A method and system for auxiliary identification of diabetic nephropathy based on CT images
By segmenting the region of interest from abdominal CT images and combining it with comprehensive analysis of radiomics and clinical data, this method solves the problems of insufficient sensitivity in the detection of diabetic nephropathy and the difficulty in identifying early microstructural damage by imaging examinations in existing technologies, thus achieving non-invasive and accurate auxiliary identification of diabetic nephropathy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANDONG UNIV QILU HOSPITAL
- Filing Date
- 2026-03-27
- Publication Date
- 2026-06-23
AI Technical Summary
In existing technologies, the detection of diabetic nephropathy relies on biomarkers with insufficient sensitivity, and routine imaging examinations are unable to identify early microstructural damage.
By segmenting multiple regions of interest from abdominal CT images, radiomics features are extracted and identified using radiomics models, clinical models, and combined models. The results are then combined with radiomics features and clinical data for comprehensive analysis.
It enables non-invasive and accurate auxiliary identification of diabetic nephropathy, improves the accuracy and reliability of detection, and provides an image-assisted diagnostic tool for clinical use.
Smart Images
Figure CN122265233A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of medical image processing and computer-aided diagnosis, and in particular to a method and system for the auxiliary identification of diabetic nephropathy based on CT images. Background Technology
[0002] Diabetic nephropathy (DKD) is one of the most common microvascular complications of diabetes and a leading cause of end-stage renal disease. Early diagnosis and timely intervention are crucial for slowing disease progression and improving patient prognosis. Currently, clinical screening for DKD mainly relies on biomarkers such as urinary albumin and serum creatinine. However, these indicators suffer from insufficient sensitivity, poor stability, and difficulty in reflecting early microstructural changes in the kidneys. While routine abdominal CT scans can reveal morphological features of the kidneys, their ability to identify early functional and microstructural damage is limited. Summary of the Invention
[0003] This invention provides a method and system for the auxiliary identification of diabetic nephropathy based on CT images, which solves the technical problems in the prior art where the detection of diabetic nephropathy relies on biomarkers, resulting in insufficient sensitivity and the difficulty of identifying early microscopic lesions by conventional imaging examinations.
[0004] To achieve the above objectives, a first aspect of the present invention provides a method for assisted identification of diabetic nephropathy based on CT images, comprising: Acquire abdominal CT images of the target object; Multiple regions of interest were segmented from the abdominal CT image, including the region of interest for the kidney, the region of interest for perirenal fat, and the region of interest for body composition. Radiomics features were extracted from each region of interest. The radiomics features are screened to obtain the target radiomics features; The target radiomics features are input into a pre-constructed radiomics model, and the first identification result of the target object having diabetic nephropathy is output.
[0005] Furthermore, the region of interest includes a two-dimensional region of interest and a three-dimensional volume of interest; The body component region of interest includes at least one of skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue; The radiomics features include first-order features and second-order features. The first-order features include histogram features and morphological features. The second-order features include metrics based on gray-level co-occurrence matrix, gray-level dependency matrix, gray-level run matrix, gray-level size region matrix, and adjacent gray-level difference matrix.
[0006] Furthermore, the two-dimensional region of interest includes: skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue at the level of the third lumbar vertebra, and kidney, perirenal adipose tissue, and renal sinus adipose tissue at the level of the renal vein; The three-dimensional volume of interest includes: skeletal muscle, subcutaneous adipose tissue and visceral adipose tissue at the level of the first to fifth lumbar vertebrae, perirenal adipose tissue along the perirenal fascia, and adipose tissue of the kidney and renal sinus from the diaphragm to the first sacral vertebra.
[0007] Furthermore, the radiomics features are screened to obtain target radiomics features including: Calculate the correlation coefficient between each radiomics feature and remove redundant features with a correlation coefficient greater than or equal to a preset threshold; A feature selection algorithm is used to select a predetermined number of features with the highest scores from the remaining features; The predetermined number of features are further filtered using a regularization algorithm to obtain candidate radiomics features; After merging the candidate radiomics features from each region of interest, regularization algorithms and cross-validation are used again for screening, and the optimal regularization parameters are determined based on the model performance indicators. The final target radiomics features were determined from the features after further screening through univariate and multivariate logistic regression analysis.
[0008] Furthermore, the method also includes: Obtain the clinical data of the target subjects; The clinical data is filtered to obtain the target clinical features; The target clinical features are input into a pre-built clinical model, and a second identification result based on clinical data is output.
[0009] Furthermore, the method also includes: The target radiomics features are fused with the target clinical features and input into a pre-constructed joint model to output a third identification result.
[0010] A second aspect of the present invention provides a CT image-based auxiliary identification system for diabetic nephropathy, comprising: The image acquisition module is used to acquire abdominal CT images of the target object; The image segmentation module is used to segment multiple regions of interest from the abdominal CT image, including the region of interest for the kidney, the region of interest for perirenal fat, and the region of interest for body composition. The feature extraction module is used to extract radiomics features from each region of interest. The feature filtering module is used to filter the radiomics features to obtain target radiomics features; The identification module is used to input the target radiomics features into a pre-constructed radiomics model and output the first identification result that the target object has diabetic nephropathy.
[0011] A third aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps in the CT image-based auxiliary identification method for diabetic nephropathy as described in the first aspect of the present invention.
[0012] A fourth aspect of the present invention provides a computer-readable storage medium having a program stored thereon that, when executed by a processor, implements the steps of the CT image-based assisted identification method for diabetic nephropathy as described in the first aspect of the present invention.
[0013] A fifth aspect of the present invention provides a computer program product comprising software code, wherein the program in the software code performs the steps of the CT image-based auxiliary identification method for diabetic nephropathy as described in the first aspect of the present invention.
[0014] Compared with existing technologies, the present invention provides a CT image-based auxiliary identification method and system for diabetic nephropathy (DKD), which has the following beneficial effects: By segmenting multiple regions of interest, such as the kidneys, perirenal fat, and body components, from abdominal CT images of the target subject, and extracting radiomics features for screening and modeling, non-invasive and accurate auxiliary identification of DKD is achieved. This method can extract microscopic imaging biomarkers related to kidney function impairment from conventional CT images, and by integrating information from multiple sites, it improves the accuracy and reliability of DKD identification, providing a novel image-assisted diagnostic tool for clinical practice. Attached Figure Description
[0015] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.
[0016] Figure 1 A flowchart of the CT image-based auxiliary identification method for diabetic nephropathy provided in Embodiment 1 of the present invention; Figure 2 A flowchart illustrating the process of constructing the inclusion and exclusion rules for subjects and the division of the research cohort, as provided in Embodiment 1 of the present invention; Figure 3 This is a core workflow diagram of constructing a diabetic nephropathy diagnostic model based on CT images, provided in Embodiment 1 of the present invention. Figure 4 This is a schematic diagram illustrating the performance of the radiomics model provided in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram illustrating the performance of the clinical model provided in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram illustrating the performance of the joint model provided in Embodiment 1 of the present invention; Figure 7 This is an architecture diagram of the CT image-based auxiliary identification system for diabetic nephropathy provided in Embodiment 2 of the present invention. Detailed Implementation
[0017] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0018] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form as well. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0019] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.
[0020] All data acquisition in this embodiment is carried out in accordance with laws and regulations and with user consent, and the data is used legally.
[0021] Example 1 like Figure 1 This embodiment provides a method for assisted identification of diabetic nephropathy based on CT images, including: S1. Obtain abdominal CT images of the target object.
[0022] Specifically, a 64-slice spiral CT scanner was used to perform abdominal CT scans on the target subjects. Scanning parameters included: tube voltage 120 kV, tube current 150 mA, pitch 1 mm, slice spacing 5 mm, slice thickness 5 mm, and tube rotation speed 0.5 s. During the scan, the target subjects lay supine and held their breath. The acquired abdominal CT images were stored and exported in a digital medical imaging and communication format. The inclusion and exclusion rules for subjects and the process of constructing the study cohort are as follows: Figure 2 As shown.
[0023] S2. Multiple regions of interest (ROIs) are segmented from the abdominal CT image. These ROIs include the region of interest for the kidney, the region of interest for perirenal fat, and the region of interest for body composition. Each ROI includes a two-dimensional region of interest (ROI) and a three-dimensional volume of interest (VOI). The body composition region of interest includes at least one of skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue.
[0024] Specifically, two radiologists with experience in abdominal radiology diagnosis used 3D Slicer software to semi-automatically delineate CT images, while maintaining clinical confidentiality. Intra-group correlation coefficient analysis was used to assess inter-observer consistency, and only regions of interest with an ICC > 0.8 were retained for subsequent analysis.
[0025] The two-dimensional region of interest includes: skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue at the level of the third lumbar vertebra, and kidney, perirenal adipose tissue, and renal sinus adipose tissue at the level of the renal vein.
[0026] The three-dimensional volume of interest includes: skeletal muscle, subcutaneous adipose tissue and visceral adipose tissue at the level of the first to fifth lumbar vertebrae, perirenal adipose tissue along the perirenal fascia, and adipose tissue of the kidney and renal sinus from the diaphragm to the first sacral vertebra.
[0027] During segmentation, tissues are identified based on their anatomical location and Henle units: the HU range for skeletal muscle is -30 to 150, the HU range for subcutaneous adipose tissue is -190 to -30, the HU range for visceral adipose tissue is -150 to -50, the HU range for the kidney is -30 to 150, and the HU range for perirenal adipose tissue and renal sinus adipose tissue is -150 to -50.
[0028] S3. Extract radiomics features from each region of interest.
[0029] Features were extracted from six 2D regions of interest and six 3D volumes of interest in each examination using radiomics feature extraction software (e.g., the UltraScholar 2.0 platform based on PyRadiomics). The radiomics features included first-order and second-order features. The first-order features included histogram features and morphological features, while the second-order features included metrics based on the gray-level co-occurrence matrix (GLCM), gray-level dependency matrix (GLDM), gray-level run-length matrix (GLRLM), gray-level size region matrix (GLSZM), and adjacent gray-level difference matrix (NGTDM).
[0030] S4. The radiomics features are screened to obtain the target radiomics features. The workflow for radiomics feature extraction and screening in this embodiment is as follows: Figure 3As shown in the flowchart, the process involves: acquiring images using a 3D Slicer and performing semi-automatic segmentation; then extracting and pre-selecting radiomics features using the UltraScholar platform; filtering the obtained features and clinical variables to establish radiomics models, clinical models, and combined models; and evaluating the models through nomograms, ROC curves, calibration curves, and decision curves.
[0031] The screening process specifically includes the following sub-steps: S41. Calculate the Pearson correlation coefficient between each radiomics feature and remove redundant features with a correlation coefficient greater than or equal to 0.95.
[0032] S42. Use the SelectKBest feature selection algorithm to select a predetermined number of features with the highest scores from the remaining features.
[0033] S43. The Least Absolute Shrinkage and Selection Operator (LASSO) regularization algorithm is used to further filter the predetermined number of features to obtain candidate radiomics features. For each region of interest module, if the number of features after LASSO screening is less than 10, all identified features are retained. If no effective variables are obtained after LASSO, the 10 features with the highest scores are selected as candidate features for the module using the SelectKBest algorithm.
[0034] S44. After merging the candidate radiomics features from each region of interest, the LASSO algorithm is used again in combination with 10-fold cross-validation for screening. The optimal regularization parameter is determined based on the area under the curve, and the feature set after further screening is obtained.
[0035] In a specific embodiment of the present invention, after merging the features selected from each sub-model, the LASSO algorithm is used again to select 34 features from the original 117 features. The specific distribution of these features is as follows: 2 from visceral adipose tissue (VAT) ROI, 3 from renal sinus adipose tissue (RSAT) ROI, 5 from kidney ROI, 2 from perirenal adipose tissue (PRAT) ROI, 3 from skeletal muscle (SM) ROI, 2 from visceral adipose tissue (VAT) VOI, 4 from subcutaneous adipose tissue (SAT) VOI, 1 from renal sinus adipose tissue (RSAT) VOI, 4 from kidney VOI, 3 from perirenal adipose tissue (PRAT) VOI, and 5 from skeletal muscle (SM) VOI. The coefficients of these 34 features are detailed in Table 1.
[0036] Table 1. LASSO Screening Characteristics and Coefficients
[0037] Note: vatarea, region of interest for visceral fat; sinusarea, region of interest for sinus fat; pratarea, region of interest for perirenal fat; rarea, region of interest for the kidney; smarea, region of interest for skeletal muscle; smv, volume of interest for skeletal muscle; vatv, volume of interest for visceral fat; satv, volume of interest for subcutaneous fat; sinusv, volume of interest for sinus fat; rv, volume of interest for the kidney; pratv, volume of interest for perirenal fat.
[0038] S45. Through univariate and multivariate logistic regression analysis, the final target radiomics features are determined from the features after further screening.
[0039] In a specific embodiment of the present invention, the 34 features selected by S44 were included in univariate and multivariate logistic regression analysis. Finally, the 5 features with p < 0.05 in the multivariate analysis were selected as the final target radiomics features, which specifically include: auto__original_shape_Elongation and auto__wavelet-HHL_glszm_SmallAreaLowGrayLevelEmphasis of the two-dimensional region of interest of renal sinus adipose tissue (RSAT), auto__lbp-3D-k_glszm_SmallAreaEmphasis of the two-dimensional region of interest of the kidney, and auto__wavelet-HHH_glszm_SmallAreaHighGrayLevelEmphasis and auto__wavelet-HLL_firstorder_Maximum of the three-dimensional volume of interest of perirenal adipose tissue (PRAT).
[0040] S5. Input the target radiomics features into a pre-constructed radiomics model and output the first identification result that the target object has diabetic nephropathy. The radiomics model is trained using classification algorithms such as logistic regression based on the target radiomics features of historical samples and their corresponding diabetic nephropathy labels.
[0041] In one specific embodiment of the present invention, the area under the curve (AUC) values of the radiomics model in the training queue, internal validation queue, and external validation queue are 0.869, 0.831, and 0.833, respectively. The specific performance characteristics of the radiomics model are as follows: Figure 4 As shown, Figure 4In the diagram, 'a' is the Lasso path plot, showing the trend of variable coefficient trajectories as the regularization penalty intensity increases; 'b' is the 10-fold cross-validation results of the Lasso model; and 'c' is the nomogram of the final radiomics features, where C represents SinusArea-auto__original_shape_Elongation, D represents SinusArea-auto__wavelet-HHL_glszm_SmallAreaLowGrayLevelEmphasis, G represents RenalArea-auto__lbp-3D-k_glszm_SmallAreaEmphasis, and AG represents PRATVo. `lume-auto__wavelet-HHH_glszm_SmallAreaHighGrayLevelEmphasis`, where AH stands for `PRATVolume-auto__wavelet-HLL_firstorder_Maximum`, and `PRAT` represents perirenal adipose tissue; `d` represents the coefficient distribution of the lasso model; `e` represents the ROC curve of the radiomics model, where the red line represents the training cohort curve, the blue line represents the internal validation cohort curve, and the green line represents the external validation cohort curve; `f` represents the calibration curve of the nomogram for the external validation cohort; `g` analyzes the clinical net benefit of the radiomics model under different decision thresholds to guide intervention measures in the external validation cohort.
[0042] Furthermore, the method also includes a step of clinical data fusion: S6. Obtain clinical data of the target population, including age, sex, blood pressure, duration of diabetes, waist circumference, smoking status, diabetes treatment methods, and laboratory test indicators (such as LDL cholesterol, HDL cholesterol, body mass index, uric acid, serum creatinine, cystatin C, urinary albumin, urinary albumin-to-creatinine ratio, neutrophil-to-lymphocyte ratio, hemoglobin, and albumin). Descriptive statistics for this population are shown in Table 2.
[0043] Table 2 Characteristics of the study population
[0044] Note: BMI, Body Mass Index; NLR, Neutrophil-to-Lymphocyte Ratio; PLR, Platelet-to-Lymphocyte Ratio; SII, Systemic Immune Inflammatory Index; a: t-test; b: Chi-square test; c: Mann-Whitney test.
[0045] The clinical data were screened, specifically by using univariate logistic regression to identify variables with p < 0.05, which were then incorporated into multivariate logistic regression to determine the target clinical features. For example, the target clinical features obtained after screening included albumin, hemoglobin, uric acid, and systolic blood pressure. The results of the logistic regression analysis of the clinical variables are detailed in Table 3.
[0046] Table 3. Results of logistic regression analysis of clinical variables
[0047] Note: BMI, Body Mass Index; NLR, Neutrophil-to-Lymphocyte Ratio; PLR, Platelet-to-Lymphocyte Ratio; SII, Systemic Immune Inflammation Index.
[0048] S7. Input the target clinical features into a pre-built clinical model and output a second identification result based on clinical data. The clinical model is trained using logistic regression based on the target clinical features of historical samples and their corresponding diabetic nephropathy labels. The specific performance of the clinical model is as follows: Figure 5 As shown, Figure 5 In the diagram, a is a nomogram; b is the ROC curve; c is the calibration curve of the external validation cohort; and d is the decision curve analysis of the external validation cohort.
[0049] S8. The target radiomics features and the target clinical features are fused and input into a pre-constructed joint model, and a third identification result is output. The joint model is trained using algorithms such as logistic regression based on the fusion features of the target radiomics features and target clinical features of historical samples and their corresponding diabetic nephropathy labels.
[0050] In one specific embodiment of the present invention, the performance of the combined model outperforms individual radiomics models and clinical models, achieving area under the curve values of 0.913, 0.919, and 0.867 in the training cohort, internal validation cohort, and external validation cohort, respectively. The specific performance characteristics of the combined model are as follows: Figure 6 As shown, Figure 6 In the diagram, a is a nomogram; b is the ROC curve; c is the calibration curve of the external validation cohort; and d is the decision curve analysis of the external validation cohort.
[0051] As shown in Table 4, the diagnostic efficacy of the three models was further compared by calculating sensitivity, specificity, positive predictive value, negative predictive value, and Youden index.
[0052] Table 4 Diagnostic efficacy of the three models
[0053] Note: AUC, area under the ROC curve; 95% CI, 95% confidence interval.
[0054] It should be noted that the above steps S6-S8 can be executed in parallel or sequentially with S1-S5, and multiple identification results can be obtained for clinical reference.
[0055] In the specific implementation of this invention, all data acquisition is conducted in compliance with legal and ethical requirements (Research Ethics Committee of Qilu Hospital, Shandong University, Approval No.: KYLL-202411-022) and with user consent. Informed consent is waived for retrospective studies and the use of anonymized data.
[0056] Example 2 like Figure 7 As shown, this embodiment provides a CT image-based auxiliary identification system for diabetic nephropathy, including: The image acquisition module is used to acquire abdominal CT images of the target object. Specifically, the image acquisition module communicates with the CT scanner to acquire abdominal CT images acquired according to preset scanning parameters, including: tube voltage 120kV, tube current 150 mA, pitch 1 mm, slice spacing 5 mm, slice thickness 5 mm, and tube rotation speed 0.5 s.
[0057] The image segmentation module is used to segment multiple regions of interest (ROIs) from the abdominal CT image. These ROIs include a renal ROI, a perirenal fat ROI, and a body composition ROI. Each ROI comprises a two-dimensional region of interest (2D) and a three-dimensional volume of interest (3D). The body composition ROI includes at least one of skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue. The image segmentation module uses a semi-automatic delineation method to identify tissues based on anatomical location and Henle units. Specifically, the segmentation range includes: skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue at the level of the third lumbar vertebra, and the kidney, perirenal adipose tissue, and renal sinus adipose tissue at the level of the renal vein as 2D ROIs; skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue at the levels of the first to fifth lumbar vertebrae, perirenal adipose tissue along the perirenal fascia, and the kidney and renal sinus adipose tissue from the diaphragm to the first sacral vertebra as 3D volumes of interest. The image segmentation module also assesses inter-observer consistency and ensures that the intra-group correlation coefficient of the ROIs used for subsequent analysis is greater than 0.8.
[0058] The feature extraction module is used to extract radiomics features from each region of interest. The radiomics features include first-order and second-order features. The first-order features include histogram features and morphological features, while the second-order features include metrics based on gray-level co-occurrence matrix, gray-level dependency matrix, gray-level run-length matrix, gray-level size region matrix, and adjacent gray-level difference matrix. Specifically, the feature extraction module extracts features from six two-dimensional regions of interest and six three-dimensional volumes of interest.
[0059] The feature selection module is used to filter the radiomics features to obtain target radiomics features. Specifically, the feature selection module is used to: calculate the Pearson correlation coefficient between each radiomics feature and remove redundant features with a correlation coefficient greater than or equal to 0.95; use the SelectKBest feature selection algorithm to select a predetermined number of features with the highest scores from the remaining features; use the LASSO regularization algorithm to further filter the predetermined number of features to obtain candidate radiomics features; merge the candidate radiomics features from each region of interest, and then use the LASSO algorithm again combined with 10-fold cross-validation for further filtering, determining the optimal regularization parameter based on the area under the curve; and use univariate and multivariate logistic regression analysis to determine the final target radiomics features from the features after further filtering.
[0060] In one specific embodiment, the feature screening module ultimately determined five target radiomics features, specifically including: the auto__original_shape_Elongation feature and the auto__wavelet-HHL_glszm_SmallAreaLowGrayLevelEmphasis feature of the two-dimensional region of interest in the renal sinus adipose tissue; the auto__lbp-3D-k_glszm_SmallAreaEmphasis feature of the two-dimensional region of interest in the kidney; and the auto__wavelet-HHH_glszm_SmallAreaHighGrayLevelEmphasis feature and the auto__wavelet-HLL_firstorder_Maximum feature of the three-dimensional volume of interest in the perirenal adipose tissue.
[0061] The identification module is used to input the target radiomics features into a pre-constructed radiomics model and output the first identification result that the target object has diabetic nephropathy.
[0062] Furthermore, the system also includes: The clinical data acquisition module is used to acquire clinical data of the target subjects, including age, gender, blood pressure, duration of diabetes, waist circumference, smoking status, diabetes treatment methods, and laboratory test indicators.
[0063] The clinical feature screening module is used to filter the clinical data to obtain target clinical features. Specifically, through univariate and multivariate logistic regression analysis, variables with p < 0.05 are identified as target clinical features, including albumin, hemoglobin, uric acid, and systolic blood pressure.
[0064] The clinical model recognition module is used to input the target clinical features into a pre-built clinical model and output a second recognition result based on clinical data. The joint model recognition module is used to fuse the target radiomics features with the target clinical features, input them into a pre-constructed joint model, and output a third recognition result.
[0065] The system also includes a model evaluation module, which is used to evaluate the performance of each model through receiver operating characteristic (ROC) curves and area under the curve, calibration curves and decision curves, and to calculate sensitivity, specificity, positive predictive value, negative predictive value and Youden index to evaluate its clinical utility.
[0066] Example 3 Embodiment 3 of the present invention provides an electronic device.
[0067] An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor. The processor includes, but is not limited to, at least one of a central processing unit (CPU), a graphics processing unit (GPU), a neural network processor (NPU), a tensor processor (TPU), or an artificial intelligence acceleration chip. When executing the program, the program implements the steps in the CT image-based assisted identification method for diabetic nephropathy as described in Embodiment 1 of the present invention.
[0068] The detailed steps are the same as those of the CT image-based assisted identification method for diabetic nephropathy provided in Example 1, and will not be repeated here.
[0069] Example 4 Embodiment 4 of the present invention provides a computer-readable storage medium.
[0070] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the CT image-based auxiliary identification method for diabetic nephropathy as described in Embodiment 1 of the present invention.
[0071] The detailed steps are the same as those of the CT image-based assisted identification method for diabetic nephropathy provided in Example 1, and will not be repeated here.
[0072] Example 5 Embodiment 5 of the present invention provides a computer program product.
[0073] A computer program product includes software code, wherein the program in the software code performs the steps of the CT image-based auxiliary identification method for diabetic nephropathy as described in Embodiment 1 of the present invention.
[0074] The detailed steps are the same as those of the CT image-based assisted identification method for diabetic nephropathy provided in Example 1, and will not be repeated here.
[0075] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages. For example, in one implementation, the methods and systems can be developed based on deep learning frameworks (such as TensorFlow, PyTorch, etc.) and using the Python language. Those skilled in the art will understand that other suitable programming languages or tools can also be used for implementation without departing from the core ideas of the present invention.
[0076] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0077] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0078] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0079] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.
Claims
1. A method for auxiliary identification of diabetic nephropathy based on CT images, characterized in that, include: Acquire abdominal CT images of the target object; Multiple regions of interest were segmented from the abdominal CT image, including the region of interest for the kidney, the region of interest for perirenal fat, and the region of interest for body composition. Radiomics features were extracted from each region of interest. The radiomics features are screened to obtain the target radiomics features; The target radiomics features are input into a pre-constructed radiomics model, and the first identification result of the target object having diabetic nephropathy is output.
2. The method according to claim 1, characterized in that, The region of interest includes a two-dimensional region of interest and a three-dimensional volume of interest; The body component region of interest includes at least one of skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue; The radiomics features include first-order features and second-order features. The first-order features include histogram features and morphological features. The second-order features include metrics based on gray-level co-occurrence matrix, gray-level dependency matrix, gray-level run matrix, gray-level size region matrix, and adjacent gray-level difference matrix.
3. The method according to claim 2, characterized in that, The two-dimensional region of interest includes: skeletal muscle, subcutaneous adipose tissue and visceral adipose tissue at the level of the third lumbar vertebra, and kidney, perirenal adipose tissue and renal sinus adipose tissue at the level of the renal vein; The three-dimensional volume of interest includes: skeletal muscle, subcutaneous adipose tissue and visceral adipose tissue at the level of the first to fifth lumbar vertebrae, perirenal adipose tissue along the perirenal fascia, and adipose tissue of the kidney and renal sinus from the diaphragm to the first sacral vertebra.
4. The method according to claim 1, characterized in that, The radiomics features were screened to obtain the target radiomics features, including: Calculate the correlation coefficient between each radiomics feature and remove redundant features with a correlation coefficient greater than or equal to a preset threshold; A feature selection algorithm is used to select a predetermined number of features with the highest scores from the remaining features; The predetermined number of features are further filtered using a regularization algorithm to obtain candidate radiomics features; After merging the candidate radiomics features from each region of interest, regularization algorithms and cross-validation are used again for screening, and the optimal regularization parameters are determined based on the model performance indicators. The final target radiomics features were determined from the features after further screening through univariate and multivariate logistic regression analysis.
5. The method according to claim 1, characterized in that, The method further includes: Obtain the clinical data of the target subjects; The clinical data is filtered to obtain the target clinical features; The target clinical features are input into a pre-built clinical model, and a second identification result based on clinical data is output.
6. The method according to claim 5, characterized in that, The method further includes: The target radiomics features are fused with the target clinical features and input into a pre-constructed joint model to output a third identification result.
7. A CT image-based auxiliary identification system for diabetic nephropathy, characterized in that, include: The image acquisition module is used to acquire abdominal CT images of the target object; The image segmentation module is used to segment multiple regions of interest from the abdominal CT image, including the region of interest for the kidney, the region of interest for perirenal fat, and the region of interest for body composition. The feature extraction module is used to extract radiomics features from each region of interest. The feature filtering module is used to filter the radiomics features to obtain target radiomics features; The identification module is used to input the target radiomics features into a pre-constructed radiomics model and output the first identification result that the target object has diabetic nephropathy.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps of the CT image-based auxiliary identification method for diabetic nephropathy according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the CT image-based auxiliary identification method for diabetic nephropathy according to any one of claims 1 to 6.
10. A computer program product, comprising software code, characterized in that, The program in the software code executes the steps of the CT image-based auxiliary identification method for diabetic nephropathy according to any one of claims 1 to 6.