A varicocele blood vessel classification method based on intelligent image analysis technology

Abstract

The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical field of medical image intelligent processing. The application discloses a varicocele blood vessel classification method based on intelligent image analysis technology, and relates to the technical

Description

Technical Field

[0001] This invention relates to the field of intelligent medical image processing technology, and in particular to a method for classifying varicocele vessels based on intelligent image analysis technology. Background Technology

[0002] The clinical diagnosis of varicocele primarily relies on ultrasound imaging assessment, with color Doppler ultrasound being the preferred method due to its non-invasiveness and real-time capability. Conventional methods involve measuring vein diameter, observing blood reflux, and assessing severity using the Dubin grading system. Current techniques generally employ fundamental-wave ultrasound imaging to capture vascular morphology and structure, supplemented by pulsed Doppler to detect hemodynamic parameters, forming a combined morphological-functional diagnostic system. Technological advancements in this field focus on improving image resolution and blood flow sensitivity, aiming to provide clinicians with more precise anatomical and blood flow information.

[0003] Despite the high maturity of existing technologies, there are still two areas for optimization. First, fundamental wave imaging and hemodynamic parameters are mostly analyzed independently, lacking dynamic coupling modeling in the spatial-temporal dimension, which may overlook the correlation between vascular wall micro-vibration and blood flow abnormalities. Second, diagnostic conclusions are easily influenced by physicians' subjective judgment, especially in borderline cases (such as between grade II and III), where traditional methods do not integrate dynamic confidence mechanisms to quantify the reliability of grading. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a varicocele vessel classification method based on intelligent image analysis technology to solve the problems of insufficient multimodal data collaboration and the dependence of hierarchical stability on experience.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for classifying varicocele vessels based on intelligent image analysis technology. The method includes: acquiring raw radio frequency signal data and preprocessing it to obtain a fundamental wave video stream and a harmonic video stream; extracting vessel wall contour feature points from the fundamental wave video stream, simultaneously performing phase-locked amplification on these feature points in the harmonic video stream, calculating vibration energy values, and generating a harmonic energy heatmap; extracting spatial structural and temporal features from the harmonic energy heatmap and the fundamental wave video stream to obtain a dynamic coupling feature tensor, and performing time-shifted recombination on the dynamic coupling feature tensor to calculate the probability of varicocele grading; calculating a confidence index based on the dynamic coupling feature tensor, and dynamically testing the probability of varicocele grading based on the confidence index to determine the clinical grade and generate a clinical grading decision spectrum; performing an affine transformation on the harmonic energy heatmap and marking high-energy risk areas to generate an enhanced heatmap, and simultaneously coupling the enhanced heatmap with the clinical grading decision spectrum to generate an intelligent classification report for varicocele.

[0008] As a preferred embodiment of the varicocele vascular classification method based on intelligent image analysis technology described in this invention, the original radio frequency signal data includes a timestamp sequence, modulation amplitude value, and original phase value;

[0009] The preprocessing includes bandpass filtering, quadrature demodulation, envelope detection, scan conversion, and motion compensation.

[0010] As a preferred embodiment of the varicocele vessel classification method based on intelligent image analysis technology described in this invention, the steps for calculating the vibration energy value are as follows:

[0011] Based on the fundamental video stream, the spermatic vein lumen region is extracted to obtain a binarized mask image;

[0012] The binarized mask image is subjected to morphological skeletonization to generate the initial blood vessel centerline. Based on the initial blood vessel centerline, the contour feature points of the blood vessel wall are selected, and the initial coordinate set of the feature points is output.

[0013] Based on the initial coordinate set of feature points, the movement trajectory of the feature points of the blood vessel wall contour is tracked to obtain the trajectory coordinate sequence;

[0014] Based on the trajectory coordinate sequence, pixel intensity values ​​are extracted from the harmonic video stream to generate a time-intensity two-dimensional matrix. The time-intensity two-dimensional matrix is ​​then subjected to bandpass filtering to calculate the vibration energy value.

[0015] As a preferred embodiment of the varicocele vascular classification method based on intelligent image analysis technology described in this invention, the generation of harmonic energy heat map refers to mapping vibration energy values ​​to a three-dimensional grid and filling the spatial gaps through bilinear interpolation to generate a harmonic energy heat map.

[0016] As a preferred embodiment of the varicocele vessel classification method based on intelligent image analysis technology described in this invention, the steps for obtaining the dynamically coupled feature tensor are as follows:

[0017] The three-dimensional coordinates and vibration intensity values ​​of each pixel in the harmonic energy heatmap are extracted, and the blood vessel diameter measurement and blood flow direction identification are collected based on the fundamental wave video stream to generate a joint vibration-morphology dataset.

[0018] Based on the vibration-morphology joint dataset, the vibration energy growth rate and vasodilation rate are calculated, and the number of abrupt changes in blood flow direction is counted to obtain the dynamic coupling feature tensor.

[0019] As a preferred embodiment of the varicocele vascular classification method based on intelligent image analysis technology described in this invention, the steps for calculating the varicocele grading probability are as follows:

[0020] The dynamic coupling feature tensor is subjected to cyclic shifting to generate a time-shifted coupling feature tensor.

[0021] Based on time-shift coupled feature tensors and preset clinical rules, the probability of varicocele grading is calculated.

[0022] As a preferred embodiment of the varicocele vascular classification method based on intelligent image analysis technology described in this invention, the calculation of the confidence index refers to extracting the effective time window proportion parameter in the dynamic coupling feature tensor and calculating the confidence index through a linear weighting method.

[0023] As a preferred embodiment of the varicocele vessel classification method based on intelligent image analysis technology described in this invention, the steps for generating the clinical grading decision spectrum are as follows:

[0024] Based on the confidence index, the probability of varicocele grading is dynamically corrected using linear interpolation, and the corrected grading probability set is output.

[0025] Based on the corrected grading probability set, a threshold determination is made for the probability peak distribution to generate a clinical grading decision spectrum.

[0026] As a preferred embodiment of the varicocele vessel classification method based on intelligent image analysis technology described in this invention, the steps for generating the enhanced heatmap are as follows:

[0027] The harmonic energy heatmap is mapped to the anatomical coordinate system of the fundamental video stream through affine transformation, generating a spatially corrected heatmap matrix.

[0028] Based on the spatially corrected heatmap matrix, the dynamic vibration energy threshold is calculated, and high-energy risk areas are marked according to the dynamic vibration energy threshold to generate an enhanced heatmap.

[0029] As a preferred embodiment of the varicocele vascular classification method based on intelligent image analysis technology described in this invention, the generation of the intelligent classification report for varicocele refers to superimposing and fusing the enhanced heat map with the key frames of the fundamental video stream to generate a joint view, and embedding the clinical grading decision spectrum in the joint view to generate the intelligent classification report for varicocele.

[0030] The beneficial effects of this invention are as follows: By extracting the spatial structure and time-frequency features of the harmonic energy heatmap and the fundamental wave video stream, the dynamic coupling feature tensor is obtained and time-shifted and recombined, realizing the spatiotemporal dynamic fusion of vascular structure and vibration energy, thereby calculating the probability of varicocele grading and improving the accuracy of pathological feature capture and the consistency of grading; at the same time, based on the confidence index calculated by the dynamic coupling feature tensor, the grading probability is dynamically tested and a clinical grading decision spectrum is generated, constructing a quantitative and reliable grading verification mechanism, and enhancing the credibility and anti-interference ability of clinical decision-making. Attached Figure Description

[0031] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart of a varicocele vein classification method based on intelligent image analysis technology;

[0033] Figure 2 Flowchart for generating a harmonic energy thermogram;

[0034] Figure 3 The flowchart shows the calculation of dynamically coupled feature tensors and hierarchical probabilities.

[0035] Figure 4 This is a flowchart for clinical triage decision spectrum and intelligent report generation. Detailed Implementation

[0036] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0037] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0038] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0039] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for classifying varicocele vessels based on intelligent image analysis technology, including the following steps:

[0040] S1. Acquire raw radio frequency signal data and perform preprocessing to obtain fundamental and harmonic video streams;

[0041] The raw radio frequency signal data includes a timestamp sequence, modulation amplitude values, and raw phase values;

[0042] It should be noted that the timestamp sequence is a time axis coordinate sequence generated by recording the transmission / reception time points of each frame of radio frequency signal using the clock of the ultrasound equipment at 0.1μs intervals; the modulation amplitude value is the amount of modulation of the carrier amplitude by tissue reflection obtained by extracting the envelope information of the original radio frequency signal through orthogonal demodulation processing; the original phase value is the original phase offset captured by directly reading the instantaneous phase difference between the transmitted wave and the echo signal through a phase-locked loop circuit.

[0043] Preprocessing includes bandpass filtering, quadrature demodulation, envelope detection, scan conversion, and motion compensation;

[0044] It should be noted that bandpass filtering refers to frequency screening of the original radio frequency signal to separate the fundamental and harmonic components; quadrature demodulation refers to mixing the filtered fundamental / harmonic signals with a reference carrier (a standard sine / cosine signal with the same frequency as the transmitted ultrasound, synchronously generated by the clock generator of the ultrasound equipment) to generate a complex signal containing real and imaginary parts; envelope detection refers to directly extracting the signal amplitude envelope from the complex signal output by quadrature demodulation to generate polar coordinate echo data (containing amplitude and angle information); scan conversion refers to mapping the polar coordinate echo data to the Cartesian coordinate system; motion compensation refers to spatiotemporal registration of the fundamental and harmonic video streams to eliminate inter-frame offset caused by probe displacement.

[0045] S2. Based on the fundamental wave video stream, extract the contour feature points of the blood vessel wall, and simultaneously perform phase-locked amplification processing on the contour feature points of the blood vessel wall in the harmonic video stream to calculate the vibration energy value and generate a harmonic energy heat map.

[0046] Based on the fundamental video stream, the spermatic vein lumen region is extracted to obtain a binarized mask image;

[0047] Furthermore, in the first frame of the fundamental video stream, a rectangular region of interest (ROI) containing the spermatic vein is generated using the testicular mediastinum as an anatomical landmark. The ROI coordinate range is output, and the grayscale image of the corresponding region is extracted based on the ROI coordinate range. A Gaussian weighted histogram of the pixel grayscale values ​​within the ROI coordinate range is calculated, and the grayscale value corresponding to the valley of the bimodal distribution is selected as the segmentation threshold. The image region within the ROI coordinate range is classified into pixels at the pixel level according to the segmentation threshold, generating an initial binary mask image. The initial binary mask image is then processed sequentially with noise reduction and region growing to fill gaps, and an optimized binary mask image is output.

[0048] The expression for calculating the Gaussian-weighted histogram of pixel grayscale values ​​is:

[0049]

[0050] Where H(g) is the weighted frequency statistical value of the target gray level g; g is the target gray level; I c σ is the grayscale value of the c-th pixel within the ROI coordinate range; σ is the Gaussian kernel width; N is the grayscale value of I within the ROI coordinate range. c The number of pixels; It is the Gaussian distribution normalization factor; π is an irrational number representing the ratio of the circumference to the diameter of a circle.

[0051] The binarized mask image is subjected to morphological skeletonization to generate the initial blood vessel centerline. Based on the initial blood vessel centerline, the contour feature points of the blood vessel wall are selected, and the initial coordinate set of the feature points is output.

[0052] Furthermore, the binarized mask image is first subjected to morphological skeletonization: boundary pixels are eliminated through 8-neighbor iteration until a connected skeleton with a single pixel width is obtained, generating an initial binary image of the vessel centerline. The initial binary image of the vessel centerline is then scanned, and the spatial coordinates of all pixels with a value of 1 (in the binary mask image, pixels with a value of 1 represent the vessel lumen region, while pixels with a value of 0 represent the background or other tissues) are recorded, generating a disordered set of vessel centerline coordinate points. Then, a depth-first search is performed on the disordered set of vessel centerline coordinate points to establish the topological relationship of the vessel branches. The ordered coordinate points are resampled along the main branch path at fixed arc length intervals, and the ordered coordinate sequence of the vessel centerline is output. Based on the coordinate points of each point in the ordered coordinate sequence of the vessel centerline, the boundary points of the binary mask image are searched in the vertical direction to both sides to obtain the contour point pairs on the left and right sides, and finally the initial coordinate set of feature points is output.

[0053] It should be noted that 8-neighbor iterative elimination refers to the topology optimization process in the process of thinning a binary image, which involves detecting the connectivity of 8 adjacent pixels (top, bottom, left, right and 4 diagonal points) within a 3×3 window centered on the target pixel, and gradually deleting boundary pixels through multiple rounds of scanning until a single-pixel wide skeleton is generated.

[0054] Based on the initial coordinate set of feature points, the movement trajectory of the feature points of the blood vessel wall contour is tracked to obtain the trajectory coordinate sequence;

[0055] Furthermore, based on the initial coordinate set of feature points, spatial position coordinates, gray-level gradient vectors, and local radius of curvature are extracted for each feature point to generate a feature point motion parameter table. According to the feature point motion parameter table, the displacement change of each feature point is tracked between consecutive frames of the fundamental video stream, and a single-frame displacement increment sequence is output. Based on the single-frame displacement increment sequence, the probe displacement and the blood vessel's own motion are separated. The global displacement is corrected through spatial transformation, and the compensated feature point local motion trajectory fragment is output. The compensated feature point local motion trajectory fragment is then processed by Kalman filtering to eliminate high-frequency jitter noise. At the same time, the trajectory data of all frames are integrated according to the timestamp to output the trajectory coordinate sequence of the blood vessel wall contour feature points.

[0056] Based on the trajectory coordinate sequence, pixel intensity values ​​are extracted from the harmonic video stream to generate a time-intensity two-dimensional matrix. The time-intensity two-dimensional matrix is ​​then subjected to bandpass filtering to calculate the vibration energy value.

[0057] The expression for calculating the vibration energy value is:

[0058]

[0059] Where E is the vibrational energy value; I t μ is the pixel grayscale value within the time window t; N is the total number of frames in the analysis period; μ I It is within N frames I t The average value; I t -μ I It is the difference between the instantaneous intensity and the mean.

[0060] Furthermore, based on the trajectory coordinate sequence of the feature points of the blood vessel wall contour, the pixel intensity value is extracted from the corresponding timestamp and spatial position of the harmonic video stream to generate the original time-intensity two-dimensional matrix. The original time-intensity two-dimensional matrix is ​​then subjected to 3-7Hz bandpass filtering to retain the target frequency band signal and output the filtered time-intensity two-dimensional matrix. At the same time, the vibration energy value of each feature point is calculated based on the filtered time-intensity two-dimensional matrix.

[0061] Vibration energy values ​​are mapped to a three-dimensional grid, and spatial gaps are filled by bilinear interpolation to generate a harmonic energy heat map.

[0062] Furthermore, based on the vibration energy values ​​and trajectory coordinate sequences, a three-dimensional grid coordinate system covering the spatial range and time series of blood vessels is established, generating an empty three-dimensional grid matrix. The vibration energy values ​​are then filled into the corresponding nodes of the empty three-dimensional grid matrix according to their corresponding time-space locations, forming a non-uniform grid matrix containing discrete vibration energy values. Based on the blank nodes in the non-uniform grid matrix, the vibration energy values ​​of the four adjacent known nodes around the blank node are extracted and weighted and fused according to the spatial proportion of their relative positions to fill the blank nodes in the grid, generating a complete three-dimensional energy grid matrix. The three-dimensional energy grid matrix is ​​then projected onto the space-time plane and converted through color mapping to output the final harmonic energy heat map.

[0063] S3. Extract the spatial structure and time-frequency features of the harmonic energy heat map and the fundamental video stream, obtain the dynamic coupling feature tensor, and perform time-shifted recombination of the dynamic coupling feature tensor to calculate the grading probability of varicocele.

[0064] The three-dimensional coordinates and vibration intensity values ​​of each pixel in the harmonic energy heatmap are extracted, and the blood vessel diameter measurement and blood flow direction identification are collected based on the fundamental wave video stream to generate a joint vibration-morphology dataset.

[0065] Furthermore, the three-dimensional coordinates and corresponding timestamps of each pixel in the harmonic energy heatmap are extracted, and a four-dimensional data table is generated by combining the vibration energy values. Within the ROI coordinate range, blood vessel edge detection and color Doppler analysis are performed on keyframes of the fundamental wave video stream: the distance between the two edges is measured along the vertical direction of the initial blood vessel centerline to obtain blood vessel diameter data, and color Doppler pixel values ​​are extracted from the keyframes of the fundamental wave video stream. The blood flow direction is identified by the spatial distribution of red and blue pixels (forward: red pixel clustering > blue; reverse: blue > red; turbulent: red and blue discrete mixture), and a blood vessel morphology parameter table is output. The four-dimensional data table and the blood vessel morphology parameter table are aligned by matching timestamps and spatial coordinates, and blood vessel diameter and blood flow direction data are supplemented to generate a vibration-morphology joint dataset.

[0066] It should be noted that color Doppler analysis is a blood flow imaging technique based on the ultrasound Doppler effect. It realizes the detection of intravascular hemodynamics by encoding blood flow velocity and direction in real time (red: flow towards the probe, blue: away from the probe). The keyframe of the fundamental wave video stream refers to the image at a specific time point in the fundamental wave video stream where the rate of change of blood vessel diameter or abrupt change in blood flow direction reaches a local extreme value (based on the frequency distribution data of the rate of change of blood vessel diameter and abrupt change in blood flow from multi-center clinical studies, the 90th percentile value is used as the boundary for pathological abnormality judgment).

[0067] Based on the vibration-morphology joint dataset, the vibration energy growth rate and vasodilation rate are calculated, and the number of abrupt changes in blood flow direction is counted to obtain the dynamic coupling feature tensor.

[0068] The expression for calculating the vibration energy growth rate is:

[0069]

[0070] Where R is the vibration energy growth rate; E t It is the average vibrational energy of the current time window t; E t-Δt It is the average vibration energy of the previous time window t; Δt is the time difference between adjacent time windows;

[0071] The expression for calculating the rate of vasodilation is:

[0072]

[0073] Where V is the rate of vasodilation; D t D is the median diameter of blood vessels in the current time window t; t-Δt It is the median diameter of blood vessels in the previous time window t;

[0074] Furthermore, based on the vibration-morphology joint dataset, the data is first segmented into 200ms time windows, and the vibration energy sequence and vessel diameter sequence within each time window are extracted to generate a time window dataset. Then, based on the change in vibration energy values ​​between adjacent time windows, the energy growth rate is calculated to generate a vibration energy growth rate sequence. Simultaneously, the vessel dilation rate is calculated based on the ratio of the change in vessel diameter to the time window interval to generate a vessel dilation rate sequence. The number of times the blood flow direction changes from forward to reverse or turbulent within each time window is counted, and a blood flow direction abrupt change event table is output. Finally, the vibration energy growth rate sequence, vessel dilation rate sequence, and blood flow direction abrupt change event table are integrated according to time windows to construct a dynamically coupled feature tensor containing time window number, spatial location, energy growth rate, dilation rate, and number of abrupt changes.

[0075] The dynamic coupling feature tensor is subjected to cyclic shifting to generate a time-shifted coupling feature tensor.

[0076] Furthermore, based on the dynamic coupling feature tensor, the total number of time windows and spatial points are extracted to generate a tensor metadata table. The cyclic displacement is calculated based on the total number of time windows in the tensor metadata table to ensure that the minimum displacement is one time window (ensuring coverage of at least one complete cardiac cycle segment; clinical validation shows that displacements shorter than 200ms cannot effectively capture pathologically related phase differences). A cyclic right shift operation is performed on the dynamic coupling feature tensor in the time window dimension according to the cyclic displacement to generate an initial time-shift coupling feature tensor. The initial time-shift coupling feature tensor is then matched and validated with the time window dataset for timestamp and spatial location matching, and the validated time-shift coupling feature tensor is output.

[0077] The expression for calculating the cyclic displacement is:

[0078]

[0079] Where S is the cyclic displacement; f is the sampling frequency coefficient; and O is the patient's real-time heart rate.

[0080] Based on time-shift coupled feature tensors and preset clinical rules, the probability of varicocele grading is calculated.

[0081] The expression for calculating the probability of varicocele grading is as follows:

[0082]

[0083] Where P is the probability of varicocele grading; F is the frequency of blood flow direction abrupt change; T1 is the vibration energy threshold; T2 is the critical value for vasodilation; T3 is the threshold for blood flow direction abrupt change; α is the contribution weight of vibration energy feature; β is the contribution weight of vasodilation feature; γ is the contribution weight of blood flow direction feature; C is the correction coefficient for the proportion of effective data; and E is the vibration energy value.

[0084] Furthermore, based on the time-shift coupling feature tensor, the varicocele grading criteria are first loaded from the pre-stored clinical rules, and the vibration energy threshold, vasodilation critical value, and blood flow direction change frequency threshold are extracted to generate a quantified threshold table. The time-shift coupling feature tensor is then matched item by item with the quantified threshold table, and feature combinations that meet the diagnostic conditions for each grade are labeled, outputting a matching result label matrix. The features in the matching result label matrix are weighted and accumulated according to preset weights to calculate the probability of varicocele grading.

[0085] It should be noted that the preset clinical rules are quantitative standards based on the international consensus on the diagnosis and treatment of varicocele, including vibration energy threshold, vasodilation threshold, and blood flow direction mutation frequency threshold; the preset weights are fixed coefficients allocated according to the contribution of clinicopathological features (vibration energy weight 0.6, vasodilation weight 0.3, blood flow mutation weight 0.1), and have been verified and optimized through multi-center studies.

[0086] S4. Based on the dynamic coupling feature tensor, calculate the confidence index, and perform dynamic testing on the probability of varicocele grading according to the confidence index to determine the clinical grade and generate a clinical grading decision spectrum.

[0087] Extract the effective time window proportion parameter from the dynamic coupling feature tensor and calculate the confidence index using the linear weighting method;

[0088] The expression for calculating the confidence index is:

[0089]

[0090] Where M is the confidence index; Q is the percentage of valid data; M1 is the confidence threshold (determined through multicenter clinical studies, it is the minimum valid percentage threshold for determining data integrity, with an exemplary value range of [0.8, 0.95]);

[0091] Furthermore, the integrity of vibration energy values, vasodilation rates, and blood flow direction data within each time window in the scanning dynamic coupling feature tensor is verified. Valid and invalid time windows are marked, a list of valid time window markers is generated, and the number of valid windows is counted based on the list. The percentage of valid data is calculated. The percentage of valid data is then transformed using a linear weighting formula, and the final confidence index is output.

[0092] Based on the confidence index, the probability of varicocele grading is dynamically corrected using linear interpolation, and the corrected grading probability set is output.

[0093] Furthermore, based on the confidence index and the probability of varicocele grading, the confidence index is mapped to a probability correction coefficient using linear interpolation (for example: coefficient = 1.0 when C ≥ 0.9; coefficient = C / 0.9 when 0.8 ≤ C < 0.9; coefficient = 0.8 when C < 0.8). The probability values ​​of each grade in the varicocele grading probability are scaled using the probability correction coefficient to achieve dynamic probability decay. Finally, the decayed probability values ​​are normalized to output the corrected grading probability set. The specific mathematical formula is as follows.

[0094]

[0095] Where G is the corrected hierarchical probability set; P1 is the conservative probability benchmark;

[0096] It should be noted that the conservative probability benchmark is the probability safety distribution of varicocele grading, which is based on the probability distribution characteristics of asymptomatic and mild cases through multicenter clinical studies and has been reviewed and confirmed by an expert consensus meeting.

[0097] Based on the corrected grading probability set, a threshold determination is made on the probability peak distribution to generate a clinical grading decision spectrum;

[0098] Furthermore, based on the corrected grading probability set, the maximum probability value of each level is extracted to generate a probability peak table. The probability peak table is then compared with the preset clinical threshold table to output a threshold judgment result table. If there are multiple level over-threshold conflicts in the threshold judgment result table (for example, both level III probability 88% and level II probability 78% exceed the threshold), the conflicts are resolved according to the priority of level III > level II > level I > level 0 to generate a clinical grading decision spectrum.

[0099] It should be noted that the preset clinical threshold table is a varicocele grading probability threshold established based on multicenter clinical validation. It includes the standards of Grade III ≥85%, Grade II ≥70%, Grade I ≥60%, and Grade 0 with no threshold (the Grade III probability threshold (85%) was selected through receiver operating characteristic curve analysis of multicenter clinical studies, the Grade II threshold (70%) with the maximum Youden index, and the Grade I threshold (60%) with the sensitivity >90%).

[0100] S5. Perform an affine transformation on the harmonic energy heatmap and mark high-energy risk areas to generate an enhanced heatmap. At the same time, couple the enhanced heatmap with the clinical grading decision spectrum to generate an intelligent classification report for varicocele.

[0101] The harmonic energy heatmap is mapped to the anatomical coordinate system of the fundamental video stream through affine transformation, generating a spatially corrected heatmap matrix.

[0102] Furthermore, based on the initial coordinate set of feature points, a right-handed anatomical coordinate system is established with the center point of the spermatic vein entrance as the origin, the main vein axis direction as the Z-axis unit vector, and the testicular mediastinal tangent as the X-axis unit vector, and the anatomical coordinate system parameters are output. The image row and column coordinates of each pixel in the harmonic energy heatmap are analyzed and converted into physical coordinates through the probe spatial calibration parameters, and the heatmap physical coordinate sequence is output. According to the anatomical coordinate system parameters and the heatmap physical coordinate sequence, the high-energy region (energy value > 85 percentile) is selected as the priority alignment target, and the optimal spatial alignment is performed to generate an affine transformation matrix. The heatmap physical coordinate sequence is input into the affine transformation matrix for spatial transformation to generate a spatially corrected heatmap matrix.

[0103] It should be noted that the probe spatial calibration parameters are the core variables for generating the affine transformation matrix, obtained through factory calibration and clinical validation. These include the probe tilt compensation coefficient, the sound velocity tissue attenuation correction factor, and the translation amount, which are used to map the image pixel coordinates to physical space coordinates.

[0104] Based on the spatially corrected heatmap matrix, the dynamic vibration energy threshold is calculated, and high-energy risk areas are marked according to the dynamic vibration energy threshold to generate an enhanced heatmap;

[0105] The expression for calculating the dynamic vibration energy threshold is:

[0106]

[0107] Where K is the dynamic vibration energy threshold; E1 is the set of energy values ​​of all pixels in the heatmap slice; E2 is the vibration energy threshold; L1 is the percentile threshold for normal energy scenarios; L2 is the percentile threshold for high energy scenarios; max(E1) is the maximum energy value in E1; percentile(,) is the percentile function.

[0108] Furthermore, based on the risk time zone timestamps in the clinical grading decision spectrum, heatmap slices corresponding to the time period are extracted from the spatially corrected heatmap matrix to generate time window heatmap slices. The vibration energy values ​​of all pixels in the time window heatmap slices are then scanned to extract the maximum energy value. Simultaneously, the maximum energy value is compared with a preset vibration energy threshold: if the maximum energy value is less than or equal to the vibration energy threshold, the percentile threshold of the conventional energy scene is used (based on energy distribution analysis of healthy subjects in a multi-center clinical study, the 85th percentile value with a specificity >90% is selected as the conventional energy field). If the threshold value exceeds the threshold value, a high-energy scene threshold is used (based on the energy saturation correction of pathological samples; to prevent mislabeling of high-energy areas, the 90th percentile value when the sensitivity is >95% is used as the high-energy scene threshold). The dynamic vibration energy threshold is calculated; pixels with vibration energy values ​​exceeding the vibration energy threshold are marked as high-energy risk areas in the time window heatmap slice, and the original enhanced heatmap is output; finally, the coordinates of the high-energy risk areas in the original enhanced heatmap are spatially topologically verified with the initial coordinate set of feature points to remove isolated noise points and generate the verified enhanced heatmap.

[0109] The enhanced heatmap is overlaid and fused with key frames of the fundamental video stream to generate a joint view. A clinical grading decision spectrum is then embedded in the joint view to generate an intelligent classification report for varicocele.

[0110] Furthermore, based on the risk time zone timestamps in the clinical grading decision spectrum, keyframe images of the corresponding time points are extracted from the fundamental wave video stream to generate a fundamental wave keyframe sequence. The validated enhanced heatmap is spatially registered with the fundamental wave keyframe sequence. High-energy risk areas are superimposed onto the fundamental wave keyframes with 30% transparency (the optimal balance point validated by multi-center clinical trials, which ensures clear display of fundamental wave anatomical structures while making high-risk areas of the heatmap more significantly visible) to generate a combined view sequence. The clinical grading decision spectrum is embedded in the combined view sequence to generate a combined view with grading labels. The combined view with grading labels is then encapsulated according to the DICOM SR standard to generate an intelligent classification report for varicocele.

[0111] This embodiment also provides a computer device applicable to the classification method of varicocele vessels based on intelligent image analysis technology, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the classification method of varicocele vessels based on intelligent image analysis technology as proposed in the above embodiment.

[0112] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0113] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the varicocele vessel classification method based on intelligent image analysis technology as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0114] In summary, this invention achieves spatiotemporal dynamic fusion of vascular structure and vibration energy by extracting the spatial structure and time-frequency features of harmonic energy heatmaps and fundamental wave video streams, obtaining dynamically coupled feature tensors, and time-shifting and recombining them. This enables the calculation of the grading probability of varicocele, improving the accuracy of pathological feature capture and the consistency of grading. Simultaneously, based on the confidence index calculated by the dynamically coupled feature tensor, the grading probability is dynamically tested and a clinical grading decision spectrum is generated, constructing a quantitative and reliable grading verification mechanism, thereby enhancing the credibility and anti-interference ability of clinical decisions.

[0115] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for classifying varicocele vessels based on intelligent image analysis techniques, characterized by: The application relates to a method for intelligent classification of varicocele, comprising the following steps: Collecting original radio frequency signal data and performing pretreatment to obtain a fundamental wave video stream and a harmonic wave video stream; According to the fundamental wave video stream, extracting a blood vessel wall contour feature point, performing phase-locked amplification processing on the blood vessel wall contour feature point in the harmonic wave video stream, calculating a vibration energy value, and generating a harmonic energy heat map; Extracting spatial structure features and time-frequency features of the harmonic energy heat map and the fundamental wave video stream, obtaining a dynamic coupling feature tensor, and the steps are as follows, Extracting three-dimensional coordinates and vibration intensity values of each pixel point in the harmonic energy heat map, collecting blood vessel diameter measurement values and blood flow direction identifiers according to the fundamental wave video stream, and generating a vibration-morphology combined dataset; According to the vibration-morphology combined dataset, calculating a vibration energy growth rate and a blood vessel expansion rate, and counting the number of blood flow direction mutations to obtain a dynamic coupling feature tensor; Time-shifting and reorganizing the dynamic coupling feature tensor to calculate a varicocele grading probability, and the steps are as follows, Performing cyclic displacement processing on the dynamic coupling feature tensor to generate a time-shifted coupling feature tensor; Based on the time-shifted coupling feature tensor and a preset clinical rule, a varicocele grading probability is calculated; According to the dynamic coupling feature tensor, a confidence index is calculated, and the varicocele grading probability is dynamically tested according to the confidence index to determine a clinical grade and generate a clinical grading decision spectrum; Performing affine transformation on the harmonic energy heat map and marking a high-energy risk area to generate an enhanced heat map, and coupling the enhanced heat map with the clinical grading decision spectrum to generate an intelligent classification report of varicocele.

2. The method of classifying varicocele vessels based on intelligent image analysis techniques as claimed in claim 1, wherein: The original radio frequency signal data comprises a timestamp sequence, a modulation amplitude value and an original phase value; The pretreatment comprises band-pass filtering, quadrature demodulation, envelope detection, scan conversion and motion compensation.

3. The method of classifying varicocele vessels based on intelligent image analysis techniques as claimed in claim 1, wherein: The calculation of the vibration energy value comprises the following steps, According to the fundamental wave video stream, a spermatic vein lumen region is extracted to obtain a binary mask image; The binary mask image is subjected to morphological skeletonization processing to generate an initial blood vessel centerline, and the initial blood vessel centerline is used to select blood vessel wall contour feature points to output an initial coordinate set of the feature points; Based on the initial coordinate set of the feature points, the moving track of the blood vessel wall contour feature points is tracked to obtain a track coordinate sequence; According to the track coordinate sequence, the pixel intensity value in the harmonic wave video stream is intercepted to generate a time-intensity two-dimensional matrix, and the time-intensity two-dimensional matrix is subjected to band-pass filtering processing to calculate the vibration energy value.

4. The method of classifying varicocele vessels based on intelligent image analysis techniques as claimed in claim 3, wherein: The generation of the harmonic energy heat map refers to mapping the vibration energy value to a three-dimensional grid and filling the spatial gap through bilinear interpolation to generate the harmonic energy heat map.

5. The method of classifying varicocele vessels based on intelligent image analysis techniques as claimed in claim 1, wherein: The calculation of the confidence index refers to extracting an effective time window proportion parameter in the dynamic coupling feature tensor, and calculating the confidence index through a linear weighting method.

6. The method of classifying varicocele vessels based on intelligent image analysis techniques as claimed in claim 5, wherein: The generation of the clinical grading decision spectrum comprises the following steps, According to the confidence index, the varicocele grading probability is dynamically corrected through a linear interpolation method to output a corrected grading probability set; According to the corrected grading probability set, threshold determination is performed on the probability peak value distribution to generate a clinical grading decision spectrum.

7. The method of classifying varicocele vessels based on intelligent image analysis techniques as claimed in claim 1, wherein: The generation of the enhanced heat map comprises the following steps, mapping the harmonic energy heat map to the anatomical coordinate system of the fundamental video stream through affine transformation to generate a spatial correction heat map matrix; based on the spatial correction heat map matrix, calculating a dynamic vibration energy threshold, and marking a high energy risk area according to the dynamic vibration energy threshold to generate an enhanced heat map.

8. The method of classifying varicocele vessels based on intelligent image analysis techniques as claimed in claim 7, wherein: The generating of the intelligent classification report of varicocele refers to superimposing and fusing the enhanced heat map with the key frame of the fundamental video stream to generate a joint view, and embedding a clinical grading decision spectrum in the joint view to generate the intelligent classification report of varicocele.

Images