Abstract: Methods and systems for artificial intelligence based medical image segmentation are disclosed. In a method for autonomous artificial intelligence based medical image segmentation, a medical image of a patient is received. A current segmentation context is automatically determined based on the medical image and at least one segmentation algorithm is automatically selected from a plurality of segmentation algorithms based on the current segmentation context. A target anatomical structure is segmented in the medical image using the selected at least one segmentation algorithm.

Abstract: Systems and methods are provided for automatic segmentation of a vessel. A sequence of image slices containing a vessel is acquired. Features maps are generated for each of the image slices using a trained fully convolutional neural network. A trained bi-directional recurrent neural network generates a segmented image based on the feature maps.

Abstract: A computer-implemented method for editing image processing results includes performing one or more image processing tasks on an input image using an iterative editing process. The iterative editing process is executed until receiving a user exit request. Each iteration of the iterative editing process comprises using a first machine learning model to generate a plurality of processed images. Each processed image corresponds to a distinct set of processing parameters. The iterative editing process further comprises presenting the plurality of processed images to a user on a display and receiving a user response comprising (i) an indication of acceptance of one or more of the processed images, (ii) an indication of rejection of all of the processed images, or (iii) the user exit request. Following the iterative editing process clinical tasks are performed using at least one of the processed images generated immediately prior to receiving the user exit request.

Abstract: A method and apparatus for machine learning based detection of vessel orientation tensors of a target vessel from a medical image is disclosed. For each of a plurality of voxels in a medical image, such as a computed tomography angiography (CTA), features are extracted from sampling patches oriented to each of a plurality of discrete orientations in the medical image. A classification score is calculated for each of the plurality of discrete orientations at each voxel based on the features extracted from the sampling patches oriented to each of the plurality of discrete orientations using a trained vessel orientation tensor classifier. A vessel orientation tensor at each voxel is calculated based on the classification scores of the plurality of discrete orientations at that voxel.

Abstract: A method and apparatus for extracting centerline representations of vascular structures in medical images is disclosed. A vessel orientation tensor for each of a plurality of voxels associated with the target vessel, such as a coronary artery, in a medical image, such as a coronary tomography angiography (CTA) image, using a trained vessel orientation tensor classifier. A flow field is estimated for the plurality of voxels associated with the target vessel in the medical image based on the vessel orientation tensor estimated for each of the plurality of voxels. A centerline of the target vessel is extracted based on the estimated flow field for the plurality of vessels associated with the target vessel in the medical image by detecting a path that carries maximum flow.

Abstract: A method and apparatus for learning based classification of vascular branches to distinguish falsely detected branches from true branches is disclosed. A plurality of overlapping fixed size branch segments are sampled from branches of a detected centerline tree of a target vessel extracted from a medical image of a patient. A plurality of 1D profiles are extracted along each of the overlapping fixed size branch segments. A probability score for each of the overlapping fixed size branch segments is calculated based on the plurality of 1D profiles extracted for each branch segment using a trained deep neural network classifier. The trained deep neural network classifier may be a convolutional neural network (CNN) trained to predict a probability of a branch segment being fully part of a target vessel based on multi-channel 1D input.

Abstract: A method and apparatus for machine learning based detection of vessel orientation tensors of a target vessel from a medical image is disclosed. For each of a plurality of voxels in a medical image, such as a computed tomography angiography (CTA), features are extracted from sampling patches oriented to each of a plurality of discrete orientations in the medical image. A classification score is calculated for each of the plurality of discrete orientations at each voxel based on the features extracted from the sampling patches oriented to each of the plurality of discrete orientations using a trained vessel orientation tensor classifier. A vessel orientation tensor at each voxel is calculated based on the classification scores of the plurality of discrete orientations at that voxel.

Abstract: A method and apparatus for extracting centerline representations of vascular structures in medical images is disclosed. A vessel orientation tensor for each of a plurality of voxels associated with the target vessel, such as a coronary artery, in a medical image, such as a coronary tomography angiography (CTA) image, using a trained vessel orientation tensor classifier. A flow field is estimated for the plurality of voxels associated with the target vessel in the medical image based on the vessel orientation tensor estimated for each of the plurality of voxels. A centerline of the target vessel is extracted based on the estimated flow field for the plurality of vessels associated with the target vessel in the medical image by detecting a path that carries maximum flow.

Abstract: A method and apparatus for learning based classification of vascular branches to distinguish falsely detected branches from true branches is disclosed. A plurality of overlapping fixed size branch segments are sampled from branches of a detected centerline tree of a target vessel extracted from a medical image of a patient. A plurality of 1D profiles are extracted along each of the overlapping fixed size branch segments. A probability score for each of the overlapping fixed size branch segments is calculated based on the plurality of 1D profiles extracted for each branch segment using a trained deep neural network classifier. The trained deep neural network classifier may be a convolutional neural network (CNN) trained to predict a probability of a branch segment being fully part of a target vessel based on multi-channel 1D input.

Abstract: Phase unwrapping is provided for phase contrast magnetic resonance (MR) imaging. The velocity values are unaliased. For a given location over time, a path over time through a directed graph of possible velocities at each time is determined by minimization of derivatives over time. The possible velocities are based on the input velocity, the input velocity wrapped in a positive direction, and the input velocity wrapped in a negative direction, so the selection to create the minimum cost path represents unaliasing of any aliased velocities.

Abstract: Disclosed herein is a framework for facilitating waveform parameter estimation. In accordance with one aspect, time-based waveforms are generated based on analysis planes positioned along a centerline of the vessel. A surface may be fitted to upslope regions of the waveforms to determine one or more waveform parameters based on intersection of the surface with the upslope regions.

Abstract: Disclosed herein is a framework for facilitating waveform parameter estimation. In accordance with one aspect, time-based waveforms are generated based on analysis planes positioned along a centerline of the vessel. A surface may be fitted to upslope regions of the waveforms to determine one or more waveform parameters based on intersection of the surface with the upslope regions.

Abstract: A boundary in a medical image is segmented. To increase reproducibility, a multi-level segmentation approach is used. A boundary is detected based on a seed point. The boundary is used to construct a banded graph. Local segmentation is performed using the banded graph. Based on the local segmentation, a new seed point is found. The local segmentation identifies a consistent location for the seed point. The boundary detection is performed again using the new seed point.

Abstract: A method of deriving blood flow parameters from a moving three-dimensional (3D) model of a blood vessel includes determining a reference vascular cross-sectional plane through a location of a lumen in a moving 3D model of the blood vessel at one time within the model, determining a plurality of target vascular cross-sectional planes at multiple times via temporal tracking of the reference plane based on a displacement field, determining a plurality of contours based on an intersection of the target vascular cross-sectional planes with the moving 3D vessel model at multiple times within the model, and determining a blood flow parameter of the vessel from intersections of each contour of a given one of the times with a phase contrast magnetic resonance (PC-MRI) image of the blood vessel from the corresponding time.

Abstract: A boundary in a medical image is segmented. To increase reproducibility, a multi-level segmentation approach is used. A boundary is detected based on a seed point. The boundary is used to construct a banded graph. Local segmentation is performed using the banded graph. Based on the local segmentation, a new seed point is found. The local segmentation identifies a consistent location for the seed point. The boundary detection is performed again using the new seed point.

Abstract: Methods and systems for modeling cerebral aneurysm and their incoming and outgoing vessels from 3D image data are disclosed. Aneurysms and vessels are segmented from their background using a graph-cuts method. Begin and end of vessels are determined. Construction of a centerline of the incoming and outgoing vessels using a measure of vesselness in calculating a minimum cost path in a graph with nodes being representation of pixels is also disclosed. Vessel surface models are constructed from sub-voxel cross-sectional segmentation. The interpolation of vessels inside an aneurysm based on smooth continuity is disclosed. Selection of endo-vascular stents based on interpolation results is also provided.

Abstract: A method for extracting a centerline of a tubular structure in a digital medical image includes providing a 3-dimensional (3D) digitized medical image having a segmented tubular structure, finding a path in the image between a starting point and every other point in the tubular structure that minimizes an accumulative cost function, wherein the minimum accumulative cost ?(x) at a point x is a minimum of (?(x?)+Px,x?) over all nearest neighbors x? wherein Px,x? is a cost of propagation obtained from the inverse of a medialness measure computed in a plane orthogonal to a line between x and x? that is centered at a mid-point of the line, the medialness measure m(x) computed in a circular region C(x, R) centered at point x on the line, with radius R, given by m ? ( x ) = max R ? { 1 N ? ? i = 0 N - 1 ? f ( x , R ? u ? ? ( 2 ? ? ? ? i / N ) ) } , ?wherein {right arrow over (u)}(?)=sin(?){right arrow over (u)}1+cos(?){right arrow over (u)}2 and {right arrow over (u

Abstract: A method of coronary vessel segmentation and visualization includes providing a digitized coronary image, placing a plurality of seed points along an estimated centerline of a coronary vessel, selecting a seed point and constructing a cyclic graph around the seed point in a plane perpendicular to the centerline at the seed point, performing a multi-scale-mean shift filtering in the perpendicular plane to estimate image gradient values, detecting a vessel boundary using a minimum-mean-cycle optimization that minimizes a ratio of a cost of a cycle to a length of a cycle, constructing a sub-voxel accurate vessel boundary about a point on the centerline, and refining the location of the centerline point from the sub-voxel accurate boundary, where the steps of constructing a sub-voxel accurate vessel boundary and refining the centerline point location are repeated until convergence.

Abstract: A method for extracting a local center-axis representation of a vessel, includes: placing first and second seed points in an image that includes the vessel, wherein the first and second seed points are placed near a beginning and an end of a centerline of the vessel; representing the image as a discrete graph having nodes and edges, wherein the first seed point is a source node and the second seed point is a goal node; and finding a minimum-cost path between the first and second seed points by computing a cost of edges between the first and second seed points, wherein the cost of each edge is reciprocal to a vesselness measure of the edge.

Abstract: A method of deriving blood flow parameters from a moving three-dimensional (3D) model of a blood vessel includes determining a reference vascular cross-sectional plane through a location of a lumen in a moving 3D model of the blood vessel at one time within the model, determining a plurality of target vascular cross-sectional planes at multiple times via temporal tracking of the reference plane based on a displacement field, determining a plurality of contours based on an intersection of the target vascular cross-sectional planes with the moving 3D vessel model at multiple times within the model, and determining a blood flow parameter of the vessel from intersections of each contour of a given one of the times with a phase contrast magnetic resonance (PC-MRI) image of the blood vessel from the corresponding time.