AI-assisted partial renal resection perioperative period complication prediction system based on CT radiomics
By combining CT radiomics with clinical features, a deep learning system was developed to address the issues of incomplete information, low automation, and limited predictive performance during partial nephrectomy. This system enables efficient and interpretable prediction of perioperative complications, improving prediction accuracy and the reliability of clinical applications.
Patent Information
- Application Number
- CN202511257030.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-04
- Publication Date
- 2025-12-30
AI Technical Summary
Existing technologies for the perioperative period of partial nephrectomy, and for the diagnosis, classification, and prognostic prediction of kidney tumors, suffer from incomplete information, lack of automation, limited predictive performance, and insufficient clinical interpretability. In particular, research on the prediction of perioperative complications is insufficient, and high-dimensional imaging features and clinical data have not been effectively integrated.
A deep learning prediction system based on CT radiomics and clinical features is adopted, including data acquisition, preprocessing, feature extraction, fully connected deep neural network model and interpretability analysis, to achieve automated segmentation, feature extraction and prediction, and generate interpretable analysis results.
It significantly improves the performance of perioperative complication prediction, automates the entire process from image import to report generation, improves the accuracy and clinical reliability of prediction, and has good generalization and clinical deployment capabilities.
Smart Images
Figure CN121237399A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to a kind of based on CT imaging genomics and clinical feature deep learning prediction system of perioperative complication of partial nephrectomy, specifically related to a kind of integrated automatic segmentation, feature extraction, deep learning modeling and explainable analysis whole process AI assisted clinical prediction technology. BACKGROUND
[0002] At present, mainly rely on RENAL, PADUA etc. Anatomical scoring system is diagnosed, tumor typing, prognosis prediction, these scoring systems mostly rely on a small amount of artificial measurement index such as tumor size, position, and relationship with hilum. Although there have been researches in urology surgery to apply deep learning to diagnosis, tumor typing, prognosis prediction, but there is still insufficient research on perioperative complication prediction, and most of them do not integrate high-dimensional image features and clinical data, and do not form a whole process method that can be clinically deployed.
[0003] The prediction system and research for perioperative complication prediction currently have the following shortcomings:
[0004] (1) incomplete information: traditional anatomical model only contains size and position parameters, lacks quantification of high-order image information such as morphology and texture;
[0005] (2) lack of automation: existing processes rely on manual ROI drawing, and the feature extraction efficiency is low and subjective;
[0006] (3) limited prediction performance: multi-modal fusion deep learning modeling is not used, making it difficult to capture complex nonlinear relationships;
[0007] (4) lack of clinical interpretability: lack of explanation and visual output of model decision basis, reducing clinical acceptance. SUMMARY
[0008] The purpose of the present application is to provide a deep learning prediction system combining CT imaging genomics features and clinical features, which can provide perioperative complication occurrence probability, risk classification and explainable analysis, and has high generalizability and clinical deployment capability.
[0009] To achieve the above purpose, the first aspect of the technical scheme of the present application discloses an AI-assisted perioperative complication prediction system for partial nephrectomy based on CT imaging genomics, characterized in that it comprises:
[0010] A data acquisition module configured to acquire preoperative enhanced CT images and clinical data of a patient;
[0011] A data preprocessing module configured to automatically segment the kidney and tumor ROI after preprocessing the CT images obtained by the data acquisition module;
[0012] The feature extraction module is configured to extract radiomics features and depth image features from the CT images output by the data preprocessing module, and at the same time, extract clinical features based on the clinical data obtained by the data acquisition module.
[0013] The perioperative complication prediction model is configured to use the features extracted by the feature extraction module as the output. Based on a fully connected deep neural network, it predicts the perioperative complications and risk levels of patients before partial nephrectomy. The fully connected deep neural network is used in conjunction with Dropout and BatchNorm for regularization.
[0014] The prediction results output module can be configured to output the probability of complication occurrence, risk level and confidence interval, and automatically generate interpretability analysis results and clinical reports.
[0015] Preferably, the prediction result output module generates a key feature contribution heatmap as the interpretability analysis result through methods such as Feature importance based on permutation, thereby improving clinical credibility.
[0016] Preferably, the data preprocessing module includes:
[0017] The image resampling unit is configured to resample CT images to a uniform voxel size.
[0018] The standardization processing unit is configured to standardize CT images to reduce differences in scanning conditions.
[0019] The automatic segmentation unit for kidney and tumor ROI is configured to automatically segment and obtain kidney and tumor ROI from CT images processed by the image resampling unit and the normalization processing unit;
[0020] The ROI volume generation unit is configured to generate ROI volumes based on the kidney and tumor ROIs.
[0021] Preferably, when constructing the perioperative complication prediction model, the radiomics features extracted by the feature extraction module are the original features, and the features extracted by the feature extraction module are filtered using the feature filtering module. The retained features are used as the features extracted by the feature extraction module when the system is finally deployed.
[0022] Preferably, when constructing the perioperative complication prediction model, the perioperative complication prediction model is trained using a model training module.
[0023] Preferably, when the model training module trains the perioperative complication prediction model, it uses weighted cross-entropy as the loss function to handle class imbalance.
[0024] Preferably, a 95% confidence interval estimate is obtained by using bootstrap 1000 resampling operations.
[0025] Preferably, the generated clinical report includes an image heatmap, key clinical features, and risk warnings.
[0026] The second aspect of the technical solution of the present invention discloses an electronic device, comprising:
[0027] One or more processors;
[0028] A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute software to implement the perioperative complication prediction system described above.
[0029] The third aspect of the present invention discloses a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to implement the perioperative complication prediction system described above.
[0030] The perioperative complication prediction system for partial nephrectomy provided by this invention is used to predict the perioperative complications and risk classification of patients before partial nephrectomy. It can be integrated into hospital PACS or independent AI platforms to achieve: individualized risk assessment; assisting surgeons in developing surgical strategies and perioperative management; reducing the incidence of complications and optimizing patient safety and prognosis.
[0031] Compared with existing technical solutions, the present invention has the following beneficial effects:
[0032] (1) Compared with the existing anatomical scoring systems RENAL and PADUA, the prediction system disclosed in this invention can significantly improve the prediction performance of complications, and the prediction performance is better (AUC 0.87 vs 0.68 / 0.69, p<0.001);
[0033] (2) It can automate the entire process from DICOM image import, segmentation, feature extraction to prediction and report generation, with short processing time;
[0034] (3) The system disclosed in this invention can output probability, confidence interval and interpretability feature contribution, thereby enhancing clinical credibility and facilitating clinical application;
[0035] (4) The system disclosed in this invention can adapt to multi-center, cross-device data and has good generalization, scalability and promotion application value. Attached Figure Description
[0036] Figure 1This is a system block diagram of the system disclosed in this invention;
[0037] Figure 2 This is a schematic diagram of segmentation and image extraction;
[0038] Figure 3 This is a schematic diagram illustrating the importance of features. Detailed Implementation
[0039] Various aspects and features of the present invention are described herein with reference to the accompanying drawings.
[0040] It should be understood that various modifications can be made to the embodiments described herein. Therefore, the above description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope and spirit of this disclosure will be apparent to those skilled in the art.
[0041] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the invention and, together with the general description of the invention given above and the detailed description of the embodiments given below, serve to explain the principles of the invention.
[0042] These and other features of the invention will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.
[0043] It should also be understood that although the invention has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of the invention, which have the features described in the claims and are therefore all within the scope of protection defined herein.
[0044] The above and other aspects, features and advantages of the invention will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.
[0045] Specific embodiments of the invention are described below with reference to the accompanying drawings; however, it should be understood that the claimed embodiments are merely examples of the invention, which may be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that would obscure this disclosure. Therefore, the specific structural and functional details claimed in this invention are not intended to be limiting, but are merely intended as the basis and representative basis for the claims to teach those skilled in the art to use this disclosure in a variety of substantially any suitable detailed structures.
[0046] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in still another embodiment,” all of which may refer to one or more of the same or different embodiments according to this disclosure.
[0047] Combination Figure 1The first aspect of this invention discloses an AI-assisted perioperative complication prediction system for partial nephrectomy based on CT radiomics, comprising:
[0048] The data acquisition module is configured to acquire preoperative enhanced CT images and clinical data of the patient. In a preferred embodiment of the present invention, the CT images acquired by the data acquisition module are in DICOM format.
[0049] The data preprocessing module is configured to preprocess the CT images obtained by the data acquisition module and then automatically segment them to obtain the kidney and tumor ROI.
[0050] In a preferred embodiment of the present invention, the data preprocessing module further includes:
[0051] The image resampling unit is configured to resample CT images to a uniform voxel size, for example: 1×1×1mm. 3 ;
[0052] The standardization processing unit is configured to standardize CT images to reduce differences in scanning conditions; further, in a preferred embodiment of the present invention, the standardization processing unit adopts z-score intensity standardization.
[0053] The automatic kidney and tumor ROI segmentation unit is configured to automatically segment and obtain the kidney and tumor ROI from CT images processed by the image resampling unit and the normalization unit, such as... Figure 2 As shown; further, in a preferred embodiment of the present invention, the automatic segmentation unit for kidney and tumor ROI is implemented based on nnU-Net, with a Dice coefficient of 0.83;
[0054] The ROI volume generation unit is configured to generate ROI volumes based on the kidney and tumor ROIs for feature extraction and model training.
[0055] The feature extraction module is configured to extract radiomics and depth image features from the CT images output by the data preprocessing module, and simultaneously extract clinical features based on the clinical data obtained by the data acquisition module.
[0056] The perioperative complication prediction model is configured to use the features extracted by the feature extraction module as the output. Based on a fully connected deep neural network, it predicts the perioperative complications and risk grading of patients before partial nephrectomy. The fully connected deep neural network is used in conjunction with Dropout and BatchNorm for regularization.
[0057] In this embodiment of the invention, multimodal fusion data is used as input to the perioperative complication prediction model, which can comprehensively capture imaging and clinical feature information. At the same time, the perioperative complication prediction model adopts deep learning technology, which can identify nonlinear and high-dimensional feature patterns.
[0058] In a preferred embodiment of this invention, when constructing the perioperative complication prediction model, the radiomics features extracted by the feature extraction module include 107 original features covering morphology, first-order statistics, GLCM, GLRLM, and other texture features, and multiple clinical features are also extracted. The features extracted by the feature extraction module are then filtered using a feature filtering module, and the retained features are used as the features extracted by the feature extraction module during the final system deployment.
[0059] In one preferred embodiment of the present invention, the feature selection module uses LASSO regression for feature selection.
[0060] In another preferred embodiment of the present invention, the feature screening module ultimately retains 23 imaging features and 2 clinical features as the features extracted by the feature extraction module when the final system is deployed.
[0061] Meanwhile, when constructing the perioperative complication prediction model, the model training module is used to train the perioperative complication prediction model.
[0062] In one preferred embodiment of the present invention, the model training module constructs a training set using 1214 patients (7:3 ratio of training set) and constructs a validation set using an external validation queue of 260 patients.
[0063] In another preferred embodiment of the present invention, when the model training module trains the perioperative complication prediction model, it uses weighted cross-entropy as the loss function to handle class imbalance.
[0064] In another preferred embodiment of the present invention, when the model training module validates the perioperative complication prediction model, AUC, accuracy, sensitivity, specificity, and F1 score are used as statistical indicators.
[0065] Based on a sample of 1214 patients and 260 external validations, the constructed perioperative complication prediction model had an AUC of 0.87 for predicting complications and an AUC of 0.83 for external validation.
[0066] The prediction results output module can be configured to output the probability of complication occurrence, risk level and confidence interval, and automatically generate interpretability analysis results and clinical reports.
[0067] In a preferred embodiment of the invention, a 95% confidence interval estimate is obtained by using bootstrap with 1000 resampling iterations. This embodiment of the invention uses bootstrap to enhance statistical robustness.
[0068] In another preferred embodiment of the present invention, the generated clinical report includes an image heatmap, contributions of key clinical features, and risk warnings. Furthermore, in another preferred embodiment of the present invention, the prediction result output module uses a feature importance based on permutation method to visualize feature contributions, thereby automatically generating interpretability analysis results and improving clinical credibility. Figure 3 As shown.
[0069] In this embodiment of the invention, the interpretability of the output of the prediction result output module ensures the clinical acceptability of the model.
[0070] The aforementioned perioperative complication prediction system can be deployed as a Docker container and accessed via a PACS interface. During operation, the patient's CT images are input into the system before surgery. The system automatically performs segmentation, feature extraction, and deep learning prediction (prediction time per case <3 seconds), outputting an individualized risk PDF report for the surgical team's decision-making.
[0071] A second aspect of the present invention is to provide an electronic device, including at least a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-described perioperative complication prediction system when executing the computer program in the memory.
[0072] In some embodiments, the processor executing a computer program may be a processing device that includes one or more general-purpose processing devices, such as a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), etc. More specifically, the processor may be a Complex Instruction Set Computing (CISC) microprocessor, a Reduced Instruction Set Computing (RISC) microprocessor, a Very Long Instruction Word (VLIW) microprocessor, a processor that runs other instruction sets, or a processor that runs a combination of instruction sets. The processor may also be one or more special-purpose processing devices, such as an Application-Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), a System-on-a-Chip (SoC), etc.
[0073] The memory may be a read-only memory (ROM), random access memory (RAM), phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), electrically erasable programmable read-only memory (EEPROM), other types of random access memory (RAM), flash drives or other forms of flash memory, cache, registers, static memory, optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD) or other optical storage, magnetic tape cassette or other magnetic storage device, or any other possible non-transitory medium used to store information or instructions that can be accessed by computer equipment.
[0074] A third aspect of the present invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described perioperative complication prediction system.
[0075] The computer-readable storage medium of this invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. In this invention, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device; for example, it can be the aforementioned memory.
[0076] The computer programs of embodiments of the present invention can be organized into one or more computer-executable components or modules. Various aspects of this disclosure can be implemented with any number and combination of such components or modules. For example, aspects of this disclosure are not limited to the specific computer-executable instructions or specific components or modules shown in the drawings and described herein. Other embodiments may include different computer-executable instructions or components having more or fewer functions than those shown and described herein.
[0077] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.
Claims
1. An AI-assisted perioperative complication prediction system for partial nephrectomy based on CT imaging omics, characterized in that, The method comprises the following steps: a data acquisition module configured to acquire preoperative enhanced CT images of a patient and clinical data; a data preprocessing module configured to automatically segment the kidney and tumor ROI after preprocessing the CT images obtained by the data acquisition module; a feature extraction module configured to extract the radiomics features and deep image features of the CT images output by the data preprocessing module, and extract clinical features based on the clinical data obtained by the data acquisition module; a perioperative complication prediction model configured to take the features extracted by the feature extraction module as output, and predict the perioperative complications and risk classification of the patient before partial nephrectomy based on a fully connected deep neural network, wherein the fully connected deep neural network is regularized by Dropout and BatchNorm; a prediction result output module configured to output the complication occurrence probability, risk classification and confidence interval, and automatically generate an explainability analysis result and a clinical report. 2.The CT image-based AI-assisted partial nephrectomy perioperative complication prediction system of claim 1, wherein, The prediction result output module generates a key feature contribution heat map as the explainability analysis result. 3.The CT image-based AI-assisted partial nephrectomy perioperative complication prediction system of claim 1, wherein, The data preprocessing module comprises: an image resampling unit configured to resample the CT images to a uniform voxel size; a standardization processing unit configured to standardize the CT images to reduce differences in scanning conditions; a kidney and tumor ROI automatic segmentation unit configured to automatically segment the kidney and tumor ROI from the CT images processed by the image resampling unit and the standardization processing unit; an ROI volume generation unit configured to generate an ROI volume based on the kidney and tumor ROI. 4.The CT image-based AI-assisted partial nephrectomy perioperative complication prediction system of claim 1, wherein, In constructing the perioperative complication prediction model, the radiomics features extracted by the feature extraction module are the original features, and the feature selection module is used to select the features extracted by the feature extraction module, and the remaining features are used as the final features extracted by the feature extraction module when the system is deployed. 5.The CT image-based AI-assisted partial nephrectomy perioperative complication prediction system of claim 1, wherein, In constructing the perioperative complication prediction model, the model training module is used to train the perioperative complication prediction model. 6.The CT image-based AI-assisted partial nephrectomy perioperative complication prediction system of claim 4, wherein, When the model training module trains the perioperative complication prediction model, weighted cross-entropy is used as the loss function to handle class imbalance. 7.The CT image-based AI-assisted partial nephrectomy perioperative complication prediction system of claim 1, wherein, Bootstrap 1000 times resampling is used to obtain 95% confidence interval estimation. 8.The CT image-based AI-assisted partial nephrectomy perioperative complication prediction system of claim 1, wherein, The generated clinical report contains image heat map, main clinical feature contribution and risk prompt.
9. An electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute a software implementation of the perioperative complication prediction system according to claim 1.
10. A computer-readable storage medium having executable instructions stored thereon, which instructions, when executed by a processor, cause the processor to execute an implementation of the perioperative complication prediction system according to claim 1.