A method for assessing renal function in a patient with kidney disease using diffusion tensor imaging

Abstract

The application relates to the technical field of image processing, in particular to a method for evaluating renal function of a nephropathy patient through diffusion tensor imaging. The method comprises the following steps: obtaining a kidney diffusion tensor original image through single-excitation echo planar imaging acquisition; then comparing adjacent pixels, extracting signal intensity difference and gradient direction difference; then judging local geometric distortion according to the difference information, and carrying out correction processing pixel by pixel to obtain a corrected high-quality diffusion tensor image; finally, calculating a fitting tensor, extracting diffusion parameters of a renal cortex and a renal medulla region, and comparing and quantifying the diffusion parameters with preset renal function indexes to output an objective quantitative result of renal function of the patient. Through the pixel-by-pixel distortion correction and direction feature extraction, the precision of the diffusion tensor image and the accuracy of the kidney region division are improved, and objective quantitative analysis of the renal function is realized.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method for assessing renal function in patients with kidney disease using diffusion tensor imaging. Background Technology

[0002] Diffusion tensor imaging (DTI), an important branch of magnetic resonance imaging (MRI), can reflect the anisotropic characteristics of water molecule diffusion within biological tissues and has high application value in imaging neurological diseases and organ function. However, existing DTI techniques still have the following shortcomings in kidney imaging and functional assessment: First, because DTI typically uses single-excitation echo-plane imaging, the images are highly susceptible to interference from magnetic field inhomogeneities and patient respiratory movements, leading to geometric distortions in local areas. Most existing correction methods are based on overall adjustments, making it difficult to correct individual pixels individually, resulting in loss of detail and affecting imaging accuracy. Second, kidney tissue partitioning relies on grayscale or simple texture features, making it susceptible to noise and blurred boundaries, leading to inaccurate division of the renal cortex and medulla, and consequently, unstable diffusion parameter extraction. Summary of the Invention

[0003] Therefore, it is necessary to provide a method for assessing renal function in patients with kidney disease using diffusion tensor imaging to address at least one of the aforementioned technical problems.

[0004] To achieve the above objectives, a method for assessing renal function in patients with kidney disease using diffusion tensor imaging is provided, the method comprising the following steps:

[0005] Step S1: Acquire the raw image of the patient's renal diffusion tensor, wherein the raw image of the renal diffusion tensor is acquired by single-excitation echo-plane imaging;

[0006] Step S2: Compare adjacent pixels in the original image of the kidney diffusion tensor to obtain the differences in signal intensity and gradient direction between pixels;

[0007] Step S3: Determine whether there is local geometric distortion based on the differences in signal intensity and gradient direction between pixels, and perform pixel-by-pixel correction based on the local geometric distortion to obtain the corrected diffusion tensor image;

[0008] Step S4: Calculate the fitted tensor of the diffusion tensor image, extract the diffusion parameters of the renal cortex and renal medulla based on the fitted tensor, and perform quantitative analysis with preset renal function indicators to obtain the quantitative results of the patient's renal function indicators.

[0009] The present invention has the following beneficial effects:

[0010] I. By calculating the differences in signal intensity and gradient direction at the pixel level and performing pixel-by-pixel correction based on local geometric distortion, this method effectively eliminates local deformations caused by factors such as magnetic susceptibility effects and motion artifacts. The dual correction of distorted pixels in both intensity and direction, along with the continuous transition and superposition of the correction region, results in a final diffusion tensor image with higher accuracy and consistency, laying a reliable data foundation for subsequent kidney tissue zoning and functional analysis.

[0011] Second, by constructing a dominant difference direction and a reference vector, and combining angle calculation and directional deviation assessment, this method can capture signal intensity changes while resolving spatial directional features between pixels, thus more accurately distinguishing the renal cortex and renal medulla. Compared with existing methods that solely rely on grayscale or texture features, this approach is more sensitive to local tissue directional changes, significantly improving the accuracy and stability of kidney region segmentation.

[0012] Third, by extracting diffusion parameters through tensor fitting and comparing them with preset renal function indicators to measure differences, complex imaging data can be transformed into intuitive quantitative results. This method breaks through the traditional qualitative analysis model that relies on experience-based judgment and provides an objective and repeatable means of functional evaluation. Attached Figure Description

[0013] Figure 1 A schematic diagram of the steps involved in a method for assessing renal function in patients with kidney disease using diffusion tensor imaging.

[0014] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2.

[0015] Figure 3 This is a comparison image of diffusion tensor image correction for a method of diffusion tensor imaging to assess renal function in patients with kidney disease according to the present application.

[0016] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0019] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] To achieve the above objectives, please refer to Figures 1 to 3 A method for assessing renal function in patients with kidney disease using diffusion tensor imaging, the method comprising the following steps:

[0021] Step S1: Acquire the raw image of the patient's renal diffusion tensor, wherein the raw image of the renal diffusion tensor is acquired by single-excitation echo-plane imaging;

[0022] In one embodiment, the patient is first positioned supine on the magnetic resonance imaging (MRI) scanning table, and a fixation device is used to reduce movement of the kidneys, ensuring stable image acquisition. The scanner used is a 3.0 Tesla or 1.5 Tesla MRI instrument equipped with a multi-channel abdominal coil to enhance the signal-to-noise ratio and cover the entire renal organ region.

[0023] In some embodiments, the raw image acquisition parameters for renal diffusion tensor (DST) include, but are not limited to: echo time (TE): 50 ms–90 ms; repetition time (TR): 2000 ms–4000 ms; slice thickness: 3 mm–5 mm; slice gap: 0–1 mm; scan matrix: 128×128–256×256; diffusion sensitivity b-value: 0 s / mm², 500 s / mm², 1000 s / mm², with higher b-values ​​available upon clinical need. During the scan, respiratory triggering or respiratory gating techniques are used to synchronize the patient's respiratory cycle to reduce the impact of respiratory motion artifacts on the renal DST image.

[0024] In some embodiments, after acquisition, the original image is stored in DICOM format and contains complete spatial location information, gradient direction and b-value information, providing a data foundation for subsequent diffusion tensor reconstruction and analysis.

[0025] It is important to note that, to ensure the accuracy of the original diffusion tensor image, the scan slice thickness, b-value, and respiratory synchronization parameters can be adjusted appropriately according to the patient's body size, kidney location, and respiratory rate to obtain original data that has both a sufficient signal-to-noise ratio and reflects the microscopic water molecule diffusion characteristics of the kidney.

[0026] Step S2: Compare adjacent pixels in the original image of the kidney diffusion tensor to obtain the differences in signal intensity and gradient direction between pixels;

[0027] In one embodiment, the raw images of the kidney diffusion tensor are preprocessed, including but not limited to noise suppression processing, such as using Gaussian filtering or bilateral filtering, to reduce random noise generated during magnetic resonance imaging; artifact correction to reduce image distortion caused by breathing, heartbeat or scanner drift; and spatial registration of multiple raw images to ensure that the same pixel corresponds to the same anatomical location at different times or in different scanning directions.

[0028] In some embodiments, each pixel in the image is traversed, and that pixel is compared with its neighboring pixels. The neighboring pixels can be a 3×3, 5×5, or 7×7 pixel square window, with an appropriate neighborhood selected based on the image resolution and kidney size. For each pixel, the signal intensity difference between it and its neighboring pixels is calculated. The signal intensity difference can be obtained by comparing the average intensity, maximum value, or minimum value of the center pixel with that of its neighboring pixels, and the difference can be normalized to ensure comparability of data under different scanning conditions.

[0029] For each pixel, the gradient direction difference is further calculated. In some embodiments, spatial analysis of the intensity changes of pixels in the neighborhood is performed to determine the direction and magnitude of the pixel intensity changes, and a gradient vector for each pixel is generated. In consecutive frame images, the gradient direction can be temporally smoothed, for example, by using a sliding window averaging or exponential weighted averaging method, to reduce the impact of occasional signal fluctuations.

[0030] When analyzing gradient direction, the gradient can be decomposed into components in different directions to accurately reflect the distribution of pixel intensity changes in the horizontal, vertical, and depth directions. If the gradient direction change of adjacent pixels exceeds a preset threshold, the pixel may be marked as being in a region with significant boundary or structural changes, thus assisting in subsequent region segmentation or feature extraction.

[0031] In some embodiments, signal intensity differences and gradient direction differences are combined to generate a local gradient feature map for identifying the motion region or diffusion region of the kidney. The local gradient features may include the gradient magnitude and direction of each pixel, the mean and variance of the intensity difference with neighboring pixels, and a gradient direction consistency index to determine the continuity of the local structure.

[0032] It is important to note that to ensure the accuracy of gradient calculation and signal difference determination, subpixel interpolation processing can be performed on the image before calculation, such as using bilinear or cubic interpolation, to reduce the bias caused by image discretization. Simultaneously, a dynamic threshold can be set based on the overall signal intensity distribution of the kidney to adapt to different patients or scanning conditions.

[0033] In some embodiments, the gradient amplitude can be weighted according to the main anatomical direction of the kidney, so that the signal changes along the main diffusion direction have a higher weight in the analysis, thereby enhancing the recognition accuracy of kidney diffusion features.

[0034] Step S3: Determine whether there is local geometric distortion based on the differences in signal intensity and gradient direction between pixels, and perform pixel-by-pixel correction based on the local geometric distortion to obtain the corrected diffusion tensor image;

[0035] In one embodiment, based on the signal intensity difference and gradient direction difference information of each pixel obtained in step S2, a local geometric distortion analysis is performed on the original image of the entire kidney diffusion tensor. Local geometric distortion can manifest as pixel position shift, local stretching or compression, and abnormal changes in gradient direction between adjacent pixels.

[0036] Each pixel in the image is traversed, and its signal intensity and gradient direction differences are compared with those of neighboring pixels. If the intensity change or gradient direction of a pixel deviates significantly from a preset threshold compared to its neighboring pixels, it is determined that there is a local geometric distortion in the region where that pixel is located. In some embodiments, a sliding window or adaptive neighborhood method can be used to determine the distortion, ensuring that the detected distortion includes both small offsets and larger deformed areas.

[0037] After identifying a pixel with local geometric distortion, a pixel-by-pixel correction process is performed. In some embodiments, the correction method includes an interpolation method based on neighboring pixels or a local affine transformation to fine-tune the pixel's position information so that its gradient direction is consistent with that of neighboring pixels, while maintaining the continuity of signal strength.

[0038] For the correction of each pixel, information from multiple neighborhood levels around the pixel can be considered, including nearest neighbor and second nearest neighbor pixels, to improve the smoothness and accuracy of the correction. During the correction process, coarse adjustment can be performed first to eliminate obvious offsets, followed by fine adjustment to ensure the continuity of local gradient directions and the stability of the diffuse tensor structure.

[0039] In some embodiments, the correction process also incorporates overall kidney structural features and the expected distribution pattern of the diffusion tensor to constrain the consistency of the correction results. For example, when a pixel gradient direction is detected to suddenly deviate from the main direction in a specific region, the correction process prioritizes maintaining the continuity of the overall tissue structure and fine-tunes the abnormal pixel based on the average gradient direction of the surrounding pixels.

[0040] It is important to note that to improve pixel-by-pixel correction accuracy, subpixel-level interpolation can be performed on the image before processing, such as using cubic convolution interpolation or bicubic interpolation methods, to make the corrected pixel positions smoother in continuous space. Simultaneously, during the correction process, the neighborhood size and threshold can be dynamically adjusted to adapt to the signal intensity and gradient characteristics of different regions, thereby effectively reducing the impact of local distortions on diffusion tensor analysis.

[0041] Step S4: Calculate the fitted tensor of the diffusion tensor image, extract the diffusion parameters of the renal cortex and renal medulla based on the fitted tensor, and perform quantitative analysis with preset renal function indicators to obtain the quantitative results of the patient's renal function indicators.

[0042] In one embodiment, based on the corrected dispersion tensor image obtained in step S3, tensor fitting processing is first performed on each pixel. During the tensor fitting process, by analyzing the signal attenuation in each direction, the local dispersion tensor information of the pixel is obtained, reflecting the anisotropic diffusion characteristics of water molecules in that region.

[0043] After tensor fitting, the diffuse tensor image is segmented into two main tissue regions: the renal cortex and the renal medulla. Region segmentation can be based on pixel intensity, gradient direction, and morphological features, using automatic or semi-automatic segmentation algorithms. For example, thresholding combined with edge detection can be used to extract the renal cortex contour, and the remaining region can be labeled as the renal medulla.

[0044] For the segmented renal cortex and renal medulla regions, diffusion parameters are extracted, including but not limited to the average diffusion coefficient (ADC), anisotropy index (FA), and eigenvalues ​​of the diffusion tensor in each principal direction. During the extraction process, statistical analysis can be performed on the pixel values ​​of each region to obtain the regional mean, standard deviation, and distribution, thereby improving the reliability and repeatability of the indicators.

[0045] After obtaining the diffusion parameters, they are quantitatively analyzed against preset renal function indicators. These preset renal function indicators may include reference ranges for healthy individuals or clinical standards, such as glomerular filtration rate and the ratio of the renal cortex to the renal medulla diffusion coefficient. In some embodiments, the diffusion parameters can be graded according to their deviation from the preset standards, for example, into three levels: low, medium, and high, to indicate the severity of renal function decline.

[0046] Quantitative analysis results can be represented numerically, graphically, or color-coded to generate quantitative results of patients' renal function indicators. For example, it can output the average ADC value and FA value of the renal cortex and renal medulla, as well as the regional ratio and deviation grade, and generate a visual report.

[0047] It is important to note that, to improve analytical accuracy, spatial smoothing, noise reduction, and outlier removal can be incorporated into the fitting and region segmentation processes to ensure that the extracted diffusion parameters accurately reflect the physiological state of kidney tissue. Through these steps, the final quantitative results of patient renal function indicators can be used for disease grading and treatment efficacy evaluation.

[0048] As an example of the present invention, reference is made to... Figure 2 As shown, step S2 in this example includes:

[0049] Step S21: In the original image of the kidney diffusion tensor, for each pixel to be processed, determine a fixed-size neighborhood centered on that pixel.

[0050] Step S22: Traverse each neighboring pixel in the neighborhood, calculate the difference in signal strength between each neighboring pixel and the center pixel, and record the magnitude of the difference to determine the signal strength difference between the pixels.

[0051] Step S23: Calculate the gradient direction between adjacent pixels and the center pixel within the neighborhood to obtain the relative direction change data;

[0052] Step S24: Classify the signal intensity difference and direction change data in the neighborhood by comparing them with a preset threshold, and generate the local difference distribution of the center pixel; extract the direction information with the largest change amplitude from the local difference distribution as the dominant difference direction of the center pixel;

[0053] Step S25: Determine the gradient direction difference of each pixel based on the dominant difference direction of the center pixel.

[0054] In one embodiment, for each pixel to be processed in the original image of the kidney diffusion tensor, this pixel is first marked as the center pixel. A neighborhood region of a fixed size is determined around the center pixel, such as a square neighborhood window of 3×3 or 5×5 pixels centered on the center pixel. The choice of neighborhood size can be adjusted according to the image resolution and noise characteristics to ensure that the neighborhood contains sufficient local information while avoiding an excessively large neighborhood that would cause local features to be smoothed.

[0055] The algorithm iterates through each neighboring pixel in the neighborhood, comparing its signal intensity value with that of the center pixel and calculating the difference. The difference information includes amplitude and sign, and is recorded in a signal intensity difference dataset. The amplitude reflects the magnitude of the local signal intensity change, and the sign indicates the directionality of the change. If the difference exceeds a preset threshold (e.g., 10–50 grayscale units), it is considered a significant change.

[0056] The gradient direction of each neighboring pixel relative to the center pixel is calculated, and the gradient direction of the neighboring pixels relative to the center pixel is represented as an angle value. The direction change of each pixel is recorded, and a direction change dataset is generated. If the gradient direction change exceeds a set threshold (e.g., 10–30 degrees), it is marked as a local direction anomaly. The direction difference is used to reflect the local continuity of the kidney tissue structure in this region.

[0057] Signal intensity difference and gradient direction difference data are compared with preset thresholds to generate a local difference distribution map of the center pixel. Through local difference distribution analysis, the direction information with the largest change amplitude or the most significant anomaly is extracted and used as the dominant difference direction of the center pixel. This dominant direction is used for subsequent pixel-by-pixel local geometric distortion determination. The preset threshold can be determined based on the scanner noise level and imaging resolution to ensure the sensitivity and reliability of the detection results.

[0058] Based on the dominant difference direction, the gradient direction difference of each pixel in the neighborhood relative to the center pixel is calculated, and a pixel gradient direction difference dataset is generated. This dataset is used for subsequent pixel-by-pixel local geometric distortion judgment and correction. If the direction difference between adjacent pixels is continuous and significant, it indicates a local geometric distortion region; if the direction difference is small and consistent, the region can be regarded as a stable region.

[0059] During implementation, the original image can be preprocessed, such as by Gaussian filtering or median filtering, to suppress the influence of random noise on the calculation of difference and direction. It is important to note that the neighborhood size, signal difference threshold, and direction difference threshold should be dynamically adjusted based on the specific kidney image resolution, scanning equipment characteristics, and the patient's physiological condition to ensure that the distribution of local differences and the dominant difference direction accurately reflect the kidney tissue structure.

[0060] Step S25 includes:

[0061] Determine the reference vector corresponding to the dominant difference direction based on the dominant difference direction of the center pixel;

[0062] For each neighboring pixel in the neighborhood of the center pixel, extract the local orientation information of that neighboring pixel;

[0063] Calculate the angle between the local orientation information of adjacent pixels and the reference vector;

[0064] The included angle is used as the directional deviation of the adjacent pixel relative to the center pixel, and the directional deviation is recorded.

[0065] By iterating through the directional deviation of each pixel, the gradient direction difference of each pixel can be obtained.

[0066] In one embodiment, for each center pixel in the original image of the kidney diffusion tensor, a reference vector consistent with the dominant difference direction extracted in step S24 is first determined based on that center pixel. The reference vector can be represented as a unit vector in two-dimensional space, with its direction consistent with the dominant difference direction and its length normalized to 1. The reference vector serves as a benchmark for measuring the directional deviation of adjacent pixels within the neighborhood.

[0067] Within the neighborhood defined by the center pixel (e.g., a 3×3 or 5×5 pixel window), local orientation information is extracted for each neighboring pixel, including the signal intensity gradient direction and the pixel grayscale change direction. Orientation information can be obtained by calculating the difference vector of the pixel's grayscale value relative to the center pixel. For edge pixels, boundary expansion or mirror filling methods are used to ensure neighborhood integrity.

[0068] The local orientation vector of each neighboring pixel within the neighborhood is compared with the reference vector, and the angle between them is calculated. The angle can be calculated by obtaining the cosine value through vector dot product, and then obtaining the angle through the inverse cosine function. The calculation result is recorded as an angle value between 0° and 180°, which is used to quantify the degree of deviation of the orientation of neighboring pixels from the dominant orientation of the center pixel.

[0069] The angle between each adjacent pixel and the reference vector of the center pixel is used as the orientation deviation, generating an orientation deviation dataset. The orientation deviation not only reflects the orientation consistency between pixels but can also be used to identify local geometric distortion regions. A larger deviation indicates a more significant orientation difference, while a smaller deviation indicates that the local orientation tends to be consistent. Steps S251–S254 are repeated for each center pixel and its neighboring pixels in the image to obtain the orientation deviation distribution for the entire image. By statistically analyzing the orientation deviation, a gradient orientation difference map for each pixel can be generated. This difference map is used for subsequent pixel-by-pixel judgment and correction of local geometric distortion.

[0070] During implementation, to reduce the impact of noise, the directional deviation can be smoothed, for example, using neighborhood mean filtering or weighted smoothing filtering. For continuous frame images, trend analysis can be performed in conjunction with time series data to distinguish between transient directional changes and true geometric distortion. The neighborhood size, angle calculation accuracy, and smoothing parameters should be dynamically adjusted according to the imaging resolution, scanning equipment performance, and kidney structural characteristics.

[0071] The reference vector corresponding to the dominant difference direction is determined based on the dominant difference direction of the center pixel, including:

[0072] Obtain the dominant difference direction angle information of the center pixel;

[0073] The angle information of the dominant difference direction is compared with the preset coordinate system reference direction to determine the angular position of the center pixel in the coordinate system.

[0074] Based on the angular position, calculate the component values ​​of the center pixel on the horizontal and vertical axes, and combine the component values ​​to obtain the direction vector;

[0075] The direction vector is normalized to obtain the reference vector corresponding to the dominant difference direction.

[0076] In one embodiment, for each central pixel in the original image of the kidney diffusion tensor, the dominant difference direction angle information of the central pixel is first obtained based on the magnitude and direction distribution of its grayscale variation in the neighborhood. The angle information is typically expressed in degrees or radians, measured relative to the horizontal axis of the image, and ranges from 0° to 180°. The dominant difference direction angle information of the central pixel is compared with a preset reference direction of the image coordinate system. The reference direction of the coordinate system can be the positive direction of the horizontal axis. Based on the comparison result, the absolute direction of the angular position of the central pixel in the coordinate system is determined, ensuring that the reference vector direction is consistent with the spatial coordinate system of the image.

[0077] Based on the determined angular position, the component values ​​of the center pixel in the horizontal (x-axis) and vertical (y-axis) directions are calculated. Specifically, the horizontal component equals the magnitude of the direction vector multiplied by the cosine of the angle, and the vertical component equals the magnitude of the direction vector multiplied by the sine of the angle. This component value reflects the spatial distribution of the pixel's dominant direction in the image coordinate system. The calculated horizontal and vertical components are combined into a two-dimensional vector to form the direction vector. Subsequently, the direction vector is normalized, with its length standardized to 1, to eliminate the influence of amplitude on subsequent calculations. This normalized direction vector is the reference vector corresponding to the dominant difference direction of the center pixel, used for angle comparison with the directions of adjacent pixels in the neighborhood and for calculating the direction deviation.

[0078] In practical applications, to improve the stability of direction vectors, temporal smoothing can be performed on the direction vectors of the same pixel in consecutive frames of images. For pixels with incomplete image edges or neighborhoods, mirror expansion or zero-padding can be used to maintain neighborhood integrity. Normalization is not only used to maintain quantization consistency but also facilitates subsequent gradient direction difference statistics and geometric distortion judgment.

[0079] The angle information of the dominant difference direction is compared with the preset coordinate system reference direction to determine the angular position of the center pixel in the coordinate system, including:

[0080] The angle difference between the dominant difference direction angle information and the reference direction of the preset coordinate system is calculated one by one to obtain the difference between the center pixel and the reference direction.

[0081] The direction of the center pixel is determined by the sign of the difference, and it is determined whether it is on the positive or negative side of the coordinate system relative to the reference direction. Combined with the absolute value of the difference, the specific angular position of the center pixel in the coordinate system is determined.

[0082] In one embodiment, for each central pixel in the original image of the kidney diffusion tensor, the dominant difference direction angle information of that pixel is first obtained. This angle information is measured relative to the horizontal axis of the image, ranging from 0° to 180°. The angle difference between the dominant difference direction angle information of the central pixel and the reference direction of a preset coordinate system is calculated one by one. The angle difference calculation can be performed by taking the minimum difference between two angles, considering periodicity, so that the result is between -180° and +180°, thereby obtaining the magnitude of the difference between the central pixel and the reference direction.

[0083] The orientation of the center pixel is determined by the sign of the angle difference. If the difference is positive, the center pixel is determined to be on the positive side of the coordinate system relative to the reference direction; if the difference is negative, it is determined to be on the negative side of the coordinate system.

[0084] By combining the directional bias information with the absolute value of the angle difference, the specific angular position of the center pixel in the coordinate system is further determined. This specific angular position is used for subsequent calculations of the direction vector components and the normalized reference vector, ensuring spatial consistency between the reference vector and the image coordinate system.

[0085] In practice, the angle difference can be appropriately smoothed to reduce the impact of noise or local abnormal pixels on the angle position determination. Meanwhile, for edge pixels or incomplete neighborhoods, mirror expansion or neighborhood compensation strategies can be used to ensure the stability of the angle position calculation.

[0086] Preferably, the pixel-by-pixel correction processing based on local geometric distortion in step S3 includes:

[0087] A pixel set is obtained based on local geometric distortion, where the pixel set includes distorted pixels and normal pixels;

[0088] For each distorted pixel, determine the signal strength and gradient direction data of its adjacent normal pixels;

[0089] The signal intensity of the distorted pixel is weighted and averaged with the signal intensity of the adjacent normal pixel to generate a correction intensity value.

[0090] The gradient direction of the distorted pixel is vector-differed from the gradient direction of the adjacent normal pixel to generate a correction direction value.

[0091] The corresponding distorted pixel data is updated based on the correction intensity value and correction direction value to obtain the corrected diffusion tensor image.

[0092] In one embodiment, for each pixel to be processed in the original image of the kidney diffusion tensor, firstly, based on the pixel signal intensity difference and gradient direction difference data calculated in step S2, local geometric distortion regions are identified, and a set of pixels containing distorted pixels and adjacent normal pixels is obtained. Distorted pixels refer to pixels whose signal intensity or gradient direction deviates from that of neighboring normal pixels by more than a preset threshold, while normal pixels are pixels with stable signals and consistent gradient directions.

[0093] For each distorted pixel, the signal intensity and gradient direction data of its neighboring normal pixels are extracted from its neighborhood. The neighborhood range can be a 3×3, 5×5, or 7×7 pixel window, selected according to the local image resolution and noise level, to ensure that there are enough normal pixels in the neighborhood that can represent the local features.

[0094] The signal intensity of the distorted pixel is calculated by weighting it with the signal intensity of the normal pixels in its neighborhood. The weights are determined based on the distance from the center pixel and the stability of the signal intensity; pixels that are closer to the center pixel or have more stable signals have a higher weight, thus obtaining the corrected pixel intensity value.

[0095] The gradient direction of the distorted pixel is vector-differed with the gradient direction of the normal pixels in the neighborhood to calculate the direction correction. The vector difference operation takes into account the periodicity of the direction angle and limits the direction adjustment to the range of -180° to +180° to ensure direction continuity and generate the corrected gradient direction value.

[0096] The calculated correction intensity and correction direction values ​​are updated to the corresponding distorted pixel data, replacing the signal intensity and gradient direction information of the original pixels, thus completing pixel-by-pixel correction.

[0097] The above correction steps are sequentially applied to all identified distorted pixels in the image to obtain the corrected diffusion tensor image. This image is more continuous in local geometry, with smoother gradient directions, reducing signal distortion caused by local distortions, which is beneficial for the accurate extraction and quantitative analysis of subsequent kidney function indicators.

[0098] Preferably, updating the corresponding distorted pixel data based on the correction intensity value and correction direction value includes:

[0099] The difference between the correction intensity value and the signal intensity of the distorted pixel is calculated, and the correction weight for updating the intensity of the pixel is determined based on the magnitude of the difference.

[0100] The angle between the correction direction value and the gradient direction of the distorted pixel is calculated, and the correction magnitude of the direction update of the pixel is determined according to the size of the angle.

[0101] The intensity and orientation of the distorted pixels are updated item by item based on the correction weight and correction magnitude, respectively, to obtain the updated distorted pixels;

[0102] The updated pixels are sequentially superimposed according to their spatial positions to obtain a continuously transitioning correction region, which is then merged with the original kidney diffusion tensor image to obtain the corrected diffusion tensor image.

[0103] In one embodiment, for each identified distorted pixel in the original image of the kidney diffusion tensor, the difference between the corrected intensity value calculated in the aforementioned steps and the original signal intensity of the pixel is first calculated to obtain intensity deviation information. The correction weight of the pixel in the update process is determined based on the magnitude of the deviation; a larger weight indicates a more severe deviation in the original signal of the pixel, requiring a greater degree of correction.

[0104] For the gradient direction of a pixel, the angle between the corrected direction value and the original gradient direction of the pixel is calculated to obtain the direction deviation. Based on the magnitude of the direction deviation, the correction magnitude for updating the pixel's gradient direction is determined to ensure that the direction adjustment corrects distortion while maintaining local directional continuity. Based on the obtained correction weights and correction magnitudes, the signal intensity and gradient direction of distorted pixels are updated item by item. Specifically, the new signal intensity of the pixel is equal to the weighted combination of the original intensity and the corrected intensity according to the correction weights; the new gradient direction of the pixel is equal to the combination of the original direction and the corrected direction according to the correction magnitude, thus obtaining the updated pixel data.

[0105] All updated distorted pixels are sequentially superimposed according to their spatial positions in the image to form a continuous correction region. This correction region achieves a smooth transition in both intensity and direction, reducing discontinuities caused by local distortions.

[0106] The corrected region with a continuous transition is merged with the original renal diffusion tensor image, replacing the original distorted pixel regions, to obtain the finally corrected diffusion tensor image. The corrected image is locally geometrically continuous with smooth gradient directions, providing a reliable basis for subsequent extraction of diffusion parameters from the renal cortex and renal medulla.

[0107] Preferably, the spatial location of a pixel includes: the pixel's position coordinates in the row direction, the pixel's position coordinates in the column direction, and the pixel's index coordinates in the layer slice direction.

[0108] In one embodiment, the spatial position of each pixel is represented by three-dimensional coordinates, including: row direction coordinates: representing the row index of the pixel in the image matrix, used to determine the position of the pixel in the horizontal direction; column direction coordinates: representing the column index of the pixel in the image matrix, used to determine the position of the pixel in the vertical direction; and slice direction number: representing the position of the pixel in the slice sequence of the kidney diffusion tensor image, used to distinguish pixels with the same plane coordinates on different slices.

[0109] By recording the three-dimensional spatial position of each pixel, each distorted pixel can be precisely located during pixel-by-pixel correction, ensuring that the correction intensity and direction values ​​are correctly applied to the corresponding pixels. Simultaneously, the three-dimensional coordinate information facilitates a seamless transition when merging the corrected region with the original image, preventing misalignment or discontinuity in gradient direction and signal intensity between different slices. This improves the spatial consistency of the overall diffusion tensor image and the accuracy of subsequent kidney function parameter extraction.

[0110] Preferably, the extraction method for the renal cortex and renal medulla region in step S4 includes:

[0111] Candidate image regions containing the entire kidney region are determined based on the boundaries of the original image of the kidney diffusion tensor.

[0112] Within the candidate image region, based on grayscale distribution and texture features, the renal cortex region and the renal medulla region are initially divided to obtain preliminary division results;

[0113] Based on the preliminary segmentation results, adjacent pixel consistency correction was performed to obtain the renal cortex and renal medulla regions.

[0114] In one embodiment, a candidate image region encompassing the entire kidney is determined based on the boundary information of the original kidney diffusion tensor image. Specifically, this involves: performing edge detection on the image, identifying the kidney's outline, and generating a minimum rectangular or polygonal region that surrounds the entire kidney as the initial boundary for subsequent region segmentation.

[0115] Within the candidate image region, the renal cortex and renal medulla are initially divided based on pixel grayscale distribution and texture features. Grayscale distribution features include the mean and variance of pixel intensity; texture features include local grayscale co-occurrence matrix statistics or local gradient direction distribution. By analyzing these features, the kidney region is divided into high-signal regions (renal cortex) and low-signal regions (renal medulla), yielding preliminary segmentation results.

[0116] The initial segmentation results undergo neighbor pixel consistency correction. Specifically, for each pixel within the initially segmented region, its category is checked against its surrounding pixels. If isolated pixels or small anomalous areas exist, they are reclassified to the category with the highest proportion of neighboring pixels to eliminate local noise and discontinuities. This step yields continuous and smooth segmentation results for the renal cortex and renal medulla. The final output of the renal cortex and renal medulla regions can be used for subsequent diffusion parameter extraction and quantitative analysis of renal function, providing a reliable basis for quantitative assessment of kidney function.

[0117] In another embodiment, edge detection is performed on the original kidney diffusion tensor image to obtain the kidney contour. A bounding rectangle of this contour is used as a candidate image region, for example, 120 pixels wide and 150 pixels high. All pixels within this region are used for subsequent analysis. Within the candidate region, the mean and standard deviation of grayscale values ​​for each pixel are calculated. For example, the grayscale values ​​of pixels in the renal cortex are concentrated in the range of 150-220, and those in the renal medulla are concentrated in the range of 80-140. Simultaneously, local texture features are extracted, such as the contrast and entropy values ​​in the 3×3 neighborhood grayscale co-occurrence matrix. Based on the grayscale threshold and texture features, pixels with higher grayscale values ​​and contrast and entropy values ​​that meet the threshold conditions are labeled as renal cortex, and the rest are labeled as renal medulla. Subsequently, the proportion of pixels of the same category within the 3×3 neighborhood of each pixel is checked. If the proportion is less than 50%, the pixel category is adjusted to the category with the largest neighborhood proportion. At the same time, isolated spots with an area less than 5 pixels are merged or deleted to ensure the continuous smoothness of the renal cortex and renal medulla regions. Finally, a continuous mask image of the renal cortex and renal medulla is generated, with a gray value of 180 for the renal cortex and 100 for the renal medulla. This mask image is then combined with the original diffusion tensor image for subsequent extraction and quantitative analysis of renal function parameters.

[0118] Preferably, the quantitative analysis of kidney function indicators in step S4 includes:

[0119] Obtain preset renal function indicators, which include reference parameters characterizing the state of kidney function;

[0120] The diffusion parameters were compared with the preset renal function indicators one by one, and the proportional relationship between the two was calculated.

[0121] By accumulating the differences in the proportional relationships, a comprehensive measure of the differences between the renal cortex and renal medulla regions is obtained.

[0122] Numerical quantification was performed based on a comprehensive difference measure to obtain the quantitative results of the patient's renal function indicators.

[0123] In one embodiment, preset renal function indicators are obtained, including reference parameters characterizing the state of kidney function, such as the average diffusion coefficient range of the renal cortex and renal medulla, reference values ​​of the anisotropy index, and the characteristic range of each vector of the diffusion tensor. These preset indicators may be derived from statistical data of healthy subjects or provided by clinical trial databases.

[0124] Based on the corrected diffusion tensor image obtained in step S3, diffusion parameters of the renal cortex and renal medulla are extracted, including the local average diffusion coefficient, anisotropy index, and principal gradient direction information. Each parameter corresponds to a preset reference value of renal function index, forming a one-to-one data pair.

[0125] The aforementioned diffusion parameters were compared with their corresponding preset renal function indicators, and the proportional relationship between the two was calculated. This proportional relationship can be expressed as the ratio, difference, or standardized difference between the corrected parameter value and the reference parameter value. The degree of difference for each parameter was calculated independently, providing a quantitative basis for subsequent comprehensive analysis.

[0126] The differences in each parameter are accumulated to form a comprehensive measure of renal cortical difference, and the differences in multiple parameters in the renal medullary region are accumulated to form a comprehensive measure of renal medullary difference. The accumulation method can be weighted summation or weighted average, with the weights set according to the importance of different parameters to renal function.

[0127] Based on a comprehensive measurement of differences between the renal cortex and renal medulla, the numerical differences are mapped to quantifiable renal function indicators, resulting in quantitative results for the patient's renal function. These quantitative results may include functional scores for individual regions, overall renal function scores, and specific deviations of each parameter from reference values, facilitating clinical assessment of the patient's renal function status.

[0128] In another embodiment, suppose the diffusion parameters of a patient's kidney diffusion tensor image obtained through step S3 are as follows:

[0129] area Parameter type Corrected parameter values Reference kidney function indicators Difference Calculation Method renal cortex Average diffusion coefficient 1.20×10⁻³mm² / s 1.35×10⁻³mm² / s Ratio: 1.20 / 1.35 ≈ 0.889 renal cortex Anisotropy Index 0.42 0.50 Difference: 0.50 - 0.42 = 0.08 renal cortex Major gradient direction deviation 5° 0° Angle difference: 5° Renal medulla Average diffusion coefficient 1.80×10⁻³mm² / s 1.90×10⁻³mm² / s Ratio: 1.80 / 1.90 ≈ 0.947 Renal medulla Anisotropy Index 0.35 0.40 Difference: 0.40 + 0.35 = 0.05 Renal medulla Major gradient direction deviation 3° 0° Angle difference: 3°

[0130] Weights are assigned to each parameter based on their importance (e.g., average diffusion coefficient 0.5, anisotropy index 0.3, gradient direction deviation 0.2), and the overall difference between the renal cortex and renal medulla is calculated. Overall difference of renal cortex = 0.5 × (1-0.889) + 0.3 × 0.08 + 0.2 × (5 / 180) ≈ 0.081 + 0.024 + 0.006 ≈ 0.111; Overall difference of renal medulla = 0.5 × (1-0.947) + 0.3 × 0.05 + 0.2 × (3 / 180) ≈ 0.0265 + 0.015 + 0.0033 ≈ 0.0448.

[0131] Map the comprehensive difference measure to a functional score (0-1, where 0 indicates complete normality and 1 indicates the most severe abnormality):

[0132] area Comprehensive Difference Measurement Functionality rating renal cortex 0.111 0.11 Renal medulla 0.0448 0.045 Overall renal function - 0.0775

[0133] This quantitative result clearly shows the functional status of the renal cortex and renal medulla.

[0134] Most importantly, updating the intensity and orientation of distorted pixels item by item based on the correction weight and correction magnitude also includes:

[0135] The intensity components of distorted pixels are updated item by item based on the corrected weights to generate preliminary intensity correction results.

[0136] The orientation components of the distorted pixels are updated item by item based on the correction magnitude to generate preliminary orientation correction results;

[0137] The initial intensity correction result and the initial orientation correction result are jointly corrected according to the preset pixel bidirectional fine-tuning law to obtain the bidirectional corrected distorted pixel points;

[0138] The local consistency of the bidirectionally corrected distorted pixels is compared. If the difference between the correction and the surrounding pixels exceeds the preset micro-difference range, a second iteration is performed through the pixel bidirectional fine-tuning law to generate the updated distorted pixels.

[0139] In one embodiment, firstly, for the detected distorted pixel, the system uses the signal strength of its neighboring normal pixels as a reference benchmark, and updates the intensity components of the distorted pixel item by item according to a preset correction weight. Specifically, the current intensity value of the distorted pixel is compared item by item with the weighted reference value of the neighboring normal pixels, and gradually moves closer to the reference value according to the proportion of the correction weight to generate a preliminary intensity correction result.

[0140] Secondly, for the directional components of distorted pixels, the system updates them item by item according to a preset correction range. In this process, the directional correction does not adopt a one-time jump method, but fine-tunes the directional components of distorted pixels according to the correction range, so that they gradually tend to be consistent with the direction of adjacent normal pixels, thereby generating a preliminary directional correction result.

[0141] Subsequently, the obtained preliminary intensity correction results and preliminary orientation correction results are jointly corrected according to a preset pixel bidirectional fine-tuning law. The core of this joint correction process is to simultaneously consider the coordination of the two components, intensity and orientation, to avoid over-correction of a single component and resulting in overall deviation, thereby obtaining bidirectionally corrected distorted pixels.

[0142] After obtaining the bidirectional correction result, a local consistency comparison needs to be performed on the distorted pixel. Specifically, the corrected distorted pixel is compared with several surrounding normal pixels item by item. If the correction differences of its intensity and orientation components are both within the preset micro-difference range, the correction is confirmed to be complete; if the difference of any component exceeds the micro-difference range, the system triggers the pixel bidirectional fine-tuning law for a second iteration update.

[0143] During the second iteration, the intensity and orientation of the distorted pixels are readjusted with finer step increments to ensure a high degree of consistency between the corrected result and neighboring pixels, ultimately generating the updated distorted pixels. This method not only enables item-by-item updates of distorted pixels but also achieves more accurate correction results across multiple iterations.

[0144] In another embodiment, after detecting a distorted pixel, the intensity and orientation of the distorted pixel are updated item by item based on the correction weight and correction magnitude. For example, when the intensity distribution of neighboring normal pixels is concentrated at a grayscale value of approximately 151 and the orientation is concentrated at approximately 11°, the initial intensity of the central distorted pixel is only 120, and the orientation deviates to 45°. The system first extracts a neighboring reference value, sets the intensity correction weight to a relatively large proportion (e.g., close to 70% of the neighboring reference value), and sets the orientation correction magnitude to a fixed micro-step (e.g., 5°). In the first update, the intensity of the central pixel is adjusted from 120 to approximately 141, and the orientation is adjusted from 45° to 40°. Subsequently, the system compares the correction result with the surrounding pixels for consistency and finds that there is still a significant difference in intensity and orientation, triggering a second iteration. In the second iteration, the intensity is further updated to 146, and the orientation is updated to 35°. The difference has narrowed, but it still does not meet the micro-difference threshold, so it continues to the third iteration. After the third iteration, the intensity reached 149, which was basically consistent with the neighborhood value, while the direction was updated to 30°. Although there was still a certain gap with the neighborhood direction, it had been significantly improved. If the direction deviation still exceeded the strictly set 1° threshold, the system could choose to continue iterating or trigger a region-level correction. If convergence was allowed within an engineering range of 3° to 5°, the correction could be considered complete. Throughout the process, the system not only updated the intensity and direction item by item, but also performed joint correction through a bidirectional fine-tuning law to avoid mutual interference between intensity and direction corrections. At the same time, a local consistency comparison was performed after each iteration to ensure that the correction result remained consistent with the overall trend of the neighborhood. If the target was not achieved after several consecutive iterations, the system marked the pixel as an "invalid correction point" and handed it over to the region-level fitting or manual review. The entire correction process was recorded in the composite compensation log, including the initial value, the number of iterations, the intensity and direction results after each correction, and the judgment conclusion, for subsequent parameter optimization and system adaptive calibration.

[0145] Of particular importance, the preset methods for pixel bidirectional fine-tuning also include:

[0146] Using the initial intensity deviation and orientation deviation values ​​of the distorted pixels as the rule benchmark, fine-tuning reference curves for the intensity dimension and orientation dimension are established respectively.

[0147] A set of complementary weight coefficients is preset between the intensity dimension and the direction dimension. When the intensity correction magnitude is large, the direction correction weight is automatically reduced, and when the direction correction magnitude is large, the intensity correction weight is automatically reduced.

[0148] The correction order is specified to be performed in an alternating manner of strength priority and direction compensation, until the correction difference between the two dimensions is less than the set micro-difference threshold.

[0149] In one embodiment, the initial intensity deviation and orientation deviation values ​​of the distorted pixels are used as rule benchmarks to establish fine-tuning reference curves for the intensity dimension and the orientation dimension, respectively. The reference curves use the magnitude of the deviation value as the input variable and the corresponding correction step size as the output variable. They can be in the form of piecewise linear functions or decreasing curves to achieve rapid correction when the deviation is large and gradual convergence when the deviation is small.

[0150] Based on this, the system presets a set of complementary weight coefficients between the intensity and orientation dimensions. These complementary weight coefficients satisfy the constraint that their sum is constant, and are used to dynamically allocate between the two types of corrections: when the intensity correction magnitude is large, the weight of the orientation correction is automatically reduced to avoid excessive orientation shift; when the orientation correction magnitude is large, the weight of the intensity correction is automatically reduced to ensure the continuity and stability of the overall image signal structure.

[0151] To ensure the orderliness and convergence of the correction process, the correction order is specified to be performed in an alternating manner of "intensity priority - orientation compensation". That is, in each iteration, the intensity of the distorted pixels is first corrected according to the intensity reference curve, and then orientation compensation is performed according to the orientation reference curve. This alternation avoids the overcorrection problem caused by single-dimensional correction.

[0152] During the correction process, the system calculates the remaining correction difference between the intensity and orientation dimensions in real time. The fine-tuning process terminates when the correction difference in both dimensions is less than a set differential threshold. This differential threshold can be dynamically set based on image resolution and noise level to ensure effective distortion elimination without introducing new calculation errors. Through this bidirectional pixel fine-tuning law, the intensity and orientation of distorted pixels achieve a dynamically balanced progressive correction effect in both dimensions. This results in a significantly improved signal intensity continuity and gradient orientation consistency in the final corrected image, thereby enhancing the accuracy and robustness of diffusion tensor imaging in renal function assessment.

[0153] In another embodiment, for example, when the initial intensity deviation of a distorted pixel is 15 and the orientation deviation is 12 degrees, the system first uses this initial deviation value as a rule benchmark to establish fine-tuning reference curves for the intensity and orientation dimensions, respectively. A set of complementary weight coefficients is preset between intensity and orientation, initially allocated as an intensity weight of 0.7 and an orientation weight of 0.3. Thus, in the first round of correction, the intensity is updated with a correction of 10.5, reducing the intensity deviation to 4.5, while the orientation is updated with a correction of 3.6 degrees, reducing the orientation deviation to 8.4 degrees. In the second round of correction, as the intensity deviation gradually approaches the threshold, the system automatically adjusts the weight allocation. With an intensity of 0.4 and a direction of 0.6, the intensity is further corrected by 1.8, reducing the deviation to 2.7. The direction correction is 5.04 degrees, reducing the direction deviation to 3.36 degrees. In the third round of correction, since both the intensity and direction are close to the preset micro-difference threshold, the system adjusts the weights to intensity 0.5 and direction 0.5. After the intensity is corrected by 1.35, the deviation drops to 1.35, and after the direction is corrected by 1.68 degrees, the deviation drops to 1.68 degrees. Finally, both the intensity and direction deviations are below the set threshold range, satisfying the convergence condition. This achieves an alternating correction mode that prioritizes intensity and compensates for direction, gradually reducing and stabilizing the two-dimensional deviation of the distorted pixels.

[0154] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0155] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for assessing renal function in patients with kidney disease using diffusion tensor imaging, characterized in that, Includes the following steps: Step S1: Acquire the raw image of the patient's renal diffusion tensor, wherein the raw image of the renal diffusion tensor is acquired by single-excitation echo-plane imaging; Step S2: Compare adjacent pixels in the original kidney diffusion tensor image to obtain the signal intensity difference and gradient direction difference between pixels; Step S2 includes the following steps: Step S21: In the original image of the kidney diffusion tensor, for each pixel to be processed, determine a fixed-size neighborhood centered on that pixel. Step S22: Traverse each neighboring pixel in the neighborhood, calculate the difference in signal strength between each neighboring pixel and the center pixel, and record the magnitude of the difference to determine the signal strength difference between the pixels. Step S23: Calculate the gradient direction between adjacent pixels and the center pixel within the neighborhood to obtain the relative direction change data; Step S24: Classify the signal intensity difference and direction change data in the neighborhood by comparing them with a preset threshold, and generate the local difference distribution of the center pixel; extract the direction information with the largest change amplitude from the local difference distribution as the dominant difference direction of the center pixel; Step S25: Determine the gradient direction difference of each pixel based on the dominant difference direction of the center pixel; Step S3: Determine whether there is local geometric distortion based on the differences in signal intensity and gradient direction between pixels, and perform pixel-by-pixel correction based on the local geometric distortion to obtain the corrected diffusion tensor image; Step S4: Calculate the fitted tensor of the diffusion tensor image, extract the diffusion parameters of the renal cortex and renal medulla based on the fitted tensor, and perform quantitative analysis with preset renal function indicators to obtain the quantitative results of the patient's renal function indicators.

2. The method for assessing renal function in patients with kidney disease using diffusion tensor imaging according to claim 1, characterized in that, Step S25 includes: Determine the reference vector corresponding to the dominant difference direction based on the dominant difference direction of the center pixel; For each neighboring pixel in the neighborhood of the center pixel, extract the local orientation information of that neighboring pixel; Calculate the angle between the local orientation information of adjacent pixels and the reference vector; The included angle is used as the directional deviation of the adjacent pixel relative to the center pixel, and the directional deviation is recorded. By iterating through the directional deviation of each pixel, the gradient direction difference of each pixel can be obtained.

3. The method for assessing renal function in patients with kidney disease using diffusion tensor imaging according to claim 2, characterized in that, The reference vector corresponding to the dominant difference direction is determined based on the dominant difference direction of the center pixel, including: Obtain the dominant difference direction angle information of the center pixel; The angle information of the dominant difference direction is compared with the preset coordinate system reference direction to determine the angular position of the center pixel in the coordinate system. Based on the angular position, calculate the component values ​​of the center pixel on the horizontal and vertical axes, and combine the component values ​​to obtain the direction vector; The direction vector is normalized to obtain the reference vector corresponding to the dominant difference direction.

4. The method for assessing renal function in patients with kidney disease using diffusion tensor imaging according to claim 3, characterized in that, The angle information of the dominant difference direction is compared with the preset coordinate system reference direction to determine the angular position of the center pixel in the coordinate system, including: The angle difference between the dominant difference direction angle information and the reference direction of the preset coordinate system is calculated one by one to obtain the difference between the center pixel and the reference direction. The direction of the center pixel is determined by the sign of the difference, and it is determined whether it is on the positive or negative side of the coordinate system relative to the reference direction. Combined with the absolute value of the difference, the specific angular position of the center pixel in the coordinate system is determined.

5. The method for assessing renal function in patients with kidney disease using diffusion tensor imaging according to claim 1, characterized in that, Step S3, which involves pixel-by-pixel correction based on local geometric distortion, includes: A pixel set is obtained based on local geometric distortion, where the pixel set includes distorted pixels and normal pixels; For each distorted pixel, determine the signal strength and gradient direction data of its adjacent normal pixels; The signal intensity of the distorted pixel is weighted and averaged with the signal intensity of the adjacent normal pixel to generate a correction intensity value. The gradient direction of the distorted pixel is vector-differed from the gradient direction of the adjacent normal pixel to generate a correction direction value. The corresponding distorted pixel data is updated based on the correction intensity value and correction direction value to obtain the corrected diffusion tensor image.

6. The method for assessing renal function in patients with kidney disease using diffusion tensor imaging according to claim 5, characterized in that, The data for updating the corresponding distorted pixels based on the correction intensity value and correction direction value includes: The difference between the correction intensity value and the signal intensity of the distorted pixel is calculated, and the correction weight for updating the intensity of the pixel is determined based on the magnitude of the difference. The angle between the correction direction value and the gradient direction of the distorted pixel is calculated, and the correction magnitude of the direction update of the pixel is determined according to the size of the angle. The intensity and orientation of the distorted pixels are updated item by item based on the correction weight and correction magnitude, respectively, to obtain the updated distorted pixels; The updated pixels are sequentially superimposed according to their spatial positions to obtain a continuously transitioning correction region, which is then merged with the original kidney diffusion tensor image to obtain the corrected diffusion tensor image.

7. The method for assessing renal function in patients with kidney disease using diffusion tensor imaging according to claim 6, characterized in that, The spatial location of a pixel includes: the pixel's position coordinates in the row direction, the pixel's position coordinates in the column direction, and the pixel's index coordinates in the layer slice direction.

8. The method for assessing renal function in patients with kidney disease using diffusion tensor imaging according to claim 1, characterized in that, The extraction methods for the renal cortex and renal medulla in step S4 include: Candidate image regions containing the entire kidney region are determined based on the boundaries of the original image of the kidney diffusion tensor. Within the candidate image region, based on grayscale distribution and texture features, the renal cortex region and the renal medulla region are initially divided to obtain preliminary division results; Based on the preliminary segmentation results, adjacent pixel consistency correction was performed to obtain the renal cortex and renal medulla regions.

9. The method for assessing renal function in patients with nephropathy using diffusion tensor imaging according to claim 1, characterized in that, Step S4 involves quantitative analysis against preset renal function indicators, including: Obtain preset renal function indicators, which include reference parameters characterizing the state of kidney function; The diffusion parameters were compared with the preset renal function indicators one by one, and the proportional relationship between the two was calculated. By accumulating the differences in the proportional relationships, a comprehensive measure of the differences between the renal cortex and renal medulla regions is obtained. Numerical quantification was performed based on a comprehensive difference measure to obtain the quantitative results of the patient's renal function indicators.

Images