Laser speckle blood flow imaging method and system

Abstract

The application discloses a laser speckle blood flow imaging method and system, and the method comprises the following steps: acquiring a laser speckle image sequence to be processed; processing the laser speckle image sequence to be processed by using a non-overlapping window space contrast algorithm to obtain a low-resolution space contrast image in a logarithmic domain to be processed; inputting the low-resolution space contrast image in the logarithmic domain to be processed into a pre-trained image super-resolution model to obtain a predicted high-resolution speckle contrast image in the logarithmic domain; and calculating a laser speckle blood flow image according to the predicted high-resolution speckle contrast image in the logarithmic domain. The application adopts a lightweight image super-resolution network to improve the spatial resolution, is easy to process in real time, and is easy to realize laser speckle blood flow imaging.

Description

Technical Field

[0001] This invention relates to the field of biomedical image processing technology, and in particular to a laser speckle blood flow imaging method and system. Background Technology

[0002] Laser speckle contrast imaging offers advantages such as non-contact operation, full-field measurement, fast imaging speed, high resolution, simple equipment, and low cost. It can measure blood flow velocity while acquiring vascular morphology and has been widely applied to blood flow monitoring in areas such as the brain, fundus, and skin. Rapid, high-resolution blood flow imaging technology based on laser speckle contrast imaging plays a crucial role in scientific research such as transient hemodynamics and in the clinical treatment of related diseases. Traditional laser speckle imaging methods mainly include spatial contrast algorithms, temporal contrast algorithms, and spatiotemporal contrast algorithms. Spatial contrast algorithms calculate contrast using a spatial sliding pane on a single frame of raw laser speckle image, achieving high temporal resolution but reducing spatial resolution. Conversely, temporal contrast algorithms use multiple frames for contrast calculation, reducing temporal resolution but achieving relatively high spatial resolution. Spatiotemporal contrast methods use a combined spatiotemporal pane to calculate contrast, striking a trade-off between spatiotemporal resolution. Traditional laser speckle imaging algorithms struggle to achieve high spatiotemporal resolution blood flow monitoring, requiring a trade-off between temporal and spatial resolution, which limits their further development and application.

[0003] Laser speckle contrast imaging typically employs graphics processing units (GPUs) for acceleration, but inter-frame temporal contrast algorithms require significant computational resources, resulting in poor acceleration. In contrast, intra-frame spatial contrast algorithms offer inherent parallel computing advantages, easily achieving significant acceleration; the larger the spatial sliding pane, the higher the parallel speedup ratio, but the lower the spatial resolution. Another approach is to reduce the number of frames required for temporal contrast algorithms by using filtering algorithms for noise reduction. However, to achieve high temporal resolution, a smaller number of frames is usually used, leading to excessive statistical noise and difficulty in recovering image details. Therefore, no effective solution has yet been proposed for high spatiotemporal resolution laser speckle blood flow imaging. Summary of the Invention

[0004] The purpose of this invention is to overcome the defects of the prior art and provide a laser speckle blood flow imaging method and system.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] In a first aspect, the present invention provides a laser speckle blood flow imaging method, comprising:

[0007] Acquire the sequence of laser speckle images to be processed;

[0008] For the laser speckle image sequence to be processed, the non-overlapping pane spatial contrast algorithm is used to process it to obtain the logarithmic domain low-resolution spatial contrast image to be processed. The logarithmic domain low-resolution spatial contrast image to be processed is input into the pre-trained image super-resolution model to obtain the predicted logarithmic domain high-resolution speckle contrast image.

[0009] Based on the predicted logarithmic domain high-resolution speckle contrast image, the laser speckle blood flow image is calculated.

[0010] As a further improvement of the present invention, the training method of the pre-trained image super-resolution model includes:

[0011] Acquire laser speckle image sequences;

[0012] Based on the first frame of the original laser speckle image in the laser speckle image sequence, an initial low-resolution spatial contrast image is calculated.

[0013] An initial low-resolution temporal contrast image is calculated based on the image frames of the laser speckle image sequence;

[0014] Based on the initial low-resolution temporal contrast image, a high-resolution temporal contrast image is obtained through noise reduction processing.

[0015] Based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image, the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image are calculated.

[0016] Repeat the above steps to collect a large number of log-domain low-resolution spatial contrast images and log-domain high-resolution temporal contrast images of different samples to form a dataset; use the log-domain low-resolution spatial contrast images in the dataset as source domain images and the corresponding log-domain high-resolution temporal contrast images as target images.

[0017] The source domain image and the corresponding target domain image are input in pairs into a lightweight image super-resolution network for training, resulting in an image super-resolution model; the lightweight image super-resolution network is a sub-pixel convolutional network.

[0018] As a further improvement of the present invention, an initial low-resolution spatial contrast image is calculated based on the first frame of the original laser speckle image of the laser speckle image sequence, including:

[0019] The original laser speckle image sequence is set as I(x,y,t), where I represents the image grayscale value, (x,y) represents the pixel coordinates, and t represents the image frame number. The image size is set to [W,H], the value range of x is [1,W], the value range of y is [1,H], and the value range of t is [1,Nt].

[0020] Select a non-overlapping pane of size M×M and crop the first frame of the original laser speckle image I(x,y,1) so that its size is divisible by M. The cropped image size is [Wcrop,Hcrop]. Further divide it into M×M squares. The non-overlapping pane traverses the cropped first frame of the original laser speckle image and calculates the ratio of the standard deviation to the mean of each square to obtain the initial low-resolution spatial contrast image Ks.

[0021] As a further improvement of the present invention, an initial low-resolution temporal contrast image is calculated based on the image frames of the laser speckle image sequence, including:

[0022] The original Nt laser speckle images are cropped to [Wcrop, Hcrop]; the ratio of the standard deviation to the mean of the pixel time series at the corresponding position is calculated to obtain the initial low-resolution temporal contrast image Kt.

[0023] As a further improvement of the present invention, the method for calculating the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image is as follows:

[0024] K log =—ln(K) 2 ) / N

[0025] Among them, K log For the logarithmic domain contrast image, ln is the natural logarithm, and N is the normalization coefficient.

[0026] As a further improvement of the present invention, the subpixel convolutional network includes a feature extraction layer, an upsampling layer, and a feature reconstruction layer;

[0027] The feature extraction layer is a convolutional layer containing multiple convolutional kernels to capture image features at different levels;

[0028] The upsampling layer is a sub-pixel convolutional layer, which rearranges pixels through special convolution operations to increase pixel size. The scale of the sub-pixel convolution determines the upsampling layer's multiplier.

[0029] The upsampling layer's multiplier is the same as the side length of the non-overlapping pane, which is M times.

[0030] The feature reconstruction layer consists of multiple convolutional layers;

[0031] During the training of the subpixel convolutional network, hyperbolic tangent is used as the activation function for all convolutional layers, mean squared error and structural similarity loss are used as the loss function, and Adam is used as the optimizer.

[0032] As a further improvement of the present invention, the method for calculating the laser speckle blood flow image (BFI) based on the predicted logarithmic domain high-resolution speckle contrast image is as follows:

[0033]

[0034] Among them, K log For the logarithmic domain contrast image, N is the normalization coefficient.

[0035] In a second aspect, the present invention provides a laser speckle blood flow imaging system, comprising:

[0036] The acquisition module is used to acquire the sequence of laser speckle images to be processed;

[0037] The prediction module is used to process the laser speckle image sequence to be processed using a non-overlapping pane spatial contrast algorithm to obtain a low-resolution spatial contrast image in the logarithmic domain. The low-resolution spatial contrast image in the logarithmic domain is then input into a pre-trained image super-resolution model to obtain a predicted high-resolution speckle contrast image in the logarithmic domain.

[0038] The calculation module is used to calculate the laser speckle blood flow image based on the predicted logarithmic domain high-resolution speckle contrast image.

[0039] In the prediction module, the training method of the pre-trained image super-resolution model includes:

[0040] Acquire laser speckle image sequences;

[0041] Based on the first frame of the original laser speckle image in the laser speckle image sequence, an initial low-resolution spatial contrast image is calculated.

[0042] An initial low-resolution temporal contrast image is calculated based on the image frames of the laser speckle image sequence; a high-resolution temporal contrast image is obtained by noise reduction processing based on the initial low-resolution temporal contrast image.

[0043] Based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image, the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image are calculated.

[0044] Repeat the above steps to collect a preset number of different logarithmic domain low-resolution spatial contrast images and logarithmic domain high-resolution temporal contrast images to form a dataset; use the logarithmic domain low-resolution spatial contrast images in the dataset as source domain images and the corresponding logarithmic domain high-resolution temporal contrast images as target images.

[0045] The source domain image and the corresponding target domain image are input in pairs into a lightweight image super-resolution network for training, resulting in an image super-resolution model; the lightweight image super-resolution network is a sub-pixel convolutional network.

[0046] In the prediction module, an initial low-resolution spatial contrast image is calculated based on the first frame of the original laser speckle image sequence, including:

[0047] The original laser speckle image sequence is set as I(x,y,t), where I represents the image grayscale value, (x,y) represents the pixel coordinates, and t represents the image frame number. The image size is set to [W,H], the value range of x is [1,W], the value range of y is [1,H], and the value range of t is [1,Nt].

[0048] Select a non-overlapping pane of size M×M and crop the first frame of the original laser speckle image I(x,y,1) so that its size is divisible by M. The cropped image size is [Wcrop,Hcrop]. Divide it into M×M squares. Traverse the cropped first frame of the original laser speckle image with the non-overlapping pane and calculate the ratio of the standard deviation to the mean of each square to obtain the initial low-resolution spatial contrast image Ks.

[0049] In the prediction module, an initial low-resolution temporal contrast image is calculated based on the image frames of the laser speckle image sequence, including:

[0050] The original Nt laser speckle images are cropped to [Wcrop, Hcrop]; the ratio of the standard deviation to the mean of the pixel time series at the corresponding position is calculated to obtain the initial low-resolution temporal contrast image Kt.

[0051] In the prediction module, the method for calculating the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image is as follows:

[0052] K log =—ln(K) 2 ) / N

[0053] Among them, K log For the logarithmic domain contrast image, ln is the natural logarithm, and N is the normalization coefficient.

[0054] In the prediction module, the subpixel convolutional network includes a feature extraction layer, an upsampling layer, and a feature reconstruction layer;

[0055] The feature extraction layer is a convolutional layer containing multiple convolutional kernels, used to capture image features at different levels;

[0056] The upsampling layer is a sub-pixel convolutional layer, which rearranges pixels through special convolution operations to increase pixel size. The scale of the sub-pixel convolution determines the upsampling layer's multiplier.

[0057] The feature reconstruction layer consists of multiple convolutional layers;

[0058] During the training of the subpixel convolutional network, hyperbolic tangent is used as the activation function for all convolutional layers, mean squared error and structural similarity loss are used as the loss function, and Adam is used as the optimizer.

[0059] In the calculation module, the method for calculating the laser speckle blood flow image (BFI) based on the predicted logarithmic domain high-resolution speckle contrast image is as follows:

[0060]

[0061] Among them, K log For the logarithmic domain contrast image, N is the normalization coefficient.

[0062] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the laser speckle blood flow imaging method.

[0063] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the laser speckle blood flow imaging method.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] To address the problem of insufficient frame counts in existing technologies leading to excessive statistical noise and difficulty in recovering image details, this invention proposes a high spatiotemporal resolution laser speckle blood flow imaging method based on deep learning. The laser speckle image sequence to be processed employs a non-overlapping pane spatial contrast algorithm, significantly improving computational speed. After processing, a pre-trained image super-resolution model is used for prediction to obtain a high-resolution speckle contrast image in the logarithmic domain, which is then used to calculate the laser speckle blood flow image. A lightweight image super-resolution network is used to improve spatial resolution, facilitating real-time processing and enabling easy implementation of laser speckle blood flow imaging. Attached Figure Description

[0066] Figure 1 This is a flowchart of a laser speckle blood flow imaging method according to the present invention;

[0067] Figure 2 This is a flowchart of a laser speckle blood flow imaging method according to an embodiment of the present invention;

[0068] Figure 3 This is a schematic diagram of the laser speckle imaging system of the present invention;

[0069] Figure 4 This is a schematic diagram of the non-overlapping pane spatial contrast algorithm processing of the present invention, wherein (a) is a schematic diagram of the non-overlapping pane spatial contrast algorithm and (b) is a schematic diagram of the time contrast algorithm.

[0070] Figure 5 This is a schematic diagram of the lightweight image super-resolution model network structure of the present invention;

[0071] Figure 6 This is a flowchart of the real-time acquisition and image processing of the laser speckle system of the present invention;

[0072] Figure 7 The images are: (a) a low-resolution spatial contrast image of the logarithmic domain obtained by the method of the present invention in an animal experiment; and (b) a high spatiotemporal resolution laser speckle blood flow image.

[0073] Figure 8 This is a schematic diagram of the laser speckle blood flow imaging system of the present invention;

[0074] Figure 9 This is a schematic diagram of the electronic device structure according to a preferred embodiment of the present invention.

[0075] In the diagram, 1 is the laser; 2 is the beam expander; 3 is the biological tissue to be tested; 4 is the lens; and 5 is the camera. Detailed Implementation

[0076] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0077] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0078] like Figure 1 As shown, the first objective of this invention is to provide a laser speckle blood flow imaging method, comprising:

[0079] S1, acquire the laser speckle image sequence to be processed;

[0080] S2, For the laser speckle image sequence to be processed, the non-overlapping pane spatial contrast algorithm is used to process it to obtain the logarithmic domain low-resolution spatial contrast image to be processed. The logarithmic domain low-resolution spatial contrast image to be processed is input into the pre-trained image super-resolution model to obtain the predicted logarithmic domain high-resolution speckle contrast image.

[0081] S3. Based on the predicted logarithmic domain high-resolution speckle contrast image, the laser speckle blood flow image is calculated.

[0082] This method accelerates the computation by employing a non-overlapping pane spatial contrast algorithm for the laser speckle image sequence, significantly improving computation speed. An image super-resolution model is used for prediction, addressing the issue of low measurement accuracy due to limited data. High-resolution speckle contrast images in the logarithmic domain are obtained before calculating the laser speckle blood flow image. A lightweight image super-resolution network is used to improve spatial resolution, facilitating the realization of high-resolution laser speckle blood flow imaging.

[0083] In addition, this invention also mainly provides a training method for the above-mentioned pre-trained image super-resolution model, which specifically includes the following steps:

[0084] S4, acquire laser speckle image sequence;

[0085] S5, calculate the initial low-resolution spatial contrast image based on the first frame of the original laser speckle image of the laser speckle image sequence;

[0086] S6, calculate the initial low-resolution temporal contrast image based on the image frames of the laser speckle image sequence;

[0087] S7. Based on the initial low-resolution time-contrast image, a high-resolution time-contrast image is obtained through noise reduction processing.

[0088] S8. Based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image, the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image are calculated.

[0089] S9. Repeat the above steps to collect a large number of different samples of logarithmic domain low-resolution spatial contrast images and logarithmic domain high-resolution temporal contrast images to form a dataset; use the logarithmic domain low-resolution spatial contrast images in the dataset as source domain images and the corresponding logarithmic domain high-resolution temporal contrast images as target images.

[0090] S10, input the source domain image and the corresponding target domain image in pairs into the lightweight image super-resolution network for training; to obtain the image super-resolution model; the lightweight image super-resolution network is a sub-pixel convolutional network.

[0091] The high spatiotemporal resolution laser speckle blood flow imaging method provided by this invention can be used for precise real-time measurement of blood flow velocity in subcutaneous blood vessels, cerebral blood vessels, and intestinal blood vessels. Based on this invention, medical devices for high spatiotemporal resolution precise measurement of blood flow velocity can be developed, offering significant economic benefits. This invention provides blood flow velocity measurement technology that improves the spatiotemporal resolution of laser speckle blood flow imaging.

[0092] The method of the present invention will be described in detail below with reference to specific embodiments and accompanying drawings:

[0093] This invention provides a laser speckle blood flow imaging method that employs a non-overlapping pane spatial contrast algorithm and utilizes a graphics processing unit for acceleration to quickly acquire low-resolution spatial contrast images in the logarithmic domain. A trained lightweight image super-resolution network is then used for rapid processing, which improves the resolution of the low-resolution spatial speckle contrast images in the logarithmic domain. Combining these two methods achieves laser speckle blood flow imaging with higher spatiotemporal resolution compared to traditional laser speckle imaging.

[0094] like Figure 2 The diagram shown is a flowchart of a laser speckle blood flow imaging method according to the present invention. The following is a description of the method in conjunction with... Figure 2 The present invention will now be described in detail.

[0095] The method specifically includes the following steps:

[0096] Step S100: Measure biological tissue using a laser speckle contrast imaging system and acquire laser speckle image sequences;

[0097] Furthermore, in S100, the method for measuring biological tissue and acquiring laser speckle image sequences using a laser speckle imaging system is as follows:

[0098] S101, turn on the camera of the laser speckle imaging system, adjust the camera lens focal length until it is precisely focused on the surface of the biological tissue, turn on the laser, and the laser shines on the surface of the biological tissue after passing through the beam expander and the reflector. The backscattered light from the surface of the biological tissue and its interior forms the original laser speckle image. Adjust the camera exposure time, denoted as T ms, to prevent overexposure.

[0099] S102: Set the camera acquisition time, start acquiring and saving the original laser speckle image sequence.

[0100] Step S200: Calculate the initial low-resolution spatial contrast image based on the first frame of the original laser speckle image sequence;

[0101] Furthermore, in S200, the method for calculating the initial low-resolution spatial contrast image based on the first frame of the original laser speckle image sequence is as follows:

[0102] S201, set the original laser speckle image sequence as I(x,y,t), where I represents the image grayscale value, (x,y) represents the pixel coordinates, t represents the image frame number, and set the image size as [W,H], the value range of x is [1,W], the value range of y is [1,H], and the value range of t is [1,Nt].

[0103] S202, Select a non-overlapping pane of size M×M, crop the size of the first frame of the original laser speckle image I(x,y,1) so that its size is divisible by M. The size of the cropped image is [Wcrop,Hcrop]. Further divide it into squares of size M×M. The non-overlapping pane traverses the cropped first frame of the original laser speckle image and calculates the ratio of the standard deviation to the mean of each square to obtain the initial low-resolution spatial contrast image Ks.

[0104] Step S300: Calculate the initial low-resolution temporal contrast image based on the image frames of the laser speckle image sequence;

[0105] Furthermore, in S300, the algorithm for calculating the initial low-resolution temporal contrast image based on the image frames of the laser speckle image sequence is as follows: Similar to the image cropping method in S201, the Nt original laser speckle images are cropped to [Wcrop, Hcrop]. The ratio of the standard deviation to the mean of the corresponding pixel time series is calculated to obtain the initial high-resolution temporal contrast image Kt.

[0106] Step S400: Calculate the corresponding high-resolution temporal contrast image based on the initial low-resolution temporal contrast image;

[0107] Furthermore, in S400, the algorithm for calculating the high-resolution time contrast image based on the initial low-resolution time contrast image is as follows: an image filtering algorithm is selected to perform noise reduction processing on the initial low-resolution time contrast image.

[0108] Preferably, the image filtering algorithm for noise reduction is a three-dimensional block matching collaborative filtering algorithm.

[0109] Step S500: Based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image, calculate the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image.

[0110] Furthermore, in S500, the method for calculating the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image is as follows:

[0111] K log =—ln(K) 2 ) / N

[0112] Among them, K log For the logarithmic contrast image, ln is the natural logarithm, and N is the normalization coefficient;

[0113] Step S600: Repeat steps S100-S500 to obtain a large number of log-domain low-resolution spatial contrast images and log-domain high-resolution temporal contrast images of different samples. Use the log-domain low-resolution spatial contrast images as source domain images and the log-domain high-resolution temporal contrast images as target images.

[0114] Step S700: Input the source domain image and the corresponding target domain image in pairs into a lightweight image super-resolution network for training; to obtain the image super-resolution model; the lightweight image super-resolution network is a sub-pixel convolutional network;

[0115] As a specific embodiment of the present invention, the efficient subpixel convolutional network includes a feature extraction layer, an upsampling layer, and a feature reconstruction layer.

[0116] The feature extraction layer is a convolutional layer containing multiple convolutional kernels to capture image features at different levels. The upsampling layer is a subpixel convolutional layer that rearranges pixels through special convolution operations to increase pixel size; the scale of the subpixel convolution determines the upsampling layer's multiplier.

[0117] For example, the upsampling factor is M times the side length of the non-overlapping pane.

[0118] The feature reconstruction layer consists of multiple convolutional layers.

[0119] Preferably, the specific method for training the image super-resolution model is as follows: using hyperbolic tangent as the activation function for all convolutional layers, using a combination of mean squared error and structural similarity loss as the loss function, and using the Adam deep learning optimizer as the optimizer during the training process.

[0120] Step S800: For the laser speckle image sequence to be processed, a low-resolution spatial contrast image in the logarithmic domain is obtained by processing it using a non-overlapping pane spatial contrast algorithm. This low-resolution spatial contrast image in the logarithmic domain is then input into the image super-resolution model to obtain the predicted high-resolution speckle contrast image in the logarithmic domain. Specifically:

[0121] A laser speckle imaging system was used to reacquire and preprocess the laser speckle image sequence. The non-overlapping pane spatial contrast algorithm of the graphics processing unit was used to obtain a low-resolution spatial contrast image in the logarithmic domain. This image was then input into the trained image super-resolution model to obtain a high-resolution speckle contrast image in the logarithmic domain.

[0122] Step S900: Calculate a high-resolution laser speckle blood flow image based on the logarithmic domain high-resolution speckle contrast image output by the image super-resolution model.

[0123] Furthermore, in S900, the method for calculating the laser speckle blood flow image based on the logarithmic domain high-resolution speckle contrast image output by the image super-resolution model is as follows:

[0124]

[0125] Preferably, the size of the non-overlapping pane is 8×8.

[0126] Preferably, the calculation of spatial contrast images is accelerated using a graphics processing unit.

[0127] Preferably, the image super-resolution model is a lightweight and efficient subpixel convolutional network.

[0128] Figure 3 This is a schematic diagram of the laser speckle imaging system structure in a corresponding embodiment of the laser speckle blood flow imaging method of the present invention.

[0129] Figure 4 This is a schematic diagram of the non-overlapping pane spatial contrast algorithm and temporal contrast algorithm of the present invention. Figure 4 (a) is a schematic diagram of the non-overlapping pane spatial contrast algorithm. By calculating the ratio of the standard deviation to the mean of the pixels in the first frame of the original laser speckle image within the non-overlapping pane, a low-resolution spatial contrast image is obtained. This algorithm uses non-overlapping panes, which is naturally suitable for accelerating graphics processing units and greatly improves the calculation speed. Figure 4(b) is a schematic diagram of the time contrast algorithm, which uses multiple frames of original laser speckle image sequences to calculate the ratio of the standard deviation to the mean of the corresponding pixel time series to obtain a high-resolution time contrast image.

[0130] Figure 5 This is a schematic diagram of the network structure of the lightweight image super-resolution model of the present invention, which includes a feature extraction layer, an upsampling layer, and a feature reconstruction layer. The feature extraction layer is used to extract features from non-overlapping pane spatial contrast images and typically contains multiple convolutional kernels to capture image features at different levels. The upsampling layer uses sub-pixel convolutional layers to upsample the image, converting the low-resolution spatial contrast image in the logarithmic domain into a high-resolution temporal contrast image in the logarithmic domain. Sub-pixel convolution rearranges pixels through special convolution operations to increase the image size. The feature reconstruction layer typically consists of one or more convolutional layers to further improve image quality, help restore lost details and textures, and adjust the contrast and sharpness of the upsampled image.

[0131] Figure 6 This is a flowchart illustrating the real-time acquisition and image processing of a laser speckle system for a laser speckle blood flow imaging method. The original laser speckle image sequence acquired by the laser speckle imaging system is processed using a non-overlapping pane spatial contrast algorithm, accelerated by a graphics processing unit, to quickly obtain a low-resolution spatial contrast image in the logarithmic domain. This image is then rapidly processed by a trained lightweight image super-resolution network to obtain a high-resolution speckle contrast image in the logarithmic domain, thereby achieving a high spatiotemporal resolution laser speckle blood flow image.

[0132] Figure 7 The results of laser speckle blood flow imaging were presented. Figure 7 Image (a) is a low-resolution spatial contrast blood flow image, which has low resolution, high noise, and low image quality. Figure 7 Image (b) is a high spatiotemporal resolution laser speckle blood flow image obtained using the method of the present invention. It has high resolution, low noise, and good image quality.

[0133] like Figure 8 As shown, a second objective of this invention is to provide a laser speckle blood flow imaging system, comprising:

[0134] The acquisition module is used to acquire the sequence of laser speckle images to be processed;

[0135] The prediction module is used to process the laser speckle image sequence to be processed using a non-overlapping pane spatial contrast algorithm to obtain a low-resolution spatial contrast image in the logarithmic domain. The low-resolution spatial contrast image in the logarithmic domain is then input into a pre-trained image super-resolution model to obtain a predicted high-resolution speckle contrast image in the logarithmic domain.

[0136] The calculation module is used to calculate the laser speckle blood flow image based on the predicted logarithmic domain high-resolution speckle contrast image.

[0137] like Figure 9 As shown, a third objective of this invention is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the laser speckle blood flow imaging method.

[0138] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the laser speckle blood flow imaging method.

[0139] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0140] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0141] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0142] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A laser speckle blood flow imaging method, characterized in that, include: Acquire the sequence of laser speckle images to be processed; For the laser speckle image sequence to be processed, the non-overlapping pane spatial contrast algorithm is used to process it to obtain the logarithmic domain low-resolution spatial contrast image to be processed. The logarithmic domain low-resolution spatial contrast image to be processed is input into the pre-trained image super-resolution model to obtain the predicted logarithmic domain high-resolution speckle contrast image. Based on the predicted logarithmic domain high-resolution speckle contrast image, the laser speckle blood flow image is calculated; The training method for the pre-trained image super-resolution model includes: Acquire laser speckle image sequences; Based on the first frame of the original laser speckle image in the laser speckle image sequence, an initial low-resolution spatial contrast image is calculated. An initial low-resolution temporal contrast image is calculated based on the image frames of the laser speckle image sequence; a high-resolution temporal contrast image is obtained by noise reduction processing based on the initial low-resolution temporal contrast image. Based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image, the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image are calculated. Repeat the above steps to collect a preset number of logarithmic domain low-resolution spatial contrast images and logarithmic domain high-resolution temporal contrast images to form a dataset; use the logarithmic domain low-resolution spatial contrast images in the dataset as source domain images and the corresponding logarithmic domain high-resolution temporal contrast images as target images. The source domain image and the corresponding target domain image are input in pairs into a lightweight image super-resolution network for training, resulting in an image super-resolution model; the lightweight image super-resolution network is a sub-pixel convolutional network.

2. The laser speckle blood flow imaging method according to claim 1, characterized in that, Based on the first frame of the original laser speckle image in the laser speckle image sequence, an initial low-resolution spatial contrast image is calculated, including: The original laser speckle image sequence is defined as I(x, y, t), where I represents the image grayscale value, (x, y) represents the pixel coordinates, and t represents the image frame number. The image size is defined as [W, H], the value range of x is [1, W], the value range of y is [1, H], and the value range of t is [1, Nt]. Select a non-overlapping pane of size M×M and crop the first frame of the original laser speckle image I(x, y, 1) so that its size is divisible by M. The cropped image size is [Wcrop, Hcrop]. Divide it into M×M squares. Traverse the cropped first frame of the original laser speckle image with the non-overlapping pane and calculate the ratio of the standard deviation to the mean of each square to obtain the initial low-resolution spatial contrast image Ks.

3. The laser speckle blood flow imaging method according to claim 1, characterized in that, Based on the image frames of the laser speckle image sequence, an initial low-resolution temporal contrast image is calculated, including: The original Nt laser speckle images are cropped to [Wcrop, Hcrop]; the ratio of the standard deviation to the mean of the pixel time series at the corresponding position is calculated to obtain the initial low-resolution temporal contrast image Kt.

4. The laser speckle blood flow imaging method according to claim 2, characterized in that, The method for calculating the logarithmic domain low-resolution spatial contrast image and the logarithmic domain high-resolution temporal contrast image based on the initial low-resolution spatial contrast image and the corresponding high-resolution temporal contrast image is as follows: Among them, K log For the logarithmic domain contrast image, ln is the natural logarithm, and N is the normalization coefficient.

5. The laser speckle blood flow imaging method according to claim 1, characterized in that, The subpixel convolutional network includes a feature extraction layer, an upsampling layer, and a feature reconstruction layer; The feature extraction layer is a convolutional layer containing multiple convolutional kernels, used to capture image features at different levels; The upsampling layer is a sub-pixel convolutional layer, which rearranges pixels through special convolution operations to increase pixel size. The scale of the sub-pixel convolution determines the upsampling layer's multiplier. The feature reconstruction layer consists of multiple convolutional layers; During the training of the subpixel convolutional network, hyperbolic tangent is used as the activation function for all convolutional layers, mean squared error and structural similarity loss are used as the loss function, and Adam is used as the optimizer.

6. The laser speckle blood flow imaging method according to claim 1, characterized in that, The method for calculating the laser speckle blood flow image (BFI) based on the predicted logarithmic domain high-resolution speckle contrast image is as follows: Among them, K log For the logarithmic domain contrast image, N is the normalization coefficient.

7. A laser speckle blood flow imaging system, implementing the laser speckle blood flow imaging method according to any one of claims 1-6, characterized in that, include: The acquisition module is used to acquire the sequence of laser speckle images to be processed; The prediction module is used to process the laser speckle image sequence to be processed using a non-overlapping pane spatial contrast algorithm to obtain a low-resolution spatial contrast image in the logarithmic domain. The low-resolution spatial contrast image in the logarithmic domain is then input into a pre-trained image super-resolution model to obtain a predicted high-resolution speckle contrast image in the logarithmic domain. The calculation module is used to calculate the laser speckle blood flow image based on the predicted logarithmic domain high-resolution speckle contrast image.

8. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the laser speckle blood flow imaging method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the laser speckle blood flow imaging method according to any one of claims 1-6.

Images