Method and apparatus for detecting flow instability
Through AR modeling and Doppler ultrasound bandwidth imaging technology, the non-invasive problem of flow instability detection in blood vessels is solved, efficient and accurate flow instability detection is achieved, and new for atherosclerotic diseases are provided. Diagnostic tools enhance understanding of plaque progression.
Patent Information
- Application Number
- CN202411133633.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2018-09-07
- Filing Date
- 2019-09-06
- Publication Date
- 2024-11-26
AI Technical Summary
The prior art is difficult to detect and identify flow instability in blood vessels non-invasively, especially in terms of flow instability caused by narrowing of blood vessels. Traditional methods such as Doppler ultrasound have the ability to track flow velocity fluctuations in multiple spatial locations. Inadequate, the time resolution is difficult to adapt to rapid changes, and the limitations of data acquisition lead to the risk of fake artifacts for identification of unstable flows.
Ultrasonic pulses are used to impart the area of interest to the acoustic waves, and the Doppler bandwidth is derived from the processed RF data through AR modeling. High frame frequency ultrasonic and Doppler ultrasonic bandwidth imaging (DUBI) methods are used, combined with color flow imaging (CFI) and ultrasound enhanced contrast (CEUS) to detect flow instability in a non-invasive manner, improve the accuracy and consistency of Doppler bandwidth estimation through AR modeling, generate Doppler bandwidth maps and flow spot maps to visualize flow. stability.
It realizes efficient and accurate detection of flow instability in blood vessels without relying on multiple cardiac cycle measurements, improves the fine spectral resolution and consistency of Doppler bandwidth estimation, reduces artifact risk, and can Tracking spatiotemporal changes in flow instability in real time provides new insights into the mechanisms of atherosclerotic plaque formation, enhancing the ability to early diagnosis of atherosclerotic diseases.
Smart Images

Figure CN119014902A_ABST
Abstract
Description
This application is a divisional application of the invention application with application date of September 6, 2019, application number 201980058594.2, and name “Method and device for detecting flow instability”. Priority Claims and Cross-References
[0001] This application claims priority to U.S. Provisional Patent Application No. 62 / 728,271 filed on September 7, 2018, the entire contents of which are incorporated herein by reference. Background Art
[0002] Flow instabilities, such as those in blood vessels due to vessel narrowing, have been shown to pose a risk for future cardiovascular and cerebrovascular events in humans and animals. However, non-invasively detecting and identifying flow instabilities in blood vessels remains challenging. BRIEF DESCRIPTION OF THE DRAWINGS
[0003] When with Figure 1 When read together, various aspects of the present disclosure can be best understood from the following detailed description. Note that, in accordance with standard practice in the industry, various features are not drawn to scale. In fact, the size of various features may be arbitrarily increased or reduced for clarity.
[0004] Figure 1 is a flow chart of a method for detecting flow instability according to some embodiments;
[0005] Figure 2 is a schematic diagram of an apparatus for detecting flow instability according to some embodiments;
[0006] FIG3 is an illustration of a method of detecting flow instability according to some embodiments;
[0007] Figure 4A is a side view of a nozzle according to some embodiments;
[0008] Figure 4B is an image of a nozzle-based unsteady flow model according to some embodiments;
[0009] Figures 5A-5C is an image of detected flow instability according to some embodiments using a contrast enhanced ultrasound (CEUS) method;
[0010] Figures 5D-5F is an image of detected flow instabilities according to some embodiments using the CFI (Doppler variance mapping) method;
[0011] Figures 6A-6C It is an image of unstable flow measured by contrast-enhanced ultrasound (CEUS) method;
[0012] Figures 7A-7D is a graph of sensitivity and specificity analysis of methods for detecting flow instability according to some embodiments;
[0013] Figures 8A-8O is an image of a result of a method of detecting flow instabilities according to some embodiments;
[0014] Figures 9A-9B is an image of the Doppler bandwidth of a stenosis model according to some embodiments. DETAILED DESCRIPTION
[0015] The following disclosure provides many different embodiments or examples for implementing the different features of the provided subject matter. Specific examples of components and arrangements are described below to simplify the present disclosure. Of course, these are merely examples and are not intended to be limiting. For example, in the following description, forming a first feature on or above a second feature may include an embodiment in which the first feature and the second feature are directly contacted to form, and may also include an embodiment in which an additional feature is formed between the first feature and the second feature so that the first feature and the second feature may not be in direct contact. In addition, the present disclosure may repeat reference numerals and / or letters in various examples. This repetition is for the purpose of simplicity and clarity and does not necessarily indicate the relationship between the various embodiments and / or configurations discussed. Methods for detecting flow instabilities
[0016] In some embodiments, the present disclosure is directed to a method of detecting flow instability.
[0017] In some embodiments, the present disclosure is directed to a method of detecting flow instability in a blood vessel of a human or animal. In some embodiments, the present disclosure is directed to a method of detecting flow instability in a blood vessel caused by vascular narrowing.
[0018] In some cases, flow instability is caused by narrowing of blood vessels (vascular stenosis). An example of vascular stenosis is carotid atherosclerotic stenosis, a narrowing of the carotid artery, the main blood vessel in the neck that supplies blood to the brain, neck, and face. Carotid atherosclerotic stenosis can cause a stroke or a stroke-like episode.
[0019] In other approaches, risk stratification of carotid atherosclerotic stenosis is based on measurements of luminal narrowing and peak systolic velocity. However, atherosclerotic plaques with the same degree of stenosis may differ substantially in their associated risk, prompting diagnosis beyond the immediate site of stenosis. From a hemodynamic perspective, the presence of stenosis disrupts blood flow, and stenosis leads to flow instability. This phenomenon has established the foundation for physical examination of carotid atherosclerosis via the detection of a carotid bruit using a stethoscope. Unstable blood flow (i.e., non-laminar flow, including turbulent flow) has also been associated with an increased risk of atherosclerosis and plaque progression, as well as thrombosis and embolism. Therefore, monitoring blood flow instability may also provide new clinical insights into the mechanisms of atherosclerotic plaque formation.
[0020] Flow instabilities are characterized by fluctuations in flow velocity in time and space. Medical imaging methods have been exploited to derive flow turbulence indices, which describe the variation in blood flow velocity for a particular cardiac situation between multiple consecutive cardiac cycles. For example, phase-contrast magnetic resonance imaging has demonstrated the feasibility of mapping turbulent regions from measurements of mean velocity variations within voxels. Similarly, Doppler ultrasound can be used to calculate turbulence indices by measuring the standard deviation of flow velocity over consecutive cardiac cycles. However, fluctuations in heart rate and stroke volume present a significant drawback to the turbulence index approach due to its sensitivity to inter-cardiac cycle variations.
[0021] Rather than relying on measurements of multiple cardiac cycles, unstable flow can be detected by identifying high-frequency velocity fluctuations at specific moments in the cardiac cycle. In some approaches, instantaneous blood flow velocity fluctuations are measured using an intravascular catheter, but this approach is highly invasive. In some approaches, Doppler ultrasound has been used to detect flow instabilities in a non-invasive manner in the form of Doppler spectrum broadening (as a result of flow velocity fluctuations) based on the characteristics of the Doppler spectrum bandwidth. This approach has demonstrated initial success in assessing plaque risk. However, Doppler ultrasound, which operates only on a single range gate, lacks the ability to track flow velocity fluctuations in multiple spatial locations.
[0022] A variation on the single-gated data acquisition paradigm of Doppler ultrasound is to perform Doppler-based color flow imaging (CFI), which provides a color-coded rendering of mean axial velocity estimates or velocity variance estimates. Using CFI, a panoramic version of single-gated Doppler ultrasound, unstable flow regions can be visually identified as bubble-like patches or mosaic color patterns in the CFI frame. However, there are multiple caveats when using CFI for flow instability analysis. First, although CFI is within the range of video display ( 20fps) produces real-time frame rates, but in some cases the temporal resolution of CFI struggles to adapt to the rapidly changing nature of unsteady flow. Second, CFI data acquisition involves multiple triggers on each scan line in the image view, so each CFI frame does not capture a coherent spatial snapshot of the unsteady flow. Third, because the size of each slow-time collection used for Doppler processing is limited (8-16 samples) under the constraints of real-time requirements, the CFI derivation of flow estimates at pixel locations is prone to significant fluctuations and inaccuracies. Given all these issues, there is a high risk that a bubble-like or mosaic-like coloring pattern in a CFI frame is simply a fake artifact rather than a true indication of unsteady flow, especially when CFI parameters are not adjusted correctly.
[0023] In some embodiments, a method 100 for detecting flow instabilities includes: insonating a region of interest with an ultrasonic pulse 110; acquiring radio frequency (RF) data from echo pulses of the ultrasonic pulse 120; processing the RF data 130; and deriving a Doppler bandwidth from the processed RF data by AR modeling 140.
[0024] In some embodiments, insonating the region of interest with an ultrasound pulse train 110 includes insonating the region of interest with an unfocused plane wave, a diverging spherical wave, a weakly focused wide beam wave, or a focused converging wave. In some embodiments, insonating the region of interest includes insonating with a wide viewing angle. In some embodiments, the ultrasound is a high frame rate ultrasound.
[0025] In some embodiments, acquiring RF data 120 includes acquiring RF data from each array channel of a plurality of array channels.
[0026] In some embodiments, processing the RF data 130 includes beamforming the RF data to generate a panoramic image of the region of interest. In some embodiments, performing beamforming of the echo data includes performing parallel beamforming.
[0027] In some embodiments, processing RF data 130 further comprises forming a 3D data matrix by stacking multiple panoramic images along a slow time dimension; and applying clutter filtering to the 3D data matrix. In some embodiments, processing RF data 130 further comprises applying clutter filtering to the 3D data matrix.
[0028] In some embodiments, the method 100 of detecting flow instability is a method of detecting flow instability in a blood vessel. In some embodiments, the blood vessel is an artery.
[0029] In some embodiments, the region of interest is repeatedly insonated 110 with ultrasound pulses at different phases of the cardiac cycle, radio frequency (RF) data is acquired 120 from echo pulses of the ultrasound pulses, and the RF data is processed 130 to capture temporal information at different phases of the cardiac cycle.
[0030] In some embodiments, deriving the Doppler bandwidth 140 of the processed RF data by AR modeling includes estimating the Doppler bandwidth at a given pixel location over a short time period by first deriving an AR model of the slow time set, and then estimating the spectral power of the signal model. In some embodiments, the Doppler bandwidth is determined as the full width at half maximum (FWHM) of the power spectrum. In some embodiments, the estimation of the Doppler bandwidth process is repeated for each image pixel to generate a Doppler ultrasound bandwidth imaging (DUBI) frame as a depiction of flow instabilities over an image view. In some embodiments, to track the evolution of the Doppler bandwidth over time, the entire estimation procedure is repeated at subsequent time points (spaced M slow time samples apart) to generate a time series of DUBI frames over an image view (see Figure 3B ).
[0031] In some embodiments, the Doppler bandwidth 140 of the processed RF data is derived by AR modeling, including modeling the nth sample in the slow time set with N samples according to the P-order complex AR model representation of Equation 1 below:
[0032] Among them, in formula 1, a P,k is the kth complex AR parameter of the model, and e[n] is the nth sample in the complex modeling error.
[0033] In some embodiments, since the Burg method works by iteratively minimizing the average of the forward prediction error and the backward prediction error, the Burg method is used to iteratively estimate the AR parameters {a P,k} to produce good spectrum estimation reliability.
[0034] In some embodiments, deriving the Doppler bandwidth of the processed RF data by AR modeling 140 further includes calculating the Doppler power spectrum S of the set x[n] constructed from the AR model by parametric spectrum fitting as defined by Equation 2: AR [f]:
[0035] In Equation 2, f is the unit frequency, Δt is the pulse interval, and σ P 2 is the average of the average powers of the forward prediction error and the backward prediction error.
[0036] Reference Figure 3B In some embodiments, when S AR [f] Normalized to S AR After the maximum power of [f], the Doppler bandwidth of the slow-time ensemble is determined as the full width at half maximum (FWHM) S AR [f].
[0037] Deriving the Doppler bandwidth from the processed RF data via AR modeling is desirable because, by using AR modeling, the Doppler bandwidth estimation process of DUBI becomes a task of finding the FWHM on a smoothed Doppler power spectrum. In turn, DUBI's Doppler bandwidth estimation process is less susceptible to random spectral spikes that may appear in the Doppler power spectrum originating from the classical periodogram method, thus improving the consistency of the Doppler bandwidth estimate compared to other methods. Another advantage of using the AR modeling approach to derive the Doppler power spectrum is that the spectrum can be reconstructed with a finer spectral resolution, rather than being limited by the number of samples in the slow-time signal. As a result, the resulting Doppler bandwidth estimates are more accurate because they are less susceptible to discretization noise.
[0038] Without wishing to be bound by theory, the inventors believe that flow instabilities naturally give rise to a wide range of flow velocities over the sample volume in the region of interest. Since flow velocity is related to the magnitude of the Doppler shift, a wide range of flow velocities in turn produces a larger Doppler bandwidth estimate.
[0039] Thus, in some embodiments, when the derived Doppler bandwidth is equal to or greater than a predetermined value, it is determined that flow instability exists. In some embodiments, when the derived Doppler bandwidth is less than a predetermined value, it is determined that flow instability does not exist.
[0040] In some embodiments, the samples of RF data used to derive the Doppler bandwidth are samples from a time window having a duration significantly shorter than the duration of a cardiac cycle.
[0041] Since the duration of the cardiac cycle in humans is about 0.6-1 second, in some embodiments, the duration of insonation is 100 milliseconds (ms) or less, such as 50 ms or less, 40 ms or less, 30 ms or less, 20 ms or less, 10 ms or less, 5 ms or less, 2 ms or less, or 1 ms or less. If the duration of insonation is greater than 100 ms, changes in flow velocity caused by changes in the phase of the cardiac cycle may introduce noise.
[0042] In some embodiments, the method of detecting flow instability further comprises comparing the derived Doppler bandwidth to a predetermined value. In some embodiments, comparing the derived Doppler bandwidth to the predetermined value comprises comparing the derived Doppler bandwidth at one or more pixels of the panoramic image to the predetermined value.
[0043] In some embodiments, the method 100 of detecting flow instabilities further includes displaying a map 150 of the Doppler bandwidth synchronously with displaying the flow trajectory or the flow speckle pattern of the B-mode image.
[0044] Using the Doppler bandwidth estimates derived at each pixel location at different times, the DUBI frame can be formed into a three-layer display scheme to facilitate visualization of flow instabilities in the image view over time. Figure 3C As shown, DUBI is able to simultaneously display: (a) Doppler bandwidth maps annotating flow instabilities at each pixel location; (b) flow speckle patterns revealing flow trajectories; and / or, (c) B-mode images showing anatomical structures.
[0045] Flow speckle co-visualization can be performed by color-coded speckle imaging, where flow speckle values can be derived at different pixel locations by calculating the power of the corresponding slow time set after clutter filtering. A description of flow speckle co-visualization can be found in the article "High-frame-rateultrasound color-encoded speckle imaging of complex flow dynamics" published by the inventors in Ultrasound Med Biol, the entire contents of which are incorporated herein by reference.
[0046] In some embodiments, to form a three-layer display of DUBI, a flow speckle map is first overlaid on the flow area of the B-mode image and displayed as a base layer. Then, the Doppler bandwidth estimate is mapped to a hot color with brighter colors corresponding to higher Doppler bandwidths. This color map is then overlaid using the Alpha compositing principle. For non-flow areas, transparency is set to 100% to show the anatomical structure. For flow areas, transparency is set to 70% to show the flow speckle pattern and color-coded Doppler bandwidth estimates. This rendering strategy is repeated at different time points, and the image frames are stacked together to form an image playback system. Device for detecting flow instability
[0047] In some embodiments, the present disclosure is directed to a device 200 for detecting flow instability. The device 200 for detecting flow instability includes: a transmitter 210 configured to cause an ultrasonic pulse to act on a region of interest; a receiver 220 configured to receive an echo pulse in the ultrasonic pulse; a scanner 230 configured to acquire radio frequency (RF) data from the echo pulse; and a processor 240 configured to: process the RF data; and derive a Doppler bandwidth from the processed RF data through AR modeling.
[0048] In some embodiments, transmitter 210 or receiver 220 includes an ultrasound array transducer.
[0049] In some embodiments, processor 240 is configured to process the RF data or derive the Doppler bandwidth from the processed RF data through AR modeling in a manner similar to that described in the previous section.
[0050] In some embodiments, processor 240 is further configured to use the RF data acquired by the receiver to generate a flow speckle pattern revealing flow trajectories or a B-mode image.
[0051] In some embodiments, processor 240 is further configured to generate a map of Doppler bandwidth in synchronization with a flow speckle pattern or B-mode image to be displayed by a display. In some embodiments, processor 240 is configured to generate a map of Doppler bandwidth based on synchronization with a flow speckle pattern or B-mode image in a manner similar to those described in the previous section. In some embodiments, the display is included in device 200. In some embodiments, device 200 is configured to be connected to an external display.
[0052] In some embodiments, processor 240 is further configured to compare the Doppler bandwidth to a predetermined value. In some embodiments, a Doppler bandwidth equal to or greater than the predetermined value indicates flow instability, and a Doppler bandwidth less than the predetermined value indicates stable flow. Example
[0053] Example 1: Implementation of DUBI
[0054] Example 1-1: Imaging Hardware and Data Acquisition
[0055] DUBI is implemented on a programmable research platform that builds on innovations in ultrasound flow imaging described in the inventors' published articles "High-frame-rate ultrasound color-encoded speckle imaging of complex flow dynamics" and "Vector Projectile image: time-resolved dynamic visualization of complex flow pattern". Both articles are incorporated herein by reference in their entirety. The platform includes a 128-channel programmable transmission front end (Sonix Touch; Analogic Ultrasound, Peabody, MA, USA), a pre-beamformed DAQ tool with a 40 MHz sampling rate and 12-bit resolution, and a high-speed processing platform (GTX 1080; NVidia Corporation, Santa Clara, CA, USA) based on graphics processing unit (GPU) technology for beamforming and signal processing. Our research platform is programmed to perform high frame rate data acquisition as needed. Using an L14-5 linear array (0.3048 mm pitch; simulated ultrasound), wide-view acquisition at 10 kHz was achieved with unsteered plane wave excitation (0° transmission; 5 MHz center frequency, 5-cycle pulses). Therefore, raw data frames were acquired at a rate of 10,000 fps. Again, based on the established formula, the transmission pulse shape produced an effective axial range of 0.77 mm. In each acquisition, the raw channel-domain data was stored on the DAQ tool until the internal 16 GB storage buffer was filled (5 cm imaging depth for 3 s). The data was then streamed offline to the GPU platform for processing.
[0056] Example 1-2: Plane Wave Image Formation
[0057] For each frame of the acquired dataset, parallel beamforming was performed on a GPU computing platform using a codec programmed in Matlab (R2016a; Mathworks Inc., Natick, MA, USA) with calls to the GPU-accelerated parallel beamforming library. The codec first applied a 3-7 MHz bandpass filter to the received RF data on a per-channel basis to suppress out-of-band white noise. The filter was implemented in Matlab as a finite impulse response (FIR) filter with a minimum filter order (30 taps) formulated using the Parks-McClellan iso-ripple design algorithm. The acquired data was then obtained in analytical form using an FIR-based Hilbert transformer (50th order) as described previously. With the resolved RF data, the image frames (with a pixel pitch of 0.2 mm) were finally parallel beamformed using our GPU-based delay-and-sum algorithm (64 array channels were used with Hanning apodization). This three-stage process is repeated for data at different slow-time sampling instants, thereby generating a stack of image frames in slow time for Doppler signal processing and bandwidth estimation. Note that for a 64-channel receive aperture configuration, the effective lateral width is estimated to be between 0.31 and 0.77 mm for an imaging depth range of 2-5 cm, as determined based on well-known formulas. Furthermore, based on an internal point target simulation, using its apodization curve, it is found that the maximum sidelobe size of its receive beamformer (occurring at a depth of 2 cm) is 27.6 dB lower than the main lobe.
[0058] Example 1-3: Signal Processing for Doppler Bandwidth Estimation
[0059] Doppler bandwidth estimation was performed at various slow-time instants on a per-pixel basis. First, a Doppler clutter filter was applied to suppress tissue echoes; this filter was implemented as a FIR high-pass filter with a normalized cutoff frequency of 0.05 (i.e., 250 Hz for a 10 kHz slow-time sampling rate), and its filter order was optimized to 135 taps using an equiripple filter design algorithm. For the flow region, Doppler bandwidth estimation was performed for N = 100 samples in each slow-time set, equivalent to a window of 10 ms observation period. This relatively short duration was chosen to limit spectral broadening caused by rapid acceleration (and deceleration) of blood flow. The corresponding AR-based Doppler spectrum was then derived as described. In our implementation, an 8th-order AR model was chosen because its performance was similar to higher-order models (up to 16th order) in earlier work. In addition, the power spectrum of the AR modeling was formed with a spectral resolution of 10 Hz (i.e., 0.001 normalized frequency relative to the slow-time sampling rate) to avoid spectral discretization noise.
[0060] To speed up the derivation of each DUBI frame, the Doppler bandwidth estimation of multiple pixels was performed concurrently by designing a GPU-based parallel computing kernel for AR modeling. This GPU kernel was implemented using the C++ programming language and the compute unified device architecture application programming interface (ver.7.5; NVidia Corporation). As described in the Appendix, its formulation is a public domain computing algorithm based on the Berg method. After completing the Doppler bandwidth estimation process for each DUBI frame, the process is repeated at other slow time sampling moments to generate a time series of Doppler bandwidth maps. Specifically, the repetition is performed after moving the observation window by 25 samples along slow time (i.e., M=25; for N=100, there is 75% overlap). The effective frame rate of the Doppler bandwidth map is 400fps (10,000fps raw data frame rate divided by 25). As described above, the resulting Doppler bandwidth map is finally plotted.
[0061] Example 2: Experimental Testing Method
[0062] Example 2-1: Unsteady flow model based on nozzle
[0063] To evaluate the performance of DUBI in identifying unstable flow regions, a nozzle flow device was designed to produce flow conditions ranging from laminar to turbulent. The flow condition is characterized by its Reynolds number Re, defined as Re = uD / m, where u is the average flow velocity, D is the nozzle diameter, and m is the fluid kinematic viscosity. Since it is known that the average velocity term u is equal to the flow rate Q divided by the cross-sectional area A (i.e., u = Q / A), the Reynolds number can be easily rewritten as Re = 4Q / πDm. Based on this relationship, different Re values are achieved by changing the flow rate. In turn, a series of flow conditions with increasing degrees of progression of the flow disturbance are generated by increasing the flow rate. For each of these flow conditions, stable flow regions and unstable flow regions are identified with the help of CEUS (to be discussed in Examples 2-3).
[0064] As for the nozzle design, it is formed into a curved funnel shape to gradually reduce the flow channel diameter from 10.6mm to 1.5mm over a 15mm channel (see Figure 4A), thereby gradually increasing the flow rate. The base of the model was elongated 15 mm so that a flow connector (EW-06361-61; Cole-Parmer, Vernon Hills, IL, USA) could be securely fastened to the base end. The nozzle was then inserted into a phantom made of polyvinyl alcohol (PVA) (fabrication details discussed in Example 2-2) to conduct the flow to discharge into a straight tube flow channel with a diameter of 15 mm. The phantom provides an acoustic window to image the flow discharged from the nozzle. Note that this nozzle flow geometry serves as an ideal model for stenosis (in this case, the diameter can be reduced by 90%).
[0065] Example 2-2: Construction of nozzle flow model and flow path setting
[0066] 3-D printing was used to physically construct the nozzle. First, its physical dimensions were drafted on computer-aided design (CAD) software (SolidWorks; Dassault Systems, Waltham, MA, USA). To compile the 3-D printing instructions, the CAD model (saved in STL stereolithography file format) was imported into a slicer software (KISSlicer; ver 1.5). The instructions were then downloaded to a fused deposition modeling (FDM) system (Model DX; CreatBot 3D Printer, Zhengzhou, China) with a nozzle size of 0.25 mm to create the physical model. The thickness of the layer and skin used were 0.1 and 0.5 mm, respectively.
[0067] A flow phantom was fabricated based on the investment casting principle according to the protocol reported in previous work. In this study, a PVA-based wallless flow phantom with a 15 mm diameter flow channel throughout the tissue-mimicking slab was fabricated. The flow channel was formed by first embedding a straight rod in the PVA solution and then removing the rod after the solution solidified. The straight rod had a diameter of 15 mm and a length of 280 mm and served as a replica negative mold of the flow channel. In order for the phantom to grip tightly over the inlet nozzle and outlet flow connector, the 30 mm diameter at both ends of the rod was reduced to 9 mm (the effective length of the flow channel became 220 mm). Similar to the nozzle, the inner core was drawn on SolidWorks and physically constructed using an FDM system with the same settings. Next, the rod was lightly sanded using sandpaper (400 grit) and then suspended in the phantom box (80 × 295 × 70 mm) by mounting it to two side panels. 3 The tissue-mimicking material was cast around the straight rod by (a) pouring the PVA solution into the phantom box, and (b) performing three freeze-thaw cycles (freezing at -20 °C for 24 h and then thawing at 4 °C for 24 h).
[0068] Note that the PVA mixture consisted of 10% PVA (341584; Sigma-Aldrich, St Louis, MO, USA), 3% graphite (282863; Sigma-Aldrich), 0.3% potassium sorbate (85520; Sigma-Aldrich), and 86.7% distilled water. As previously reported, the acoustic attenuation and velocity of this tissue-mimicking material were 0.229 dB / (cm·MHz) and 1535 m / s, respectively. After the thermal cycling process was completed, the flow channel was set up by simply sliding the straight rod out of one end of the phantom.
[0069] After the flow phantom was fabricated, the nozzle was fixed into the inlet flow connector. The device was then connected to a programmable flow pump (details described elsewhere) that supplied the blood-mimicking fluid at a constant flow rate according to the parameters listed in Table I to generate a range of flow conditions as described previously. Note that the blood-mimicking fluid was fabricated using a standardized formula based on Orgasol and a laboratory procedure described previously. Its dynamic viscosity (4.1 mPa·s) and density (1037 kg / m 3 ) matches human blood. Figure 4B Shown is a B-mode image of the inlet segment of the assembled flow phantom captured using a clinical scanner (SonixTouch; Analogic Ultrasound). Table I - Traffic and ROI size for each traffic condition
[0070] Example 2-3: Identifying unstable flow areas using CEUS
[0071] In order to facilitate the identification of areas with unstable flow, microbubble contrast agents can be used to track the flow trajectory. The basic principle behind this is that laminar flow occurs when the fluid flows in parallel layers without interruptions between the layers. In this way, microbubbles in stable areas will move along a straight path. In contrast, unstable flow will correspond to a situation where the trajectories of microbubbles are observed to have mutually intersecting paths. According to this concept, a microbubble contrast agent (USphere Prime; Trust BioSonics, Hsinchu, Taiwan, China) was manually and slowly injected into the inlet of the flow phantom for CEUS imaging.
[0072] Plane wave data acquisition was repeated for all flows using the same data acquisition scheme described in Example 1-1, but using 2-cycle pulses at 5 MHz at 50% of the original transmit power instead of exciting the microbubbles in the stable cavitation regime. High frame rate CEUS images were then generated by the same image formation method as described in Example 1-2. To highlight the flow tracks, high persistence B-mode images were rendered by averaging the beamformed RF signal magnitude over multiple frames before logarithmic compression. This resulted in high echogenic microbubbles producing streaks along their tracks, representing the flow routes. This process was repeated in an overlapping manner every K frames to generate an image replay of the flow routes for analysis by the playback system.
[0073] Image persistence and overlap (i.e., effective frame rate) were adjusted according to the flow rate to normalize the retention time of microbubbles (expressed in number of frames) and thus generate consistent trajectories. Table II summarizes the CEUS rendering parameters for three representative flow conditions. The regions where microbubbles were observed to cross the path were identified and classified as unstable flow regions. On the other hand, the region before the breakdown of the laminar layer was classified as being in a stable flow regime. The intermediate boundary where the flow transitioned from laminar to turbulent was also identified for all flow conditions. The region of interest (ROI) was manually selected using Matlab built-in functions; the height of the ROI was set to 1.5 mm to match the nozzle diameter. The length of the ROI varied depending on the location of the intermediate boundary; in order to avoid ambiguity in the selection of ROIs for performance analysis, a 2.5 mm margin was kept from the intermediate boundary for stable and unstable regions, which was a conservative stance. The rightmost two columns of Table I summarize the ROI size for each region at different flow rates. Table II - Contrast-enhanced ultrasound cine ring rendering parameters
[0074] Example 2-5: ROC analysis of Doppler bandwidth measurement
[0075] To evaluate the sensitivity and specificity of DUBI in determining flow instability, ROC analysis was performed. The procedure involved the following key steps. First, the Doppler bandwidth estimates measured in the ROI for all image frames were classified as belonging to the stable flow (negative) group or the unstable flow (positive) group based on the CEUS reference data. Next, a bandwidth threshold was set to classify the measured bandwidth into its predicted condition (stable or unstable); Doppler bandwidths exceeding the threshold were classified as positive (i.e., unstable) and vice versa. True negatives (TN) and false positives (FP) were calculated from the stable group, while true positives (TP) and false negatives (FN) were calculated from the unstable group. This process was repeated at different bandwidth thresholds in the range of 0.1 to 10 kHz in increments of 0.1 kHz, and each corresponding set of TN, FN, TP, and FP values was calculated. Using these data, the sensitivity [TP / (TP+FN)] and specificity [TN / (TN+FP)] of the test were derived to plot the ROC curve. The area under the curve was calculated as a summary measure of the ROC. In addition, the Youden index (sensitivity + specificity - 1) was calculated for all points on the ROC curve to identify the optimal cutoff value that maximizes sensitivity and specificity with equal weighting.
[0076] Example 2-5: Comparative analysis with CFI
[0077] The performance of DUBI was compared with that of conventional CFI. To facilitate this comparison, CFI frames were computed by applying a CFI scanline-based imaging paradigm to reprocess the original channel-domain dataset as described in Example 1-1. Specifically, the GPU beamformer of our platform was reconfigured to perform four-line parallel receive beamforming for each frame of the channel-domain data. The full image view was divided into 48 regions, each of which contained four beams at adjacent lateral positions. Four-line beamforming of 10 consecutive pulse events was performed on each region before advancing to the next region. Therefore, at each pixel position in the CFI scanline, the slow-time set was the size of 10 samples with a 10kHz sampling rate, resulting in an observation period of 10ms (i.e., the same as that of DUBI). The effective CFI frame rate was 20.8fps. For each slow-time set, tissue clutter was suppressed using a first-order infinite impulse response high-pass filter (0.05 normalized cutoff; using projection initialization), and then the mean flow velocity and velocity variance were estimated by the Kasai autocorrelation algorithm. The flow estimates at different pixels in the CFI frame are mapped to hot and cold tones to present the average flow velocity information. They are also mapped to three color tones to present the average velocity and variance information.
[0078] Example 2-6: Demonstration using an anthropomorphic phantom
[0079] To further demonstrate the efficacy of DUBI for detecting flow instabilities under physiologically relevant conditions, a series of imaging experiments were performed on a healthy carotid bifurcation model, a bifurcation with moderate stenosis (50% eccentric stenosis relative to the internal carotid artery diameter as defined based on NASCET criteria), and a bifurcation with severe stenosis (75% eccentric stenosis). These geometries have well-studied flow characteristics obtained from Doppler ultrasound and particle image velocimetry.
[0080] The bifurcation phantom was fabricated using the same investment casting procedure. The vascular cores (healthy, 50% eccentric stenosis, and 75% eccentric stenosis) were identical to the core geometries reported previously, with unstenotic diameters of 6.0 mm, 4.2 mm, and 3.5 mm for the common carotid artery, internal carotid artery, and external carotid artery, respectively. The vascular cores were first drawn using CAD software (SolidWorks) and physically fabricated using the FDM system mentioned above. The physical configuration of the vascular cores was then embedded into a phantom box (80 × 295 × 70 mm). 3 , width×length×height) in a PVA solution, and three freeze-thaw cycles were performed to solidify the PVA solution. Finally, the geometry of the blood vessel was established by removing the core (by fastening the core at the bifurcation site and sliding out the fastened parts from both ends).
[0081] During the experiments, the bifurcated phantom was connected to a programmable flow pump that drove a pulsatile flow profile (20 ml / s systolic flow; 60 bpm). Plane wave imaging was performed using a custom-made PVA coupling wedge with the transducer surface tilted 20° relative to the phantom surface. Note that the use of a tilted coupling wedge was inspired by another study that used tilted gel pads to create a more favorable beam angle when performing clinical Doppler ultrasound examinations. With this experimental configuration, the DUBI image playback system was acquired using the same protocol as described in the previous subsection.
[0082] Example 3: Discovery of the Nozzle Fluid Model
[0083] Example 3-1: DUBI is effective in depicting unstable flow regions
[0084] Still frames with average Doppler bandwidth were obtained by averaging over 0.5 s under three flow conditions. The results are shown in FIG. 5A to FIG. 5C In the figure, the intermediate boundary between stable and unstable flow is indicated by the white arrow (determined according to CEUS). FIG. 5D to FIG. 5FComparative results derived from the CFI variance map (i.e., average Doppler variance) are also shown. One general observation to note is that the maximum value of the DUBI average Doppler bandwidth map increases from 2.1 kHz to 3.1 kHz and 5.2 kHz as Re increases from 375 to 750 and 1125. More importantly, for the case of Re = 1125, the development of unstable flow corresponds to the spatial peak region in the average Doppler bandwidth map derived from DUBI. The 5.2 kHz spatial peak in the unstable flow region is significantly higher than the 2.4 Hz spatial maximum in the upstream flow jet near the nozzle. This visualization is not clearly highlighted in the average Doppler variance map because the upstream flow jet was found to produce similar average Doppler variance values (spatial maximum: 6.7 kHz) as the unstable flow region (spatial maximum: 7.1 kHz).
[0085] Example 3-2: DUBI’s findings are similar to CEUS
[0086] As a further analysis, FIG. 6A to FIG. 6C Selected CEUS frames are shown with dashed lines drawn to indicate intermediate boundaries between stable and unstable flows. The locations of these boundaries in the high Re case (middle and right frames) are found to be similar to FIG. 5A to FIG. 5C The position where the Doppler bandwidth estimate starts to rise in is very close, indicating that the Doppler bandwidth can be a reliable indicator for distinguishing flow instability.
[0087] Example 3-3: DUBI shows strong ROC performance in mapping flow instabilities
[0088] Fig. 7A A graph including the ROC curve (dark line) for DUBI with samples collected from all ROIs at three flow rates. It can be observed that the ROC curve (area under the curve of 0.85) lies above the diagonal line (grey dashed line), indicating that DUBI has a positive predictive ability in determining flow instability. Fig. 7A (Gray line) The Youden index was also plotted as an overall indicator of sensitivity and specificity. The optimal Youden index was found to be 0.54 when the sensitivity and specificity were 0.72 and 0.83, respectively. Figure 7B As shown in the dual population histogram shown, this optimal point corresponds to a bandwidth threshold of 2.4 kHz. These ROC results represent a significant improvement over those obtained with Doppler variance mapping. Figure 7C and Fig.7D As shown, the best sensitivity and specificity of Doppler variance mapping are 0.68 and 0.66, respectively, and their maximum Youden index is 0.34 with a bandwidth threshold of 4.0 kHz. The area under the ROC curve of Doppler variance mapping (0.72) is also smaller than that of DUBI.
[0089] Example 4: Findings from the Carotid Bifurcation Experiment
[0090] Example 4-1: DUBI highlights unstable flow in a narrow area
[0091] When DUBI was applied to a pulsating carotid bifurcation phantom, it was found to be able to highlight the blood flow instability generated downstream of the stenosis in the diseased bifurcation.
[0092] It is important to note that the Doppler bandwidth range in the stenotic vessel (upper branch) increases due to the narrowing of the lumen at the site of stenosis. This observation is consistent with the results from the flow nozzle model, where increasing Re will lead to a larger Doppler bandwidth (in the bifurcation experiment, the flow distribution remains unchanged, but the "nozzle" narrows as the degree of stenosis increases). The Doppler bandwidth in the healthy model is low (<1.4kHz) throughout the cardiac cycle. In contrast, high Doppler bandwidths (>2.4kHz) were observed in both diseased models, especially in the flow jet region during the flow systolic and dicrotic phases of the cardiac cycle.
[0093] FIG. 8A to FIG. 8O Selected frames of DUBI are shown in Figure 2 to facilitate further interpretation of the information provided by DUBI at specific time points of interest. This figure depicts the Figure 8B , Figure 8G and Figure 8I ), end-systolic period ( Figure 8C , Figure 8H and Figure 8M ), end diastole ( Fig.8D , Figure 8I and Figure 8N ) is the systolic peak at the moment of the peak Doppler bandwidth measured in Fig. 8A , Figure 8F and Figure 8K ) frames, and the corresponding Doppler spectrogram of the stenosis site ( Fig. 8E , Figure 8J and Fig.8O ). Three main observations can be made. First, the maximum Doppler bandwidth increases with increasing stenosis, which is reflected in the brighter thermal tones in the DUBI frames. Second, the peak Doppler bandwidth is found in the jet tail [ Figure 8K and Figure 8L ], where the flow disturbance is the strongest. Third, for the 50% stenosis model ( FIG. 8F to FIG. 8J ), a significant increase in Doppler bandwidth was found only during flow systole, whereas for the 75% stenosis model, the elevated Doppler bandwidth persisted throughout the cardiac cycle ( Figure 8K to Figure 8O ). The time and location of the peak Doppler bandwidth are generally consistent with those measured using turbulence intensity under similar flow conditions reported previously.
[0094] Example 4-2: Maximum Doppler bandwidth is related to the degree of stenosis
[0095] As a further analysis of DUBI, Fig. 9A and Fig. 9B The time traces of the Doppler bandwidth measured at the stenotic jet region (dark line) and its tail (grey line) of two diseased bifurcation phantoms are shown. A dashed line indicating the systolic peak is also included in this figure. For the 75% stenosis model, the maximum Doppler bandwidth at the jet tail is significantly higher (7.7kHz) compared to the 50% stenosis model (3.3kHz). In addition, in the 75% stenosis model, the Doppler bandwidth at the jet tail shows large temporal fluctuations throughout the cardiac cycle and shows a decreasing trend at the end of systole (100ms after the systolic peak) because flow deceleration naturally favors the restoration of stable flow conditions. Another point worth noting is that in both diseased models, the Doppler bandwidth at the flow jet is lower than the Doppler bandwidth at the jet tail. This finding, as expected, indicates that flow instability does not occur primarily at the flow jet, but rather downstream of the jet. For the 75% stenosis model, the Doppler bandwidth was found to show a greater degree of fluctuation at the stenotic flow jet. This trend may be because for the 75% stenosis model, the higher temporal variation of the jet velocity naturally favors the transition between the stable and unstable flow regimes.
[0096] Example 5: DUBI as a framework for mapping flow instabilities
[0097] Visualizing unstable flows in a non-invasive manner is not an easy task. In particular, two practical flow characteristics must be addressed when designing new flow instability mapping frameworks: (a) unstable flow patterns can vary spatially at a given time point due to their dissipative properties; and (b) flow conditions can vary temporally during the cardiac cycle due to the pulsatility of blood flow. DUBI is specifically designed to visualize and track these spatiotemporal dynamics. It has three key features that collectively make visualization of unstable flows possible. First, DUBI is able to track the spatial variations of flow instabilities (measured via local Doppler bandwidth) across the entire image view at high frame rates that exceed the range of video display [ Figure 3A ]. Second, DUBI uses an AR modeling approach to consistently derive Doppler bandwidth estimates [ Figure 3B ]. Third, DUBI's triple display method can simultaneously present flow instability information (Doppler bandwidth), flow trajectory (flow speckle), and anatomical background. Figure 3C ].
[0098] At least Figure 4A and Figure 4BThe effectiveness of DUBI for identifying unstable flows in a range of flow conditions from laminar to turbulent was demonstrated in the first study of a nozzle flow device ( Figure 4A and Figure 4B ) to evaluate the performance of DUBI, using the CEUS images obtained as the baseline reference ( FIG. 6A to FIG. 6C ). It is found that the unstable flow region corresponds to the high Doppler bandwidth region in DUBI ( FIG. 5A to FIG. 5C This correspondence has been widely found to be sensitive and specific compared to conventional CFI mapping of Doppler variance ( 7A to 7D The practical value of DUBI was also confirmed through a series of carotid bifurcation experiments ( Figures 8A to 8O DUBI was found to be effective in identifying unstable flow in the jet tail downstream of the stenosis ( Fig. 9A and Fig. 9B ).
[0099] DUBI represents the first image-based, non-invasive flow instability mapping framework with fine temporal resolution. From a clinical diagnostic standpoint, this framework unlocks new potential to improve the management of atherosclerotic disease. For example, the emergence of unstable blood flow can indicate the onset of plaque formation, and thus DUBI may help facilitate early diagnosis of atherosclerosis. Additionally, because unstable flow has been shown to contribute to the progression of atherosclerotic plaques, new insights into plaque progression may be gained by correlating the intensity and size of areas of flow instability. Furthermore, given that nozzle flow devices have demonstrated initial potential to detect flow instabilities emerging from stenotic flow outlets, DUBI may be further developed into a new tool for valvular stenosis diagnosis to complement other emerging ultrasound technologies.
[0100] Example 6: GPU-based implementation of Berg method
[0101] For the GPU computation kernel for Doppler bandwidth estimation, a thread block is allocated to process the Doppler bandwidth estimation of the slow time set of a pixel in the DUBI frame. Note that the corresponding slow time set x[n] is first transferred to the shared memory of the GPU for fast data access. Then, a thread block containing N threads (i.e., the same size as the slow time set) is continued to derive the P-order AR parameter value set for the slow time set by executing a fast implementation of the Berg method involving P iterations from p=1 to p=P. In each iteration, two computation steps are performed. First, the computation thread is tasked with computing the forward prediction error set and the backward prediction error set, denoted as e[n], respectively. f P [n] and e b P[n]. Specifically, for the pth iteration, the task of the nth thread is to compute the nth sample in the following error sets (each set has N samples):
[0102] Note that for the pth iteration, the average of (A1) and (A2) corresponds to the error term e[n] in the AR model in (1).
[0103] In the second calculation step of the same iteration, the task of the subgroup of threads is to calculate the intermediate AR parameter a P,k The second calculation step obtains the intermediate AR parameters by updating the intermediate AR parameters using the following formula:
[0104] where i = 1, 2, ..., p-1, a o,o = 1, * indicates complex conjugate, and
[0105] After repeating the above calculation steps for P iterations, the AR model coefficient set is finally determined.
[0106] The foregoing summarizes the features of several embodiments so that those skilled in the art can better understand the various aspects of the present disclosure. Those skilled in the art will appreciate that the present disclosure can be easily used as a basis for designing or modifying other processes and structures to achieve the same purposes and / or achieve the same advantages as the embodiments described herein. Those skilled in the art will also recognize that such equivalent constructions do not depart from the spirit and scope of the present disclosure, and that various changes, substitutions and modifications may be made without departing from the spirit and scope of the present disclosure.
Claims
1. A method for detecting flow instability, comprising: insonate the area of interest with ultrasonic pulses; acquiring radio frequency (RF) data from the echo pulses of the ultrasonic pulses; processing the RF data; and Doppler bandwidth is estimated from processed RF data by autoregressive AR modeling; Also included is: repeating the estimation of the Doppler bandwidth for each image pixel to generate a Doppler ultrasound bandwidth imaging frame as a description of the flow instability over the image view.
2. The method of claim 1, further comprising comparing the Doppler bandwidth to a predetermined value, wherein a Doppler bandwidth equal to or greater than the predetermined value indicates flow instability and a Doppler bandwidth less than the predetermined value indicates stable flow.
3. The method of claim 1, wherein deriving the Doppler bandwidth of the processed RF data by AR modeling comprises estimating the Doppler bandwidth at a given pixel position over a short time period by first deriving an AR model of a slow-time set, and then estimating the spectral power of the signal model.
4. The method of claim 1, wherein the Doppler bandwidth is determined as the full width at half maximum of the power spectrum.
5. The method of claim 1, further comprising displaying the map of the Doppler bandwidth synchronously with displaying a flow trajectory or a flow speckle pattern of a B-mode image.
6. The method of claim 5, further comprising: The flow spot map is overlaid on the flow area of the B-mode image and displayed as a base layer; Map the Doppler bandwidth estimate to a warm tone with brighter colors corresponding to higher Doppler bandwidths, overlay this colormap using the alpha compositing principle.
7. The method as claimed in claim 6, overlaying the color map using the Alpha synthesis principle comprises: For non-flow areas, transparency was set to 100% to show the anatomy; For flow areas, set transparency to 70% to show the flow speckle pattern and color-coded Doppler bandwidth estimates.
8. The method of claim 7, further comprising: The rendering strategy of overlaying this color map using the Alpha synthesis principle is repeated at different time points, and the image frames are stacked together to form an image playback system.
9. A device for detecting flow instability, comprising: a transmitter configured to cause ultrasound pulses to act on the area of interest; a receiver configured to receive an echo pulse from the ultrasonic pulse; a scanner configured to acquire radio frequency (RF) data from the echo pulses; and A processor configured to: processing the RF data; and Estimate Doppler bandwidth from processed RF data using autoregressive AR modeling; Also included is: repeating the estimation of the Doppler bandwidth for each image pixel to generate a Doppler ultrasound bandwidth imaging frame as a description of the flow instability over the image view.
10. The apparatus of claim 9, the processor further configured to: further include comparing the Doppler bandwidth to a predetermined value, wherein a Doppler bandwidth equal to or greater than the predetermined value indicates flow instability, and a Doppler bandwidth less than the predetermined value indicates stable flow.
11. The apparatus of claim 9, wherein deriving the Doppler bandwidth of the processed RF data by AR modeling comprises estimating the Doppler bandwidth at a given pixel position in a short time by first deriving an AR model of a slow time set, and then estimating the spectral power of the signal model.
12. The apparatus of claim 9, wherein the Doppler bandwidth is determined as the full width at half maximum of the power spectrum.
13. The apparatus of claim 9, wherein the processor is further configured to generate a map of the Doppler bandwidth, and to display the map synchronously with the flow speckle pattern or the B-mode image.
14. The apparatus of claim 13, wherein the processor is further configured to: The flow spot map is overlaid on the flow area of the B-mode image and displayed as a base layer; Map the Doppler bandwidth estimate to a warm tone with brighter colors corresponding to higher Doppler bandwidths, overlay this colormap using the alpha compositing principle.
15. The apparatus of claim 14, wherein overlaying the color map using an Alpha synthesis principle comprises: For non-flow areas, transparency was set to 100% to show the anatomy; For flow areas, set transparency to 70% to show the flow speckle pattern and color-coded Doppler bandwidth estimates.
16. The apparatus of claim 15, further comprising: The rendering strategy of overlaying this color map using the Alpha synthesis principle is repeated at different time points, and the image frames are stacked together to form an image playback system.
