Automated detection of arterial input function and/or venous output function voxels in medical imaging
AssigneeThe Board of Trustees of the Leland Stanford Junior University
InventorsRoland Bammer, Matus Straka, Gregory Albers
Life of this patent
Estimating the term…
Abstract
Automated AIF/VOF identification is useful in connection with providing rapid and automatic quantitative imaging results. Examples include CBF (cerebral blood flow) and CBV (cerebral blood volume) perfusion maps. In the present work, AIF/VOF regions of images are identified by making use of a predefined anatomic vasculature template to provide numerical weighting values for voxels. A score is computed for each voxel that includes normalized signal fit parameters and the numerical weighting. AIF/VOF voxels are identified as N voxels having the highest score, where N can depend on the image resolution.
Claims10
A method for identifying arterial input function (AIF) and/or venous-output function (VOF) voxels in perfusion imaging of a brain, the method comprising:
- a) acquiring one or more CT or MR perfusion images;
- b) automatically selecting some or all of the voxels of the perfusion images for AIF/VOF processing;
- c) for each selected voxel, automatically fitting a Gaussian to its signal(time), where the parameters of the Gaussian are amplitude A, width σ, and time offset to;
- d) automatically normalizing each of the parameters separately to zero mean and unity standard deviation to provide normalized amplitude A′, normalized width σ′ and normalized time offset to′;
- e) automatically excluding all voxels having A′<0, σ′>0, and to′>0, to provide a voxel subset;
- f) automatically computing a score S=A′·(−to′)·(−σ′) for each voxel in the voxel subset;
- g) automatically identifying the AIF/VOF voxels as being N voxels in the voxel subset having the highest scores, where N is a predetermined integer that depends on spatial resolution of the one or more CT or MR perfusion images.
An apparatus for real-time processing of medical image data, wherein the apparatus comprises a processor configured to:
- a) acquire one or more CT or MR perfusion images;
- b) automatically select some or all of the voxels of the perfusion images for AIF/VOF processing;
- c) for each selected voxel, automatically fit a Gaussian to its signal(time), where the parameters of the Gaussian are amplitude A, width σ, and time offset to;
- d) automatically normalize each of the parameters separately to zero mean and unity standard deviation to provide normalized amplitude A′, normalized width σ′ and normalized time offset to′;
- e) automatically exclude all voxels having A′<0, σ′>0, and to′>0, to provide a voxel subset;
- f) automatically compute a score S=A′·(−to′)·(−σ′) for each voxel in the voxel subset;
- g) automatically identify the AIF/VOF voxels as being N voxels in the voxel subset having the highest scores, where N is a predetermined integer that depends on spatial resolution of the one or more CT or MR perfusion images.
Dependencies are inferred from the claim text.