Patent Application
20220386999
The following generally relates to ultrasound and more particularly to super resolution ultrasound imaging.
The literature indicates that the diffraction limit of conventional ultrasound imaging is about half a wavelength. Clinical ultrasound imaging applications have used wavelengths between 200 microns (200 μm) and one millimeter (1 mm), which precludes imaging of smaller structures with diameters less than 100 μm such as the microvasculature. Super resolution ultrasound imaging produces an image with a resolution that is beyond the diffraction limit of conventional ultrasound imaging, providing for visualization of microvascular vessels with diameters down to 8 μm.
Unfortunately, super resolution ultrasound imaging is an invasive procedure that includes continuously administering a microbubble based contrast agent intravenously during the examination. In general, with super resolution ultrasound imaging, centers of the individual microbubbles are tracked over time in ultrasound images. Since individual microbubbles are tracked, the density of the microbubbles in the administered contrast agent is sparse as it might not be possible to distinguish individual microbubbles in dense populations. However, with a sparse density it can take several minutes (e.g., 7 minutes) for the microbubbles to disperse through the full circulation.
During the examination, the subject needs to remain still with a precision of around 50 μm for several minutes. Unfortunately, it is difficult for a person to remain that still for several minutes. In addition to voluntary movement, the acquisition is also subject to effects from involuntary movement such as breathing, the heart beating, etc. Unlike a conventional scan that can be completed in a single breath hold, it is not reasonable to ask the subject to hold their breath for a several minute examination to mitigate effects from breathing. To compensate for both voluntary and involuntary movement, the processing includes motion compensation. The long acquisition time also precludes real time display.
The microbubbles in a microbubble based contrast agent tend to be fragile and burst if exposed to certain levels of acoustic pressure. To prevent them from bursting, acquisition sequences have been limited to a mechanical index (MI) (which is used as the index of cavitation bio-effects) of 0.05 to 0.20. The United States Food and Drug Administration (US FDA) limits diagnostic ultrasound imaging to an MI of 1.9. Unfortunately, the lower MI used with super resolution ultrasound imaging reduces the transmitted energy and thus the signal-to-noise ratio (SNR) and penetration depth.
In view of at least the foregoing, there is an unresolved need for an improved approach to super resolution ultrasound imaging.
Aspects of the application address the above matters, and others.
In one aspect, an apparatus includes a processing pipeline. The processing pipeline includes a stationary structure motion corrector configured to motion correct stationary structure in a series of ultrasound images for subject motion, wherein the series of ultrasound images includes the stationary structure and flowing structure. The processing pipeline further includes a stationary structure remover configured to remove the stationary structure from the motion corrected series of ultrasound images thereby producing flow images of the flowing structure. The processing pipeline further includes a flowing structure detector configured to detect peaks of flow of the flowing structure in the flow images over time to generate images of detected flow peaks. The images of detected flow peaks are accumulated over time to generate a high resolution ultrasound image.
In another aspect, a method includes motion correcting stationary structure in a series of ultrasound images for subject motion, wherein the series of ultrasound images includes the stationary structure and flowing structure. The method further includes removing the stationary structure from the motion corrected series of ultrasound images thereby producing flow images of the flowing structure. The method further includes detecting peaks of flow of the flowing structure in the flow images over time to generate images of detected flow peaks. The method further includes accumulating the images of detected flow peaks over time to generate a high resolution ultrasound image.
In yet another aspect, a computer-readable storage medium storing instructions that when executed by a computer cause the computer to: motion correct stationary structure in a series of ultrasound images for subject motion, wherein the series of ultrasound images includes the stationary structure and flowing structure, remove the stationary structure from the motion corrected series of ultrasound images thereby producing flow images of the flowing structure, detect peaks of flow of the flowing structure in the flow images over time to generate images of detected flow peaks, and accumulate the images of detected flow peaks over time to generate a high resolution ultrasound image.
Those skilled in the art will recognize still other aspects of the present application upon reading and understanding the attached description.
The application is illustrated by way of example and not limited by the figures of the accompanying drawings, in which like references indicate similar elements and in which:
The following describes a non-invasive super resolution ultrasound imaging approach that mitigates one or more of the above-noted shortcoming of microbubble based super resolution ultrasound imaging. In general, this approach includes a processing pipeline configured to process ultrasound images to create images of only/mainly flowing structure such as erythrocytes and display an accumulation of peaks of flowing structure detected over time in the images to produce a super resolution ultrasound image.
The probe 104 includes a transducer array 114 with one or more transducer elements 116. The transducer array 114 includes a one dimensional (1-D), matrix or row-column array, a linear, curved or otherwise shaped, a fully populated or sparse, etc. array. The transducer elements 116 are configured to convert excitation electrical pulses into an ultrasound pressure field and to convert received ultrasound pressure fields (echoes) into an electrical (e.g., a radio frequency (RF)) signal. The echoes are generated in response to the transmitted pressure field interacting with matter, e.g., erythrocytes, tissue, etc.
The console 106 includes transmit circuitry (TX) 118 configured to generate the excitation electrical pulses that excite the transducer elements 116 and receive circuitry (RX) 120 configured to receive the RF signals produced by the transducer elements 116. In one embodiment, the RX 120 (or other circuitry) is configured to also condition or preprocess the RF signal, e.g., amplify, digitize, etc. In the illustrated embodiment, a switch (SW) 122 is configured to switch between the TX 118 and RX 120 for transmit and receive operations. In an alternative embodiment, separated switches are employed.
In one embodiment, the TX 118 and RX 120 are controlled to concurrently acquire all image lines in each emission. For example, a subset (i.e. one or a subgroup) of the elements can be excited to simultaneously produce pressure fields that together emit a focused beam, and all of the elements can be used to receive echoes. This can be repeated for multiple different subsets, where each emission/reception provides data to generate a lower resolution image, and a higher resolution image can be generated by combining lower resolution images. An example of a suitable sequence is described in Jensen et al., “Synthetic aperture ultrasound imaging,” Ultrasonics, vol. 44, pp. e5-e15, 2006. Another example using plane waves is mentioned in Tanter et al., “Ultrafast imaging in biomedical ultrasound, IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 2014, 61, 1, pp. 102-119.
In another embodiment, the TX 118 and RX 120 are controlled to sequentially acquire one image line at a time. For instance, a subset (i.e. one or a subgroup) of the elements can be excited to simultaneously produce pressure fields that together to emit a focused beam, and the subset can then receive a single line of echoes in response thereto. This is repeated for multiple different subsets to sequentially acquire multiple image lines which together form an image. It is to be understood that the above three sequences are non-limiting and other acquisition sequences, including known sequences for 2-D, 3-D and/or 4-D imaging, are also contemplated herein.
The console 106 further includes a processing pipeline 124. The processing pipeline 124 can include one or more processors (e.g., a central processing unit (CPU), graphics processing unit (GPU), a microprocessor, etc.) configured to execute computer readable instructions encoded or embedded on computer readable storage medium such as memory 126 to perform the acts describe herein. In general, the processing pipeline 124 is configured to process the RF signals from the processing pipeline 124 to create a set of images to combine to generate a super resolution ultrasound image.
As described in greater detail below, in one instance the processing pipeline 124 generates a series of high resolution images with the RF signals, motion corrects the series of high resolution images for stationary structure motion due to subject movement (voluntary and/or involuntary), removes the stationary structure from the motion corrected series of high resolution images to produce flow images of the flowing structure (e.g., erythrocytes), detects positions of flow peaks of the flowing structure in the flow images, and accumulates the images of peak detections over time.
It one instance, this may include producing an image of the microvasculature with diameters less than 100 μm, e.g., down to 2 μm. The density of erythrocytes in circulation is ≈5 million per mm3 and the approach does not track individual erythrocytes, but rather their flow as a whole with examination times in the seconds, allowing for real time display, relaxing motion compensation requirements and subject movement constraints, and allowing for an MI up to the US FDA limit for diagnostic ultrasound imaging and thus not compromising SNR or depth penetration. In addition, the approach described herein is non-invasive, unlike microbubble based approaches.
The console 106 further includes a scan converter 128 and a display 130. The scan converter 128 is configured to scan convert each image for display, e.g., by converting the images to the coordinate system of the display 130. A (2-D and/or 3-D) super resolution ultrasound image is then built up over time by adding the current processed image to the currently displayed image. The image can be further processed, e.g., smoothed with a Gaussian or other kernel to account for inaccuracies in the motion estimation, sub-sample registration, and/or sparseness of detection, and/or otherwise processed.
The console 106 further includes a user interface 132, which includes one or more input devices (e.g., a button, a touch pad, a touch screen, etc.) and one or more output devices (e.g., a display screen, a speaker, etc.). The console 106 further includes a controller 134 configured to control one or more of the transmit circuitry 118, the receive circuitry 120, the switch 122, the processing pipeline 124, the scan converter 128, the display 130, and/or the user interface 132.
The illustrated processing pipeline 124 includes a beamformer 202. The beamformer 202 is configured to beamform the RF signals and output a series of images. A non-limiting example of suitable beamforming is described in Stuart et al., “Real-time volumetric synthetic aperture software beamforming of row-column probe data,” IEEE Trans. Ultrason., Ferroelec., Freq. Contr., Apr. 8, 2021. Other beamforming approaches are contemplated herein.
The illustrated processing pipeline 124 further includes a stationary structure motion corrector 204. The stationary structure motion corrector 204 is configured to correct the images in the series for motion of stationary structure due to, e.g., subject involuntary and/or voluntary motion. In one instance, this includes rotating, translating, etc. images in the series to align the stationary structure image to image. A suitable approach, which estimates stationary structure motion from the envelope data using speckle correlation, is described in Trahey et al., “Angle independent ultrasonic detection of blood flow,” IEEE Trans. Biomed. Eng., vol. BME-34, no. 12, pp. 965-967, 1987.
With this approach, for local motion estimation, each image is divided into a plurality of partially overlapping sub-regions, with one of the images being identified as a reference image. Examples of suitable sub-region sizes include, but are not limited to, 1×1 millimeters square (mm2), 10×10 mm2, larger, smaller, non-square, etc. Examples of a suitable reference image the first B-mode image, the middle B-mode image, the last B-mode image, or other B-mode image of the series of B-mode images. In another instance, more than one reference image is utilized.
Another suitable approach, which determines both the axial and the lateral components using transverse oscillation, is described in U.S. Pat. No. 6,859,659 B1 to Jensen, filed Nov. 9, 2001, and entitled “Estimation of vector velocity,” which is incorporated herein by reference in its entirety. Another approach, which uses directional beamforming, is described in U.S. Pat. No. 6,725,076 B1 to Jensen, filed Jan. 25, 2002, and entitled “Vector velocity estimation using directional beam forming and cross-correlation,” which is incorporated herein by reference in its entirety.
The motion estimation can alternatively be performed for 3-D data as described in US patent application publication number 2016/0206285 to Christiansen et al., filed Jan. 19, 2015, and entitled “3-d flow estimation using row-column addressed transducer arrays,” which is incorporated herein by reference in its entirety. With these approaches, the motion is determined relative to a reference image across the full acquisition. All images are then co-registered to the reference frame and aligned, e.g., via interpolation, such as spline interpolation, etc., for having a sequence of images aligned to the same spatial position over time.
Each sub-region in the reference image is cross-correlated with the corresponding sub-region in the other images, and motion in the axial and lateral directions is estimated. The estimated motion for each sub-region is assigned to a center (and/or other location) of the corresponding sub-region, and the collection of motion estimates for an image provides a discrete motion field through that image. The estimated displacements vary spatially and temporally. Motion can be estimated at any point in any image (i.e. space and/or time) using interpolation such a spline, etc. on the motion field. All the images are then compensated with this motion to align their content with the reference frame prior to the stationary structure remover 206.
The illustrated processing pipeline 124 further includes a stationary structure remover 206. The stationary structure remover 206 is configured to remove the stationary structure from the motion corrected images, producing images of just or mainly the flowing structure, e.g., the erythrocytes. In general, this can be achieved by subtracting the stationary structure from the images, which leaves the flowing structure.
Suitable approaches for removing the stationary structure include singular valued decomposition (SVD), filtering, Principal component analysis (PCA), and/or other approach. An example of a suitable approach is described in Demene et al., “Spatiotemporal clutter filtering of ultrafast ultrasound data highly increases Doppler and fUltrasound sensitivity,” IEEE Trans. Med. Imag., vol. 34, no. 11, pp. 2271-2285, 2015.
With one approach, a Casarotti matrix can be formed from the data with a size of NzNx×Nt, where Nz is a number of samples in the axial direction, Nx is a number of lateral lines, and Nt is a number of time samples. The number of calculations to process is proportional to O((NzNx)Nt2). The images can be divided into overlapping patches of pixels (e.g., 180×180 pixels) where a border of pixels overlaps adjacent patches. The singular values of these patches are then calculated. The singular Values representing stationary structure can be set to zero. Singular Values representing mostly noise can also be removed.
Images of flowing structure can then be reconstructed from the remaining non-zero singular values. An example of a suitable reconstructor is described in Baranger et al., “Adaptive spatiotemporal SVD clutter filtering for ultrafast Doppler imaging using similarity of spatial singular vectors,” IEEE Trans. Med. Imag., vol. 37, no. 7, pp. 1574-1586, July 2018. In another instance, the stationary tissue structures are removed using other echo canceling approaches using a filter as described in Torp, “Clutter rejection filters in color flow imaging: A theoretical approach,” IEEE Trans. Ultrason., Ferroelec., Freq. Contr., vol. 44, pp. 417-424, 1997, and/or other similar methods as described in Yu et al., “Eigen-based clutter filter design for ultrasound color flow imaging: A review,” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 57(5), 5456258, 2010.
The illustrated processing pipeline 124 further includes a flowing structure detector 208. The flowing structure detector 208 is configured to detect flowing structure (e.g. the erythrocytes) in the flow images. In one instance, the flowing structure detector 208 detects positions of local peaks in each of the patches. There will be many erythrocytes, and the complex summation of all scatterers within the vessel will have a peak position within the vessel. Individual erythrocytes cannot and need not be detected, unlike the microbubble approach, which requires detecting individual microbubbles.
For this approach, the envelope data can be log compressed and normalized with respect to the standard deviation of all the patches in the first image. Peaks above a threshold of minus 30 decibel (−30 dB) from the maximum in the first (or other) image are then identified as the location of the flow of interest (i.e. the erythrocytes). A subsample location of the peak can be found from interpolation, such as a polynomial interpolation around the peak position. Detections outside of a predetermined window (e.g., 100×100 pixels) about a central region of the patch can be discarded with only detections inside the window being kept. This can be performed for all patches and for all frames in the sequence.
Again, the output of the processing pipeline 124 is a series of images of detected peaks of flow, which, in this example, corresponds to the flowing erythrocytes. As discussed herein, the scan converter 128 scan converts each image for display, and a super resolution ultrasound image is generated on the display by accumulating the individual images over time.
Variations are discussed next.
In one variation, the processing pipeline 124 is part of a different computing apparatus. With this embodiment, RF signals stored in memory and/or beamformed images stored in memory can be loaded and processed with the processing pipeline 124 to generates the images of peak detections as described herein. Such an apparatus can include a scan converter and display for constructing and visually displaying a super resolution ultrasound image. Alternatively, or additionally, the images of peak detections can be conveyed to the system 102, another ultrasound system, and/or another computer apparatus to construct and visually display a super resolution ultrasound image.
In another variation, the processing pipeline 124 further includes a flow tracker 302, as shown in
In another variation, the processing pipeline 124 further includes the flow tracker 302 and a velocity estimator 402, as shown in
In another variation, the system 102 includes a vessel width estimator 502, as shown in
In another variation, another component(s) of blood (e.g., leukocytes, thrombocytes, plasma, etc.) is tracked using the approach described herein.
The ordering of the following acts is for explanatory purposes and is not limiting. As such, one or more of the acts can be performed in a different order, including, but not limited to, concurrently. Furthermore, one or more of the acts may be omitted and/or one or more other acts may be added.
At 802, ultrasound data is acquired, as described herein and/or otherwise.
At 804, the ultrasound data is beamformed to produce images, as described herein and/or otherwise.
It is to be appreciated that acts 802 and 804 can be omitted. For example, in another instance previously acquired and stored images are retrieved from memory.
At 806, the stationary structure in the images is corrected for subject motion, as described herein and/or otherwise.
At 808, the stationary structure is removed from the images producing flow images, as described herein and/or otherwise.
At 810, positions of peaks of the flowing structure are detected in the flow images, as described herein and/or otherwise.
At 812, a super resolution ultrasound image is generated and displayed by summing images of the detected peaks of flow, as described herein and/or otherwise.
Optionally, flow information such as direction, volume flow, etc., and/or derived quantities such as pressure gradients, resistive index, turbulence and/or perfusion, etc. can be estimated and visualized, as described herein and/or otherwise.
Optionally, velocity information such as mean velocity, a peak velocity, etc. can be estimated and visualized, as described herein and/or otherwise.
In one embodiment, the above is implemented for 2-D ultrasound imaging. Alternatively, or additionally, the above is implemented for 3-D ultrasound imaging, e.g., to yield all the quantities in a volume using e.g., matrix probes and/or row-column arrays for volumetric imaging.
The above may be implemented by way of computer readable instructions, encoded or embedded on the memory 126 (i.e., the computer readable storage medium, which excludes transitory medium), which, when executed by a computer processor(s) cause the processor(s) to carry out acts described herein. Additionally, or alternatively, at least one of the computer readable instructions is carried by a signal, carrier wave or other transitory medium (which is not computer readable storage medium).
The application has been described with reference to various embodiments. Modifications and alterations will occur to others upon reading the application. It is intended that the invention be construed as including all such modifications and alterations, including insofar as they come within the scope of the appended claims and the equivalents thereof.
