Questions about how individual parts of claim 1 might be changed. A question may describe something different from what the claim language recites, or a version of it; no view is offered on which.
1.1Diverging wave acquisition
What ifWhat if a curved transmit wavefront from a virtual source behind the array illuminates the imaging region on each pulse?
A phased array applies transmit delays calculated from the virtual source position so that each pulse spreads across the field of view. Receive beamforming uses the diverging transmit path and element-to-pixel receive paths to form a complex image for each pulse.
1.2Spatial and temporal matrix factorization
What ifWhat if the processor applies singular value decomposition to a pixel-by-frame matrix?
Claim language“performing a CP decomposition”
The processor flattens each complex image into one column and decomposes the resulting matrix into spatial vectors, singular values, and temporal vectors. It suppresses the leading coherent tissue modes and reshapes the reconstructed matrix into an image sequence.
1.3Temporal spectral cutoff
What ifWhat if a temporal spectral test sets the boundary for removing tissue components?
Claim language“searching a curvature inflection point”
The processor calculates the fraction of energy within ±30 Hz for each temporal factor at a slow-time sampling rate of 1 kHz. In the ordered component list, the first factor with less than 80 percent of its energy in that band marks the start of the retained components.
1.4Motion compensated background subtraction
What ifWhat if the processor subtracts a motion-compensated tissue estimate directly from the complex image sequence?
Claim language“truncating a low-rank signal”
The processor estimates tissue displacement by matching envelope image patches across frames and uses the displacement field to align the complex samples. It estimates the background with a running complex average in the aligned coordinates and subtracts that estimate to produce residual IQ frames.
1.5Frequency domain doppler estimation
What ifWhat if a windowed Fourier transform estimates Doppler flow from the filtered IQ samples?
Claim language“performing a fixed delay correlation calculation”
At each pixel, the processor applies a cosine taper to 128 consecutive filtered IQ samples and computes their discrete Fourier transform. It sums squared spectral magnitudes over the selected Doppler frequency band to form a flow power image.
Section 2 · machine-generated
Variations
Further arrangements of the same subject matter, described here as of the publication date at the foot of this page.
Other field of use
Flow imaging in opaque process pipes
An ultrasonic linear array couples to an acoustic window in a pipe carrying a suspension of scattering particles. Ultrafast plane-wave acquisitions form a complex tensor with axial, lateral, and slow-time dimensions. Canonical polyadic decomposition separates the tensor into components ordered by energy, and reconstruction excludes the components before the accepted curvature breakpoint to suppress stationary pipe echoes. A slow-time filter and fixed-delay correlation produce a map of the axial flow component.
Component substitution
Tucker decomposition for stationary clutter removal
A higher-order singular value decomposition replaces canonical polyadic decomposition and produces spatial and temporal factor matrices and a core tensor. The temporal singular values form a descending curve, and a discrete curvature estimator identifies a breakpoint subject to a preset minimum drop. The reconstruction sets the core slices associated with temporal modes before the breakpoint to zero and multiplies the core by the three factor matrices. The resulting in-phase and quadrature data pass through a slow-time filter and fixed-delay correlation.
Parameter or scale variation
Long ensembles with overlapping flow updates
The receiver beamforms complex frames at 2,000 frames per second and groups 192 consecutive frames into each processing window. Consecutive windows advance by 48 frames, which sets the output interval to 24 milliseconds. Each window starts with 32 rank-one terms and adds eight terms per retry when the curvature breakpoint fails the preset acceptance test. The reconstructed and filtered signal is correlated at a lag of four frames, corresponding to 2 milliseconds.
Combination with existing systems
Flow overlay on grayscale ultrasound
A duplex ultrasound scanner uses the same array to acquire a grayscale anatomical frame immediately before each plane-wave ensemble. The beamformer maps both acquisitions to the same spatial grid, and the processor applies rank decomposition, curvature-based truncation, and in-phase and quadrature reconstruction to the plane-wave data. Fixed-delay correlation of the filtered reconstruction supplies a flow color for each pixel whose correlation magnitude exceeds a preset threshold. The display overlays those colors on the corresponding grayscale frame.
Cardiac gated ultrasound flow imaging
An electrocardiographic monitor and an ultrasound receiver timestamp cardiac trigger events and beamformed plane-wave frames against a shared clock. The processor selects a contiguous slow-time window beginning at a preset delay after each trigger and rejects windows interrupted by the next trigger. It applies rank decomposition and curvature-based truncation within each window, then filters and correlates the reconstructed complex data at a fixed lag. The display averages the correlation maps from windows acquired at the same trigger delay.
Specific version of a part
Three angle plane wave acquisition
A linear array transmits plane waves at −6, 0, and +6 degrees in a repeating sequence. A receive beamformer aligns and coherently sums the three complex images in each cycle to produce a compounded IQ frame.
Complex alternating least squares fit
The processor reads a component count of 24 and arranges the IQ samples in a depth-by-lateral-position-by-frame tensor. It fits 24 rank-1 terms by updating one factor matrix at a time while holding the other two fixed.
Curvature test with endpoint bounds
The processor sorts the normalized rank-1 terms by descending weight and searches for an interior sign change in the discrete curvature of the weight curve. It accepts the point when at least two components lie on each side and otherwise repeats the decomposition with a larger component count.
High pass and lag one correlation
A slow-time high-pass filter with a 25 Hz cutoff processes each reconstructed IQ sequence sampled at 1 kHz. The processor averages lag-one conjugate products over 64 sample pairs and maps the magnitude to Doppler flow intensity.
Section 3 · quoted from USPTO records
Other work
Passages reproduced word for word from two earlier-published U.S. patent documents, placed beside the language of claim 1. No characterization is offered. Each passage is marked A or B for the document it comes from; ¶ numbers are the paragraph numbers in the USPTO’s published full text of that document.
Unlike the conventional coded excitation techniques, by which long modulated signals are transmitted and the received signals are decoded by the designed pulse compression filter, MW imaging transmits successively multiple tilted plane waves encoded by a Hadamard matrix with the short time interval for each transmission-acquisition event.
B · US 2021/0000450 A1 · ¶ 5
1.2step two, reading an amount of decomposed rank-1 tensors and performing a CP decomposition for the ultrasonic ultrafast plane wave data;
In an example embodiment, Progressive Canonical Parallel (PCP) tensor decomposition is employed to factorize each block into a set of rank-one tensors, each of which can be written as an outer product of three vectors as in {circumflex over (χ)}j=Σr=1Rjλj,r(bj,r(1)°bj,r(2)°bj,r(3)).
A · US 9,681,147 B2 · ¶ 25
1.3step three, searching a curvature inflection point for a rank-n tensor curve obtained by the CP decomposition, and determining whether the curvature inflection point satisfies a preset constraint condition; and executing step four when yes, otherwise executing the step two;
The error is compared to a threshold at 33. If the error is higher than a threshold, another rank-one tensor is added at 34 to represent the given data block.
A · US 9,681,147 B2 · ¶ 28
1.4step four, truncating a low-rank signal before the curvature inflection point in the rank-n tensor curve, and reconstructing IQ signal data; and
The tissue clutter filter is the similar to, in which a singular value decomposition (SVD) filter was applied for spatial-temporal processing of the blood and tissue information. The cutoff singular value was 60.
B · US 2021/0000450 A1 · ¶ 200
1.5step five, filtering the IQ signal data and performing a fixed delay correlation calculation on the IQ signal data filtered, so as to obtain an ultrasonic Doppler blood flow signal.
In addition to the B-mode imaging, blood dynamics displayed as power Doppler images were obtained from the in vivo human beating heart. 70 post-compounded images were acquired to generate one power Doppler image. The tissue clutter filter is the similar to, in which a singular value decomposition (SVD) filter was applied for spatial-temporal processing of the blood and tissue information.
B · US 2021/0000450 A1 · ¶ 200
An empty cell means only that no passage was reproduced there.
Statements in these documents
Thus, the goal is to develop a compression approach for image ensembles while achieving full random access.
A · US 9,681,147 B2 · ¶ 4
The proposed tensor-based framework can access any image within an ensemble at different levels of quality (and corresponding scalable bitrates) without the need to reconstruct or access any other image from the same ensemble.
A · US 9,681,147 B2 · ¶ 4
Therefore, methods to increase the SNR without changing the amplitude of the transmitted signals need to be investigated.
B · US 2021/0000450 A1 · ¶ 5
Recently, a multiplane wave (MW) imaging method using Hadamard encoding was proposed to increase the SNR of ultrafast imaging without compromising the frame rate.
B · US 2021/0000450 A1 · ¶ 5
Cite aspatent.fyi, US 2026/0354787 A1, claim 1, first published 2026-11-19T03:31:31-05:00, https://patent.fyi/US20260354787/, SHA-256 73428bd8cb297faec1b4f45f018d3752c540b449173354995b7624dd104fa7a5