Ultrasonic image processing method and ultrasonic imaging device

Abstract

Ultrasonic image processing method and ultrasonic imaging device The present invention relates to an ultrasonic image processing method comprising at least: - decomposition (204) into singular values ​​of each image of at least one initial sequence of ultrasonic images, calculating spatial eigenvectors and / or temporal eigenvectors of each of the images; - partitioning (206) of said spatial eigenvectors and / or said temporal eigenvectors, distributing said spatial eigenvectors and / or said temporal eigenvectors into several groups of spatial eigenvectors and / or several groups of temporal eigenvectors on the basis respectively of spatial and / or temporal similarities; - calculation (208) of at least one final sequence of images by concatenation of said spatial eigenvectors and / or said temporal eigenvectors of one of the groups of vectors. Figure for the summary: Fig. 2

Description

Title of the invention: Ultrasonic image processing method and ultrasonic imaging device. Technical field

[0001] The present description relates to the field of ultrasound imaging, and more particularly to the processing of ultrasound images, i.e. to images obtained by an ultrasound imaging device. Previous technique

[0002] In the field of ultrasound imaging, or Doppler-based ultrasound imaging, the implementation of a singular value decomposition (SVD) of an ultrasound image sequence makes it possible to distinguish different types of echoes present in the images and, for example, to obtain images containing only the desired echoes. For example, in the medical field, the implementation of an SVD on an ultrasound image sequence can be used to distinguish signals reflected by tissues, those representative of blood flow, and those related to noise. It then becomes possible to isolate the signals related to blood flow from the other signals.

[0003] To achieve this distinction, the spatiotemporal eigenvectors from an SVD are ordered by decreasing energy and separated into several groups, or batches. For example, in the vascular imaging example, the spatiotemporal eigenvectors can be divided into three batches to distinguish tissues, blood flow, and noise, since tissues generate the most energetic signals, blood flow generates signals of intermediate energy, and finally, noise generates the lowest energy signals.

[0004] The distribution of these eigenvectors within different batches is achieved by first defining one or more energy thresholds corresponding to the values ​​separating the different batches. These thresholds can be determined arbitrarily or algorithmically. The choice of these thresholds is crucial to the quality of the final image sequence. In the example described above, only the average energy vectors relating to blood flow can be retained so that the final image sequence shows only this blood flow.

[0005] There are adaptive methods for determining optimal threshold values, for example by correlating spatial or temporal vectors with similar energy levels, or by analyzing the energy profile, as described in the paper by J. Baranger et al., “Adaptive Spatiotemporal SVD Clutter Filtering for Ultrafast Doppler Imaging Using Similarity of Spatial Singular Vectors”. IEEE Trans Med Imaging. 2018 Jul; 37(7). However, with such a method, if the eigenvectors belonging to the same set are separated, from an energetic point of view, by other eigenvectors from another set, it is not possible to group them correctly.

[0006] In the case of blood flow imaging, the classification and distribution of eigenvectors according to their energy are suitable when the blood flow signal is clearly distinct from that related to tissues, as is the case, for example, when this flow is that observed in an artery. However, in the case of extensive vascularization with slow flows, or heterogeneous vascularization including, for example, arteries and venules, the information related to tissues and flows may have comparable energies, and it is then no longer possible to distinguish the corresponding sets of eigenvectors.

[0007] Similar problems also arise in fields other than medical imaging, such as fluid detection by ultrasonic imaging applied to industry. Summary of the invention

[0008] There is a need to propose a method for processing ultrasonic images, as well as an ultrasonic imaging device implementing such a method, which does not present the disadvantages previously described.

[0009] One embodiment overcomes all or part of the drawbacks of known solutions and proposes an ultrasonic image processing method comprising at least: - decomposition into singular values ​​of each image of at least one initial sequence of ultrasonic images, calculating spatial eigenvectors and / or temporal eigenvectors of each of the images; - partitioning of said spatial eigenvectors and / or said temporal eigenvectors, distributing said spatial eigenvectors and / or said temporal eigenvectors into several groups of spatial eigenvectors and / or several groups of temporal eigenvectors on the basis of spatial and / or temporal similarities respectively; - calculation of at least one final sequence of images by concatenation of said spatial eigenvectors and / or said temporal eigenvectors of one of the groups of vectors.

[0010] According to a particular embodiment: - the initial sequence comprises Ni ultrasonic images, each comprising Np pixels; - the singular value decomposition calculates Ni unitary spatial eigenvectors of Np values ​​each, and Ni temporal eigenvectors of Ni values ​​each; with Ni and Np corresponding to non-nuisance positive integers.

[0011] According to a particular embodiment, the number of spatial eigenvector groups and / or the number of temporal eigenvector groups is between 2 and 4.

[0012] According to a particular embodiment, the spatial and / or temporal similarities are representative of correlations between pixel values ​​of the images of the initial sequence of ultrasonic images.

[0013] According to a particular embodiment, the ultrasonic image processing method further comprises, before the implementation of the decomposition into singular values, at least one acquisition of the initial sequence of ultrasonic images.

[0014] According to a particular embodiment, the initial sequence of ultrasound images corresponds to a sequence of vascular ultrasound images.

[0015] According to a particular embodiment, the number of spatial eigenvector groups and / or the number of temporal eigenvector groups is equal to 2 or 3 and such that the final image sequence corresponds to a blood flow image sequence, or the number of spatial eigenvector groups and / or the number of temporal eigenvector groups is equal to 3 and such that a first final image sequence corresponds to an image sequence of a first blood flow and a second final image sequence corresponds to an image sequence of a second blood flow faster than the first blood flow.

[0016] According to a particular embodiment, the ultrasonic images of the initial sequence and / or the final sequence are in two dimensions or in three dimensions.

[0017] According to a particular embodiment, the ultrasonic image processing method further comprises, after the calculation of the final image sequence by concatenation of the spatial eigenvectors and / or the temporal eigenvectors of one of the vector groups, the implementation of an algorithm for recognizing geometric shapes within the images of the final sequence.

[0018] According to a particular embodiment, the geometric shape recognition algorithm corresponds to a tubular shape recognition algorithm.

[0019] An ultrasonic imaging device is also proposed configured to implement an ultrasonic image processing method according to a particular embodiment.

[0020] According to a particular embodiment, the ultrasonic imaging device includes an ultrasonic transducer circuit configured to perform an acquisition of the initial sequence of ultrasonic images by pulsed Doppler imaging.

[0021] According to a particular embodiment, the ultrasound imaging device further comprises: - a data processing circuit configured to implement the ultrasonic image processing method; - a control circuit electrically coupled to the ultrasonic transducer circuit and the data processing circuit, configured to transmit ultrasonic emission control signals to the ultrasonic transducer circuit, to receive electrical measurement signals transmitted by the ultrasonic transducer circuit and to transmit electrical measurement signals to the data processing circuit. Brief description of the drawings

[0022] These features and advantages, as well as others, will be described in detail in the following description of particular embodiments, given by way of non-limiting example, in relation to the accompanying figures, among which:

[0023] - Figure 1 schematically represents an ultrasound imaging device according to a particular embodiment;

[0024] - Figure 2 represents, in the form of a schematic diagram, the steps of a method for processing ultrasonic images according to a particular embodiment;

[0025] - Figures 3 to 5 represent images obtained by implementing a method for processing ultrasonic images according to a particular embodiment;

[0026] - Figures 6 and 7 represent MIP images obtained by implementing a ultrasonic image processing method according to a particular embodiment. Description of the implementation methods

[0027] The same elements have been designated by the same reference numerals in the different figures. In particular, the structural and / or functional elements common to the different embodiments may have the same reference numerals and may have identical structural, dimensional and material properties.

[0028] For the sake of clarity, only the steps and elements necessary for understanding the described embodiments have been shown and are detailed. In particular, various components (ultrasonic transducer circuit, control circuit, processing circuit, etc.), various steps implemented (processing of acquired images, calculation of final images, details of the calculations performed, etc.), and the coding necessary for implementing the different steps of the process are not detailed. A person skilled in the art will be able to implement these components in detail from the functional description given here.

[0029] Unless otherwise specified, when referring to two interconnected elements, this means directly connected without intermediate elements other than conductors, and when referring to two elements connected or coupled (in English "coupled") to each other, it means that these two elements can be connected or linked through one or more other elements.

[0030] In the following description, when reference is made to absolute positional qualifiers, such as "front," "back," "top," "bottom," "left," "right," etc., or relative positional qualifiers, such as "above," "below," "superior," "inferior," etc., or to orientational qualifiers, such as "horizontal," "vertical," etc., unless otherwise specified, reference is made to the orientation of the figures in a normal operating position. However, these terms do not imply the actual position and orientation of the device during use.

[0031] Unless otherwise specified, the expressions "approximately", "roughly", and "in the order of" mean within 10%, preferably within 5%.

[0032] Similarly, unless otherwise indicated, the ranges of values ​​indicated include the limits of these ranges.

[0033] An ultrasonic imaging device 100 according to a particular embodiment is described below in relation to [Fig.1].

[0034] The device 100 comprises at least one ultrasonic transducer circuit 102, a control circuit 104, and a data processing circuit 106. In the schematic example shown in [Fig. 1], the control circuit 104 is electrically coupled to the ultrasonic transducer circuit 102 and is thus capable of transmitting signals to the ultrasonic transducer circuit 102, controlling the emission of ultrasound for the measurements to be performed, and also capable of receiving electrical measurement signals transmitted by the ultrasonic transducer circuit 102. The control circuit 104 is also coupled to the data processing circuit 106 in order, in particular, to transmit to the data processing circuit 106 the measurement results obtained from the ultrasonic transducer circuit 102.

[0035] The circuit 102 may include a plurality of ultrasonic transducers enabling the acquisition of at least one initial sequence of ultrasonic images. The ultrasonic transducers are, for example, arranged in a matrix or in another manner adapted to the captures to be performed. The number of ultrasonic transducers in the circuit 102 may depend on the dimensions of the capture surface of the device 100.

[0036] In the described embodiment, the device 100 corresponds to a medical imaging device intended for acquiring images of blood flow. Alternatively, the device 100 could be configured to perform other types of ultrasonic image acquisition, for example for flow detection in pipes or any other application, for example industrial or domestic.

[0037] The circuit 102 is configured here to perform image acquisition by pulsed wave Doppler imaging, or PWD. In a In this configuration, the transducer circuit 102 is configured to emit a series of ultrasonic pulses. The responses obtained in this configuration do not correspond to a change in the received Fonde frequency, but rather to a pseudo-Doppler effect, that is, a change in the time interval between the received echoes relative to the time interval between the emitted waves. In the example device 100 described, pulsed Doppler imaging is used to determine the spatial location of blood flow in biological tissues.

[0038] The ultrasonic transducer circuit 102 may, for example, include CMUT type transducers (“Capacitive Micromachined Ultrasonic Transducer” in English, or capacitive micro-machined ultrasonic transducer) or PMUT type transducers (“Piezoelectric Micromachined Ultrasonic Transducer” in English, or piezoelectric micro-machined ultrasonic transducer) or of another type.

[0039] In the example embodiment described here, the control circuit 104 can be configured to provide the ultrasonic transducer circuit 102 with electrical excitation signals causing the emission of ultrasonic waves by the ultrasonic transducer circuit 102, and to receive electrical response signals generated by the ultrasonic transducer circuit 102 under the effect of receiving reflected ultrasonic waves.

[0040] In the described embodiment, the data processing circuit 106 can be configured to analyze and process the electrical response signals received by the control circuit 104 and sent by the ultrasonic transducer circuit 102. The processing circuit 106 may include, for example, a microprocessor coupled with a memory to perform the processing of the received data.

[0041] The control circuit 104 and the data processing circuit 106 are configured to perform ultrasonic image processing by implementing the steps described below and represented in diagram form on [Fig.2].

[0042] The example of a processing method described below relates to the processing of vascular ultrasound images in order to isolate images of at least one blood flow from the biological tissues in which the blood flow occurs and from the noise. However, this method could be applied to other applications such as, for example, the detection of flow in industrial pipes or pipelines.

[0043] In the example described, the captured ultrasonic images of the initial sequence and the calculated images of the final sequence correspond to raster images, i.e., two-dimensional images. Alternatively, these images may correspond to three-dimensional, or 3D, images, for example, constructed from several two-dimensional images.

[0044] During a first step 202, the device 100 can acquire at least one initial sequence of ultrasonic images. For example, this acquisition This process may include sending an acquisition command, or a sequence of instructions, to the control circuit 104 for the successive acquisition of images, followed by the emission of a series of ultrasonic signals in the form of pulses by the transducer circuit 102, then the reception of the echoes by the transducer circuit 102, and the processing of the responses obtained to generate the initial sequence of ultrasonic images. The processing of the responses to generate the initial sequence of ultrasonic images is not described in detail, as this processing may involve various steps such as filtering, envelope detection, logarithmic compression, etc. The data relating to the acquired images is, for example, stored in a memory of the device 100, or in an external memory of the device 100 connected to the device 100 via a communication link.

[0045] In a subsequent step 204, the initial sequence of ultrasonic images is subjected to singular value decomposition, or SVD, calculating spatial eigenvectors and / or temporal eigenvectors for each of the images in the initial sequence. In the described embodiment, the singular value decomposition is implemented such that spatial eigenvectors and temporal eigenvectors are calculated for each of the ultrasonic images in the initial sequence.

[0046] According to a particular example, the initial sequence comprises Ni ultrasonic images, each containing Np pixels, and the singular value decomposition computes Ni unit spatial eigenvectors of Np values ​​each, and Ni temporal eigenvectors of Ni values ​​each, with Ni and Np corresponding to non-null positive integers. In such an example, the Ni eigenvalues ​​of the diagonalization of these vectors represent the energies associated with these vectors.

[0047] Such an SVD step may involve the implementation of an algorithm as described in the document "Spatiotemporal Clutter Filtering of Ultrafast Ultrasound Data Highly Increases Doppler and fUltrasound Sensitivity" by C. Demené et al., IEEE Transactions on Medical Imaging, vol.34, no. 11, November 2015.

[0048] In a subsequent step 206, a partitioning, or clustering, of the previously calculated spatial and / or temporal eigenvectors is implemented, distributing said vectors into several groups of spatial and / or temporal eigenvectors based respectively on spatial and / or temporal similarities. This partitioning allows the spatial and / or temporal eigenvectors to be separated and distributed into a defined number of sets, based on the spatial and / or temporal similarities between the values ​​of the vectors and not on their energy value. The spatial and / or temporal similarities may, in particular, be representative of correlations between pixel values ​​of images from the initial sequence of ultrasonic images.

[0049] According to one particular embodiment, only spatial eigenvectors can be used for implementing the partitioning, these spatial eigenvectors being distributed into groups of spatial eigenvectors according to spatial similarities between the vectors. According to a first variant, only temporal eigenvectors can be used for implementing the partitioning, these temporal eigenvectors being distributed into groups of temporal eigenvectors according to temporal similarities between the vectors.According to a second variant, spatial eigenvectors and temporal eigenvectors can be used for the implementation of partitioning, with spatial eigenvectors being distributed into groups of spatial eigenvectors according to spatial similarities between them, and temporal eigenvectors being distributed into groups of temporal eigenvectors according to temporal similarities between them.

[0050] In the described example of a processing method, which relates to the processing of vascular ultrasound images with the aim of isolating images of at least one blood flow from biological tissues and noise, partitioning can be implemented on the spatial eigenvectors so as to distribute them, for example:

[0051] - in two groups of spatial eigenvectors, one being representative of the tissues biological factors present in the images and the other being representative of blood flow in the veins and / or arteries present in the images, or

[0052] - in three groups of spatial eigenvectors, a first group being representative of the biological tissues present in the images, a second group being representative of the blood flow in the veins and / or arteries present in the images, and a third group being representative of the noise present in the images, or

[0053] - in three groups of spatial eigenvectors, a first group being representative of the biological tissues present in the images, a second group being representative of the rapid blood flows in the veins and / or arteries present in the images, and a third group being representative of the slow blood flows in the blood vessels present in the images.

[0054] By way of example, the partitioning step 206 can be implemented using a "clustering" function of a computer calculation software, such as the "Clusterdata" function of the Matlab software (registered trademark) marketed by the company Mathworks (registered trademark).

[0055] A step 208 of calculating at least one final sequence of images by concatenating the spatial eigenvectors and / or the temporal eigenvectors of one of the vector groups is then carried out. Considering the examples described above, this step 208 can output a final sequence of images representative of blood flow, or two final image sequences, one representative of rapid blood flow and the other representative of slow blood flow.

[0056] By way of example, Figures 3 to 5 represent final 3D images obtained by implementing the previously described process such that the partitioning is carried out by dividing the spatial eigenvectors into three groups. The first image shown in [Fig. 3] is calculated from the spatial eigenvectors that have been grouped into the first of the three groups, and in which the smallest blood vessels (with the lowest blood flow rates) are visible. The second image shown in [Fig. 4] is calculated from the spatial eigenvectors that have been grouped into the second of the three groups, and in which an arteriole with a high blood flow rate is visible. The third image shown in [Fig. 5] is calculated from the spatial eigenvectors that have been grouped into the third group, and in which the tissues are visible.It is possible to calculate only the image(s) of interest to the user, i.e., the first and second images in the example above.

[0057] By way of comparison, if a distribution of spatial eigenvectors were implemented based on the energy of these eigenvectors, those relating to the arteriole and those relating to smaller blood vessels would have been distributed in the same group of eigenvectors, which would have led to obtaining a single sequence of images in which the arteriole and the blood vessels would be visible on each of them.

[0058] According to another example, [Fig. 6] represents final 3D images obtained by implementing the previously described process, such that the partitioning is carried out by distributing the spatial eigenvectors into three groups. These images are of the "Maximum Intensity Projection" (MIP) type. Views a) to c) correspond to such images along a first axis, and views d) to f) correspond to the same data viewed along a second axis perpendicular to the first axis. The images in views a) and d) are calculated from the spatial eigenvectors that have been grouped into one of the three groups, and in which the vertical arterioles and venules, as well as a larger vessel at greater depth (~5 mm in this example), are visible. The images in views b) and e) are calculated from the spatial eigenvectors that have been grouped into a second of the three groups, and in which the superficial horizontal vessels are visible.The images in views c) and f) are calculated from the spatial eigenvectors that have been grouped within the third group, and in which the tissues are visible. It is possible to calculate only the image(s) of interest to the user, i.e., the first and second images in the example above.

[0059] By way of comparison, the use of a conventional method to determine an optimal threshold leads to a single image containing the two types of flow visible in views a), b), d) and e) of [Fig. 6]. However, in such a single image, it is more difficult to distinguish certain vessels, in particular the deep vessel and the thinnest vessels.

[0060] In the various examples above, it is possible that a final sequence of images retained at the end of the process corresponds to a sequence of blood flow images, or to two distinct image sequences, one representative of slow blood flow and the other representative of fast blood flow. Furthermore, compared to techniques based on eigenvector energies, eigenvector partitioning based on spatial and / or temporal similarities between eigenvectors allows for better isolation of information relating to biological tissues and noise from the desired information relating to blood flow.

[0061] In general, the number of spatial eigenvector groups and / or the number of temporal eigenvector groups is, for example, between 2 and 4. The number of eigenvector groups is chosen in particular according to the desired classification granularity for the eigenvectors determined by the SVD.

[0062] According to one example, the method may include implementing, after the computation step 208, a step 210 implementing a shape recognition algorithm within the images of the final sequence obtained, in order to automatically identify particular shapes. For example, it is possible to perform tubular structure recognition in order to locate one or more blood vessels and / or arterioles and / or arteries. For example, such recognition may include applying filters to the images obtained from the final sequence, for example, Frangi or Jerman filters when the recognition relates to tubular structures. For example, when a Jerman filter is applied, it is possible to define an index (for example, the proportion of pixels whose amplitude is greater than an arbitrary threshold) whose value allows the selection of pixel groups describing vessels.

[0063] Fig. 7 represents the same images as those previously described in relation to Fig. 6, but on which a Jerman filter is applied.

[0064] In the described method, the energies of the eigenvectors are not considered, and only the spatial and / or temporal similarities between the eigenvectors are evaluated, independently of their energy. This method makes it possible, for example, to reconstruct images of blood vessels without the presence of static elements such as artifacts, interface echoes, or even the movement of biological tissues.

[0065] Advantageously, the ultrasonic image processing method and the device 100 can be used for the purpose of detecting blood vessels, for example in the field of medicine or ultrasound.

[0066] Various embodiments and variations have been described. A person skilled in the art will understand that certain features of these various embodiments and variations could be combined, and other variations will become apparent to a person skilled in the art.

[0067] Finally, the practical implementation of the embodiments and variants described is within the reach of a person skilled in the art, based on the functional indications given above.

Claims

Demands

1. A method for processing ultrasonic images comprising at least: - decomposition (204) into singular values ​​of each image of at least one initial sequence of ultrasonic images, calculating spatial eigenvectors and / or temporal eigenvectors of each of the images; - partitioning (206) of said spatial eigenvectors and / or said temporal eigenvectors, distributing said spatial eigenvectors and / or said temporal eigenvectors into several groups of spatial eigenvectors and / or several groups of temporal eigenvectors on the basis of spatial and / or temporal similarities respectively; - calculation (208) of at least one final sequence of images by concatenation of said spatial eigenvectors and / or said temporal eigenvectors of one of the groups of vectors.

2. A method for processing ultrasonic images according to claim 1, wherein: - the initial sequence comprises Ni ultrasonic images each comprising Np pixels; - the singular value decomposition (204) calculates Ni unit spatial eigenvectors of Np values ​​each, and Ni temporal eigenvectors of Ni values ​​each; with Ni and Np corresponding to non-nuisance positive integers.

3. A method for processing ultrasonic images according to any one of the preceding claims, wherein the number of spatial eigenvector groups and / or the number of temporal eigenvector groups is between 2 and 4.

4. A method for processing ultrasonic images according to any one of the preceding claims, wherein the spatial and / or temporal similarities are representative of correlations between pixel values ​​of the images from the initial sequence of ultrasonic images.

5. Method for processing ultrasonic images according to any one of the preceding claims, further comprising, before implementation of the decomposition (204) into singular values, at least one acquisition (202) of the initial sequence of ultrasonic images.

6. Method for processing ultrasound images according to any one of the preceding claims, wherein the initial ultrasound image sequence corresponds to a vascular ultrasound image sequence.

7. An ultrasonic image processing method according to claim 6, wherein the number of spatial eigenvector groups and / or the number of temporal eigenvector groups is equal to 2 or 3 and such that the final image sequence corresponds to a blood flow image sequence, or wherein the number of spatial eigenvector groups and / or the number of temporal eigenvector groups is equal to 3 and such that a first final image sequence corresponds to an image sequence of a first blood flow and a second final image sequence corresponds to an image sequence of a second blood flow faster than the first blood flow.

8. A method for processing ultrasonic images according to any one of the preceding claims, wherein the ultrasonic images of the initial sequence and / or the final sequence are in two dimensions or in three dimensions.

9. Method for processing ultrasonic images according to any one of the preceding claims, further comprising, after the calculation (208) of the final image sequence by concatenation of the spatial eigenvectors and / or the temporal eigenvectors of one of the vector groups, the implementation of a recognition algorithm (210), within the images of the final sequence, of geometric shapes.

10. Ultrasonic image processing method according to claim 9, wherein the geometric shape recognition algorithm (210) corresponds to a tubular shape recognition algorithm.

11. Ultrasonic imaging device (100) configured to implement an ultrasonic image processing method according to any one of the preceding claims.

12. Ultrasonic imaging device (100) according to claim 11, comprising an ultrasonic transducer circuit (102) configured to perform an initial sequence acquisition of ultrasonic images by pulsed Doppler imaging.

13. Ultrasonic imaging device (100) according to claim 12, further comprising: - a data processing circuit (106) configured to implement the ultrasonic image processing method; - a control circuit (104) electrically coupled to the ultrasonic transducer circuit (102) and the data processing circuit (106), configured to transmit ultrasonic emission control signals to the ultrasonic transducer circuit (102), to receive electrical measurement signals transmitted by the ultrasonic transducer circuit (102) and to transmit the electrical measurement signals to the data processing circuit (106).

Images