Vascular pulse visualization method and system

Abstract

The invention discloses a blood vessel pulsation visualization method and system, and belongs to the field of image processing, and the method comprises the steps: collecting physiological tissue video data containing blood vessels; converting each frame of image from an RGB color space to a YIQ color space; separating a brightness signal Y and chrominance signals I and Q from the YIQ color space; performing spatial decomposition on each frame of image in the YIQ color space by using a filter bank of a complex controllable pyramid; amplitude spectrum signals Ar, theta (x, y, t) and phase spectrum signals Br, theta (x, y, t) are extracted from the sub-band sequence; constructing a vascular pulse filter (VPF) according to the pulse frequency of the user; processing the phase spectrum signals Br and theta (x, y, t) through a vascular pulsation filter VPF, and extracting a vascular pulsation phase change signal; according to the vascular pulsation phase change signal and the amplitude spectrum signals Ar and theta (x, y, t), a reconstructed vascular pulsation video is obtained through inverse transformation reconstruction. Aiming at the difficulty in visual observation caused by tiny shaking of blood vessels in the prior art, the visual sensitivity is improved.

Description

Technical Field

[0001] The present application relates to the field of image processing, and more specifically, to a method and system for visualizing vascular pulsation. Background Art

[0002] Since its inception in the early 19th century, medical endoscope technology has undergone four major technological innovations: rigid tube endoscopes, semi-curved endoscopes, fiberscopes, and finally electronic endoscopes. As clinical needs continue to escalate, endoscope technology continues to evolve to meet this growing medical need. In recent years, the penetration rate of minimally invasive surgical procedures in China has increased significantly, demonstrating the importance of minimally invasive techniques in modern medicine.

[0003] In minimally invasive surgery, the surgical visualization capabilities provided by endoscopes and video output systems are crucial. Accurately identifying and locating blood vessels within the surgical field is crucial for reducing complications and collateral damage. However, relying solely on conventional video footage and the feel of surgical instruments, surgeons often struggle to clearly observe the minute pulsations of blood vessels beneath the tissue. This visual limitation increases surgical risk and highlights the urgent need for improved visualization of vascular pulsations.

[0004] A Chinese patent application, application number CN114693653A, published on July 1, 2022, discloses a method for accurately extracting pulse waves from a video of a superficial human artery. This method addresses motion and illumination interference, a common problem in non-contact pulse wave measurement methods, and performs multiple filtering operations to accurately extract the pulse wave. The application includes the following steps: S1: acquiring a video of a superficial human artery; S2: preprocessing the video using an adaptive gamma transform; S3: amplifying the video using a motion-enhancing algorithm combining Euler amplification and FRR filtering; S4: applying an adaptive threshold inter-frame differencing method and selecting multiple feature regions as regions of interest; S5: extracting the raw pulse wave signal; S6: filtering the signal with a type II Chebyshev wavelet transform, an adaptive threshold method, and an adaptive notch filter; and S7: enhancing the pulse wave signal using modulation domain spectrum subtraction. However, this method utilizes simple Euler amplification and a traditional filter cascade, making it difficult to capture changes in acceleration and jerkiness during vascular pulsation and unable to fully characterize the multidimensional dynamic characteristics of vascular pulsation. Summary of the Invention

[0005] 1. Technical problems to be solved

[0006] In response to the difficulties in intuitive observation caused by the tiny jitters of blood vessels in the existing technology, the present application provides a method and system for visualizing blood vessel pulsations, which improves the visualization sensitivity through customized blood vessel pulsation filters and phase amplification based on the user's specific pulse frequency.

[0007] 2. Technical solution

[0008] The purpose of this application is achieved through the following technical solutions.

[0009] One aspect of an embodiment of the present specification provides a method for visualizing vascular pulsation, comprising: acquiring video data of physiological tissue containing blood vessels, the physiological tissue video data comprising a multi-frame image sequence; converting each frame of the image from an RGB color space to a YIQ color space; separating a luminance signal Y and chrominance signals I and Q from the YIQ color space; spatially decomposing each frame of the image in the YIQ color space using a filter bank of a complex controllable pyramid to obtain subband sequences of different scales and directions; and extracting an amplitude spectrum signal A from the subband sequence. r,θ (x,y,t) and phase spectrum signal B r,θ (x, y, t); construct a blood vessel pulsation filter VPF according to the user's pulse frequency; use the blood vessel pulsation filter VPF to filter the phase spectrum signal B r,θ (x, y, t) is processed to extract the vascular pulsation phase change signal; according to the vascular pulsation phase change signal and the amplitude spectrum signal A r,θ (x, y, t), reconstructed through inverse transformation to obtain the reconstructed vascular pulsation video.

[0010] Furthermore, S3 uses a filter bank of a complex controllable pyramid to perform spatial decomposition on each frame of the image in the YIQ color space to obtain subband sequences of different scales and directions; extracts amplitude spectrum signals and phase spectrum signals from the subband sequences, including: calculating the total number of spatial decomposition layers N; constructing an N-layer M-directional filter bank based on the bandpass template and direction template of the complex controllable pyramid; using the filter bank, performing N-layer spatial decomposition on each frame of the image to obtain a frequency domain filtered spectrum F of N-layer M-directional. r,θ (u,v); according to the spectrum F after frequency domain filtering r,θ (u,v), respectively extract the amplitude spectrum signal A containing amplitude information r,θ (x, y, t), and the phase spectrum signal B containing the phase r,θ (x, y, t); the formula for calculating the total number of spatial decomposition layers N is: N = floor(log2min(h, w))-2, where floor means rounding down, h means the image frame height, and w means the image frame width.

[0011] Furthermore, the filter bank is used to perform N-layer spatial decomposition on each frame image to obtain the spectrum F after N-layer M-directional frequency domain filtering. r,θ (u,v) uses the following formula: Among them, u and v represent frequency coordinates, C r,θrepresents the frequency domain filter with scale r and direction θ in the filter bank, x and y are spatial coordinates, h and w represent the width and height of the image respectively, I(x,y,t) represents the input image, t represents time, and j is the imaginary unit.

[0012] Furthermore, according to the spectrum F after frequency domain filtering r,θ (u,v), respectively extract the amplitude spectrum signal A containing amplitude information r,θ (x, y, t), and the phase spectrum signal B containing the phase r,θ (x, y, t) uses the following formula: A r,θ (x,y,t)=∥I r,θ (x,y,t)∥;B r,θ (x,y,t)=arg[I r,θ (x,y,t)]; where I r,θ (x, y, t) represents the complex spectrum of the image after filtering, ∥★∥ represents the modulo operation, which calculates the absolute value of the complex number, and arg represents the phase operation, which calculates the angle of the complex number.

[0013] Further, S4, a blood vessel pulsation filter VPF is constructed according to the user's pulse frequency; the phase spectrum signal B is filtered by the blood vessel pulsation filter VPF. r,θ (x, y, t) is processed to extract the vascular pulsation phase change signal, including: constructing the acceleration filter kernel H(x, t) and the jerk filter kernel J(x, t) according to the user's pulse frequency; constructing the vascular pulsation filter VPF according to the acceleration filter kernel H(x, t) and the jerk filter kernel J(x, t); converting the phase spectrum signal B r,θ (x, y, t) is used as input, and the vascular pulsation filter VPF is used to extract the vascular pulsation phase change signal.

[0014] Furthermore, an acceleration filter kernel H(x, t) is constructed based on the user's pulse frequency, including: obtaining the user's pulse frequency f and the video frame rate r respectively; calculating the Gaussian-Laplacian operator based on the Gaussian filter kernel; and constructing the acceleration filter kernel H(x, t) based on the user's pulse frequency f, the video frame rate r, and the Gaussian-Laplacian operator. The acceleration filter kernel is composed of the Gaussian-Laplacian operator. Given the linear characteristics of the Gaussian-Laplacian operator, the following relationship holds: Among them, I(x,t) represents the input image; G σ (t) represents the Gaussian filter kernel; Represents the convolution operation;

[0015] Based on the user's pulse frequency f, the video frame rate r, and the Gaussian-Laplacian operator, the acceleration filter kernel H(x,t) is constructed, including: Where σ represents the standard deviation, r represents the video frame rate.

[0016] Furthermore, based on the user's pulse frequency, a jerk filter kernel J(x,t) is constructed, which includes: calculating the jerk of the phase with respect to time t; converting the jerk into smoothness σ (x,t); according to smoothness σ (x, t), construct the urge filter kernel J(x, t): According to the hierarchical structure of the complex controllable pyramid, construct the urge filter kernel of each layer Among them, l represents the level of the pyramid, σ is the standard deviation, and β is set as a hyperparameter to adjust the sensitivity of the filter.

[0017] For each layer of the jerk filter kernel Perform downsampling correction: Where λ is the downsampling factor, 0<λ<1; the jerk filter kernel of each layer after downsampling correction Perform propagation correction to obtain the final jerk filter kernel of each layer: Where N is the number of pyramid levels used for correction, and the function The filter at pyramid level i is resized to the size of pyramid level l using bicubic interpolation.

[0018] To calculate the jerk of the phase with respect to time t, use the following formula:

[0019] Among them, Jerk σ (x, t) represents the jerk value of the phase at spatial position x and time t, G σ (t) is the Gaussian filter, σ is the standard deviation; is the third derivative of the phase, and r is the scale of the phase.

[0020] Convert jerk value to smoothness σ (x, t), using the following formula:

[0021]

[0022] smoothness σ (x,t)=1-nJerk σ (x,t).

[0023] Among them, nJerk σ (x, t) represents the normalized jerk value, ranging from 0 to 1, where x and t are the spatial and temporal positions of the phase, respectively;

[0024] Furthermore, based on the acceleration filter kernel H(x, t) and the jerk filter kernel J(x, t), a vascular pulsation filter VPF is constructed, including:

[0025] VPF f,σ,λ,l (x,t)=H f (x,t)×J σ,λ,l (x,t)

[0026]

[0027] Among them, VPF f,σ,λ,l (x, t) is the blood vessel pulsation filter, σ is the standard deviation, f is the pulse frequency, λ is the downsampling factor, l is the level of the pyramid, C f,l (x,t) represents the phase after amplification, Indicates the phase at a certain level of the pyramid, is the convolution operation;

[0028] Another aspect of the embodiments of this specification also provides a vascular pulsation visualization system, including: an acquisition module, which acquires physiological tissue video data containing blood vessels, wherein the physiological tissue video data includes a multi-frame image sequence; a color conversion module, which converts each frame of image from the RGB color space to the YIQ color space, and separates the luminance signal Y, and the chrominance signals I and Q from the YIQ color space; a spatial decomposition module, which uses a complex controllable pyramid filter group to spatially decompose each frame of image in the YIQ color space to obtain sub-band sequences of different scales and directions; a signal extraction module, which extracts amplitude spectrum signals and phase spectrum signals from the sub-band sequences; a filter module, which constructs a vascular pulsation filter VPF according to the user's pulse frequency, processes the phase spectrum signal using the vascular pulsation filter VPF, and extracts the vascular pulsation phase change signal; and a video reconstruction module, which amplifies the vascular pulsation phase change signal and reconstructs the vascular pulsation video based on the amplified vascular pulsation phase change signal and the amplitude spectrum signal.

[0029] 3. Beneficial effects

[0030] Compared with the existing technology, the advantages of this application are:

[0031] A vascular pulsation filter (VPF) is constructed based on the user's actual pulse frequency, achieving precise matching of the filter with individual physiological characteristics and significantly improving the sensitivity of extracting vascular pulsation signals of specific frequencies. Through the synergistic effect of the acceleration filter kernel and the jerk filter kernel, not only the displacement changes of vascular pulsation are captured, but also the changes in acceleration and jerkiness during the vascular pulsation process are extracted, comprehensively characterizing the dynamic characteristics of vascular pulsation. The jerk filter kernel is constructed and corrected for each level separately, and the synergistic effect of multi-scale filters is achieved through downsampling correction and propagation correction, ensuring that the corresponding pulsation signal can be accurately extracted at each scale. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A technical roadmap provided for the embodiments of this application;

[0033] Figure 2 A flow chart of the blood vessel pulsation visualization method provided in an embodiment of the present application;

[0034] Figure 3 A specific design scheme is provided for the phase-based vascular pulsation filter design of the algorithm adopted in this application. DETAILED DESCRIPTION

[0035] The present application is described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] refer to Figure 1 and Figure 2 As shown, an embodiment of the present application provides an endoscopic video vascular pulsation visualization method, which specifically includes three main modules: a video acquisition module; a vascular pulsation signal extraction module; and a vascular pulsation amplification and visualization module.

[0037] The video acquisition module is used to insert the endoscope into the human body through a small incision or channel to shoot, so as to obtain clear and stable physiological videos of the body, especially physiological tissue modules containing blood vessels.

[0038] Among them, the vascular signal extraction module needs to process the captured in-vivo video of the human body, mainly to obtain the decomposed information at various scales and directions, and extract the vascular pulsation change signal of interest. Specifically, it includes the following steps: performing color space conversion on each frame of the in-vivo video captured by the endoscope to obtain the brightness information and chromaticity information of each frame of the image; spatially decomposing the brightness information of each frame of the endoscopic image through a complex controllable pyramid to obtain sub-band sequences of different scales and different motion directions; and extracting the vascular pulsation signal of interest from the sub-band sequences of various scales and different motion directions.

[0039] Specifically, when processing the target blood vessel motion video, the video is first processed in frame order (from the first frame to the S-th frame, denoted as i = 1, 2, 3, ..., S). For each frame of the image, it is first converted from the RGB color space to the YIQ color space. This conversion helps to separate the image brightness information from the chromaticity information, thereby facilitating the subsequent independent processing of brightness and chromaticity. The YIQ color space is the standard adopted by the NTSC color television system, in which the Y component represents brightness information, while the I and Q components represent chromaticity information, representing color changes from orange to cyan and from purple to yellow-green, respectively. The conversion relationship between RGB and YIQ is as follows:

[0040] To reduce the effects of noise, we focus on the Y channel, or luminance information, and spatially decompose it using a multi-scale complex controllable pyramid method. This process involves filtering and downsampling each image frame, resulting in a series of subband sequences at different scales and orientations. Phase changes are then calculated for these subband sequences, where the phase changes in the complex controllable pyramid correspond to local motion within the spatial subbands of the image.

[0041] The spatial decomposition steps are as follows: Calculate the total number of image decomposition levels: Assume that the width of the t-th frame of the video image sequence input complex controllable pyramid is w and the height is h. The complex controllable pyramid is a multi-resolution representation of the image. The total number of pyramid decomposition levels N for the input image is calculated as follows: N = floor(log2min(h,w))-2, where floor represents rounding down.

[0042] Filters are spatially decomposed: A frequency domain filter is obtained by multiplying the bandpass template and the directional template of the complex controllable pyramid. When the scale of the complex controllable pyramid is 1, the bandpass template of that scale is multiplied by the M directional templates to obtain the M directional filters of that layer. Similarly, the same operation is performed on the bandpass templates of N scales, and finally an N-layer M-directional frequency domain filter bank is obtained. The calculation formula for spatially decomposing the input image I(x, y, t) using this filter bank is as follows: Where C r,θ is the frequency domain filter with scale r and direction θ in the filter bank, F r,θ (u, v) is the filtered spectrum of the input image I(x, y, t). Furthermore, a complex controllable pyramid filter bank achieves N-layer, M-directional spatial decomposition, enabling the simultaneous capture of pulsation signals at various scales (from large vessels to tiny capillaries) and in various directions, significantly enhancing the system's ability to analyze complex vascular networks. Amplitude and phase spectrum signals are separated from the subband sequence. The phase information is extremely sensitive to minute motions and provides a key information carrier for the subsequent accurate extraction of vascular pulsation.

[0043] Decomposition of amplitude and phase: After filtering the input image using the filter bank, perform inverse transformation to obtain N layers of M directions of the image spatial domain complex spectrum I r,θ (x,y,t). On this basis, the phase spectrum B containing phase information r,θ (x, y, t), and the amplitude spectrum A containing amplitude information r,θ The calculation formula for (x,y,t) is as follows:

[0044] A r,θ (x,y,t)=∥∥I r,θ (x,y,t)∥∥

[0045] B r,θ (x,y,t)=arg[I r,θ (x,y,t)]

[0046] The primary purpose of complex controllable pyramid decomposition is to ensure that localized micro-phase processing is equivalent to localized motion processing. In addition to the high-pass and low-pass residuals from the complex controllable pyramid decomposition, the phases derived from decompositions at different scales and directions are the core processing targets. The input video frame is decomposed into a series of sine waves using a Fourier series, each corresponding to a different frequency and phase. Complex controllable filters are then used to extract the phase information for each frequency and direction.

[0047] When processing each image frame, a phase-based vascular pulsation amplification algorithm is used to extract the vascular pulsation signal of interest from the original signal. Adjusting the phase information amplifies small movements, but the phase information separated by the complex controllable pyramid contains large movements. By designing a vascular pulsation filter, large movements, including linear and high-order movements, can be effectively filtered out. The phase information is input into the vascular pulsation filter. During surgery, the doctor can obtain the patient's accurate pulse rate through the ECG monitor. The maximum spectral response point of the vascular pulsation filter is set to the pulse frequency, and the filtered signal is output, which is the vascular pulsation signal in the captured video. All other noise signals are filtered out.

[0048] In one embodiment, endoscopic vascular pulsation can be represented by phase shift. For a given input signal f(x), if the displacement of the blood vessel at time t is δ(t), the signal can be decomposed into the sum of sinusoidal waves of all frequencies using Fourier series. The original signal in the video is then represented as

[0049]

[0050] The global phase information of the displacement δ(t) at frequency w is

[0051] The phase signal can be expanded using a Taylor series in the time domain of t = h:

[0052]

[0053] Where r is the scale of the complex controllable pyramid.

[0054] Among them, the first-order linear signal is regarded as the large movement of the target in the spatial domain, while the second-order signal is the small change of the target. The remaining high-order signals are also noise signals. Therefore, in the captured in-vivo video containing vascular pulsation, it is necessary to use a vascular pulsation filter to filter out large movements such as camera shake and extract small movements other than vascular pulsation.

[0055] In order to filter out other signals besides blood vessel pulsation, a blood vessel pulsation filter (VPF) is designed. Figure 2 As shown in Figure 1, the filter is mainly composed of two filter kernels: the acceleration filter kernel H(x, t) and the jerk filter kernel J(x, t).

[0056] From a physiological perspective, vascular pulsation is essentially a small displacement change of the blood vessel wall caused by the periodic flow of blood, which manifests as a periodic acceleration characteristic. In this application, the acceleration filter kernel is composed of a Gaussian-Laplacian operator (DoG). The Gaussian-Laplacian function is the second-order derivative of the Gaussian filter kernel, and the Gaussian filter kernel does not introduce false resolution. Due to the linear characteristics of the Laplace operator, the following relationship exists: in, represents convolution, G σ (t) is the Gaussian filter kernel.

[0057] Specifically, traditional methods typically focus only on displacement or velocity changes, while ignoring higher-order dynamic characteristics such as acceleration and jerk. The Laplacian of Gaussian operator, acting as an acceleration filter kernel, works in conjunction with the jerk filter kernel to establish a complete dynamic characteristic extraction chain from displacement to velocity to acceleration to jerk, capturing the full dimensional characteristics of vascular pulsation.

[0058] In order to filter out large linear motion and leave small nonlinear acceleration changes, the acceleration filter kernel is designed by the Gaussian-Laplacian operator: Specifically,

[0059] In one embodiment, when the pulse frequency is known, a second-order acceleration filter kernel can be used to accurately extract the vascular pulsation of interest, and the standard deviation σ is determined by the prior pulse frequency and is: r is the video frame rate, and f is the pulse frequency. This design creates a "tuned filter" that responds most strongly to acceleration changes at a specific frequency (the user's pulse rate) while significantly reducing its response to other frequencies. This characteristic is particularly important because the frequency range of vascular pulsation is relatively stable and predictable.

[0060] Furthermore, there are fundamental differences in jerkiness between vascular pulsation and endoscope jitter: vascular pulsation exhibits smooth acceleration changes with low jerkiness, while endoscope jitter exhibits abrupt acceleration changes with high jerkiness. The large motions caused by endoscope jitter are not limited to linear motions; they also include nonlinear, rapid motions that are output by the acceleration filter kernel. The jerk filter kernel is designed to filter out these rapid large motions.

[0061] Jerk is the third-order derivative of phase with respect to time, and is used to measure the rate of acceleration change. It can also be used to distinguish between smooth changes and rapid changes. In one embodiment, jerk can be used to distinguish between blood vessel pulsation and endoscope jitter.

[0062] The jerk is calculated as follows: Among them, the Gaussian filter kernel G σ (t) can smooth the endoscopic image and reduce noise, It is the jerk calculation of the phase.

[0063] Next, the jerkiness is converted to smoothness. First, normalization is performed. By calculating the jerkiness of each pixel and normalizing it to between 0 and 1, a value representing smoothness is obtained:

[0064]

[0065] smoothness σ (x,t)=1-nJerk σ (x,t).

[0066] Among them, smoothness σ (x, t) represents smoothness. When vascular pulsation is detected, the smoothness is close to 1, and when endoscope jitter is detected, the smoothness is close to 0. Converting jerkiness to smoothness actually establishes an adaptive weighting system based on physical properties. When smoothness is close to 1, it indicates that the detection is the gentle periodic changes unique to blood vessels; when smoothness is close to 0, it indicates that the detection is non-physiological abrupt changes that should be filtered out.

[0067] By introducing the hyperparameter β, designing the jerk filter kernel, and adjusting the weight according to the smoothness value to selectively transmit the vascular pulsation signal, we have: JAF σ (x,t)=smoothness σ(x,t) β .

[0068] Each endoscopic image frame is spatially decomposed using a complex controllable pyramid. Lower layers (near the top of the pyramid) contain coarser images, suitable for capturing large motions and low-frequency information, while higher layers (near the base of the pyramid) contain more detailed images, suitable for capturing small motions and high-frequency information. By processing each pyramid level layer by layer, filter kernels can be applied at different resolution levels to handle motion transformations of different scales. Pyramid correction includes downsampling correction and propagation correction, which ensure that the behavior of filter kernels at different pyramid levels is consistent. Downsampling correction adjusts the weights of filter kernels at different levels to adapt to changes in resolution. Propagation correction propagates high-level information (such as the identification of rapid large motion) to lower levels, helping lower-level filter kernels more accurately distinguish between subtle changes and rapid large motion.

[0069] In particular, the complex controllable pyramid decomposes the image into sub-bands of different scales and directions. In this application, blood vessels in the human body have different sizes, from the aorta to tiny capillaries, and their pulsation characteristics vary in spatial frequency. The lower layers of the pyramid structure capture the movement of large-scale blood vessels (low-frequency information), and the higher layers capture the movement of tiny blood vessels (high-frequency information). Blood vessels have different directions in tissues. Through multi-directional decomposition, changes in blood vessel pulsation in any direction can be captured, avoiding the problem of directional blind spots in traditional methods. Decomposition at different scales actually realizes frequency band segmentation, and each sub-band only contains information in a specific frequency range, which significantly improves the signal-to-noise ratio of blood vessel pulsation signals of specific scales.

[0070] In one embodiment, each frame of endoscopic image is decomposed into multiple resolution pyramids, and a jerk filter kernel is constructed for each pyramid layer to obtain: Where l represents the level of the pyramid. To ensure consistent behavior of the filter kernels at different pyramid levels, downsampling correction and propagation correction are performed. First, the weights of the filter kernels are adjusted to adapt to the scale changes of different resolution levels: Where λ is the downsampling factor, 0<λ<1.

[0071] The high-level information is then propagated to the low-level layers to help the low-level filter kernels more accurately distinguish between vascular pulsation and endoscope jitter: Where N is the number of pyramid levels, and the function Bicubic interpolation is used to adjust the filter size of pyramid level i to the size of pyramid level l. The resulting modified jerkiness filter kernel can distinguish the essential differences between vascular pulsation and rapid large endoscope motion.

[0072] like Figure 3As shown in the figure, after combination, the acceleration filter kernel and the high-order filter kernel work together to generate the vascular pulsation filter VPF, which can achieve accurate extraction and amplification of blood vessels. f,σ,λ,l (x,t)=H f (x,t)×J σ,λ,l (x,t), Among them, the main goal of the amplification visualization module is to amplify the blood vessel pulsation signal extracted by the second module, and then reconstruct the video after amplification, so that the output video can finally clearly present the pulsation state of the blood vessel.

[0073] The phase signal obtained by the blood vessel pulsation filter can reflect the pulsating movement of the blood vessel, which is tiny and invisible in the visual domain. That is, the visual effect of the blood vessel pulsation can be enhanced by amplifying the phase. Among them, α is the magnification factor, which can be adjusted by yourself or set to a fixed value for easy observation. is the phase after amplification. Specifically, phase amplification maintains the nonlinear dynamic characteristics of vascular pulsation, which is different from simple linear contrast enhancement. Phase amplification does not change the frequency characteristics of the signal, retains the temporal pattern and rhythmic characteristics of vascular pulsation, which is of great significance for diagnosis. Furthermore, the extracted vascular pulsation phase change signal is amplified by α times, so that the tiny pulsation changes that are difficult to detect with the naked eye are effectively amplified to the visible range. When reconstructing the video, the original amplitude information is retained, and only the phase signal is enhanced, avoiding the generation of artifacts caused by excessive processing and ensuring the authenticity of the visualization results. The inverse transform reconstruction method is used to ensure the temporal and spatial continuity of the processed video, so that the enhanced vascular pulsation presents a natural and smooth visual effect, which improves the intuitiveness of clinical observation.

[0074] Finally, the amplified phase information and the original amplitude signal are used to reconstruct the video signal to obtain the result of the amplified blood vessel pulsation: Specifically, amplitude mainly carries the structure, color, and brightness information of the tissue, while phase mainly carries displacement and edge information. By amplifying only the phase, the system achieves selective enhancement of motion information while retaining the original visual characteristics of the tissue. Traditional direct amplification of pixel values will amplify both noise and artifacts. In contrast, the noise growth rate of the phase amplification method is greatly reduced because noise behaves differently in the phase domain and amplitude domain, and phase amplification mainly affects structured signals (such as blood vessel pulsation) rather than random noise.

[0075] The above schematically describes the invention of the present application and its implementation methods. This description is not restrictive. Without departing from the spirit or basic features of the present application, the present application can be implemented in other specific forms. What is shown in the drawings is only one of the implementation methods of the invention of the present application. The actual structure is not limited to this. Any figure mark in the claims should not limit the claims involved. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the present invention, a structural method and embodiment similar to the technical solution without creativity should fall within the scope of protection of this patent. In addition, the word "including" does not exclude other elements or steps, and the word "one" before an element does not exclude the inclusion of "multiple" elements. The multiple elements stated in the product claim can also be implemented by one element through software or hardware. Words such as first and second are used to indicate names and do not indicate any specific order.

Claims

1. A method for visualizing blood vessel pulsation, characterized in that: include: Acquiring video data of physiological tissue containing blood vessels, wherein the video data of the physiological tissue comprises a multi-frame image sequence; Convert each frame image from RGB color space to YIQ color space; Separate the luminance signal Y and the chrominance signals I and Q from the YIQ color space; Each frame of the image in the YIQ color space is spatially decomposed using a filter bank of a complex controllable pyramid to obtain subband sequences of different scales and directions; Extract the amplitude spectrum signal A from the subband sequence r,θ (x,y,t) and phase spectrum signal B r,θ (x,y,t); Construct a vascular pulsation filter VPF according to the user's pulse frequency; The phase spectrum signal B is filtered by the vascular pulsation filter VPF. r,θ (x, y, t) is processed to extract the vascular pulsation phase change signal; According to the blood vessel pulsation phase change signal and the amplitude spectrum signal A r,θ (x, y, t), reconstructed through inverse transformation to obtain the reconstructed vascular pulsation video.

2. The method for visualizing blood vessel pulsation according to claim 1, wherein: Extracting amplitude spectrum signals and phase spectrum signals from the subband sequence, including: Calculate the total number of spatial decomposition layers N; According to the bandpass template and direction template of the complex controllable pyramid, an N-layer M-directional filter bank is constructed; Using the filter bank, each frame image is decomposed into N layers of space to obtain the spectrum F after N layers of M directions of frequency domain filtering. r,θ (u,v); According to the spectrum F after frequency domain filtering r,θ (u,v), respectively extract the amplitude spectrum signal A containing amplitude information r,θ (x, y, t), and the phase spectrum signal B containing the phase r,θ (x,y,t).

3. The method for visualizing blood vessel pulsation according to claim 2, wherein: Get the spectrum F after frequency domain filtering in N layers and M directions r,θ (u,v) uses the following formula: Among them, u and v represent frequency coordinates, C r,θ represents the frequency domain filter with scale r and direction θ in the filter bank, x and y are spatial coordinates, h and w represent the width and height of the image respectively, I(x,y,t) represents the input image, t represents time, and j is the imaginary unit.

4. The method for visualizing blood vessel pulsation according to claim 2, wherein: Extract the amplitude spectrum signal A containing amplitude information r,θ (x, y, t), and the phase spectrum signal B containing the phase r,θ (x,y,t) uses the following formula: A r,θ (x,y,t)=∥I r,θ (x,y,t)∥ B r,θ (x,y,t)=arg[I r,θ (x,y,t)] Among them, I r,θ (x, y, t) represents the complex spectrum of the image space after filtering, ∥*∥ represents the modulo operation, which calculates the absolute value of the complex number, and arg represents the phase operation, which calculates the angle of the complex number.

5. The method for visualizing blood vessel pulsation according to claim 1, wherein: The extracted vascular pulsation phase change signal includes: According to the user's pulse frequency, the acceleration filter kernel H(x, t) and the jerk filter kernel J(x, t) are constructed respectively; According to the acceleration filter kernel H(x,t) and the jerk filter kernel J(x,t), a vascular pulsation filter VPF is constructed; The phase spectrum signal B r,θ (x, y, t) is used as input, and the vascular pulsation filter VPF is used to extract the vascular pulsation phase change signal.

6. The method for visualizing blood vessel pulsation according to claim 5, wherein: Construct the acceleration filter kernel H(x,t), including: Get the user's pulse frequency f and video frame rate r respectively; Calculate the Gaussian-Laplacian operator based on the Gaussian filter kernel; The acceleration filter kernel H(x,t) is constructed based on the user's pulse frequency f, the video frame rate r, and the Gaussian-Laplacian operator.

7. The method for visualizing blood vessel pulsation according to claim 6, wherein: Based on the user's pulse frequency f, the video frame rate r, and the Gaussian-Laplacian operator, the acceleration filter kernel H(x,t) is constructed, including: Among them, t represents time; x represents location information; G σ (t) represents the Gaussian filter kernel.

8. The method for visualizing blood vessel pulsation according to claim 5, wherein: Construct the jerk filter kernel J(x,t), including: Calculate the jerk of the phase with respect to time t; Convert jerk to smoothness σ (x,t); According to smoothness σ (x, t), construct the jerk filter kernel J(x, t); According to the hierarchical structure of the complex controllable pyramid, the jerk filter kernel of each layer is constructed Where l represents the level of the pyramid, σ is the standard deviation, and β is set as a hyperparameter to adjust the sensitivity of the filter; For each layer of the jerk filter kernel Perform downsampling correction: Where λ is the downsampling factor, 0<λ<1; The jerk filter kernel of each layer after downsampling correction Perform propagation correction to obtain the final jerk filter kernel of each layer: Where N is the number of pyramid levels used for correction, and the function The filter at pyramid level i is resized to the size of pyramid level l using bicubic interpolation.

9. The method for visualizing blood vessel pulsation according to claim 8, wherein: Construct a vascular pulsation filter VPF, including: VPF f,σ,λ,l (x,t)=H f (x,t)×J σ,λ,l (x,t) Among them, VPF f,σ,λ,l (x, t) is the blood vessel pulsation filter, σ is the standard deviation, f is the pulse frequency, λ is the downsampling factor, l is the level of the pyramid, C f,l (x,t) represents the phase after amplification, Indicates the phase at a certain level of the pyramid, is the convolution operation.

10. A blood vessel pulsation visualization system, characterized in that: include: An acquisition module, which acquires physiological tissue video data including blood vessels, wherein the physiological tissue video data includes a multi-frame image sequence; The color conversion module converts each frame of image from RGB color space to YIQ color space, and separates the brightness signal Y, and the chrominance signals I and Q from the YIQ color space; The spatial decomposition module uses a filter bank of a complex controllable pyramid to perform spatial decomposition on each frame of the image in the YIQ color space to obtain subband sequences of different scales and directions; A signal extraction module extracts amplitude spectrum signals and phase spectrum signals from the sub-band sequence; The filter module constructs a vascular pulsation filter VPF according to the user's pulse frequency, processes the phase spectrum signal using the vascular pulsation filter VPF, and extracts the vascular pulsation phase change signal; The video reconstruction module amplifies the blood vessel pulsation phase change signal and reconstructs the blood vessel pulsation video based on the amplified blood vessel pulsation phase change signal and the amplitude spectrum signal.

Images