Search

CN-122004867-A - Glomerulus filtration rate estimation method and device for patient with double-renal function asymmetry

CN122004867ACN 122004867 ACN122004867 ACN 122004867ACN-122004867-A

Abstract

The invention provides a glomerular filtration rate estimation method and device for patients with asymmetric double renal functions. The method includes obtaining basic information, clinical data, laboratory test data, and imaging data of a patient, the imaging data including kidney volume, blood flow, thickness of cortical and medulla, and renal function manifestations of right and left kidneys. By analyzing the imaging data, a functional difference metric coefficient for quantifying the asymmetry of the double kidney function is calculated. And then, extracting global feature vectors of the patient by using a deep learning model, and carrying out fusion processing on the imaging data, the clinical data and the laboratory detection data to generate comprehensive feature vectors. On the basis, the functional weights of the right kidney and the left kidney are adjusted according to the functional difference measurement coefficient, and finally the adjusted functional weights and the global feature vector are input into a depth regression model, and the corrected GFR estimated value is calculated. Can provide more accurate renal function assessment according to the asymmetric condition of double renal functions.

Inventors

  • FAN ZHENLIANG
  • MA HONGZHEN
  • FAN JUNFEN
  • XIA HONG

Assignees

  • 浙江省中医院、浙江中医药大学附属第一医院（浙江省东方医院）

Dates

Publication Date
20260512
Application Date
20260128

Claims (10)

  1. 1. A glomerular filtration rate estimation method based on comprehensive data analysis, the method comprising the steps of: S1, basic information, clinical data, laboratory detection data and imaging data of a patient are acquired, wherein the imaging data at least comprise kidney volumes, blood flow, thicknesses of kidney cortex and medulla and kidney function performances of right kidney and left kidney; S2, calculating a function difference measurement coefficient for quantifying asymmetry of double kidney functions by comparing kidney volume difference, blood flow difference and difference of kidney cortex and medulla thickness of the right kidney and the left kidney based on the imaging data; s3, extracting global feature vectors of the patient through a deep learning model, wherein the global feature vectors comprise imaging feature vectors of kidneys at each side, physiological feature vectors and laboratory detection feature vectors, the imaging feature vectors are extracted from imaging data, the physiological feature vectors are extracted from clinical data, the laboratory detection feature vectors are extracted from laboratory detection data, and fusion processing is carried out on the feature vectors to generate comprehensive feature vectors; S4, adjusting the functional weights of the right kidney and the left kidney in GFR estimation according to the functional difference measurement coefficient; S5, inputting the adjusted functional weight and the global feature vector into a depth regression model, calculating to obtain a corrected GFR estimated value, and outputting the corrected GFR estimated value.
  2. 2. The method for estimating glomerular filtration rate based on comprehensive data analysis according to claim 1, wherein calculating the functional difference metric coefficient for quantifying asymmetry of double kidney function by comparing the difference in renal volume, the difference in blood flow, and the difference in renal cortex and medulla thickness of the right and left kidneys comprises: S21, performing differential calculation on the kidney volumes of the right kidney and the left kidney to obtain a volume differential coefficient; s22, performing differential calculation on blood flow of the right kidney and the left kidney to generate differential coefficients of blood flow; s23, performing difference calculation on the thicknesses of kidney cortex and medulla of the right kidney and the left kidney to generate a thickness difference coefficient; s24, obtaining a functional difference measurement coefficient by carrying out weighted fusion on the volume difference coefficient, the blood flow difference coefficient and the thickness difference coefficient.
  3. 3. The method for estimating glomerular filtration rate based on comprehensive data analysis of claim 1 wherein said deep learning model comprises a plurality of convolutional neural network layers, a fully connected neural network layer, a deep neural network layer and an attention mechanism layer, extracting global feature vectors of the patient via the deep learning model, and performing fusion processing on said feature vectors to generate a comprehensive feature vector, comprising: S31, extracting kidney volume, blood flow and thickness characteristics of cortex and medulla of the right kidney and the left kidney from the imaging data through a convolutional neural network layer to generate an imaging characteristic vector; S32, processing basic information of a patient through a fully connected neural network layer to generate a basic information feature vector, wherein the basic information comprises age and gender; S33, processing physiological characteristics in clinical data through a fully connected neural network layer to generate a clinical characteristic vector, wherein the physiological characteristics comprise blood pressure, urine volume, serum creatinine and urine protein; S34, extracting laboratory detection feature vectors from laboratory detection data through a deep neural network layer, wherein the laboratory detection data comprise biomarkers in blood and urine; S35, processing the kidney function expression data through a deep neural network layer to generate a kidney function expression characteristic vector, wherein the kidney function expression data comprises a urine examination result, a kidney function change trend and a kidney disease history; S36, the attention mechanism layer adopts a weighted fusion strategy to fuse the imaging feature vector, the basic information feature vector, the clinical feature vector, the laboratory detection feature vector and the kidney function expression feature vector to generate a comprehensive feature vector.
  4. 4. The method for glomerular filtration rate estimation based on integrated data analysis of claim 3, wherein adjusting the functional weights of the right and left kidneys in the GFR estimation based on the functional difference metric coefficients comprises: s41, inputting one or more key sub-features in the functional difference measurement coefficient and the comprehensive feature vector into a pre-constructed dynamic weight generation network; the dynamic weight generation network comprises a gating circulation unit network and a multi-layer perceptron, wherein the gating circulation unit network is used for analyzing historical function change trend implied by kidney function expression feature vectors of patients so as to capture dynamic modes of kidney function compensation or decompensation; the output of the gating circulation unit network and the output of the multi-layer perceptron are interacted at a fusion layer, and a group of intermediate weight vectors are obtained through calculation by a cross attention mechanism; The intermediate weight vector is input to a differentiable soft distribution layer, and the soft distribution layer outputs a first functional weight of the right kidney and a second functional weight of the left kidney, wherein the temperature parameter of the soft distribution layer can be learned in the model training process so as to control the confidence degree and the sharpness degree of weight distribution.
  5. 5. The method for estimating glomerular filtration rate based on comprehensive data analysis according to claim 4, wherein said depth regression model includes a weighted feature path and an original feature path, and wherein said function weight and said global feature vector after adjustment are input into the depth regression model to calculate a corrected GFR estimate, comprising: S51, carrying out Hadamard product operation on the first functional weight and the image and functional expression sub-feature vector related to the right kidney to generate a right kidney weight feature, and carrying out Hadamard product operation on the second functional weight and the corresponding sub-feature vector related to the left kidney to generate a left kidney weight feature; s52, the original feature path carries out high-order nonlinear transformation and abstraction on the comprehensive feature vector through a deep network formed by a plurality of residual blocks, and global depth semantic features related to GFR are extracted; s53, splicing the kidney specific weighted context vector and the global depth semantic feature, and inputting the kidney specific weighted context vector and the global depth semantic feature into a gating fusion module, wherein the gating fusion module learns to generate a group of dynamic fusion weights for adaptively controlling the contribution proportion of weighted context information and global semantic information in a final decision; And S54, the output of the gating fusion module is processed through a regression pre-measuring head containing a dropout layer, and finally the corrected GFR estimated value is output.
  6. 6. A glomerular filtration rate estimation device based on integrated data analysis, the device comprising: The system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring basic information, clinical data, laboratory detection data and imaging data of a patient, and the imaging data at least comprise kidney volumes, blood flow, thicknesses of kidney cortex and medulla and kidney function performances of right kidney and left kidney; The function difference quantification module is used for calculating a function difference measurement coefficient for quantifying asymmetry of double kidney functions by comparing kidney volume difference, blood flow difference and difference of kidney cortex and medulla thickness of the right kidney and the left kidney based on the imaging data; The global feature extraction and fusion module is used for extracting global feature vectors of a patient through a deep learning model, and comprises an imaging feature vector of each kidney extracted from imaging data, a physiological feature vector extracted from clinical data and a laboratory detection feature vector extracted from laboratory detection data, and carrying out fusion processing on the feature vectors to generate a comprehensive feature vector; the functional weight dynamic adjustment module is used for adjusting the functional weights of the right kidney and the left kidney in GFR estimation according to the functional difference measurement coefficient; And the correction estimation output module is used for inputting the adjusted functional weight and the global feature vector into a depth regression model, calculating to obtain a corrected GFR estimated value and outputting the corrected GFR estimated value.
  7. 7. The glomerular filtration rate estimation device of claim 6, wherein the functional difference quantification module comprises: the volume difference calculation unit is used for carrying out difference calculation on the kidney volumes of the right kidney and the left kidney so as to obtain a volume difference coefficient; a blood flow difference calculation unit for performing difference calculation on blood flows of the right kidney and the left kidney to generate a blood flow difference coefficient; the thickness difference calculation unit is used for performing difference calculation on the thicknesses of the kidney cortex and the medulla of the right kidney and the left kidney so as to generate a thickness difference coefficient; and the weighted fusion unit is used for obtaining a functional difference measurement coefficient by weighted fusion of the volume difference coefficient, the blood flow difference coefficient and the thickness difference coefficient.
  8. 8. The glomerular filtration rate estimation device based on comprehensive data analysis of claim 6, wherein the deep learning model comprises a plurality of convolutional neural network layers, a fully connected neural network layer, a deep neural network layer and an attention mechanism layer, and the global feature extraction and fusion module comprises: The imaging feature extraction unit is used for extracting kidney volume, blood flow, thickness features of cortex and medulla of the right kidney and the left kidney from imaging data through the convolutional neural network layer to generate imaging feature vectors; The basic information feature extraction unit is used for processing basic information of a patient through the fully-connected neural network layer to generate a basic information feature vector, wherein the basic information comprises age and gender; The clinical feature extraction unit is used for processing physiological features in clinical data through the fully connected neural network layer to generate clinical feature vectors, wherein the physiological features comprise blood pressure, urine volume, serum creatinine and urine protein; the laboratory detection feature extraction unit is used for extracting laboratory detection feature vectors from laboratory detection data through the deep neural network layer, wherein the laboratory detection data comprise biomarkers in blood and urine; The system comprises a function performance characteristic extraction unit, a function analysis unit and a data analysis unit, wherein the function performance characteristic extraction unit is used for processing kidney function performance data through a deep neural network layer to generate kidney function performance characteristic vectors, and the kidney function performance data comprises urine examination results, kidney function change trend and kidney disease history; and the attention fusion unit is used for fusing the imaging feature vector, the basic information feature vector, the clinical feature vector, the laboratory detection feature vector and the kidney function expression feature vector by adopting a weighted fusion strategy by the attention mechanism layer to generate a comprehensive feature vector.
  9. 9. An electronic device comprising a processor and a memory in which computer program instructions are stored, characterized in that the computer program instructions, when run by the processor, cause the processor to perform the method according to any of claims 1-5.
  10. 10. A computer-readable medium, on which computer program instructions are stored, characterized in that the computer program instructions, when executed by a processor, cause the processor to perform the method according to any of claims 1-5.

Description

Glomerulus filtration rate estimation method and device for patient with double-renal function asymmetry Technical Field The invention relates to the technical field of glomerular filtration rate calculation, in particular to a glomerular filtration rate estimation method and device for patients with double-renal-function asymmetry. Background Chronic Kidney Disease (CKD) has become a common health problem worldwide, and particularly, the incidence of CKD tends to increase year by year with the increase of chronic diseases such as diabetes, hypertension, and the like. Glomerular Filtration Rate (GFR) is the most important index for evaluating kidney function, can effectively reflect the filtration capacity of kidney, and is important for diagnosis of CKD, formulation of treatment scheme and disease monitoring. Traditionally, GFR determinations have generally relied on the clearance of exogenous markers (e.g., inulin, iohexol, etc.), which, while accurate, are complex to operate and costly and are not suitable for large-scale clinical applications. The conventional alternative method is to estimate through formulas of endogenous metabolites such as serum creatinine, urea nitrogen and the like, common formulas such as a Cockcroft-Gault formula, an MDRD formula and the like, which are convenient for clinical application, but in certain special patient groups (such as unilateral kidney excision, kidney disease, asymmetric renal function and the like), the estimated result has deviation. Existing GFR estimation methods generally assume that the double kidney function is symmetrical, whereas in clinic many patients experience an asymmetry in double kidney function due to lesions, surgery or other factors. The conventional estimation method cannot effectively cope with the influence caused by the asymmetry of the double kidney functions, and thus cannot provide accurate kidney function assessment. In order to improve the problem, a dynamic correction method based on deep learning is developed, and the difference of double kidney functions can be considered, so that the technical problem to be solved is urgent. Disclosure of Invention In view of the technical problems set forth in the background art, the present invention provides a glomerular filtration rate estimation method, apparatus, electronic device and computer readable medium based on comprehensive data analysis. A first aspect of the present invention provides a glomerular filtration rate estimation method based on comprehensive data analysis, the method comprising the steps of: S1, basic information, clinical data, laboratory detection data and imaging data of a patient are acquired, wherein the imaging data at least comprise kidney volumes, blood flow, thicknesses of kidney cortex and medulla and kidney function performances of right kidney and left kidney; S2, calculating a function difference measurement coefficient for quantifying asymmetry of double kidney functions by comparing kidney volume difference, blood flow difference and difference of kidney cortex and medulla thickness of the right kidney and the left kidney based on the imaging data; s3, extracting global feature vectors of the patient through a deep learning model, wherein the global feature vectors comprise imaging feature vectors of kidneys at each side, physiological feature vectors and laboratory detection feature vectors, the imaging feature vectors are extracted from imaging data, the physiological feature vectors are extracted from clinical data, the laboratory detection feature vectors are extracted from laboratory detection data, and fusion processing is carried out on the feature vectors to generate comprehensive feature vectors; S4, adjusting the functional weights of the right kidney and the left kidney in GFR estimation according to the functional difference measurement coefficient; S5, inputting the adjusted functional weight and the global feature vector into a depth regression model, calculating to obtain a corrected GFR estimated value, and outputting the corrected GFR estimated value. A second aspect of the present invention provides a glomerular filtration rate estimation device based on integrated data analysis, the device comprising: The system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring basic information, clinical data, laboratory detection data and imaging data of a patient, and the imaging data at least comprise kidney volumes, blood flow, thicknesses of kidney cortex and medulla and kidney function performances of right kidney and left kidney; The function difference quantification module is used for calculating a function difference measurement coefficient for quantifying asymmetry of double kidney functions by comparing kidney volume difference, blood flow difference and difference of kidney cortex and medulla thickness of the right kidney and the left kidn