Method and apparatus of ultrasound contrast imaging

Abstract

Method for imaging a region of an organ by detecting the nonlinear signal of ultrasound backscattered signal from ultrasound contrast agents (UCAs) present in the region, the method including: (a) transmitting a number of na pulse sequences of ultrasound waves by an array ultrasound probe (5) in the direction of said region such as each pulse sequence comprising successive ultrasound waves emitted at one acoustic pressure amplitude, na being an integer larger than 2; (b) acquiring raw data from backscattered ultrasound waves by said array ultrasound probe (c) generating a number of na compound images from said raw data, each compound image being associated with an acoustic pressure amplitude, said number na of compound images forming a set of in-phase / quadrature (IQ) data expressed as function of spatial variables and pressure variable; (d) shaping said IQ data into a 2D space-pressure amplitude matrix form; (e) computing the decomposition of said reshaped data into separable spatio-pressure bases; (f) removing values and vectors corresponding to tissue signal in order to determine a filtered 2D space-pressure amplitude matrix based on a predetermined threshold; (g) reshaping said filtered 2D space-pressure amplitude matrix into na filtered images; (h) generating a final image from the number of na filtered images.

Description

DescriptionTitle: Method and apparatus of ultrasound contrast imagingTechnical FieldThe present disclosure concerns methods and apparatuses of ultrasound imaging of organs and tissues by detection of nonlinear backscattered signals from a region containing nonlinear scatters such as microbubbles or gas vesicle.Background Art

[0001] Ultrasound imaging is a clinically established diagnostic tool, providing non-invasive, real-time visualization with high spatial and temporal resolution (typically < 500 pm and < 10 ms). Thanks to rapid development of electronics and related technology, ultrasound scanners designed for medical have become cheaper, easier to use, more compact and more powerful medical instrument. However, the changes of acoustic impedance occurring within the living tissue are small and the absorption of ultrasound energy by different types of tissues, as blood vessels, organs, etc. are such that some diagnostic applications remain challenging. Introduction of ultrasound contrast agents (UCAs) made from stabilized suspensions of gas microbubbles or gas-containing vesicles into the blood stream and organs to be imaged enhances the backscattered signal of the regions to be imaged, allowing the obtention of better and more useful ultrasound images of organs and surrounding tissue.

[0002] In the present disclosure, the term “ultrasound contrast agent(s) (UCA(s) designates gas microbubble of some pm diameter, nanobubbles of typical hundred nanometers diameter or gas vesicles of typical hundred nanometers diameter).

[0003] UCAs modify the behavior of ultrasound waves when it interacts with the UCAs. The UCAs act as nonlinear scatters and increase the intensity of the backscattered ultrasound beam by providing a nonlinear response. UCAs are efficient ultrasound reflectors for echography and this efficiency depends on local acoustic pressure, the nature of the gas and the properties of the capsule. The main parameter that can be controlled is the acoustic pressure or mechanical index (Ml) emitted by the ultrasound probe.

[0004] At very low pressure (Ml< 0.1), the UCAs oscillate synchronously with the ultrasonic wave. This is called linear diffusion regime and is not used. When the pressure is greater than a certain threshold which depends on the gas and capsule properties, i.e. between 0.1 < Ml < 0.2, UCA oscillations become asynchroneous with the pressure. This is non-linear regime. Finally at high acoustic pressure (Ml > 0.2), a rupture of the UCA shell, with the diffusion of the gas. Since tissue does not display nonlinear response to the same extentas UCAs, it is also possible to use the non-linear response of UCAs as backscattered signal to image the region containing contrast agents.

[0005] Recently, contrast ultrasound imaging has gained increasing capacities for molecular and cellular imaging due to the development of micro- and nanoscale contrast agents and reporter genes capable of targeting specific disease states or visualizing cellular processes. Acoustic biomolecular contrast agents, commonly known as gas vesicles, have demonstrated substantial potential in a wide range of biomedical applications. These gas vesicles are typically filled with air and encapsulated by a 2-nm-thick protein shell, possessing a diameter of approximately 85 nm and a length of 500 nm. Certain gas vesicles exhibit a strongly nonlinear response to acoustic pressure, which results in nonlinear scattering of ultrasound waves, enabling enhanced detection sensitivity. The acoustic range for nonlinear regime is comprised between 0.2 and 0.6 MPa.

[0006] Nevertheless, the detection of the nonlinear signal from UCAs remains a challenge because it is impaired by the signals originating from tissue and other nonlinear propagation artifacts.

[0007] In order to remove the linear part of the backscattered signal, one known method consists of transmitting multiple pulses into the medium being imaged that suppress the lineartissue signal when the successive backscattered signals are combined by subtraction or addition. The multi-pulse contrast sequences, such as amplitude modulation (AM), phase inversion (PI) and their combination amplitude modulated pulse inversion (AMPI), are performed by scanning line-by-line the region being imaged. For example, a first and second ultrasound pulses are amplitude modulated signals, the first ultrasound pulse differing only in sign (polarity) from the second ultrasound pulse transmitted. The backscattered signals generated by these successive pulses are combined by adding them so that the linear components cancel, leaving only the nonlinear component from the UCAs. Accordingly, since tissue generally reflects less harmonic components than UCAs, only backscattered signal from ultrasound contrast agents produce nonlinear signals that can be used to construct contrast image with great sensitivity.

[0008] This method imposes several strict requirements. Firstly, the amplitude modulation pulse sequences require multiple pulses with relative amplitudes of 14, 14 and 1 that must be pre-calibrated. However, in practice, it is very difficult if not impossible to set up the ultrasound imaging instrument in such a way that to ensure the cancel of the linear component by providing the exact transmit acoustic pressure amplitude of 14. Secondly, a time interval of several hundred microseconds must elapse between successive transmitted pulses, in order for the backscattered signals from the deepest regions of interest toreturned to the ultrasound probe before a following pulse can be transmitted. This requirement limits the framerate, which is reduced in a proportion depending on the number of pulses, preventing the monitoring of fast nonlinear events across the imaging plane, making them inadequate for imaging the fast nonlinear events, especially within deep tissues, for instance within the myocardium.

[0009] To address these challenges, an ultrafast amplitude modulation (uAM) has been described by Rabut et al. [Claire et al. Ultrafast amplitude modulation for molecular and hemodynamic ultrasound imaging. Appl. Phys. Lett. 14 June 2021 ; 118 (24): 244102.] In the ultrafast amplitude modulation (uAM) that combines amplitude modulation with multiplane wave transmission, several series of N successive tilted plane waves are repeated with modulated amplitude: two series of half amplitude by activating alternatively, the odd and even elements of the transducers, and one series of full amplitude by activating all elements of the probe. After reception, the subtraction of the two half-amplitude series from the full amplitude series allows the elimination of linear signal and the detection of nonlinear responses. The trios of modulated multiplane wave bursts are then repeated N times, and for each repetition, the polarities of the successive plane waves are given by the columns of the Hadamard matrix of order N. This imaging technique is capable of achieving ultrafast acquisition of nonlinear images after successive transmission of tilted, amplitude- modulated plane waves. Thus, amplitude modulation pulses are combined with multiplane waves transmissions and repeated N times with different polarizations.

[0010] Nevertheless, this technique faces several significant limitations, firstly the pulse sequences conventionally require either odd- or even-element transmission configurations due to amplitude modulation. This configuration induces inhomogeneities in the wavefront due to one over two elements transmission. Secondly, the complex multiplane wave transmission scheme inherently leads to prolonged durations that reduce the framerate. Thirdly, the multiplane wave transmission scheme leads high acoustic energy transmission that can destroy the microbubbles or in other ways altered between successive pulses the capsules. Endly, this method fails to solve the situation in which the tissue tends to have a nonlinear behavior as the UCAs. Indeed, some studies demonstrate that tissue may exhibit a nonlinear signal due to the nonlinear propagation of ultrasound waves within tissues.

[0011] Additionally, a nonlinear x-wave ultrasound imaging scheme has been proposed by Maresca et al. [Maresca et al., Nonlinear X-Wave Ultrasound Imaging of Acoustic Biomolecules, Physical Review X, 8, 041002, 2018] In this nonlinear X-wave amplitude modulation (xAM) imaging sequence, a tilted plane wave at an angle 0 with respect to the array was transmitted from the first half aperture (from the elements 1 to N / 2) of the linear array transducer. Then, another symmetric plane wave at an angle 0 with respect to thearray was transmitted from the second half aperture (from the elements N / 2+1 to N) of the linear array transducer. Then, the previously described two plane waves were simultaneously transmitted again using the full aperture of the transducer. This axisymmetric pulse sequence generates the amplitude modulation code along the bisector that separates the two N / 2 subapertures.

[0012] However, this technique requires the line-by-line transmission, therefore, it is not able to implement this imaging scheme in an ultrafast manner. Further to the limitation on the slow frame rate, the imaging field of view is also limited to one half of the aperture size of the probe in the lateral direction. Therefore, xAM technique is not suitable for the monitoring of the fast nonlinear events across the imaging plane at depth.

[0013] Recently, Wahyulaksana et al. [Wahyulaksana G et al. Higher Order Singular Value Decomposition Filter for Contrast Echocardiography, IEEE Trans Ultrason Ferroelectr Freq Control. 2023 Nov;70(11):1371-1383] proposes another approach to suppress the tissue signal by using a higher-order singular value (HOSVD) decomposition filter. It is known to use the singular value decomposition (SVD) to decompose the tissue, blood flow and noise into separate singular vectors assuming they have different spatiotemporal coherence. The SVD filter allows the suppression of tissue and noise singular vectors, with only the flow signal of the interest remaining. However, studies show that its performance degrades with slower flow rates and faster tissue motion since in that case the coherence between the contrast agent signal and tissue increases, thus preventing the separation into different components.

[0014] Wahyulaksan et al. proposes using higher-order singular value decomposition (HOSVD) to combine received backscattered signals from the complementary pulses in the multi-pulse contrast schemes, instead of the conventional linear signal combinations. The pulsing sequences has been added as a third dimension to the spatial and temporal information in the HOSVD. Then, removing the ranks in the pulsing dimension with dominant coherent signals from tissue allows to improve the contrast detection. Nevertheless, this HOSVD method necessitates the use of the temporal dimension and the spatial dimension to discriminate tissue signal and contrast agent signal. In other words, this method is limited only to contrast imaging of moving contrast agents. Consequently, this method becomes challenging in the smaller vessels where the motion of the ultrasound contrast agents becomes lower. It becomes also inefficient for gas vesicles expressed into tissue as these gas vesicles do not move compared to tissue. Consequently, ultrasound imaging of slowly moving gas vesicles or microbubbles moving in very small vessels is still an issue in major applications in contrast echography.

[0015] Therefore, one aim of the present disclosure is to propose an ultrafast nonlinear ultrasound imaging that does not requiring i) a complex and specific pulsing sequence, ii) the control of the amplitudes of the transmitted acoustic pressure and iii) any assumption on the non-negligible motion of contrast agents to remove the tissue signals, while achieving ultrafast acquisition of images and better contrast imaging sensitivity with lower acoustic transmission energy.Summary

[0016] The present disclosure improves the situation.

[0017] To this end, the present disclosure proposes a method for imaging a region of an organ by detecting the nonlinear signal of ultrasound backscattered signal from ultrasound contrast agents (UCAs) present in the region, the method including:(a) transmitting a number of napulse sequences of ultrasound waves by an array ultrasound probe in the direction of said region such as each pulse sequence comprising successive ultrasound waves emitted at one acoustic pressure amplitude, nabeing an integer larger than 2;(b) acquiring raw data from backscattered ultrasound waves by said array ultrasound probe(c) generating a number of nacompound images from said raw data, each compound image being associated with an acoustic pressure amplitude, said number naof compound images forming a set of in-phase / quadrature (IQ) data expressed as function of spatial variables and pressure variable;(d) shaping said IQ data into a 2D space-pressure amplitude matrix form;(e) computing the decomposition of said reshaped data into separable spatio-pressure bases ;(f) removing values and vectors corresponding to tissue signal in order to determine a filtered 2D space-pressure amplitude matrix based on a predetermined threshold ;(g) reshaping said filtered 2D space-pressure amplitude matrix into nafiltered images;(h) generating a final image from the number of nafiltered images.

[0018] In one embodiment, the step of computing of the decomposition of reshaped data uses a spatio-pressure singular value decomposition (SVD) to generate three matrices as a weighted and ordered sum of separable spatio-pressure matrices expressed by II, S and V, where II represents the spatial matrix, V represents the pressure amplitude matrix and S represents a weighting matrix.

[0019] In another embodiment, the step of computing of the decomposition of reshaped data uses a spatio-pressure independent component analysis (ICA).

[0020] The following features, can be optionally implemented, separately or in combination one with the others:

[0021] In one or more embodiments, each pulse sequence is composed of one single pulse having a number of M planar ultrasound waves emitted with identical phase at different angles to form M images, said step (c) including a synthesizing step of said M images to form a compound image of the region, M being an integer value greater than 1.

[0022] In one or more embodiments, each pulse sequence is composed of one single pulse having a number of M diverging ultrasound waves emitted with different center of curvature to form M images, said step (c) including a synthesizing step of said M images to form a compound image of the region, M being an integer value greater than 1.

[0023] In one or more embodiments, each pulse sequence is composed of one positive pulse having a number of M planar ultrasound waves emitted with identical positive phase at different angles to form M positive images and one corresponding negative pulse having a number of M planar ultrasound waves emitted with identical negative phase at different angles to form M negative images, said step (c) including a synthesizing step of said M positive and said M negative images to form a positive compound image and a negative compound image of the region respectively.

[0024] In one or more embodiments, each pulse sequence is composed of one positive pulse having a number of M diverging ultrasound waves emitted with identical positive phase at different center of curvature to form M positive images and one corresponding negative pulse having a number of M diverging ultrasound waves emitted with identical negative phase at different center curvature to form M negative images, said step (c) including a synthesizing step of said M positive and said M negative images to form a positive compound image and a negative compound image of the region respectively.

[0025] In one embodiment, the method further comprises a step of rearrangement of napositive compound images and nanegative compound images by concatenation to form a set of in-phase / quadrature (IQ) data to be reshaped in a Casorati matrix.

[0026] In one or more embodiments, the threshold is determined automatically by an adaptative SVD rejection or determined manually by a user.

[0027] In one embodiment, the adaptative SVD rejection uses a spatial similarity matrix and a square-fitting matrix.

[0028] In one or embodiments, the number naof acoustic pressure amplitude is comprised between 2 and 12, and preferably between 4 and 6.

[0029] The ultrasound contrast agents may be gas vesicles or microbubbles.

[0030] The present disclosure also concerns an apparatus (N LI APP) for imaging a region of an organ by detecting the nonlinear signal of ultrasound backscattered signal from ultrasound contrast agents (UCAs) present in the region, said apparatus including at least an array ultrasonic probe and a control processor configure to control said probe to :(a) transmit a number of napulse sequences of ultrasound waves by an array ultrasound probe in the direction of said region such as each pulse sequence comprising successive ultrasound waves emitted at one acoustic pressure amplitude, nabeing an integer larger than 2;(b) acquire raw data from backscattered ultrasound waves by said array ultrasound probe(c) generating a number naof compound images from said raw data, each compound image being associated with an acoustic pressure amplitude, said number of images forming a set of in-phase / quadrature (IQ) data expressed as function of spatial variables and pressure amplitude variable;(d) shape said IQ data into a 2D space-pressure amplitude matrix form;(e) compute the decomposition of said reshaped date into separable spatio-pressure amplitude bases;(f) remove values and vectors corresponding to tissue signal in order to determine a filtered 2D space-pressure amplitude matrix based on a predetermined threshold ;(g) reshape said filtered 2D space-pressure amplitude matrix into a nafiltered images;(h) generate a final image from said filtered images.

[0031] In one or more embodiments, the control processor is configured to compute the decomposition of reshaped data by using a spatio-pressure singular value decomposition (SVD) to generate three matrices as a weighted and ordered sum of separable spatio- pressure matrices expressed by II, S and V, where II represents the spatial matrix, V represents the pressure amplitude matrix and S represents a weighting matrix.

[0032] The following features, can be optionally implemented, separately or in combination one with the others:

[0033] In one or more embodiments, each pulse sequence is composed of one single pulse having a number of M planar ultrasound waves emitted with identical phase at different angles to form M images and said control system is configured to synthesize the M images to form a compound image of the region, M being an integer value greater than 1.

[0034] In one or more embodiments , each pulse sequence is composed of one single pulse having a number of M diverging ultrasound waves emitted with different center of curvature to form M images and said control system is configured to synthesize said M images to form a compound image of the region, M being an integer value greater than 1.

[0035] In one or more embodiments , each pulse sequence is composed of one positive pulse having a number of M planar ultrasound waves emitted with identical positive phase at different angles to form M positive images and one corresponding negative pulse having a number of M planar ultrasound waves emitted with identical negative phase at different angles to form M negative images, and said control system is configured to synthesize said M positive and M negative images to form a positive compound image and a negative compound image of the region respectively.

[0036] In embodiments of the apparatus, each pulse sequence is composed of one positive pulse having a number of M diverging ultrasound waves emitted with identical positive phase at different center of curvature to form M positive images and one corresponding negative pulse having a number of M diverging ultrasound waves emitted with identical negative phase at different center curvature to form M negative images, and said control system is configured to synthesize said M positive and said M negative images to form a positive compound image and a negative compound image of the region respectively.

[0037] In another aspect, it is proposed a computer software comprising instructions to implement at least a part of a method as defined here when the software is executed by a processor. In another aspect, it is proposed a computer-readable non-transient recording medium on which a software is registered to implement the method as defined here when the software is executed by a processor.Brief Description of Drawings

[0038] Other features, details and advantages will be shown in the following detailed description and on the figures, on which:Fig. 1

[0039] [Fig. 1] Figure 1 is a block diagram illustrating an embodiment of an apparatus according to the present disclosure.Fig. 2

[0040] [Fig. 2] Figure 2 is a block diagram illustrating a possible method of obtaining a final filtered image with the apparatus of Figure 1 .Fig. 3

[0041] [Fig. 3] Figure 3(a) illustrates an example of BM pulse sequence and Figure 3(b) illustrates a PI pulse sequence.Fig. 4

[0042] [Fig. 4] Figure 4 shows simulation of ultrasound images composed of tissue, gas vesicle, and noise signals: spatial distribution of tissue signal (a), gas vesicle signal (b), noise signal (c) and the final composed image (d).Fig. 5

[0043] [Fig. 5] Figure 5 shows an example of simulated backscattered amplitude of linear tissue signal (AL), nonlinear tissue signal (ANL), gas vesicle signal (AGV) and noise signal (AN) with respect to acoustic pressure amplitude.Fig. 6

[0044] [Fig. 6] Figure 6 shows the processing of the nonlinear singular value decomposition beamforming on images generated with the BM pulse sequence of Figure 3(a).Fig. 7

[0045] [Fig. 7] Figure 7 shows the processing of the nonlinear singular value decomposition beamforming on images generated with the PI pulse sequence of Figure 3(b).Fig. 8

[0046] [Fig. 8] Figure 8 shows the quantification of CTR, CNR and SNR in vitro GV phantom using BM, pulse inversion, nonlinear SVD beamforming on BM-mode images of the present disclosure and nonlinear SVD beamforming on pulse inversion images of the present disclosure.Fig. 9

[0047] [Fig. 9] Figure 9 shows comparison of in vitro GV phantom between uAM, xAM and nonlinear SVD beamforming of the present disclosure using the number of pressure level of 4 and 16 respectively; (a) uAM image; (b) xAM image; (c) nonlinear SVD beamforming method of the present disclosure using the number of pressure level of 4; (d) nonlinear SVD beamforming method using the number of pressure level of 16. All the images were displayed with a dynamic range of -25 dB. (e) Quantification of SNR between different methods, (f) Quantification of the number of transmissions and pulses used for each imaging acquisition.Fig. 10

[0048] [Fig. 10] Figure 10 shows comparison of ex-vivo imaging of chicken breast with GVs injection between uAM, xAM and nonlinear SVD beamforming of the present disclosure using the number of pressure level of 4 and 16 respectively; (a) uAM image (PNP=0.45 MPa); (b) xAM image (PNP=0.48 MPa); (c) nonlinear SVD beamforming method using the number of pressure level of 4 (PNP=0.45 MPa); (d) nonlinear SVD beamformingmethod using the number of pressure level of 16 (PNP=0.45 MPa; (e) Schematic diagram of ex vivo imaging of chicken breast, (f) B-mode image for this imaging plane, dotted circle indicates the GV injection site and plein circle represents the tissue region which used for SNR quantification; (g) Quantification of SNR for the corresponding sequences.Fig. 11

[0049] [Fig. 11] Figure 11 shows comparison of in vivo imaging of the left lower limb of a rat with GV injection between uAM, xAM and nonlinear SVD beamforming of the present disclosure using the number of pressure level of 4 and 16 respectively; (a) uAM image (PNP=0.55 MPa); (b) xAM image (PNP=0.59 MPa); (c) nonlinear SVD beamforming method using the number of pressure level of 4 (PNP=0.55 MPa); (d) nonlinear SVD beamforming method using the number of pressure level of 16 (PNP=0.55 MPa); (e) Schematic diagram of in vivo imaging of the left lower limb with GV injection, (f) B-mode imaging for this imaging plane, dotted ellipse indicates the GV injection site and plein ellipse represents the tissue region which used for SNR quantification; (g) Quantification of SNR for the corresponding sequences.Description of Embodiments

[0050] In the Figures, the same references denote identical or similar elements.

[0051] The present disclosure proposes a method and apparatus for improving the contrast imaging sensitivity of nonlinear ultrafast ultrasonic imaging based on the use of micro-and nanoscale contrast agents present in a vascular network of a human or animal for example. In particular, the present disclosure exploits a new nonlinear spatio-pressure singular value decomposition filtering of ultrasonic data acquired at ultrafast framerate and at different acoustic pressure amplitudes, allowing a better discrimination between the tissue signal and the nonlinear ultrasound contrast agent signal. The spatio-pressure singular value decomposition takes benefits of the different behavior of tissue and ultrasound contrast agents in terms of spatio-pressure coherence to form space singular vectors and pressure singular vectors. Thanks to this separation between the singular vectors, it is possible to select the specific modes of the corresponding singular vectors to reconstruct a final image with high contrast signals. This spatio-pressure singular value decomposition enables nonlinear ultrasound imaging in very small vessels and an accurate measure of different anatomical parameters may be computed. In addition, this nonlinear SVD filtering can be performed even though the ultrasound contrast agents are static.

[0052] The region of the vascular network may be a brain slice for 2D imaging or a 3D region of the brain for 3D imaging. The region to be imaged may be any part of an organ to be imaged.

[0053] Figure 1 is a block diagram that schematically illustrates a block diagram of a nonlinear ultrafast ultrasound imaging apparatus 1 (NLI APP) for the imaging of ultrasound contrast agents (UCAs), in accordance with an embodiment that is describes herein. Imaging apparatus 1 is typically used for producing ultrasound images of a target organ of a patient. For example, the target region may be a region of a vascular network.

[0054] The apparatus 1 comprises a processor 2 (PROC), for instance a specialized signal processing device controlled by a computer or a group of computers, possibly a group of computers including servers.

[0055] The processor 2 may include a computing module 3 (COMP MOD), the operation of which will be explained later.

[0056] The processor 2 may control an ultrasound contrast agent injection device 4 (I NJ).

[0057] The ultrasound contrast agent injection device 4 may comprise ultrasound contrast agents. The ultrasound contrast agents (UCAs) may be microbubbles, nanobubbles, gas vesicles, or equivalent ultrasound contrast agents. The UCAs typically reflect ultrasound echoes much stronger than red blood cells and produce nonlinear signal. Therefore, UCAs can be used for high-contrast imaging of blood vessels structure, as well as blood perfusion in organs. For example, prior to imaging, the UCAs are administrated intravenously by the ultrasound contrast agent injection device 4. For example, the ultrasound agent injection device 4 may be a push syringe.

[0058] For example, nonlinear GVs are prepared following the methodology outlined in [Rabut, C., et al., Ultrafast amplitude modulation for molecular and hemodynamic ultrasound imaging. Applied Physics Letters, 2021. 118(24)]. GVs are extracted from buoyant Anabaena flos-aquae cells through hypertonic lysis and subsequently purified by repeated centrifugally assisted flotation and resuspension. The outer GvpC layer of the gas vesicles is then removed through treatment with 6M urea, followed by additional round of centrifugally assisted flotation, dialysis in 1x PBS and resuspension to remove residual GvpC and urea. These GVs will be used for making a tissue-mimicking gas vesicle phantom. For example, the phantom is fabricated by casting 1 % (w / v) agarose in PBS supplemented with 0.2% (w / v) AI2O3. For static imaging, a custom 3D-printed mold was employed to create a cylindrical well with a 2 mm diameter. Gas vesicles were incubated at 42 °C for 1 minute, mixed in a 1 =1 ratio with low-melt molten agarose (at 42 °C), resulting in a final gas vesicle concentration equivalent to 3 GD500nm, and loaded into the phantom. The AI2O3 concentration was carefully selected to match the scattering echogenicity of the gas vesicle. This well was precisely centered at a depth of 9 mm.

[0059] This tissue-mimicking gas vesicle phantom will be used to demonstrate the efficiency of the nonlinear SVD filtering based on both space and pressure filtering of ultrafast data of the present disclosure.

[0060] The imaging apparatus 1 comprises an ultrasound probe 5 (PRB) and the processor 2 may be configured to control the ultrasound probe 5.

[0061] The ultrasound probe 5 may include an array 7 (ARR) of ultrasonic transducers. The array may be a linear array adapted to generate a 2D image of a slice of the region to be imaged, or a 2D array adapted to generate a 3D image of the region. When the array is a 2D array, it may be a sparse matrix of transducers, as known in the art. During imaging, the array of ultrasonic transducers is coupled to a patient body. The array of ultrasonic transducers transmits an ultrasonic signal into the tissue and receives backscattered signals or echoes from the region of the organ to be imagined. In the present disclosure, the array 7 is controlled by the processor 2 to transmit the ultrasonic signals according to specific pulse sequences as illustrated in Figure 3(a) or Figure 3(b).

[0062] The received signals are processed by the computing module 3 using the imaging method that is described herein in order to extract only the nonlinear (harmonic) signals reflected from the ultrasound contrast agents from the received signals. The proposed processing methods allow to enhance the contrast-to-tissue-Ratio (CTR), the signal-to- noise ratio (SNR), the contrast-to-noise ratio (CNR), to reconstruct an ultrasound image of the target region with a higher contrast and sensitivity.

[0063] Typical arrays of transducers may include a few tens to a few thousand of transducers. The array may also in some examples, be limited to one single transducer adapted to image only one line of the region, in the direction of the depth from the transducer or a few transducers adapted to image respectively lines of the region, in the direction of the depth from the transducer.

[0064] The following detailed description is done for the case of a linear or 2D array, so that the apparatus generates ultrasound measurements having pixels. In the case where the array would include just one transducer or a few transducers, the apparatus would generate an image limited to one line (ultrasound measurement) or a few lines in the direction of depth, the line(s) having pixels and the process would be similar except for the generation of the ultrasound measurements which would not require inclined planar waves of different angles of inclination.

[0065] During imaging, the transducer array 7 may be adapted to transmit and receive ultrasound waves having a central frequency comprised for instance between 0.5 and 100 MHz, for instance between 1 and 20 MHz.

[0066] In the embodiment of Figure 1 , the imaging processor 2 controls the ultrasound probe 5 via an interface 8 and a suitable link, which may comprise any suitable cable, typically connected at both ends, using suitable connectors. This interface exchanges signals between the imaging processor 2 and the ultrasound probe 5.

[0067] In the transmit path, the interface 8 transmits signals for controlling ultrasound wave transmission by active elements of the array 7. In particular, the interface 8 adjusts the amplitude and phases of the signals so that active elements of the array 7 emit an ultrasound plane wave toward the region. The interface may comprise a Digital to Analog Converter (DAC) (not shown) for converting digital signals produced by the processor 2 into analog signals toward the ultrasound probe 5. The ultrasound waves are typically transmitted as a sequence of pulses as illustrated in Figure 3(a) and Figure 3(b) at a predefined Pulse Repetition Frequency (PRF). In a typical imaging apparatus, the PRF is on the order of 5 kHz. Depending on the depth of the imagined region, the PRF may be lower or higher than 5 kHz.

[0068] In the receive path, the interface 8 receives backscattered signals from the ultrasound probe 5. The backscattered signals contain linear and nonlinear signals of the ultrasound waves reflected by the ultrasound contrast agents, the blood and the surrounding tissue as illustrated in Figure 4. In a similar manner, the interface comprises a sampler and a DAC for converting analog signals received from the probe 5 into a digital form. For example, the received signals are demodulated and sampled into In-Phase and Quadrature (IQ) signals of the respective transducers to produce an IQ image of the region. Thus, each IQ image comprises pixels representing the received backscattered signals from the region. In the case of a nonlinear ultrafast imaging of the present disclosure, a compound IQ image is formed from several IQ images obtained at different angles and a sequence of compound IQ images is obtained at different acoustic pressure amplitude. These several compound IQ images may be then processed by the computing module 3 to produce a final image of the region by separating nonlinear signals reflected by the ultrasound contrast agents flowing in the blood stream and signals from the tissue.

[0069] With reference to Figure 2, an embodiment of method of nonlinear ultrafast ultrasound imaging will now be described below.

[0070] The ultrasound contrast agent injection device 4 may be controlled by the processor 2 to inject S1 (UCAs NJ) intravenously ultrasound contrast agents to a region of the vascular network. The UCAs may be injected directly into an organ.

[0071] The ultrasound contrast agent injection may be a bolus injection. The ultrasound contrast agent injection may be a continuous injection. In the case gas vesicles, the UCAs may be produced directly by living tissues inside the subject.

[0072] The array 7 of transducers may be controlled by the processor 2 to acquire S2 (DATA_ACQ) the raw data of the region containing the ultrasound contrast agents.

[0073] The computing module 3 may process S3 (DATA_PROC) the acquired raw data of the region to generate an ultrasound contrast image of the region to be imaged.

[0074] In the example of Figure 2, the array 7 of transducers may be controlled by the processor 2 to acquire S2 raw data of the region containing the ultrasound contrast agents by transmitting planar ultrasonic waves according to a specific pulse sequence in the region to be imaged and by receiving the resulting backscattered ultrasonic waves, at a rate of for instance 5 kHz (Pulse Repetition Frequency PRF), i.e. every 0.2 ms. More generally, the Pulse Repetition Frequency PRF may be over 500Hz.The received signals may be registered as a set of raw data for each transmitted planar ultrasonic wave.

[0075] Figure 3 illustrates two examples of pulses sequence generated by the array of transducers 7.

[0076] Figure 3(a) illustrates a first example of positive pulses sequence. One pulse sequence comprises successive transmitted planar waves that may have propagation directions which are inclined of varying successive angles with regards to the direction of the depth z in the region to be imaged, i.e. with regards to the direction normal to the array 7. This pulse sequence is called BM pulse sequence. The direction of the plane waves is parallel to the imaging plan thus the direction of the 1 D array of transducers. M planar ultrasonic waves may be successively transmitted with different angles and at one acoustic pressure amplitude to form M images. M is an integer greater than 1. Then, the M images may be coherently added to synthesize an image of the region, which is thus a compound image. For instance, M may be 5 as illustrated in Figure 3(a) with angles taking the values -12 deg, -6 deg, 0 deg, +6 deg, +12 deg. In the case of M=5 and PRF=5 kHz, the rate of the compound images of the region (framerate) is thus 1 kHz. M may be different than 5, in which case the framerate of compound images is different. Each compound image corresponds to one acoustic pressure amplitude.

[0077] Instead of using the spatial dimension and the temporal dimension to decompose the signal, the present disclosure exploits the pressure dimension to separate the tissue signal and the ultrasound contrast agent signal. In order to decompose the acquired images with two spatial dimensions, for example (x, z) and one pressure dimension, nacompound images are obtained by varying the acoustic pressure amplitude. Therefore, the array 7 oftransducers may be controlled by the processor 2 to transmit napulse sequences by changing the acoustic pressure amplitude for each sequence, nabeing the number of acoustic pressure amplitude samples. Each of the naimages is associated with a pressure amplitude. The pressure step between two successive pressure amplitudes may be constant. For instance, namay be 4 as illustrated in Figure 6 and four compound images Im1 , Im2, Im3, Im4 are obtained. These four compound images form the IQ data that may be processed by the computing module S3 to generate a final image with a higher contrast.

[0078] Figure 3(b) illustrates a second example of pulses sequence comprising positive and negative pules. This sequence is called phase inversion (PI) pulse sequence. Both positive and corresponding negative pulses comprises successive transmitted planar waves at different angles as in the BM sequence. A number M of planar ultrasonic waves may be successively transmitted with different angles to form M IQ images. For instance, M may be 5 as illustrated in Figure 3(b) for positive and negative with angles taking the values -12 deg, -6 deg, 0 deg, +6 deg, +12 deg. In the case of M=5 and PRF=5 kHz, the rate of the compound images of the region (framerate) is thus 1 kHz. Therefore, five raw IQ images are obtained for positive and negative pulses respectively. As for the positive pulse of figure 3(a), the array 7 of transducers may be controlled by the processor 2 to transmit napulse sequences by changing the acoustic pressure amplitude for each positive and negative pulses sequence, nabeing the number of acoustic pressure amplitude samples. For instance, namay be 4 as illustrated in Figure 7. Consequently, four compound images Im1 , Im2, Im3, Im4 are obtained for positive and negative pulse sequence respectively. These eight compound images form the IQ image that may be processed by the computing module S3 to generate a final image with a high contrast sensitivity.

[0079] In both cases, the IQ image obtained after the step S2 can be described as the summation of three contributions: tissue signal ST, contrast agent signal SGV and random noise signal SnOise. The IQ image can be expressed as function of spatial and pressure variables:

[0080] [MATH 1]IQ(X, Z,p) ATissueX BTissue(p) X 57’jssue(x, z) + AQV x BGV(p) X SGy(x, Z)+ ANojsex SNoise(x,z)Where, x is the lateral dimension (along the transducers array), z is the depth in the medium in front of the ultrasonic probe, and p is the pressure, ATissue, AGV, and ANoiseare the scalars which represent the amplitude values of tissue, gas vesicle, and noise components, respectively, BGVand BTissueare the amplitude response of gas vesicles and tissue BTissueto pressure amplitude

[0081] Figure 4 represents the simulated results of these three contributions: spatial distributions of tissue signal, GV signal and noise signal are shown respectively in Figure 4(a), Figure 4(b) and Figure 4(c). The figure 4(d) shows the final composed image.

[0082] Figure 5 illustrates the simulated backscattering amplitude of noise signal (AN), gas vesicles signal (AGV), linear tissue signal (AL) and nonlinear tissue signal (ANL) with respect to various pressure amplitude. The gas vesicle shows a strong nonlinear behavior (curve AGV) whereas the tissue shows a linear behavior (curve AL and curve ANL). This result shows that tissue may exhibit a small nonlinear signal due to the nonlinear propagation of ultrasound waves within tissues. Consequently, the classical approach which consists in assuming that tissue signal is always linear does not allow the suppression of tissue signal and may reduce the contrast sensitivity.

[0083] Figure 6 illustrates in detail the processing step S3 of a pressure series of images of the region acquired with a BM pulse sequence.

[0084] After the step S2, a series of successive pressure compound images are obtained. These compound images form the IQ data that may be processed by the computing module S3 to generate a final image with a higher contrast.

[0085] Based on the successive pressure compound images of the region obtained after the step S2, filtered images of the vascular network in said region may then be computed S3 by the computing module 3. In the example of Figure 6, four successive compound images are used. For instance, namay be comprised between 2 and 12, for instance between 4 and 6. One example of usable number of successive compound images for one filtered image is 4.

[0086] The acquired in-phase / quadrature (IQ) data can be expressed as function of spatial and pressure variables IQ(x, z, a) , where x represents the lateral dimension along the transducers array, z represents the axial dimension, a represents the acoustic pressure amplitude. IQ(x, z, a) corresponds to set of nxx nzx nasamples where nx, nzand naare respectively the number of spatial samples along x direction, the number of spatial samples along z direction and the number of pressure samples.

[0087] The processing of IQ data comprises the following sub-steps:

[0088] The processing S3 comprises a step of reshaping of the IQ data under a Casorati matrix form by transforming pressure series data into a 2D space-pressure matrix form with dimension (nxx nz, na).

[0089] As illustrated in Figure 6, a stack 10 of four images with 2 spatial dimensions and one pressure dimension is reshaped in one spatio-pressure representation where all pixelsat one pressure point are arranged in one column and all pressure points for one pixel are arranged in one row.

[0090] The processing S3 comprises a step of singular vector decomposition of the reshaped IQ data IQ(x * z, a) into a weighted, ordered sum of separable spatio-pressure modes expressed by II, S, and V as below :

[0091] [Math. 2]Zt2usv tl where II and V are the corresponding matrices with respective dimensions (nzx nx, na) and (na, na) and S is the weighting matrix. II represent the spatial modes of the acquisition. V represent the pressure modes of the acquisition. Each mode has individual ranks. It is possible to select the number of singular vectors to be removed in each dimension separately, i.e. respectively in spatial and pressure dimensions. Thus, thanks to the SVD processing, the Casorati matrix is decomposed on pressure basis V and a spatial basis II.

[0092] Other decompositions into separable (or uncorrelated) spatio-pressure bases than SVD are possible. For example, it is possible to use a spatio-pressure independent component analysis (ICA) to decompose the reshaped IQ data into separable spatio- pressure bases.

[0093] The SVD decomposes the IQ data into a sum of images (each one corresponding to one II) independently modulated by the pressure signal V. In other words, every pixel from the image II behaves with the pressure fluctuations given by V.

[0094] The SVD decomposes the IQ dataset into three matrices, II is the spatial matrix, S is the weighting matrix, and V is the pressure matrix.

[0095] The processing S3 comprises a step of finding the threshold to remove singular values and vectors corresponding to tissue signal.

[0096] Based on the assumption that tissue signal is described in the first singular vectors and singular values as their high spatio-pressure coherence ensures that a large number of spatial pixels will exhibit the same pressure profile, a rejection is performed using a threshold selection on the number of singular vectors to be removed from the IQ data. On the contrary, ultrasound contrast agent signal should be found in lower singular values as they exhibit lower spatio-pressure coherence. By discarding some singular values of the SVD, the linear and nonlinear tissue signal may be filtered. For instance, the first singular values of the SVD, mostly reflecting the tissue signal of the region, may be discarded.

[0097] In one example, the threshold N may be manually tuned until the resulting image is acceptable. For example, it is possible to perform a plurality of examples and store a lookup table with predefined thresholds for a range of clinical conditions, i.e. brain, heart, etc.

[0098] In another example and with reference to figure 6, the threshold may be automatically determined by an adaptative SVD rejection, as described by Baranger et al. [Baranger, J., J. Aguet, and O. Villemain, Fast Thresholding of SVD Clutter Filter Using the Spatial Similarity Matrix and a Sum-Table Algorithm. IEEE Trans Ultrason Ferroelectr Freq Control, 2023. 70(8): p. 821-830], Thus, the spatial matrix II is used to perform the threshold selection to generate a spatial similarity matrix and a square-fitting matrix respectively, to determine the values of the threshold t1 and t2 used in SVD processing to remove the number of singular vectors from the IQ data. The spatial similarity matrix is obtained from the covariance of modules of the spatial singular vectors, II. Two square-like domains are demonstrated along the diagonal of the spatial similarity matrix, representing the tissue background and the contrast signal subspaces in our study. Then, a square-fitting matrix is converted from the spatial similarity matrix to determine the SVD threshold t1 and t2. Then determined threshold t1 and t2 are used to filter the weighting matrix, S, to generate a new weighting matrix, S2.

[0099] The original S matrix is replaced by the new matrix S2, so that the Casorati matrix ti USV' becomes a filtered new Casorati matrix ti US2 V' .

[0100] The processing comprises a step of reshaping of the filtered Casorati matrix S2 V to an image format g(x, z, o) . In the example of the Figure 6, this reshaping step gives four filtered compound images.

[0101] The processing S3 comprises a step of coherent summation of all filtered compound images at different acoustic pressure amplitude to generate a final image g(x, z) as below:

[0102] [Math. 3]

[0103] Where narepresents the number of acoustic pressure amplitude. The example of Figure 6, nais 4.

[0104] As illustrated in Figure 6, thanks to the nonlinear SVD filter beamforming, the final image compared to the initial acquired image reveals clearly the ultrasound contrast agents.

[0105] Figure 7 illustrates in detail another embodiment of processing step S3 of a pressure series of images of the region acquired with a PI pulse sequence.

[0106] In this embodiment, instead of emitting one single positive pulse as illustrated in Figure 6, the transducer array 7 is controlled to emit positive and negative polarities for various acoustic pressure amplitudes and the utilization of singular value decomposition is used in conjunction with the pulse inversion technique to enhance the contrast signals originating from UCAs.

[0107] As described above with reference to Figure 3(b), in the PI pulse sequence, both positive and corresponding negative pulses are emitted at different angles as in the BM pulse sequence. Each pulse is emitted at different acoustic pressure amplitude. After the step S2, a number of napositive compound images are obtained for positive pulse and form the positive IQ data IQ(x, z, a, pl) and a number of nanegative compound images are obtained for negative pulse and form the negative IQ data IQ(x, z, a, p2) , where x represents the lateral dimension, z represents the axial dimension, a represents the acoustic pressure amplitudes, pl and p2 represent the images acquired by transmitting the pulses with the same pressure amplitude but opposite phases for each acoustic pressure amplitude (pl = - p2).

[0108] In the example of Figure 7, the second pulse p2 is inverted in polarity with respect to the first one p1. Positive pulse p1 is emitted at four acoustic pressure amplitude and negative pulse p2 is emitted at four acoustic pressure amplitude. Consequently, four positive compound images Im1 , Im2, Im3, Im4 are obtained for positive and four negative compound images Im1*, Im2*, Im3*, Im4* are obtained for negative pulse sequence. These eight compound images form the IQ data that may be processed by the computing module S3 to generate a final image with a high contrast sensitivity.

[0109] The processing of IQ data comprises the following sub-steps:

[0110] The processing step comprises a step of rearrangement of the positive and negative IQ data by concatenation. The acquired in-phase / quadrature (IQ) data can be expressed as function of spatial and pressure variables as below:

[0111] [Math. 4]IQ(x, z, a, p) = IQ(x, z, a, pl) ffi IQ(x, z, a, p2)

[0112] where IQ x, z, a, p) represents the concatenated IQ data acquired at different acoustic pressure amplitudes, x represents the lateral dimension, z represents the axial dimension, a represents the acoustic pressure amplitudes, p represents the number of transmitted pulses (p = 2). pl and p2 represents the images acquired by transmitting the pulses with the same amplitude but opposite phases for each acoustic pressure amplitude(pl = - p2). ffi operator represents the concatenation of positive and negative frames acquired.

[0113] IQ(x, z, a, pl) corresponds to set of nxx nzx na,Pisamples where nx, nz, na,Piare respectively the number of spatial samples along x direction, the number of spatial samples along z direction, the number of pressure samples, p1 represents the images acquired by transmitting the positive pulses for each acoustic pressure amplitude.

[0114] IQ(x, z, a, p2) corresponds to set of nxx nzx na,P2 samples where nx, nz, na,P2 are respectively the number of spatial samples along x direction, the number of spatial samples along z direction, the number of pressure samples, p2 represents the images acquired by transmitting the negative pulses for each acoustic pressure amplitude

[0115] The processing comprises a step of reshaping of the IQ under a Casorati matrix form by transforming pressure series data into a 2D space-pressure marix form with dimension (nxx nz, na,pi+ na,P2) as below:

[0116] [Math. 5]IQ(x, z, a, pl) ffi IQ(x, z, a, p2) = IQ(x * z, a * pl) ffi / Q(x * z, a * p2)

[0117] As illustrated in Figure 7, a stack of eight images with 2 spatial dimensions and one pressure dimension is reshaped in one spatio-pressure representation 12 where all pixels at one pressure point are arranged in one column and all pressure points for one pixel are arranged in one row.

[0118] The processing comprises a step of decomposition of the reshaped IQ data IQ(x * z, a) into a weighted, ordered sum of separable spatio-pressure modes expressed by II, S, and V as below :

[0119] [Math. 6]where II and V are the corresponding matrices with respective dimensions (nzx nx, na,pi+ na,P2) and (na,pi+ na,P2 , na,pi+ na,P2) and S is the weighting matrix. II represent the spatial modes of the acquisition. V represent the pressure modes of the acquisition. Each mode has individual ranks. It is possible to select the number of singular vectors to be removed in each dimension separately, i.e. respectively in spatial and pressure dimensions. Thus, thanks to the SVD processing, the matrix is decomposed on pressure basis V and a spatial basis II.

[0120] The processing S3 comprises a step of finding the threshold to remove singular values and vectors corresponding to tissue signal.

[0121] As in the embodiment of Figure 6, the spatial matrix II is used to perform the threshold selection to generate a spatial similarity matrix and a square-fitting matrix respectively, to determine the values of the threshold t1 and t2 used in SVD processing to remove the number of singular vectors from the IQ data. The spatial similarity matrix is obtained from the covariance of modules of the spatial singular vectors, II. Two square-like domains are demonstrated along the diagonal of the spatial similarity matrix, representing the tissue background and the contrast signal subspaces in the present disclosure. Then, a square-fitting matrix is converted from the spatial similarity matrix to determine the SVD threshold t1 and t2. Then determined threshold t1 and t2 are used to filter the weighting matrix, S, to generate a new weighting matrix, S2.

[0122] The original S matrix is replaced by the new matrix S2, so that the Casorati matrix ti USV' becomes a filtered new Casorati matrix ti US2 V' .

[0123] The processing comprises a step of reshaping of the filtered Casorati matrix 7' to an image format g(x,z, a,p) . In the example of the Figure 7, this reshaping step gives eight filtered compound images.

[0124] The processing comprises a step of separation of the filtered positive and the corresponding negative frames from the concatenated dataset as below:

[0125] [Math. 7] g(x, z, a, p) = g(x, z, a, pl) ffi g(x, z, a, p2)

[0126] The processing comprises a step of coherent summation of all nonlinear SVD filtered images to generate the pulse inversion images at each acoustic pressure amplitude:

[0127] [Math. 8] g(x, z, a) = g(x, z, a, pl) + g(x, z, a, p2)

[0128] The processing comprises a step of coherent summation of all filtered compound images at different acoustic pressure amplitude to generate a final image g(x,z) as below:

[0129] [Math. 9]

[0130] Where narepresents the number of acoustic pressure amplitude.

[0131] As illustrated in Figure 7, thanks the nonlinear SVD filter, the final image compared to the initial acquired image reveals clearly the UCAs.

[0132] The efficiency of the nonlinear SVD decomposition based on both space and pressure filtering of ultrafast data will be demonstrated below with a phantom study.

[0133] Experimental results

[0134] Gas vesicles preparation

[0135] Nonlinear GVs are prepared following the methodology outlined in Rabut et al. [Rabut, C., et al., Ultrafast amplitude modulation for molecular and hemodynamic ultrasound imaging. Applied Physics Letters, 2021. 118(24)]. GVs are extracted from buoyant Anabaena flos-aquae cells through hypertonic lysis and subsequently purified by repeated centrifugally assisted flotation and resuspension. The outer GvpC layer of the gas vesicles was then removed through treatment with 6M urea, followed by additional round of centrifugally assisted flotation, dialysis in 1x PBS and resuspension to remove residual GvpC and urea.

[0136] In vitro-experiment

[0137] A tissue-mimicking gas vesicle phantom for imaging is fabricated by casting 1 % (w / v) agarose in PBS supplemented with 0.2% (w / v) AI2O3. For static imaging, a custom 3D- printed mold is employed to create a cylindrical well with a 2 mm diameter. Gas vesicles were incubated at 42 °C for 1 minute, mixed in a 1 : 1 ratio with low-melt molten agarose (at 42 °C), resulting in a final gas vesicle concentration equivalent to 3 GD500nm, and loaded into the phantom. The AI2O3 concentration is carefully selected to match the scattering echogenicity of the gas vesicle well. This well is precisely centered at a depth of 9 mm.

[0138] Ultrasound data acquisition

[0139] For image acquisition, ultrasound sequences are implemented and executed on a research ultrasound system (Verasonics, USA) utilizing a linear ultrasound transducer equipped with 128 elements, featuring a center frequency at 15.625 MHz and a 67% -6 dB bandwidth. All acquisition scripts and processing codes are developed in Matlab (Matworks, USA). The mechanical index is increased from 0.01 to 0.08 for the transmission pulses based on a predetermined calibration result. For all BM and PI pulse sequences, singlecycle plane waves were transmitted at a frequency of 9.6154 MHz, at a frame rate of 500 Hz.

[0140] As indicated above, in the BM pulse sequence, only positive pulses are transmitted at five different angles. In the PI pulse sequence, both positive and corresponding negativepulses are transmitted at the same five angles as in the BM pulse sequence. The angle range for both pulse sequences is 24° (from -12° to +12°).

[0141] Three parameters used in the evaluation of the image quality are contrast-to-tissue ratio (CTR), contrast-to-noise ratio (CNR), and signal-to-noise ratio (SNR), respectively, as follow:

[0142] [Math. 10]

[0143] [Math.11]

[0144] [Math. 12]

[0145] AContrastis the mean amplitude of signal within the contrast region ; ATissueis the mean amplitude of signal within the tissue region; rTissuerepresents the standard deviation of signal within the tissue region.

[0146] Figure 8 demonstrates the comparison between four methods to extract the GV signals. These methods are conventional B-mode imaging, conventional pulse inversion imaging, nonlinear SVD processing on B-mode imaging as illustrated in Figure 6, nonlinear SVD processing on pulse inversion imaging as illustrated in Figure 7 respectively. The results show that the nonlinear SVD processing on B-Mode data (called SVD-BM) and the nonlinear SVD processing on pulse inversion data (called SVD-PI) methods of the present disclosure outperform the conventional methods such as Bmode imaging, Pulse Inversion imaging or ultrafast AM imaging in terms of CTR, CNR, and SNR values. It is noted that SVD-PI surpasses SVD-BM in performance. This superiority can be attributed to two key factors. Firstly, in pulse inversion acquisition, the number of acquired images for each acoustic pressure amplitude is doubled compared to the BM-mode acquisition. Consequently, a greater pool of images is available for SVD processing, allowing for enhanced contrast and effective filtering of tissue background signals. Secondly, following SVD processing, pulse inversion acquisition presents an opportunity for further enhancement of image contrast and resolution. This is achieved through the coherent summation of corresponding positive and negative frames at each acoustic pressure amplitude.

[0147] The utilization of singular value decomposition in conjunction with the pulse inversion technique allows to augment the contrast signals originating from gas vesicles. It is anticipated that SVD-PI will enhance contrast detection, particularly in scenarios involving nonlinear tissue signals and the absence of flow, where the traditionally employed conventional SVD method may demonstrate reduced efficacy.

[0148] Instead of using the spatiotemporal information, the present disclosure proposes to replace the temporal dimension with the acoustic pressure amplitude dimension. This difference distinguishes the method of the present disclosure from previous clutter filtering means. The method of the present disclosure does not rely on temporal information, enabling the proposed imaging method to effectively capture slowly moving contrast signals within the confines of small vessels.

[0149] In addition, the method of the present disclosure requires the fewest number of pulses and transmissions when compared to the conventional method with higher-pressure levels. In the present disclosure, a total of four pressure levels are employed for data processing as show in Figure 6 and Figure 7. Firstly, the experimental results demonstrate the highest CTR value among the group. Secondly, it requires the fewest number of pulses and transmissions when compared to all the other higher-pressure levels.

[0150] Figures 9, 10 and 11 show the efficacy of nonlinear SVD beamforming through in vitro, ex vivo and in vivo experimental results by comparing the signal-to-background ratio (SNR) parameter between three imaging sequences: nonlinear SVD beamforming method of the present disclosure, conventional amplitude amplification metho using uAM pulse sequence and conventional amplitude amplification metho using xAM pulse sequence.

[0151] The ex-vivo imaging is performed on a chicken breast tissue model. The 3 OD 500 nm GV solution is injected at a depth of 5 mm within the tissue.

[0152] The in- vivo imaging is performed on one male rat. (7 weeks old, Janvier labs, France in compliance with European regulations and approved by the local ethics committee). Anesthesia is induced with 2% isoflurane and maintained using intraperitoneal medetomidine injection (80 pg / kg) . For imaging of GVs in the lower limb of the rat, the rat is placed in a prone position, with the ultrasound transducer positioned on the left lower limb tissue. Prior to imaging, GVs mixed with 42 °C 2% agarose solution for a final GV OD 500nm equal to 3 is injected at the depth of 4 mm in the left lower limb of the rat. In this experiment, the GVs are static. They have been injected into a non-vascular compartment of the rat low limb tissue and not in the vessel. This demonstrates in vivo the efficacy of nonlinear SVD beamforming method of the present disclosure in the case where the GVs are static.

[0153] The uAM pulse sequence comprises eight successive tilted plane waves with a transmission frequency of 15.625 MHz repeated three times with modulated amplitude: two pulses at half amplitude (achieved by silencing the odd and even elements of the transducer, respectively) and one pulse at full amplitude. These sets of modulated pulses are then reiterated eight times. For each repetition, the polarities of the successive plane waves were determined by the column of the Hadamard matrix of order eight.

[0154] In the xAM pulse sequence, the 128 elements of the probe is split into two subapertures, the first half sub-aperture (element 1-64) and the second sub-aperture (element 65-128). First, the first sub-aperture was used to transmit a tilted plane wave with a transmission frequency of 15.625 MHz at an angle of 19.5° with respect to the array. Then the second sub-aperture was used to transmit a symmetric plane wave at the same angle with respect to the array. Finally, the two previous two plane waves were transmitted simultaneously.

[0155] In the nonlinear SVD beamforming pulse sequence, single-cycle plane waves are transmitted at a frequency of 15.625 MHz, with 11 compounded angles for each pulse with an angle range from -10° to +10°. Then the pulses were repeated for different duty cycles. By varying the duty cycle, the transmit amplitude is modulated, enabling the acquisition of nonlinear responses at different pressure levels.

[0156] The number of pulses and transmissions between the three imaging sequences are compared in the following Table 1.

[0157] Figures 9, 10 and 11 show respectively vitro, in vivo and ex-vivo experimental results SBR between the three imaging sequences: nonlinear SVD beamforming method of the present disclosure, conventional amplitude amplification metho using uAM pulse sequence and conventional amplitude amplification metho using xAM pulse sequence.

[0158] The results show that the nonlinear SVD beamforming method exhibits a higher SNR compared to uAM and xAM methods when the number of pressure levels is equal to 4 and16. All estimated pressures are computed at 4 mm depth with 1.1 dB.MHz'1.crrr1ultrasonic absorption.

[0159] Figure 9 shows that / n vitro GV phantom results, although uAM image demonstrates a good contrast enhancement of GV signals, there ar nonlinear imaging artefacts present below the GV inclusion. The xAM image demonstrates better resolution and boundary detection. Figure 9(e) shows clearly that the nonlinear SVD beamforming method exhibits a higher SBR compared to uAM and xAM methods. In addition, when the number of pressure levels is equal to 4, the nonlinear SVD beamforming method requires 4.4 times less number of pulses to achieve such a performance. When a higher number of pressure levels (N=16) is used for nonlinear SVD beamforming to reach comparable numbers of transmit pulses, the resulting images are further optimized and provid higher contrast and more tissue background suppression compared to uAM and xAM.

[0160] Figure 10 shows that in ex-vivo imaging, the mean and standard deviation values ofSNR are 4.03 ± 1.12 dB, 10.10 ± 0.44 dB, 10.07 ± 0.96 dB, and 18.64 ± 2.72 dB, correspondingly.

[0161] Figure 11 shows that in vivo imaging, the mean and standard deviation values ofSNR are 5.43 ± 0.85 dB, 10.50 ± 0.94 dB, 11.05 ± 0.83 dB, and 18.39 ± 1.36 dB, correspondingly.

[0162] These experimental results validate the performance of nonlinear SVD beamforming method through in vitro, ex vivo and in vivo models, confirming the robustness against tissue nonlinearities and acoustic attenuation.

[0163] In addition, these results show that a total of four pressure levels are found to be sufficient to reach high quality images. When the number of transmit amplitudes of 16, similar to the number of transmit pulses compared to xAM and uAM, the SNR value of nonlinear SVD beamforming is +6dB higher compared to AM imaging method.

[0164] Instead of using the temporal coherence for tissue / blood discrimination in ultrafast ultrasound datasets based on spatio-temporal decomposition that relies on the coherence with respect to transmit amplitude, the present disclosure proposes a new nonlinear SVD beamforming method that does not require any assumption on temporal information; and therefore on the motion of contrast agents or GVs, enabling to capture contrast signal of slowly-moving and / or non-moving GVs either present in tissues or flowing in vessels. The nonlinear SVD beamforming method of the present disclosure allows to enhance contrast detection and suppress tissue background signals, even though the biological sample to be imaged involves strong nonlinear tissue signals and the absence of motion of GVs or microbubbles.

Claims

Claims

1. Method for imaging a region of an organ by detecting the nonlinear signal of ultrasound backscattered signal from ultrasound contrast agents (UCAs) present in the region, the method including:(a) transmitting a number of napulse sequences of ultrasound waves by an array ultrasound probe (5) in the direction of said region such as each pulse sequence comprising successive ultrasound waves emitted at one acoustic pressure amplitude, nabeing an integer larger than 2;(b) acquiring raw data from backscattered ultrasound waves by said array ultrasound probe(c) generating a number of nacompound images from said raw data, each compound image being associated with an acoustic pressure amplitude, said number naof compound images forming a set of in-phase / quadrature (IQ) data expressed as function of spatial variables and pressure variable;(d) shaping said IQ data into a 2D space-pressure amplitude matrix form;(e) computing the decomposition of said reshaped data into separable spatio-pressure bases ;(f) removing values and vectors corresponding to tissue signal in order to determine a filtered 2D space-pressure amplitude matrix based on a predetermined threshold ;(g) reshaping said filtered 2D space-pressure amplitude matrix into nafiltered images;(h) generating a final image from the number of nafiltered images.

2. Method according to claim 1 , wherein the step of computing of the decomposition of reshaped data uses a spatio-pressure singular value decomposition (SVD) to generate three matrices as a weighted and ordered sum of separable spatio-pressure matrices expressed by II, S and V, where II represents the spatial matrix, V represents the pressure amplitude matrix and S represents a weighting matrix.

3. Method according to claim 1 , wherein the step of computing of the decomposition of reshaped data uses a spatio-pressure independent component analysis (ICA).

4. Method according to any of claims 1 to 3, wherein each pulse sequence is composed of one single pulse having a number of M planar ultrasound waves emitted with identical phase at different angles to form M images, said step (c) including a synthesizing step of said M images to form a compound image of the region, M being an integer value greater than 1 .

5. Method according to any of claim 1 to 3, wherein each pulse sequence is composed of one single pulse having a number of M diverging ultrasound waves emitted with different center of curvature to form M images, said step (c) including a synthesizingstep of said M images to form a compound image of the region, M being an integer value greater than 1 .

6. Method according to any of claims 1 to 3, wherein each pulse sequence is composed of one positive pulse having a number of M planar ultrasound waves emitted with identical positive phase at different angles to form M positive images and one corresponding negative pulse having a number of M planar ultrasound waves emitted with identical negative phase at different angles to form M negative images, said step (c) including a synthesizing step of said M positive and said M negative images to form a positive compound image and a negative compound image of the region respectively.

7. Method according to any of claims 1 to 3, wherein each pulse sequence is composed of one positive pulse having a number of M diverging ultrasound waves emitted with identical positive phase at different center of curvature to form M positive images and one corresponding negative pulse having a number of M diverging ultrasound waves emitted with identical negative phase at different center curvature to form M negative images, said step (c) including a synthesizing step of said M positive and said M negative images to form a positive compound image and a negative compound image of the region respectively.

8. Method according to claim 6 or 7, further comprising a step of rearrangement of napositive compound images and nanegative compound images by concatenation to form a set of in-phase / quadrature (IQ) data to be reshaped in a Casorati matrix.

9. Method according to any of claims 1 to 8 and claim 2, wherein the threshold is determined automatically by an adaptative SVD rejection or determined manually by a user.

10. Metho according to claim 9, wherein the adaptative SVD rejection uses a spatial similarity matrix and a square-fitting matrix.

11. Method according to any of claims 1 to 10, wherein the number naof acoustic pressure amplitude is comprised between 2 and 12, and preferably between 4 and 6.

12. Method according to any of claims 1 to 11 , wherein the ultrasound contrast agents are gas vesicles.

13. Method according to any of claims 1 to 11 , wherein the ultrasound contrast agents are microbubbles.

14. Apparatus (NLI APP) for imaging a region of an organ by detecting the nonlinear signal of ultrasound backscattered signal from ultrasound contrast agents (UCAs) present in the region, said apparatus including at least an array ultrasonic probe (5) and a control processor (2) configure to control said probe (5) to :(a) transmit a number of napulse sequences of ultrasound waves by an array ultrasound probe in the direction of said region such as each pulse sequence comprising successive ultrasound waves emitted at one acoustic pressure amplitude, nabeing an integer larger than 2;(b) acquire raw data from backscattered ultrasound waves by said array ultrasound probe(c) generating a number naof compound images from said raw data, each compound image being associated with an acoustic pressure amplitude, said number of images forming a set of in-phase / quadrature (IQ) data expressed as function of spatial variables and pressure amplitude variable;(d) shape said IQ data into a 2D space-pressure amplitude matrix form;(e) compute the decomposition of said reshaped date into separable spatio-pressure amplitude bases;(f) remove values and vectors corresponding to tissue signal in order to determine a filtered 2D space-pressure amplitude matrix based on a predetermined threshold ;(g) reshape said filtered 2D space-pressure amplitude matrix into a nafiltered images;(h) generate a final image from said filtered images.

15. Apparatus according to claim 14, wherein the control processor (2) is configured to compute the decomposition of reshaped data by using a spatio-pressure singular value decomposition (SVD) to generate three matrices as a weighted and ordered sum of separable spatio-pressure matrices expressed by II, S and V, where II represents the spatial matrix, V represents the pressure amplitude matrix and S represents a weighting matrix.

Images