Method and system for constructing renal function automatic estimation model based on plain-scan CT (Computed Tomography)

Abstract

The invention discloses a method and a system for constructing a renal function automatic estimation model based on plain-scan CT (Computed Tomography). The method comprises the following steps: S1, collecting image data of a patient suffering from renal atrophy or hydronephrosis; s2, training a kidney parenchyma and hydronephrosis region automatic segmentation model based on UNETR by using kidney parenchyma and hydronephrosis regions manually described by a radiologist; s3, image omics features are extracted from the images, and data are screened; s4, constructing a renal function index regression model, wherein the renal function index regression model comprises a GFR regression model and an SRF regression model; s5, verifying an estimated value of the regression model; and S6, calculating GFR and SRF by using the renal function index regression model, and distinguishing kidneys in different health states. By adopting the method, the glomerular filtration rate and the lateral renal function of each kidney can be automatically estimated. A rapid, convenient and low-cost renal function evaluation method is provided for treating patients with renal atrophy or hydronephrosis.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and system for constructing an automatic renal function estimation model based on plain scan CT. Background Art

[0002] GFR (Glomerular Filtration Rate) is a key indicator of kidney function, while SRF (Split Renal Function) reflects the contribution of a single kidney to overall kidney function. Accurate assessment of these parameters is crucial in urological surgical decisions, such as the treatment of patients with renal atrophy or hydronephrosis.

[0003] Limitations of existing technologies:

[0004] 1. Single photon emission computed tomography (SPECT): As the gold standard method for measuring unilateral renal GFR and SRF, SPECT has obvious limitations. First, SPECT involves the use of radioactive isotopes, which poses a potential radiation risk to patients. Secondly, the operation process of SPECT is complicated and requires professional technicians to operate and interpret, which limits its wide application in clinical applications. Furthermore, SPECT equipment is expensive and the examination cost is high, which is not conducive to medical environments with limited resources. Finally, SPECT imaging takes a long time, which is not conducive to emergency or situations where rapid diagnosis is required.

[0005] 2. Contrast-enhanced computed tomography (CECT): As another imaging method for evaluating renal function, CECT uses iodinated contrast agents to enhance the image of the kidneys, thereby estimating GFR and SRF. However, iodinated contrast agents may cause additional nephrotoxicity risks in patients with renal insufficiency and are not suitable for patients with a history of contrast agent allergy. In addition, CECT also requires specific equipment and professional operators, which limits its clinical application.

[0006] 3. Non-contrast CT (NCCT): Non-contrast CT (i.e. CT scan without iodine contrast agent) has certain potential in evaluating renal function because it avoids the side effects and costs associated with contrast agents. However, the application of traditional non-contrast CT in evaluating renal function is limited by its low accuracy and lack of automated image processing methods: manual segmentation of the kidney area is time-consuming and subjective, and GFR is calculated only based on the renal parenchymal area. The parameters are too simple and the accuracy is not high, which limits its application in clinical practice. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for constructing an automatic renal function estimation model based on plain scan CT to solve the problems raised in the above background technology.

[0008] To achieve the above-mentioned object of the invention, one aspect of the present invention provides a method for constructing an automatic estimation model of renal function based on plain scan CT, comprising the following steps:

[0009] Step S1, collecting image data of patients with renal atrophy or hydronephrosis and preprocessing the images;

[0010] Step S2, using the renal parenchyma and hydronephrosis regions manually drawn by radiologists, training a renal parenchyma and hydronephrosis region automatic segmentation model based on UNETR (UNet Transformers);

[0011] Step S3, extracting radiomics features from the images of the renal parenchyma and hydronephrosis area obtained by the automatic segmentation model in step S2 and the original image, and screening the data;

[0012] Step S4, constructing a renal function index regression model, including a GFR regression model and a SRF regression model.

[0013] Step S5, verifying the consistency and difference between the estimated value of the regression model and the value based on SPECT measurement;

[0014] Step S6, using the renal function index regression model to calculate GFR and SRF, and distinguish kidneys in different health states.

[0015] Further, step S1 includes the following steps:

[0016] Step S101, collecting data of patients with renal atrophy or hydronephrosis, wherein the patients undergo NCCT and SPECT imaging, and the gold standard G of the patients' GFR and SRF is calculated by the SPECT workstation as the dependent variable of regression modeling;

[0017] Step S102, preprocessing the NCCT image, including clipping the range of voxel grayscale I to the renal window and normalizing it using the z-score method, which is expressed as follows:

[0018] ,

[0019] in, is to find the grayscale mean, is to find the grayscale standard deviation, is the normalized voxel grayscale.

[0020] Further, step S2 includes the following steps:

[0021] Step S201, using the renal parenchyma and hydronephrosis area manually drawn by a radiologist as training samples for a deep learning-based automatic renal segmentation model;

[0022] Step S202, training a UNETR-based automatic segmentation model for renal parenchyma and hydronephrosis region.

[0023] Further, step S3 includes the following steps:

[0024] Step S301, extracting radiomics features from the NCCT image of the renal parenchyma and hydronephrosis area and the original image obtained by the automatic segmentation model in step S2, including shape features, first-order statistical features, texture features and filtering features;

[0025] Step S302, screening the data with reference to ICC (Intra class consistency coefficient) and SCC (Spearman correlation coefficients).

[0026] Furthermore, the reference ICC was used to screen the data to exclude features with an ICC lower than 0.9 between manual and automatic segmentation to improve the reproducibility of radiomics;

[0027] Furthermore, the reference SCC is used to screen the data by calculating the SCC between one feature and all other features in each highly correlated feature pair, and features with SCC>0.9 are excluded.

[0028] Furthermore, the GFR regression model construction method includes the following steps:

[0029] Step S411: The most representative features are selected from the screened features using the ElasticNet model, and the feature weights are determined to estimate rGFR (radiomics based GFR estimation, GFR estimation based on radiomics features). The remaining features are input into ElasticNet to obtain rGFR (mL / min×1.73m 2 ) is calculated as:

[0030]

[0031] Among them, p_original_glszm_LargeAreaHighGrayLevelEmphasis is the large area high gray level emphasis value based on the gray size area matrix in the renal parenchyma area, p_original_glszm_SizeZoneNonUniformity is the area size non-uniformity based on the gray size area matrix in the renal parenchyma area, p_gradient_ngtdm_Busyness is the busyness based on the gradient neighborhood gray difference matrix in the renal parenchyma area, p_square_firstorder_TotalEnergy is the total energy in the renal parenchyma area, and h_square_ngtdm_Busyness is the busyness based on the gradient neighborhood gray difference matrix in the hydronephrosis area;

[0032] Step S412: Combine rGFR with clinical characteristics and use multivariate linear regression to obtain rcGFR (radiomics-clinical based GFR estimation, GFR estimation value combining radiomics features and clinical characteristics). rcGFR (mL / min×1.73m 2 ) is calculated as:

[0033] rcGFR = age × (-0.147) + rGFR × (1.224) - 1.303,

[0034] Where rcGFR is the estimated GFR.

[0035] Furthermore, the SRF regression model construction method includes the following steps:

[0036] Step S421, calculate the relative contribution ratio of unilateral renal rcGFR to the total rcGFR rcSRF (radiomics-clinical based SRFestimation, SRF estimation value combining radiomics features and clinical features), expressed as the following formula:

[0037] rcSRFl = rcGFRl / (rcGFRl + rcGFRr),

[0038] rcSRFr = 1 - rcSRFl,

[0039] Among them, rcGFR1 and rcGFRr are rcGFR of the left and right kidneys, respectively, and rcSRF1 and rcSRFr are rcSRF of the left and right kidneys, respectively;

[0040] Step S422, calculate the ratio of the unilateral renal parenchymal volume RPV to the total renal parenchymal volume RPV_P, the calculation formula is:

[0041] RPV_Pl = RPVl / (RPVl + RPVr),

[0042] RPV_Pr = 1 - RPV_Pl,

[0043] Among them, RPV1 and RPVr are the RPVs of the left and right kidneys, respectively, and RPV_Pl and RPV_Pr are the RPV_Ps of the left and right kidneys, respectively;

[0044] Linear regression was used to obtain the estimated SRF value pSRF based on the renal parenchyma, and the calculation formula was:

[0045] pSRF = 0.991×RPV_P+ 0.5%,

[0046] Where RPV_P is RPV_Pl or RPV_Pr, and the pSRF of the left or right kidney is calculated accordingly;

[0047] Step S423, calculating the ratio RHV_P of the unilateral hydronephrosis volume RHV to the total hydronephrosis volume, using the following formula:

[0048] RHV_Pl = RHVl / (RHVl + RHVr),

[0049] RHV_Pr = 1 - RHV_Pl,

[0050] Among them, RHVl and RHVr are the RHV of the left and right kidneys, respectively, and RHV_Pl and RHV_Pr are the RPV_P of the left and right kidneys, respectively;

[0051] Linear regression was used to obtain the estimated SRF value hSRF based on hydronephrosis, calculated as:

[0052] hSRF = (-0.223)×RHV_P + 61.1%,

[0053] Among them, RHV_P is RHV_Pl or RHV_Pr, and the hSRF of the left kidney or right kidney is calculated accordingly;

[0054] Step S424, linear regression is used to obtain the final SRF estimation value rcphSRF of the left kidney or right kidney combining rcSRF, pSRF and hSRF, and the calculation formula is:

[0055] rcphSRF = rcSRF×0.416 + pSRF×0.564 + hSRF×0.535 - 25.8%.

[0056] Furthermore, in step S6, kidneys in different health states are distinguished according to the following rules:

[0057] The damaged kidney is a single kidney with GFR ≤ 30 mL / min × 1.73 m 2 A normal kidney has a single kidney GFR>30mL / min×1.73m 2 ;

[0058] Non-functioning kidney means GFR of one kidney is less than 10 mL / min × 1.73 m 2 , functional kidney means single kidney GFR ≥ 10 mL / min × 1.73 m 2 ;

[0059] Suitable for nephrectomy is SRF <15%, and non-nephrectomy is SRF ≥15%;

[0060] Decreased renal function was defined as SRF < 45%, and no decline was defined as SRF ≥ 45%;

[0061] The non-dominant kidney has SRF ≤ 60%, and the dominant kidney has SRF > 60%.

[0062] Another aspect of the present invention provides a system for constructing a renal function automatic estimation model based on plain scan CT, comprising a collection module, a segmentation module, an extraction module, a modeling module, a verification module, and an application module, wherein:

[0063] The collection module is used to collect image data of patients with renal atrophy or hydronephrosis and pre-process the images;

[0064] The segmentation module uses the renal parenchyma and hydronephrosis areas manually drawn by radiologists to train the UNETR-based automatic segmentation model of renal parenchyma and hydronephrosis areas;

[0065] The extraction module extracts radiomic features from the images of the renal parenchyma and hydronephrosis area obtained by the automatic segmentation model of the segmentation module and the original images, and screens the data;

[0066] The modeling module is used to construct the regression model of renal function indicators, including GFR regression model and SRF regression model;

[0067] The validation module is used to verify the consistency and difference between the estimated values ​​of the regression model and the values ​​based on SPECT measurements;

[0068] The application module uses the renal function index regression model to calculate GFR and SRF and distinguish kidneys in different health states.

[0069] Compared with the prior art, the present system and method have the following advantages:

[0070] 1. The model is established using plain scan CT data, avoiding the radiation risk and nephrotoxicity risk caused by other data acquisition processes.

[0071] 2. The automatic segmentation model of renal parenchyma and hydronephrosis area based on UNETR is trained. While using transformer as encoder, UNETR maintains the U-shaped structure of U-Net, has the advantage of extracting image features of different scales, and is suitable for segmenting renal parenchyma and hydronephrosis areas of different sizes.

[0072] 3. GFR regression modeling and SRF regression modeling can automatically estimate the glomerular filtration rate and lateral renal function of each kidney. For the treatment of patients with renal atrophy or hydronephrosis, a fast, convenient and low-cost method for assessing renal function is provided. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 Flowchart of the method for constructing an automatic estimation model of renal function based on plain CT.

[0074] Figure 2 The correlation diagram between the estimated value and the SPECT measured value. The dotted line indicates complete agreement. (A) is rGFR, (B) is rcGFR, (C) is rcSRF, (D) is pSRF, (E) is hSRF, and (F) is rcphSRF.

[0075] Figure 3 Bland Altman plots of regression model estimates and SPECT-based measurements. The 95% limits of agreement (LoA) are between the dashed horizontal lines. The dashed dotted lines represent the mean differences. (A) rGFR, (B) rcGFR, (C) rcSRF, (D) pSRF, (E) hSRF, and (F) rcphSRF. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] like Figure 1 FIG. 1 is a flow chart of a method of the present invention. An embodiment of the present invention provides an artificial intelligence (AI) technology based on NCCT, which is used to automatically estimate the glomerular filtration rate (GFR) and split renal function (SRF) of each kidney. The specific steps are as follows:

[0078] Step S1, data collection and preprocessing, specifically includes the following steps:

[0079] Step S101, collecting data of patients with renal atrophy or hydronephrosis, these patients underwent NCCT and SPECT imaging, and GFR and SRF were calculated by the SPECT workstation.

[0080] Step S102, preprocessing the NCCT image, including clipping the range of voxel grayscale I to the kidney window (window width = 260 HU, window level = 60 HU) and normalizing it using the z-score method:

[0081] ,

[0082] in, is to find the grayscale mean, It is to find the grayscale standard deviation.

[0083] Step S2, automatic kidney segmentation, specifically includes the following steps:

[0084] Step S201, using the renal parenchyma and hydronephrosis area manually delineated by a radiologist as training samples for a deep learning-based automatic renal segmentation model.

[0085] Step S202, training a UNETR-based automatic segmentation model for renal parenchyma and hydronephrosis. UNETR uses transformer as an encoder while maintaining the U-Net structure. It has the advantage of extracting image features of different scales and is suitable for segmenting renal parenchyma and hydronephrosis areas of different sizes. UNETR is trained using Python 3.9 and PyTorch 1.9 on a workstation powered by NVIDIA GeForce RTX2080ti. The parameters are as follows: max epoch = 1000; loss function is diceloss; optimizer is adamw; learning rate = 0.0001; batch size = 4. Data augmentation is performed during model training, including random cropping, flipping, and rotation.

[0086] Step S3, radiomics feature extraction, specifically includes the following steps:

[0087] Step S301, extracting 2260 radiomics features from the NCCT images of the renal parenchyma and hydronephrosis area obtained by the automatic segmentation model in step S2, including 28 shape features, 36 first-order statistical features, 150 texture features and 2046 filter features. Shape features, first-order statistical features and texture features are extracted from the original image, and filter features are extracted from the wavelet filter, square filter, logarithmic filter and gradient filter images.

[0088] Step S302: To improve the reproducibility of radiomics, features with an ICC lower than 0.9 between manual and automatic segmentation were excluded. In addition, pairwise feature SCCs were calculated, and feature pairs with SCC values ​​> 0.9 were considered highly correlated. In each highly correlated feature pair, the SCC between one feature and all other features was calculated, and features with larger mean SCCs were considered redundant and excluded.

[0089] Step S4, renal function index regression modeling, specifically includes the following steps:

[0090] Step S401, GFR regression modeling, specifically includes the following steps:

[0091] In step S411, the ElasticNet model is used to select the most representative features from the screened features, and the feature weights are determined to estimate the unilateral renal GFR, i.e., rGFR. The remaining features are input into ElasticNet, and the patient's GFR calculated by the SPECT workstation is used as the dependent variable to generate the GFR (rGFR) based on imaging omics features. ElasticNet can select the most representative features and determine the feature weights to estimate the single kidney GFR. The hyperparameters of ElasticNet are determined by ten-fold cross validation. Finally, rGFR (mL / min×1.73m 2 ) is calculated as:

[0092]

[0093] Among them: p_original_glszm_LargeAreaHighGrayLevelEmphasis is the large area high gray level emphasis value based on the gray size zone matrix in the renal parenchyma region, p_original_glszm_SizeZoneNonUniformity is the region size non-uniformity based on the gray size zone matrix in the renal parenchyma region, p_gradient_ngtdm_Busyness is the busyness based on the gradient neighborhood gray difference matrix in the renal parenchyma region, p_square_firstorder_TotalEnergy is the total energy in the renal parenchyma region, and h_square_ngtdm_Busyness is the busyness based on the gradient neighborhood gray difference matrix in the hydronephrosis region.

[0094] Step S412, rGFR is combined with clinical characteristics, and multivariate linear regression (MLR) is used to obtain the GFR combining radiomics and clinical characteristics, namely rcGFR (radiomics-clinical based GFR estimation, GFR estimation value combining radiomics characteristics and clinical characteristics). rcGFR (mL / min×1.73m 2 ) is calculated as:

[0095] rcGFR = age × (-0.147) + rGFR × (1.224) - 1.303,

[0096] Among them, rcGFR is the final GFR estimation result.

[0097] Step S402, SRF regression modeling, specifically includes the following steps:

[0098] Step S421, calculate the relative contribution ratio of unilateral renal rcGFR to the total rcGFR rcSRF (radiomics-clinical based SRFestimation, SRF estimation value combining radiomics characteristics and clinical characteristics):

[0099] rcSRFl = rcGFRl / (rcGFRl + rcGFRr),

[0100] rcSRFr = 1 – rcSRFl,

[0101] Among them, rcGFRl and rcGFRr are the rcGFR of the left and right kidneys, respectively, calculated according to the method in step S412; rcSRFl and rcSRFr are the rcSRF of the left and right kidneys, respectively.

[0102] Step S422, calculate the ratio of the unilateral renal parenchymal volume RPV to the total renal parenchymal volume RPV_P:

[0103] RPV_Pl = RPVl / (RPVl + RPVr),

[0104] RPV_Pr = 1 - RPV_Pl,

[0105] Among them, RPVl and RPVr are the RPV of the left and right kidneys respectively, and RPV_Pl and RPV_Pr are the RPV_P of the left and right kidneys respectively.

[0106] Linear regression was used to obtain the renal parenchyma-based SRF estimate pSRF:

[0107] pSRF = 0.991×RPV_P+ 0.5%,

[0108] Among them, RPV_P is RPV_Pl or RPV_Pr, and the pSRF of the left kidney or right kidney is calculated accordingly.

[0109] Step S423, calculating the ratio RHV_P of the unilateral hydronephrosis volume RHV to the total hydronephrosis volume:

[0110] RHV_Pl = RHVl / (RHVl + RHVr),

[0111] RHV_Pr = 1 - RHV_Pl,

[0112] Among them, RHVl and RHVr are the RHV of the left and right kidneys respectively, and RHV_Pl and RHV_Pr are the RPV_P of the left and right kidneys respectively.

[0113] Linear regression was used to obtain an estimate of SRF based on hydronephrosis:

[0114] hSRF = (-0.223)×RHV_P + 61.1%,

[0115] Among them, RHV_P is RHV_Pl or RHV_Pr, and the hSRF of the left kidney or right kidney is calculated accordingly.

[0116] Step S424, using linear regression, taking the patient's SRF calculated by the SPECT workstation as the dependent variable, and obtaining the final SRF estimation value rcphSRF of the left or right kidney combining rcSRF, pSRF and hSRF:

[0117] rcphSRF = rcSRF×0.416 + pSRF×0.564 + hSRF×0.535 - 25.8%.

[0118] Step S5, verify the consistency and difference between the estimated values ​​of the regression model and the values ​​based on SPECT measurements, using the slope and intercept of the simplified long axis, Lin's coefficient of consistency (CCC), and Bland-Altman plot with limits of agreement (LoA).

[0119] Table 1 shows the feature distribution of the patient data sets used for model training and testing. The training set has 128 cases and the test set has 117 cases. Except for RPV and single kidney GFR, there are no significant differences in other features.

[0120] Table 1. Clinical characteristics of the patients.

[0121] Clinical characteristics Training set Test Set p gender 0.853 male 74 (57.8) 69 (59.0) female 54 (42.2) 48 (41.0) Age (years) 52.76±13.70 55.79±16.39 0.070 BMI 23.90±3.29 23.86±3.23 RPV (ml) 152.83±51.58 132.70±60.43 <0.01 RHV (ml) 34.51±67.98 47.51±91.18 0.408 <![CDATA[Single kidney GFR (mL / min×1.73m 2 )]]> 40.36±14.43 36.27±19.41 0.002

[0122] For sex, data are the number of patients, and data in brackets are percentages. Age, BMI, RPV, RHV, and single kidney GFR are mean ± standard deviation. BMI = body mass index, RPV = renal parenchymal volume, RHV = hydronephrosis volume, GFR = glomerular filtration rate. p values ​​are derived from the Mann-Whitney U test (continuous parameters) and the χ2 test (categorical variables).

[0123] The correlation between the model-estimated GFR and the SPECT-measured GFR is shown in Tables 2 and Figure 2 As shown. Pearson correlation analysis showed that rGFR had a good correlation with GFR measured by SPECT (r=0.72, P<0.001), and the paired sample t-test showed that there was no statistically significant difference between rGFR and GFR measured by SPECT (P>0.05). At the same time, the CCC of rGFR was relatively poor, at 0.60. Multiple linear regression showed that age and rGFR were independent predictors, and rcGFR was calculated as a linear combination of age and rGFR (Table 2). Compared with rGFR, rcGFR had a stronger correlation (r=0.75 vs.0.72), a smaller difference (0.34 vs.1.39), and a higher CCC value (0.70 vs.0.60). In addition, the slope of rcGFR decreased (1.46 vs. 1.85) and the intercept increased (-17.53 vs. -33.68), making the two closer to perfect consistency (slope = 1 and intercept = 0). These results indicate that rcGFR has better predictive ability. In addition, as Figure 3 The Bland-Altman plot in confirms that most of the difference between the estimated GFR and the SPECT-measured GFR was within the LoA, indicating good agreement.

[0124] The correlation between the SRF estimated by the model and the SRF measured by SPECT is shown in Tables 2 and Figure 2Results: The rcSRF was calculated based on the rcGFR and was significantly correlated with the SPECT-measured SRF (r = 0.83, p < 0.001) with a CCC of 0.78. Compared with the rcSRF, the pSRF showed similar correlation (r = 0.84 vs. 0.83) and CCC (0.79 vs. 0.78), slope (1.43 vs. 1.52) and intercept (-0.26 vs. -0.22), but with a larger mean difference (1.20 vs. 0.94). The hSRF had the weakest correlation (r = 0.73). The rcSRF, pSRF, and hSRF were identified as independent predictors of SRF by multiple linear regression and subsequently combined to form the rcphSRF. In the test set, the rcphSRF had the strongest correlation (r = 0.92, p < 0.001), the smallest mean difference of 0.51, and the highest CCC of 0.88. In addition, the slope (1.37) and intercept (-0.18) of rcphSRF were closest to perfect agreement. These results confirm that rcphSRF has the best predictive ability. Figure 3 The Bland-Altman plot in further confirmed that more than 90% of the differences between the estimated SRF and the SPECT-measured SRF were within the LoA, indicating good agreement.

[0125] Table 2. Comparison of estimated values ​​with SPECT-based measurements.

[0126] Estimated value r <![CDATA[p § ]]> Difference <![CDATA[p * ]]> CCC Slope Intercept rGFR Training set 0.67 <0.001 0.00±10.94 0.989 0.55(0.48,0.61) 1.93 -37.52 Test Set 0.72 <0.001 1.39±13.93 0.085 0.60(0.54,0.66) 1.85 -33.68 rcGFR Training set 0.68 <0.001 0.00±10.54 0.762 0.64(0.57,0.70) 1.72 -18.70 Test Set 0.75 <0.001 0.34±12.87 0.669 0.70(0.64,0.75) 1.46 -17.53 rcSRF Training set 0.85 <0.001 1.03±8.50 0.011 0.77(0.73,0.81) 1.51 -0.26 Test Set 0.83 <0.001 0.94±14.23 0.314 0.78(0.74,0.81) 1.52 -0.26 pSRF Training set 0.82 <0.001 2.13±7.81 <0.001 0.79(0.74,0.83) 1.24 -0.12 Test Set 0.84 <0.001 1.20±12.58 0.212 0.79(0.75,0.83) 1.43 -0.22 hSRF Training set 0.68 <0.001 0.04±10.08 0.801 0.64(0.57,0.70) 1.46 -2.28 Test Set 0.73 <0.001 0.64±16.87 0.507 0.53(0.48,0.59) 2.32 -0.66 rcphSRF Training set 0.90 <0.001 1.39±6.02 <0.001 0.89(0.86,0.91) 1.12 -0.06 Test Set 0.92 <0.001 0.51±9.63 0.406 0.88(0.86,0.90) 1.37 -0.18

[0127] The difference is expressed as mean ± standard deviation, and the unit of GFR is mL / min×1.73m 2 , the numerical unit of SRF is %. r is the Pearson correlation coefficient. The slope and intercept are obtained by reduced major axis regression. CCC is Lin's consistency coefficient, and the 95% confidence interval is in parentheses. GFR=glomerular filtration rate, SRF=lateral renal function, rGFR=GFR estimation based on radiomics features, rcGFR=GFR estimation based on radiomics features and clinical features. rcSRF is calculated as the relative contribution of single kidney rcGFR to total rcGFR, pSRF is estimated based on the percentage of renal parenchymal volume (RPV), hSRF is estimated based on the percentage of hydronephrosis volume (RHV), and rcphSRF is calculated by multivariate linear regression of rcSRF, pSRF and hSRF. p § Values ​​are from Pearson correlation analysis, p * Values ​​were analyzed using paired sample t test.

[0128] Step S6: To evaluate the clinical application value of the model, the cutoff values ​​reported in previous studies were used to calculate the estimated values ​​for differentiating the performance indicators of kidneys in different health states: (1) damaged kidneys (single kidney GFR ≤ 30 mL / min × 1.73 m 2 ) or normal kidney (single kidney GFR> 30mL / min×1.73m 2 ); (2) Non-functional kidney (single kidney GFR < 10 mL / min × 1.73 m 2 ) or functional kidney (single kidney GFR ≥ 10mL / min×1.73m 2 ); (3) suitable for nephrectomy (SRF < 15%) or non-nephrectomy (SRF ≥ 15%); (4) impaired renal function (SRF < 45%) or not impaired (SRF ≥ 45%); (5) non-dominant kidney (SRF ≤ 60%) or dominant kidney (SRF > 60%). The area under the receiver operating characteristic curve (AUC) and the accuracy, sensitivity, and specificity of judging kidneys in different health states were calculated.

[0129] The calculation results of the discriminative performance of rcGFR and rcphSRF for kidneys in different health states are shown in Table 3. The discrimination of damaged or normal, nonfunctional or functional kidneys was evaluated based on rcGFR, and other health states were evaluated based on rcphSRF. In general, rcGFR and rcphSRF showed high AUC, accuracy, and specificity.

[0130] Table 3. Performance of rcGFR and rcphSRF in differentiating kidneys in different health states.

[0131] Health status AUC (95% CI) Accuracy (95% CI) Sensitivity (95% CI) Specificity (95% CI) Damaged or normal Training set 0.840(0.779, 0.901) 85.2 [218 / 256](80.2, 89.3) 46.7 [28 / 60](33.7, 60.0) 96.9 [190 / 196](93.5, 98.9) Test Set 0.862(0.813, 0.910) 81.2 [190 / 234](75.6, 86.0) 71.4 [70 / 98](61.4, 80.1) 88.2 [120 / 136](81.6, 93.1) No function or function Training set 0.977(0.956, 0.998) 98.4 [252 / 256](96.0, 99.6) 0 [0 / 4](0, 60.2) 100 [252 / 252](98.5, 100) Test Set 0.911(0.856, 0.965) 91.0 [213 / 234](86.6, 94.4) 10.5 [2 / 19](1.3, 33.1) 98.1 [211 / 215](95.3, 99.5) Suitable for nephrectomy or non-nephrectomy Training set 1.000(1.000, 1.000) 100 [256 / 256](98.6, 100) 100 [2 / 2](15.8, 100) 100 [254 / 254](98.6, 100) Test Set 0.959(0.929, 0.989) 97.0 [227 / 234](93.9, 98.8) 22.2 [2 / 9](2.8, 60.0) 100 [225 / 225](98.4, 100) Is renal function low? Training set 0.938(0.910, 0.967) 86.7 [222 / 256](81.9, 90.6) 80.0 [60 / 75](69.2, 88.4) 89.5 [162 / 181](84.1, 93.6) Test Set 0.980(0.966, 0.994) 92.3 [216 / 234](88.1, 95.4) 88.2 [90 / 102](80.4, 93.8) 95.5 [126 / 132](90.4, 98.3) Nondominant or dominant kidney Training set 0.958(0.934, 0.982) 90.6 [222 / 256](86.4, 93.9) 96.6 [198 / 205](93.1, 98.6) 66.7 [34 / 51](52.1, 79.2) Test Set 0.964(0.944, 0.983) 87.2 [204 / 234](82.2, 91.2) 93.2 [136 / 146](87.8, 96.7) 77.3 [68 / 88](67.1, 85.5) average value Training set 0.943 92.2 64.7 90.6 Test Set 0.935 89.7 57.1 91.8

[0132] Where: Data are percentages, the number of kidneys is in square brackets, and the 95% confidence interval is in brackets. Kidneys with poor health (i.e., damaged, nonfunctional, nephrectomized, low renal function, or nondominant kidney) were considered positive to calculate sensitivity and specificity. The performance index for distinguishing damaged or normal, nonfunctional or functional kidneys was calculated by rcGFR, and other indexes were calculated by rcphSRF.

[0133] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing an automatic estimation model of renal function based on plain scan CT, characterized in that: The following steps are involved: Step S1, collecting image data of patients with renal atrophy or hydronephrosis and preprocessing the images; Step S2, using the renal parenchyma and hydronephrosis regions manually drawn by radiologists, training a UNETR-based automatic segmentation model for renal parenchyma and hydronephrosis regions; Step S3, extracting radiomics features from the images of the renal parenchyma and hydronephrosis area obtained by the automatic segmentation model in step S2 and the original image, and screening the data; Step S4, constructing a renal function index regression model, including a GFR regression model and a SRF regression model; Step S5, verifying the consistency and difference between the estimated value of the regression model and the value based on SPECT measurement; Step S6, using the renal function index regression model to calculate GFR and SRF, and distinguish kidneys in different health states.

2. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 1, characterized in that: Step S1 includes the following steps: Step S101, collecting data of patients with renal atrophy or hydronephrosis, wherein the patients undergo NCCT and SPECT imaging, and the gold standard G of the patients' GFR and SRF is calculated by the SPECT workstation as the dependent variable of regression modeling; Step S102, preprocessing the NCCT image, including clipping the range of voxel grayscale I to the renal window and normalizing it using the z-score method, which is expressed as follows: ， in, is to find the grayscale mean, is to find the grayscale standard deviation, is the normalized voxel grayscale.

3. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 1, characterized in that: Step S2 includes the following steps: Step S201, using the renal parenchyma and hydronephrosis area manually drawn by a radiologist as training samples for a deep learning-based automatic renal segmentation model; Step S202, training a UNETR-based automatic segmentation model for renal parenchyma and hydronephrosis region.

4. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 1, characterized in that: Step S3 includes the following steps: Step S301, extracting radiomics features from the NCCT image of the renal parenchyma and hydronephrosis area and the original image obtained by the automatic segmentation model in step S2, including shape features, first-order statistical features, texture features and filtering features; Step S302, filtering the data with reference to the ICC and SCC.

5. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 4, characterized in that: The reference ICC was used to filter the data to exclude features with an ICC lower than 0.9 between manual and automatic segmentation to improve the reproducibility of radiomics.

6. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 4, characterized in that: The reference SCC is used to screen the data by calculating the SCC between one feature and all other features in each highly correlated feature pair, and features with SCC>0.9 are excluded.

7. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 1, characterized in that: The GFR regression model construction method comprises the following steps: Step S411, the most representative features are selected from the screened features using the ElasticNet model, and the feature weights are determined to estimate rGFR. The remaining features are input into ElasticNet to obtain rGFR (mL / min×1.73m 2 ) is calculated as: Among them, p_original_glszm_LargeAreaHighGrayLevelEmphasis is the large area high gray level emphasis value based on the gray size area matrix in the renal parenchyma area, p_original_glszm_SizeZoneNonUniformity is the area size non-uniformity based on the gray size area matrix in the renal parenchyma area, p_gradient_ngtdm_Busyness is the busyness based on the gradient neighborhood gray difference matrix in the renal parenchyma area, p_square_firstorder_TotalEnergy is the total energy in the renal parenchyma area, and h_square_ngtdm_Busyness is the busyness based on the gradient neighborhood gray difference matrix in the hydronephrosis area; Step S412, combining rGFR with clinical characteristics, using multivariate linear regression to obtain rcGFR, rcGFR (mL / min×1.73m 2 ) is calculated as: rcGFR = age × (-0.147) + rGFR × (1.224) - 1.303, Where rcGFR is the estimated GFR.

8. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 7, characterized in that: The SRF regression model construction method comprises the following steps: Step S421, calculate the relative contribution ratio rcSRF of unilateral kidney rcGFR to the total rcGFR, expressed as the following formula: rcSRFl = rcGFRl / (rcGFRl + rcGFRr), rcSRFr = 1 – rcSRFl, Among them, rcGFR1 and rcGFRr are rcGFR of the left and right kidneys, respectively, and rcSRF1 and rcSRFr are rcSRF of the left and right kidneys, respectively; Step S422, calculate the ratio of the unilateral renal parenchymal volume RPV to the total renal parenchymal volume RPV_P, the calculation formula is: RPV_Pl = RPVl / (RPVl + RPVr), RPV_Pr = 1 - RPV_Pl, Among them, RPV1 and RPVr are the RPVs of the left and right kidneys, respectively, and RPV_Pl and RPV_Pr are the RPV_Ps of the left and right kidneys, respectively; Linear regression was used to obtain the estimated SRF value pSRF based on the renal parenchyma, and the calculation formula was: pSRF = 0.991×RPV_P+ 0.5%, Among them, RPV_P is RPV_Pl or RPV_Pr, and the pSRF of the left kidney or right kidney is calculated accordingly; Step S423, calculating the ratio RHV_P of the unilateral hydronephrosis volume RHV to the total hydronephrosis volume, using the following formula: RHV_Pl = RHVl / (RHVl + RHVr), RHV_Pr = 1 - RHV_Pl, Among them, RHVl and RHVr are the RHV of the left and right kidneys, respectively, and RHV_Pl and RHV_Pr are the RPV_P of the left and right kidneys, respectively; Linear regression was used to obtain the estimated SRF value hSRF based on hydronephrosis, calculated as: hSRF = (-0.223)×RHV_P + 61.1%, Among them, RHV_P is RHV_Pl or RHV_Pr, and the hSRF of the left kidney or right kidney is calculated accordingly; Step S424, linear regression is used to obtain the final SRF estimation value rcphSRF of the left kidney or right kidney combining rcSRF, pSRF and hSRF, and the calculation formula is: rcphSRF = rcSRF×0.416 + pSRF×0.564 + hSRF×0.535 - 25.8%.

9. The method for constructing an automatic renal function estimation model based on plain scan CT according to claim 1, characterized in that: In step S6, kidneys in different health states are distinguished according to the following rules: The damaged kidney is a single kidney with GFR ≤ 30 mL / min × 1.73 m 2 A normal kidney has a single kidney GFR>30mL / min×1.73m 2 ; Non-functioning kidney means GFR of one kidney is less than 10 mL / min × 1.73 m 2 , functional kidney means single kidney GFR ≥ 10 mL / min × 1.73 m 2 ; Suitable for nephrectomy is SRF <15%, and non-nephrectomy is SRF ≥15%; Decreased renal function was defined as SRF < 45%, and no decline was defined as SRF ≥ 45%; The non-dominant kidney has SRF ≤ 60%, and the dominant kidney has SRF > 60%.

10. A system for automatically estimating a renal function model based on plain scan CT, characterized in that: It includes collection module, segmentation module, extraction module, modeling module, verification module and application module, among which: The collection module is used to collect image data of patients with renal atrophy or hydronephrosis and pre-process the images; The segmentation module uses the renal parenchyma and hydronephrosis areas manually drawn by radiologists to train the UNETR-based automatic segmentation model of renal parenchyma and hydronephrosis areas; The extraction module extracts radiomic features from the images of the renal parenchyma and hydronephrosis area obtained by the automatic segmentation model of the segmentation module and the original images, and screens the data; The modeling module is used to construct the regression model of renal function indicators, including GFR regression model and SRF regression model; The validation module is used to verify the consistency and difference between the estimated values ​​of the regression model and the values ​​based on SPECT measurements; The application module uses the renal function index regression model to calculate GFR and SRF and distinguish kidneys in different health states.

Images