Helix multi-point amplification sampling photoacoustic volume imaging method and system

Abstract

The invention discloses a spiral line multi-point amplification sampling photoacoustic volume imaging method and system, and belongs to the technical field of photoacoustic volume imagines.The method comprises the steps that S1, a signal amplification training dataset is obtained, specifically, an imaging area of biological tissue or a phantom sample is adjusted to be located in a set sound field range of a three-dimensional bowl-shaped ultrasonic array, and a signal amplification training dataset is obtained; performing three-dimensional photoacoustic volume imaging on the biological tissue or the phantom sample to obtain non-rotation scanning sparse data, a photoacoustic chordal graph before amplification and a photoacoustic chordal graph after amplification; and S2, constructing and training an interpolation network: constructing a U-Net interpolation network, taking the photoacoustic chordal graph before amplification as input, taking the photoacoustic chordal graph after amplification as a real label for training, carrying out error analysis training, and storing an optimal interpolation model with a minimum error. According to the invention, a deep learning virtual amplification sampling technology is adopted to replace traditional multi-rotation-angle physical rotation scanning, so that the imaging time is remarkably shortened, high-speed three-dimensional imaging under monopulse excitation is realized, and the method is suitable for real-time dynamic monitoring of scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photoacoustic volume imaging, in particular to a spiral multi-point amplification sampling photoacoustic volume imaging method and system. BACKGROUND

[0002] Photoacoustic computed tomography (PACT) is a hybrid imaging technique that combines the high contrast of optical imaging with the deep penetration capability of ultrasound imaging, and has a wide application prospect in the field of biomedicine. By detecting the ultrasonic signals generated by biological tissues under pulsed laser irradiation, PACT can achieve high-resolution imaging of tissue structure and functional information. With its comprehensive advantages in imaging resolution, penetration depth, optical contrast and non-invasive detection in vivo, PACT has developed rapidly in recent years.

[0003] Spherical array is one of the effective ways to realize single-excitation three-dimensional photoacoustic volume imaging. Traditional PACT systems often use large hemispherical ultrasonic transducer arrays (such as 1024-element channels) to obtain high-quality images. However, with the increase in the number of element channels, the system not only has a significant increase in manufacturing cost, but also has a sharp increase in the difficulty of synchronous acquisition of multi-element channel signals and image reconstruction, resulting in a decrease in time resolution. Therefore, how to maintain sufficient sampling density under the premise of reducing the actual number of transducers to suppress limited-angle artifacts and improve image quality and time resolution has become a key problem to be solved in current photoacoustic imaging technology.

[0004] Based on this, the present application designs a spiral multi-point amplification sampling photoacoustic volume imaging method and system to solve the above problems. SUMMARY

[0005] In view of the above-mentioned shortcomings of the prior art, the present application provides a spiral multi-point amplification sampling photoacoustic volume imaging method and system.

[0006] To achieve the above purpose, the present application is realized by the following technical scheme: A spiral multi-point amplification sampling photoacoustic volume imaging method, comprising the following steps: Step S1, obtaining signal amplification training data set: adjusting the imaging region of a biological tissue or a phantom sample to be located within the set acoustic field range of a three-dimensional bowl-shaped ultrasonic array, performing three-dimensional photoacoustic volume imaging on the biological tissue or the phantom sample to obtain non-rotationally scanned sparse data, a pre-amplification photoacoustic chord diagram and a post-amplification photoacoustic chord diagram; Step S2, constructing and training an interpolation network: constructing a U-Net interpolation network, taking the pre-amplification photoacoustic chord diagram as the input and the post-amplification photoacoustic chord diagram as the true label for training, performing error analysis training, and saving the optimal interpolation model with the minimum error; Step S3, non-rotation scanning fast sparse sampling: recombining the non-rotation scanning sparse data to obtain a non-rotation scanning sparse sampled photoacoustic chordogram; Step S4, optimal interpolation model for virtual rotation interpolation: inputting the non-rotation scanning sparse sampled photoacoustic chordogram into the optimal interpolation model to obtain a virtually rotation amplified photoacoustic chordogram through the virtual rotation interpolation of the optimal interpolation model; extracting the element channel data in the virtually rotation amplified photoacoustic chordogram according to the element channel sequence number defined by the pre-generated interpolation sequence to generate the photoacoustic data of the virtual rotation angle; Step S5, three-dimensional reconstruction: recombining the non-rotation scanning sparse sampled photoacoustic chordogram and the photoacoustic data corresponding to the multiple virtual rotation angles into a photoacoustic data matrix of coordinate data according to the three-dimensional space coordinates thereof to obtain a high-resolution three-dimensional photoacoustic image equivalent to real dense sampling through a three-dimensional reconstruction algorithm.

[0007] Further, the specific operation of step S1 is as follows: S11: align the central axis of the three-dimensional bowl-shaped ultrasonic array arranged according to the Fibonacci Fermat double helix with the center of the imaging region of the biological tissue or the phantom sample, adjust the relative distance between the three-dimensional bowl-shaped ultrasonic array and the biological tissue or the phantom sample, and make the imaging region located in the set sound field range of the three-dimensional bowl-shaped ultrasonic array; S12: perform three-dimensional photoacoustic volume imaging on the biological tissue or the phantom sample to obtain a sparse photoacoustic signal data matrix and three-dimensional space coordinates corresponding to each ultrasonic element as non-rotation scanning sparse data; S13: perform multi-angle three-dimensional photoacoustic volume imaging on the same biological tissue or the phantom sample through axial rotation to obtain a photoacoustic signal data matrix corresponding to each rotation angle and three-dimensional space coordinates corresponding to each ultrasonic element as interpolation data of the rotation angle; S14: recombine the non-rotation scanning sparse data according to the rotation angle and the element channel sequence number corresponding to the pre-generated interpolation sequence to obtain a pre-amplified photoacoustic chordogram; extract the photoacoustic data of the corresponding element channel from the interpolation data of the rotation angle to interpolate and amplify the pre-amplified photoacoustic chordogram to obtain an amplified photoacoustic chordogram.

[0008] Further, in the Fibonacci Fermat double helix arrangement, the three-dimensional space coordinates corresponding to each ultrasonic element are The calculation is as follows: In the formula, R is the spherical radius of the three-dimensional bowl-shaped ultrasonic array, N is the total number of element channels, n is the element channel number , is the array coverage rotation angle, The golden ratio constant is , The first array element plane projection radius.

[0009] Further, the arrangement of the pre-generated interpolation sequence is determined by the corresponding array element channel position under the rotation angle; Specifically, any array element channel in the spiral interval is selected, which is rotated around the depth direction axis, the intersection point is obtained by the rotation track intersecting with the spiral line, the rotation angle is calculated, and the array element channels before and after rotation are staggered on the spiral line under the rotation angle, which is the arrangement of the pre-generated interpolation sequence.

[0010] Further, the U-Net interpolation network adopts an asymmetric convolution kernel structure, and the U-Net interpolation network performs down-sampling and up-sampling in the time dimension.

[0011] Further, the virtual rotation interpolation specifically operates: the virtual rotation array element channel is inserted between the non-rotating array element channels, and the position of the virtual rotation array element channel is consistent with that of the real rotation array element channel.

[0012] An imaging system utilizes a spiral multi-point amplification sampling photoacoustic volumetric imaging method, comprising a pulsed laser emission module, a rotation control module, a photoacoustic volumetric imaging probe, and a computer data processing module. The pulsed laser emission module comprises a pulsed laser and a diffuser sheet, and the output end of the pulsed laser is connected with the diffuser sheet. The rotation control module comprises a logic control circuit, a voltage stabilizer, a comparator, and a rotating motor. The photoacoustic volumetric imaging probe comprises a probe shell, a three-dimensional bowl-shaped ultrasonic transducer, a rotating connector, a bearing slip ring, an optical fiber bundle, and a diffusion lens. The three-dimensional bowl-shaped ultrasonic transducer is installed in the bowl-shaped cavity of the probe shell in a three-dimensional bowl-shaped ultrasonic array arranged according to the Fibonacci-Fermat double spiral. The center of the probe shell is provided with a one-inch internal threaded column channel, and the optical fiber bundle and the diffusion lens are fixed in the channel by an external threaded iron ring. The diffusion lens is located at the front end of the optical fiber bundle. The probe shell is connected with the rotating motor through the rotating connector, the logic control circuit is electrically connected with the rotating motor, and the rotation motor is controlled to rotate axially according to the rotation angle by the output driving signal of the logic control circuit. The computer data processing module comprises a computer, a multi-element channel data collector, the multi-element channel data collector is electrically connected with the three-dimensional bowl-shaped ultrasonic transducer, the multi-element channel data collector, the voltage stabilizer and the three-dimensional bowl-shaped ultrasonic transducer are connected with the computer, the logic control circuit is connected with the comparator, the comparator is connected with the pulsed laser, and the voltage stabilizer is connected with the logic control circuit and the rotating motor.

[0013] Further, the three-dimensional bowl-shaped ultrasonic transducer has a center frequency of 5 MHz.

[0014] Further, the fiber bundle has a core diameter of 200 mu m.

[0015] Further, the rotating control module further comprises an optical encoder coaxially installed with the rotating motor, which is used for feeding back the actual rotating angle in real time, and controlling the stepping error within ±0.05° through PID closed-loop control.

[0016] The three-dimensional bowl-shaped ultrasonic transducer is connected with the rotating motor through a rotating connector, and the motor is controlled to rotate axially according to the rotating angle through the driving signal output by the logic control circuit.

[0017] Compared with the prior art, the present application has the following advantages and beneficial effects: 1. The present application uses deep learning virtual augmented sampling technology to replace the traditional multi-rotating-angle physical rotation scanning, significantly shortens the imaging time, realizes high-speed three-dimensional imaging under single-pulse excitation, and is suitable for real-time dynamic monitoring scenes.

[0018] 2. The present application uses Fibonacci-Fermat double helix arrangement to optimize the sound field coverage, combines U-Net network for virtual interpolation, restores dense sampling features on the basis of sparse sampling, outputs high-resolution three-dimensional images, and improves spatial resolution and imaging accuracy.

[0019] 3. The system of the present application integrates hardware optimization and closed-loop control, such as optical encoders to ensure rotation accuracy and bearing slip rings to prevent fiber damage, which improves the stability and reliability of the system, while reducing mechanical loss and maintenance cost.

[0020] 4. The method and system of the present application have high flexibility and can adapt to complex imaging environments, reduce the dependence on hardware rotation through virtual rotation interpolation, and widen the application range of photoacoustic imaging in the fields of biomedicine and the like. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0022] Figure 1 Flow chart of spiral multi-point augmented sampling photoacoustic volumetric imaging method based on virtual rotation; Figure 2 Interpolation point axial rotation and spiral intersection schematic diagram; Figure 3 Rotation angle and spiral multi-point interpolation schematic diagram; Figure 4 Element arrangement diagram after superposition of multi-spiral direction interpolation; Figure 5 Photoacoustic probe structure diagram of spiral multi-point augmented sampling photoacoustic volumetric imaging method based on virtual rotation; Figure 6 Network interpolation result comparison diagram. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort belong to the scope of protection of the present application.

[0024] In the following description, "left", "right", "front", "back", "up" and "down" are oriented in the perspective direction of the front view.

[0025] Embodiment one: in some embodiments, referring to the drawings of the specification Figure 1 A spiral multi-point augmented sampling photoacoustic volumetric imaging system, comprising the following steps: Step S1, acquiring signal augmentation training data set: adjusting the imaging area of the biological tissue or phantom sample to be located within the set acoustic field range of the three-dimensional bowl-shaped ultrasonic array, performing three-dimensional photoacoustic volumetric imaging on the biological tissue or phantom sample to obtain non-rotation scanning sparse data, pre-augmentation photoacoustic chord diagram and post-augmentation photoacoustic chord diagram; Step S2, constructing and training interpolation network: constructing a U-Net interpolation network, taking the pre-augmentation photoacoustic chord diagram as the input and the post-augmentation photoacoustic chord diagram as the true label for training, performing error analysis training, and saving the optimal interpolation model with the minimum error; Step S3, fast sparse sampling without rotation scanning: recombining the sparse data without rotation scanning to obtain the sparse photoacoustic chordogram without rotation scanning; Step S4, virtual rotation interpolation by optimal interpolation model: inputting the sparse photoacoustic chordogram without rotation scanning into the optimal interpolation model to obtain the virtual rotation amplified photoacoustic chordogram by virtual rotation interpolation of the optimal interpolation model; extracting the element channel data in the virtual rotation amplified photoacoustic chordogram according to the element channel sequence number defined by the pre-generated interpolation sequence to generate the photoacoustic data of the virtual rotation angle; Step S5, three-dimensional reconstruction: recombining the sparse photoacoustic chordogram without rotation scanning and the photoacoustic data corresponding to multiple virtual rotation angles into a photoacoustic data matrix of coordinate data according to their three-dimensional space coordinates, and obtaining a high-resolution three-dimensional photoacoustic image equivalent to real dense sampling by a three-dimensional reconstruction algorithm.

[0026] The specific operation of step S1 is as follows: S11: align the central axis of the three-dimensional bowl-shaped ultrasonic array arranged according to the Fibonacci Fermat double helix with the center of the imaging region of the biological tissue or the phantom sample, adjust the relative distance between the three-dimensional bowl-shaped ultrasonic array and the biological tissue or the phantom sample, so that the imaging region is located within the set sound field range of the three-dimensional bowl-shaped ultrasonic array; S12: perform three-dimensional photoacoustic volume imaging on the biological tissue or the phantom sample to obtain a sparse photoacoustic signal data matrix and three-dimensional space coordinates corresponding to each ultrasonic element as sparse data without rotation scanning; S13: perform multi-angle three-dimensional photoacoustic volume imaging on the same biological tissue or the phantom sample by axial rotation to obtain a photoacoustic signal data matrix corresponding to each rotation angle and three-dimensional space coordinates corresponding to each ultrasonic element as interpolation data of the rotation angle; S14: recombine the sparse data without rotation scanning according to the rotation angle and the element channel sequence number corresponding to the pre-generated interpolation sequence to obtain a photoacoustic chordogram before amplification; extract the photoacoustic data of the corresponding element channel from the interpolation data of the rotation angle to interpolate and amplify the photoacoustic chordogram before amplification to obtain an amplified photoacoustic chordogram.

[0027] In the Fibonacci Fermat double helix arrangement, the three-dimensional space coordinates corresponding to each ultrasonic element The calculation is as follows: In the formula, is the spherical radius of the three-dimensional bowl-shaped ultrasonic array, is the total number of element channels, is the element channel number , is the array coverage rotation angle, is the golden ratio constant, and the value is , is the first element plane projection radius. The element channels are numbered from inside to outside according to the radius of the projection on the plane, and the number is the element channel number; by extracting elements at the number interval in the three-dimensional bowl-shaped ultrasonic array, a spiral line arrangement in different directions is formed, and the three-dimensional bowl-shaped ultrasonic array data is reorganized into a two-dimensional photoacoustic chord diagram, wherein the horizontal axis of the two-dimensional photoacoustic chord diagram is the sampling channel , and the vertical axis is the sampling time .

[0028] The arrangement mode of the pre-generated interpolation sequence is determined by the corresponding element channel position under the rotation angle; Specifically, any element channel in the spiral interval is selected, which is rotated around the depth direction axis, the intersection of the rotation track and the spiral line is obtained, the rotation angle is calculated, and the element channels before and after rotation are staggered on the spiral line under the rotation angle, that is, the arrangement mode of the pre-generated interpolation sequence.

[0029] The U-Net interpolation network adopts an asymmetric convolution kernel structure, and the U-Net interpolation network performs downsampling and upsampling in the time dimension.

[0030] The asymmetric convolution kernel structure is (3, 9); The asymmetric convolution kernel structure is adapted to different features of photoacoustic signals in the time dimension and the element channel dimension; Downsampling and upsampling are performed in the time dimension to reduce the feature map size, expand the receptive field, and capture macro features, while the integrity is maintained in the element channel dimension.

[0031] The specific operation of the asymmetric convolution kernel structure is as follows: The training set contains 500 groups of data in the signal amplification training data set, the mean square error is used as the loss function, the Adam optimizer is used for training, the learning rate is 0.001, and the optimal model with the minimum error is saved after 300 iterations.

[0032] The specific operation of step S3 is as follows: The three-dimensional bowl-shaped ultrasonic array of step S1 is used for single pulse excitation scanning in a non-rotating state, a sparse three-dimensional photoacoustic signal data matrix is obtained, and a non-rotating scanning sparse sampling photoacoustic chord diagram is reorganized.

[0033] This process shortens the data acquisition time from several minutes of traditional multi-rotation angle scanning to seconds; The specific operation of the virtual rotation interpolation is that the virtual rotation element channels are inserted between the non-rotating element channels, and the positions of the virtual rotation element channels are consistent with the positions of the real rotation element channels.

[0034] The position of the virtual rotating element channel is obtained by superimposing a rotation matrix on the element channel position, and the photoacoustic data is not actually collected but virtually generated by the U-Net interpolation network.

[0035] Thus, under the constraint of the real element channel data, the U-Net interpolation network is trained to output a signal that restores the sampling characteristics.

[0036] An imaging system comprises a pulsed laser emission module, a rotation control module, a photoacoustic volumetric imaging probe, and a computer data processing module. The pulsed laser emission module comprises a pulsed laser and a diffuser sheet, and the output end of the pulsed laser is connected to the diffuser sheet. The rotation control module comprises a logic control circuit, a voltage stabilizer, a comparator, and a rotating motor. The photoacoustic volumetric imaging probe comprises a probe shell, a three-dimensional bowl-shaped ultrasonic transducer 1, a rotating connector, a bearing slip ring, a fiber bundle 2, and a diffusion lens 3. The three-dimensional bowl-shaped ultrasonic transducer 1 is installed in the bowl-shaped cavity of the probe shell in a three-dimensional bowl-shaped ultrasonic array arranged in a Fibonacci-Fermat double helix. The center of the probe shell is provided with a one-inch internal threaded column channel, and the fiber bundle 2 and the diffusion lens 3 are fixed in the channel by an external threaded iron ring. The diffusion lens 3 is located at the front end of the fiber bundle 2. The probe shell is connected to the rotating motor through the rotating connector, and the logic control circuit is electrically connected to the rotating motor, and the rotation angle of the rotating motor is controlled by the output driving signal of the logic control circuit. The computer data processing module comprises a computer and a multi-element channel data collector, the multi-element channel data collector is electrically connected to the three-dimensional bowl-shaped ultrasonic transducer, the multi-element channel data collector, the voltage stabilizer, and the three-dimensional bowl-shaped ultrasonic transducer 1 are all connected to the computer, the logic control circuit is connected to the comparator, the comparator is connected to the pulsed laser, and the voltage stabilizer is connected to the logic control circuit and the rotating motor.

[0037] The logic control circuit is responsible for outputting the start-stop and rotation pulse signals of the rotating motor, and the comparator is used to control the laser pulse of the pulsed laser; the voltage stabilizer controls the logic control circuit, the voltage stabilizer, and the rotating motor to output a set voltage.

[0038] The laser emitted by the pulsed laser passes through the diffuser sheet and is converted from a collimated Gaussian beam into a flat-top beam with relatively uniform energy distribution.

[0039] Thus, the energy distribution of the laser transmitted in the fiber bundle 2 is more uniform.

[0040] The three-dimensional bowl-shaped ultrasonic transducer 1 has a central frequency of 5 MHz; The fiber bundle 2 has a core diameter of 200 μm; The fiber bundle 2 is embedded and fixed in the bearing slip ring to prevent rigid distortion damage caused by excessive rotation; The diffusion lens 3 is fixed at the front end of the fiber bundle 2 and used for large-area excitation of the imaging area.

[0041] The rotation control module further comprises an optical encoder coaxially installed with the rotating motor, which is used for real-time feedback of the actual rotation angle and controls the stepping error within ±0.05° through PID closed-loop control.

[0042] Specific case: as shown in Figures 2-4 Firstly, the central axis of the three-dimensional bowl-shaped ultrasonic array arranged according to the Fibonacci-Fermat double helix is aligned with the center of the imaging area of the phantom blood vessel sample, the bowl-shaped array of the three-dimensional bowl-shaped ultrasonic array has a bowl-shaped array spherical radius R=50 mm, a total number of array elements N=256 and an array coverage angle θ=130°, the relative distance between the array and the phantom blood vessel sample is adjusted to 10 mm, and it is ensured that the imaging area of the phantom blood vessel sample is located within the best sound field range of the array. Three-dimensional photoacoustic volume imaging is performed on the phantom blood vessel sample, and a sparse three-dimensional photoacoustic signal data matrix and three-dimensional spatial coordinates corresponding to each channel ultrasonic element are obtained as original sparse data.

[0043] Subsequently, the intersection points of the helix lines of the helix and the rotation tracks of each channel in the interval are selected by 13-element intervals, and the required rotation angles 304.5°, 110.7°, 55.4°, etc. are calculated in reverse, multi-angle axial rotation scanning is performed on the same sample, and the photoacoustic signal data after rotation is obtained as interpolation data. Based on the positions of the elements in the interval, the interpolation sequence of each channel is determined, the original sparse data is recombined, the photoacoustic chord diagram before expansion is obtained, 13 different photoacoustic chord diagrams before expansion with a size of 19 channels x 1024 time sampling points are obtained; and the corresponding channel data is extracted from the rotation interpolation data, interpolation expansion is performed, the photoacoustic chord diagram after expansion is generated, and 13 different photoacoustic chord diagrams after expansion with a size of 73 channels x 1024 time sampling points are obtained, which are used for U-Net interpolation network training.

[0044] ​With the adoption of U-Net as the network framework, the input is the photoacoustic chordogram before amplification, and the output target is the photoacoustic chordogram after amplification. U-Net uses an asymmetric convolution kernel (3, 9) as the network to adapt to the case that the time dimension of the photoacoustic chordogram is much larger than the channel dimension, and expand the time receptive field; at the same time, only the downsampling and upsampling are performed on the time dimension to keep the channel dimension complete. The training set contains 500 groups of data, the mean square error is used as the loss function, the Adam optimizer (learning rate 0.001) is used for training, and the optimal model with the minimum error is saved after 300 iterations.

[0045] Then, non-rotation scanning fast sparse sampling is performed. For the new imaging area, the three-dimensional bowl-shaped ultrasonic array is directly used for single-pulse excitation scanning in a non-rotation state to obtain a sparse three-dimensional photoacoustic signal data matrix, and the sparse sampling photoacoustic chordogram is recombined.

[0046] Optimal interpolation model is used for virtual rotation interpolation. The sparse sampling photoacoustic chordogram obtained in step S03 is input into the trained optimal interpolation model to infer and generate amplified photoacoustic chordograms with virtual rotation angles of 304.5°, 110.7°, 55.4°, etc. Subsequently, according to the angle and channel sequence of the pre-generated interpolation sequence, the photoacoustic data corresponding to the virtual rotation angle is reversely generated.

[0047] The photoacoustic data corresponding to the multiple virtual rotation angles of 304.5°, 110.7°, 55.4°, etc. are recombined into a coordinate-data matrix according to their three-dimensional space coordinates, and three-dimensional reconstruction is performed through a filtered back-projection algorithm to output a high-resolution three-dimensional photoacoustic image.

[0048] In order to verify the effectiveness of the present application, network interpolation test is performed on a phantom blood vessel sample. As shown in Figure 6 the network interpolation result comparison shows that the present application is highly consistent with the real dense sampling data. The present application completes imaging under single-pulse excitation, and the time resolution is improved to seconds, which is suitable for live dynamic monitoring scenarios.

[0049] The present application significantly improves the imaging speed and resolution by replacing physical rotation with virtual amplification sampling, and is suitable for live dynamic monitoring scenarios.

[0050] The spiral multi-point amplification sampling photoacoustic volume imaging probe structure based on virtual rotation is as follows: a plurality of three-dimensional bowl-shaped ultrasonic transducers 1 are fixedly installed in the hemispherical groove of the shell 4 according to the Fibonacci-Fermat double spiral arrangement; the optical fiber bundle 2 and the diffusion lens 3 are installed in the mounting hole of the shell 4, and the diffusion lens 3 is located between the optical fiber bundle 2 and the three-dimensional bowl-shaped ultrasonic transducer 1.

[0051] The above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A spiral line multi-point extended sampling photoacoustic volumetric imaging method, characterized in that: The method comprises the following steps: Step S1, obtaining signal expansion training data set: adjusting the imaging area of the biological tissue or phantom sample to be within the set sound field range of the three-dimensional bowl-shaped ultrasonic array, performing three-dimensional photoacoustic volume imaging on the biological tissue or phantom sample to obtain non-rotation scanning sparse data, pre-expansion photoacoustic chord diagram and post-expansion photoacoustic chord diagram; Step S2, constructing and training the interpolation network: constructing a U-Net interpolation network, taking the pre-expansion photoacoustic chord diagram as the input and the post-expansion photoacoustic chord diagram as the true label for training, performing error analysis training, and saving the optimal interpolation model with the minimum error; Step S3, non-rotation scanning fast sparse sampling: recombining the non-rotation scanning sparse data to obtain the photoacoustic chord diagram of non-rotation scanning sparse sampling; Step S4, optimal interpolation model for virtual rotation interpolation: inputting the photoacoustic chord diagram of non-rotation scanning sparse sampling into the optimal interpolation model to obtain the virtual rotation post-expansion photoacoustic chord diagram through the virtual rotation interpolation of the optimal interpolation model; extracting the element channel data in the virtual rotation post-expansion photoacoustic chord diagram according to the element channel sequence number defined by the pre-generated interpolation sequence to generate the photoacoustic data of the virtual rotation angle; Step S5, three-dimensional reconstruction: recombining the photoacoustic chord diagram of non-rotation scanning sparse sampling and the photoacoustic data corresponding to multiple virtual rotation angles into a photoacoustic data matrix of coordinate data according to the three-dimensional space coordinates, and obtaining a high-resolution three-dimensional photoacoustic image equivalent to true dense sampling through a three-dimensional reconstruction algorithm.

2. The spiral multi-point extended sampling photoacoustic volumetric imaging method of claim 1, wherein, The specific operation of step S1 is as follows: S11: align the central axis of the three-dimensional bowl-shaped ultrasonic array arranged according to the Fibonacci Fermat double helix with the center of the imaging area of the biological tissue or phantom sample, and adjust the relative distance between the three-dimensional bowl-shaped ultrasonic array and the biological tissue or phantom sample so that the imaging area is within the set sound field range of the three-dimensional bowl-shaped ultrasonic array; S12: perform three-dimensional photoacoustic volume imaging on the biological tissue or phantom sample to obtain a sparse photoacoustic signal data matrix and three-dimensional space coordinates corresponding to each ultrasonic element as non-rotation scanning sparse data; S13: perform multi-angle three-dimensional photoacoustic volume imaging on the same biological tissue or phantom sample through axial rotation to obtain photoacoustic signal data matrices corresponding to each rotation angle and three-dimensional space coordinates corresponding to each ultrasonic element as interpolation data of the rotation angle; S14: recombine the non-rotation scanning sparse data according to the rotation angle and the element channel sequence number corresponding to the pre-generated interpolation sequence to obtain the pre-expansion photoacoustic chord diagram; extract the photoacoustic data of the corresponding element channel from the interpolation data of the rotation angle to interpolate and expand the pre-expansion photoacoustic chord diagram to obtain the post-expansion photoacoustic chord diagram.

3. The spiral multi-point extended sampling photoacoustic volumetric imaging method of claim 2, wherein, In the Fibonacci Fermat double-helix arrangement, the three-dimensional space coordinates corresponding to each ultrasonic array element The calculation is as follows: In the formula, is the spherical radius of the three-dimensional bowl-shaped ultrasonic array, is the total number of channels of the array element, is the channel number of the array element , is the array coverage rotation angle, is the golden ratio constant, and the value is , is the first array element is the plane projection radius.

4. The spiral multi-point extended sampling photoacoustic volumetric imaging method of claim 3, wherein, The arrangement mode of the pre-generated interpolation sequence is determined by the element channel position under the rotation angle; Specifically, any element channel in the helical interval is selected, which is rotated around the depth direction axis, the intersection of the rotation track and the helix is obtained, the rotation angle is calculated, and the element channels before and after rotation are staggered on the helix under the rotation angle, which is the arrangement mode of the pre-generated interpolation sequence.

5. The spiral multi-point extended sampling photoacoustic volumetric imaging method of claim 4, wherein, The U-Net interpolation network adopts an asymmetric convolution kernel structure and performs down-sampling and up-sampling in the time dimension.

6. The spiral multi-point extended sampling photoacoustic volumetric imaging method of claim 5, wherein, The virtual rotation interpolation specifically inserts a virtual rotation channel between non-rotation channels.

7. An imaging system using the spiral multi-point extended sampling photoacoustic volumetric imaging method according to any one of claims 1-6. The system comprises a pulse laser emission module, a rotation control module, a photoacoustic volumetric imaging probe, and a computer data processing module. The pulse laser emission module comprises a pulse laser and a diffusion sheet, and the output end of the pulse laser is connected with the diffusion sheet. The rotation control module comprises a logic control circuit, a voltage stabilizer, a comparator, and a rotating motor. The photoacoustic volumetric imaging probe comprises a probe shell, a three-dimensional bowl-shaped ultrasonic transducer (1), a rotating connector, a bearing slip ring, a fiber bundle (2), and a diffusion lens (3). The three-dimensional bowl-shaped ultrasonic transducer (1) is arranged in a Fibonacci-Fermat double helix three-dimensional bowl-shaped ultrasonic array in a bowl-shaped cavity of the probe shell. The center of the probe shell is provided with a one-inch internal threaded column channel, and the fiber bundle (2) and the diffusion lens (3) are fixed in the channel by an external threaded iron ring. The diffusion lens (3) is located at the front end of the fiber bundle (2). The probe shell is connected with the rotating motor through the rotating connector, and the logic control circuit is electrically connected with the rotating motor. The computer data processing module comprises a computer and a multi-channel data acquisition device, the multi-channel data acquisition device is electrically connected with the three-dimensional bowl-shaped ultrasonic transducer, the multi-channel data acquisition device, the voltage stabilizer, and the three-dimensional bowl-shaped ultrasonic transducer (1) are connected with the computer, the logic control circuit is connected with the comparator, the comparator is connected with the pulse laser, and the voltage stabilizer is connected with the logic control circuit and the rotating motor.

8. The imaging system of claim 7, wherein, The three-dimensional bowl-shaped ultrasonic transducer (1) has a center frequency of 5 MHz.

9. The imaging system of claim 7, wherein, The fiber bundle (2) has a core diameter of 200 μm.

10. The imaging system of claim 7, wherein, The rotation control module further comprises an optical encoder coaxially installed with the rotating motor, which is used for real-time feedback of the actual rotation angle and controls the stepping error within ±0.05° through PID closed-loop control.

Images