [Claim 5] The method according to 4, further comprising the step of purging the cached patient health information in plain text by the client in response to the user logging out. [Claim 6] The method according to claim 1 or 2, wherein the anonymization filter is a streaming data filter. [Claim 7] The method according to claim 1 or 2, further comprising the step of removing the patient health information before allowing the client data to proceed through the communication pipeline.

Images