Bladder cancer level evaluation method and device based on radiomics and storage medium

Abstract

The invention discloses a bladder cancer level evaluation method and device based on radiomics and a storage medium. The bladder cancer level evaluation method and device are used for improving bladder cancer level evaluation accuracy and reliability. Obtaining a multi-parameter MRI image of the bladder cancer, wherein the multi-parameter MRI image comprises a T2 weighted image, a diffusion weighted image and a dynamic contrast enhanced image; performing tumor focus part sketching on the multi-parameter MRI image; respectively carrying out feature extraction on the delineated T2 weighted image, diffusion weighted image and dynamic contrast enhanced image by using a radiomics package to generate a bladder cancer feature set; performing feature difference screening on the bladder cancer feature set according to the clinical diagnosis tag; carrying out dimension reduction processing on the screened bladder cancer feature set; constructing a single-parameter MRI classification model and a multi-parameter joint classification model according to the bladder cancer feature set after dimension reduction; and using the single-parameter MRI classification model and the multi-parameter joint classification model to jointly evaluate the bladder cancer level of the to-be-detected data.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of lesion assessment, and in particular to a radiomics-based bladder cancer grade assessment method, device, and storage medium. Background Art

[0002] Bladder cancer is the most common type of urinary tract malignancy and is primarily categorized as high-grade or low-grade. High-grade bladder cancer, due to its aggressive nature and rapid growth, typically carries a higher recurrence rate and mortality rate, making its treatment and management more complex and urgent. Accurate grading of bladder cancer is crucial not only for understanding the biological behavior of the tumor but also for developing effective treatment plans and predicting the patient's long-term prognosis. Accurate grading can help physicians assess the tumor's aggressiveness, sensitivity to treatment, and potential risk of metastasis.

[0003] In terms of technology, radiomics is an advanced analytical technique that can extract a large number of features from multimodal medical images such as CT and MRI. This technology provides rich information for the diagnosis, staging, grading, and prognosis of diseases by identifying and quantifying various patterns and features in images. By applying machine learning algorithms, radiomics can link these imaging features to specific manifestations of the disease, thereby assisting clinical decision-making without invasive examinations. Combining high-throughput data analysis and machine learning technologies, radiomics has opened up new possibilities for the clinical application of medical imaging, showing great potential in improving the accuracy of disease diagnosis and promoting personalized medicine.

[0004] Multiparametric MRI, which can capture richer information through different imaging parameters, has become a valuable diagnostic tool for studying cancer lesions. Multiparametric MRI encompasses a variety of sequences, such as T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), apparent diffusion coefficient (ADC), and dynamic contrast enhancement (DCE). Each sequence captures specific features of different cancers, and combining multiple complementary modalities provides comprehensive characterization. Exploring potential correlations among multiparametric MRI sequences can be a means of improving cancer diagnosis.

[0005] In the existing technology, after obtaining a multi-parameter MRI of bladder cancer, it is necessary to manually outline the cancerous area, and then the features in the multi-parameter MRI need to be screened and reduced in dimension, and then a hierarchical classification model is constructed. In the traditional method, the multi-parameter MRI used usually only has one or two items, and bladder cancer is a disease that is extremely difficult to classify, so it is necessary to select a suitable multi-parameter MRI according to its characteristics to construct a classification model. In addition, in the screening and dimensionality reduction process, the traditional method uses the RFE (Recursive Feature Elimination) algorithm for screening and the LASSO (Least Absolute Shrinkage and Selection Operator) algorithm for dimensionality reduction. These two methods are usually based on the patient's routine condition information for screening and dimensionality reduction, and are not based on the patient's bladder cancer's deep clinical diagnostic labels for screening and dimensionality reduction. This reduces the accuracy and reliability of the assessment of bladder cancer grading in the existing technology. Summary of the Invention

[0006] The present application discloses a radiomics-based bladder cancer grade assessment method, device, and storage medium, which are used to improve the accuracy and reliability of bladder cancer grade assessment.

[0007] The first aspect of the present application discloses a radiomics-based method for assessing the grade of bladder cancer, comprising: Acquire multiparametric MRI images of bladder cancer, including T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging; Delineate tumor lesions on multi-parametric MRI images; A radiomics package was used to extract features from the delineated T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images to generate a bladder cancer feature set. Perform feature differential screening for bladder cancer feature sets based on clinical diagnostic labels; Perform dimensionality reduction on the screened bladder cancer feature set; A single-parameter MRI classification model and a multi-parameter joint classification model were constructed based on the bladder cancer feature set after dimension reduction. A single-parameter MRI classification model and a multi-parameter joint classification model are used to jointly evaluate the bladder cancer grade of the test data.

[0008] Optionally, the bladder cancer feature set includes T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features; The steps of constructing a single-parameter MRI classification model and a multi-parameter joint classification model based on the bladder cancer feature set after dimension reduction include: Single-parameter MRI classification models were constructed based on the T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features in the bladder cancer feature set. The single-parameter MRI classification models included T2WI classification model, DWI classification model, and DCE classification model. Strong correlation feature analysis is performed on T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features to generate grayscale dependency matrix features, grayscale size region features, and first-order features. The first-order features include bladder tumor heterogeneity features and bladder tumor necrosis region features. A multi-parameter joint classification model is constructed through grayscale dependency matrix features, grayscale size region features and first-order features.

[0009] Optionally, the step of performing feature difference screening for the bladder cancer feature set based on the clinical diagnostic label includes: Normalizing the bladder cancer features in the bladder cancer feature set; Grouping and feature difference analysis were performed according to clinical diagnostic labels, and bladder cancer features were screened based on the difference analysis results.

[0010] Optionally, the clinical diagnostic label includes a bladder cancer grade label, an epithelial cell differentiation degree label, and a differentiation morphology label; The steps for differential feature screening of the bladder cancer feature set based on clinical diagnostic labels include: generating a strong correlation coefficient for each bladder cancer feature in the bladder cancer feature set according to the bladder cancer grade label, the epithelial cell differentiation degree label, and the differentiation morphology label in the clinical diagnosis label; performing feature normalization processing based on strong correlation weights on the bladder cancer features in the bladder cancer feature set according to the strong correlation coefficients; The normalized bladder cancer features were grouped according to the bladder cancer grade label, epithelial cell differentiation degree label, and differentiation morphology label; Perform feature difference analysis on the bladder cancer features after grouping, and assign differential values ​​to each bladder cancer feature based on the feature difference analysis results; Bladder cancer features are screened out from the bladder cancer feature set according to the differential assignment of each bladder cancer feature.

[0011] Optionally, the step of delineating the tumor lesion portion on the multi-parameter MRI image includes: T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were preliminarily delineated using an automatic delineation tool; The preliminary outlined T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were manually adjusted.

[0012] Optionally, after the step of acquiring a multi-parameter MRI image of bladder cancer and the step of delineating a tumor lesion portion of the multi-parameter MRI image, the bladder cancer grade assessment method further comprises: Obtain clinical diagnostic labels for T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images, including epithelial cell differentiation degree labels and differentiation morphology labels; The characteristics of bladder cancer lesions on T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging were obtained from the database. Based on the epithelial cell differentiation degree label, differentiation morphology label and bladder cancer lesion characteristics, bladder cancer enhancement features are generated for T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced images respectively; The bladder lesion area of ​​T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced imaging was enhanced using bladder cancer enhancement features.

[0013] Optionally, the step of performing dimensionality reduction processing on the screened bladder cancer feature set includes: The LASSO regression analysis method was used to reduce the dimension of the screened bladder cancer feature set.

[0014] The second aspect of the present application discloses a radiomics bladder cancer grade assessment device, comprising: A first acquisition unit is used to acquire a multi-parameter MRI image of bladder cancer, where the multi-parameter MRI image includes T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced image; A delineation unit, used for delineating the tumor lesion portion of the multi-parameter MRI image; A first generation unit is used to use an imaging omics package to extract features from the delineated T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging, respectively, to generate a bladder cancer feature set; a screening unit, configured to perform feature difference screening for a bladder cancer feature set based on clinical diagnostic labels; A dimensionality reduction unit, used for performing dimensionality reduction processing on the screened bladder cancer feature set; A construction unit, used to construct a single-parameter MRI classification model and a multi-parameter joint classification model based on the bladder cancer feature set after dimension reduction; The division unit is used to jointly evaluate the bladder cancer grade of the test data using a single-parameter MRI classification model and a multi-parameter joint classification model.

[0015] Optionally, the bladder cancer feature set includes T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features; The building blocks include: Single-parameter MRI classification models were constructed based on the T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features in the bladder cancer feature set. The single-parameter MRI classification models included T2WI classification model, DWI classification model, and DCE classification model. Strong correlation feature analysis is performed on T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features to generate grayscale dependency matrix features, grayscale size region features, and first-order features. The first-order features include bladder tumor heterogeneity features and bladder tumor necrosis region features. A multi-parameter joint classification model is constructed through grayscale dependency matrix features, grayscale size region features and first-order features.

[0016] Optionally, the screening unit comprises: Normalizing the bladder cancer features in the bladder cancer feature set; Grouping and feature difference analysis were performed according to clinical diagnostic labels, and bladder cancer features were screened based on the difference analysis results.

[0017] Optionally, the clinical diagnostic label includes a bladder cancer grade label, an epithelial cell differentiation degree label, and a differentiation morphology label; The screening unit includes: generating a strong correlation coefficient for each bladder cancer feature in the bladder cancer feature set according to the bladder cancer grade label, the epithelial cell differentiation degree label, and the differentiation morphology label in the clinical diagnosis label; performing feature normalization processing based on strong correlation weights on the bladder cancer features in the bladder cancer feature set according to the strong correlation coefficients; The normalized bladder cancer features were grouped according to the bladder cancer grade label, epithelial cell differentiation degree label, and differentiation morphology label; Perform feature difference analysis on the bladder cancer features after grouping, and assign differential values ​​to each bladder cancer feature based on the feature difference analysis results; Bladder cancer features are screened out from the bladder cancer feature set according to the differential assignment of each bladder cancer feature.

[0018] Optionally, the delineation unit includes: T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were preliminarily delineated using an automatic delineation tool; The preliminary outlined T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were manually adjusted.

[0019] Optionally, after the first acquisition unit and the delineation unit, the bladder cancer grade assessment device further comprises: a second acquisition unit, configured to respectively acquire clinical diagnostic labels of T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images, wherein the clinical diagnostic labels include epithelial cell differentiation degree labels and differentiation morphology labels; a third acquisition unit, configured to acquire bladder cancer lesion features of T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging from a database; The second generating unit is used to generate bladder cancer enhancement features for T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced images according to the epithelial cell differentiation degree label, the differentiation morphology label and the bladder cancer lesion feature; The enhancement unit is used for performing lesion area enhancement processing on the bladder area of ​​T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced image by using bladder cancer enhancement features.

[0020] Optionally, the dimension reduction unit is configured to include: The LASSO regression analysis method was used to reduce the dimension of the screened bladder cancer feature set.

[0021] The third aspect of the present application provides a radiomics bladder cancer grade assessment device, comprising: processor, memory, input and output units, and buses; The processor is connected to the memory, input and output units, and the bus; The memory stores a program, and the processor calls the program to execute the bladder cancer grade assessment method according to the first aspect and any optional method according to the first aspect.

[0022] In a fourth aspect, the present application provides a computer-readable storage medium having a program stored thereon. When the program is executed on a computer, the program executes the bladder cancer grade assessment method according to the first aspect and any optional method according to the first aspect.

[0023] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages: In this application, a multi-parameter MRI image of bladder cancer is first obtained, wherein the multi-parameter MRI image includes T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced images, and each group of multi-parameter MRI images includes a T2-weighted imaging, a diffusion-weighted imaging and a dynamic contrast-enhanced image. Next, the tumor lesion portion is outlined for the T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced images respectively to reduce the workload of subsequent feature extraction and increase the quality of subsequent feature extraction. The imaging omics package is used to extract features from the outlined T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced images respectively to generate a bladder cancer feature set. The bladder cancer features extracted from these three types of images can capture the specific features of bladder cancer and can provide comprehensive bladder cancer lesion features in combination with multiple complementary modalities. Next, the bladder cancer feature set is screened for feature differences based on clinical diagnostic labels. The screened bladder cancer feature set is subjected to dimensionality reduction processing. Next, a single-parameter MRI classification model and a multi-parameter joint classification model were constructed based on the reduced bladder cancer feature set. Through screening and dimensionality reduction, the corresponding bladder cancer features of T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were obtained. These features were then combined to form a multi-parameter joint classification model. Finally, the single-parameter MRI classification model and the multi-parameter joint classification model were used to jointly assess the bladder cancer grade of the test data.

[0024] The clinical diagnostic label is a diagnostic label for bladder cancer lesions obtained after a multi-faceted feature analysis of bladder cancer in the same set of multi-parameter MRI images. The clinical diagnostic label is used to screen the feature differences of the bladder cancer feature set, so that each bladder cancer feature in the bladder cancer feature set can be screened for feature differences with other bladder cancer features based on its clinical diagnostic label, and the bladder cancer features whose differences meet the requirements for building a classification model are screened. Next, three independent classification models based on a single imaging sequence were constructed, and a joint classification model was further developed that integrates three parameter features that are strongly correlated with bladder cancer. The aim is to improve the accuracy and reliability of bladder cancer grading assessments by integrating multimodal imaging data, and to provide more accurate decision support for clinicians. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0026] Figure 1 This is a schematic diagram of an embodiment of the radiomics-based bladder cancer grade assessment method of the present application; Figure 2 A schematic diagram of an embodiment of a method for constructing a bladder cancer MRI classification model for this application; Figure 3 A schematic diagram of an embodiment of the method for feature difference screening of the present application; Figure 4 A schematic diagram of an embodiment of the method for feature difference screening of the present application; Figure 5 A schematic diagram of an embodiment of a method for outlining a tumor lesion portion of the present application; Figure 6 A schematic diagram of an embodiment of a method for enhancing multi-parameter MRI features of the present application; Figure 7 This is a schematic diagram of an embodiment of the method for bladder cancer feature dimensionality reduction in this application; Figure 8 A schematic diagram of an embodiment of a bladder cancer grade assessment device based on radiomics in the present application; Figure 9 A schematic diagram of another embodiment of the imaging genomics bladder cancer grade assessment device of the present application; Figure 10 Flowchart of the bladder cancer grade assessment method for this application; Figure 11 A schematic diagram of T2-weighted imaging of bladder cancer in this application; Figure 12 A schematic diagram of the diffusion-weighted imaging of bladder cancer in this application; Figure 13 A schematic diagram of the structure of dynamic enhancement imaging of bladder cancer in this application; Figure 14 A schematic diagram of the strongly correlated bladder cancer features and their corresponding weights in this application; Figure 15 A schematic diagram of the ROC curve of the experiment in this application. DETAILED DESCRIPTION

[0027] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0028] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0029] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0030] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0031] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0032] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0033] In the existing technology, after obtaining a multi-parameter MRI of bladder cancer, it is necessary to manually outline the cancerous area, and then the features in the multi-parameter MRI need to be screened and reduced in dimension, and then a hierarchical classification model is constructed. In the traditional method, the multi-parameter MRI used usually only has one or two items, and bladder cancer is a disease that is extremely difficult to classify, so it is necessary to select a suitable multi-parameter MRI according to its characteristics to construct a classification model. In addition, in the screening and dimensionality reduction process, the traditional method uses the RFE (Recursive Feature Elimination) algorithm for screening and the LASSO (Least Absolute Shrinkage and Selection Operator) algorithm for dimensionality reduction. These two methods are usually based on the patient's routine condition information for screening and dimensionality reduction, and are not based on the patient's bladder cancer's deep clinical diagnostic labels for screening and dimensionality reduction. This reduces the accuracy and reliability of the assessment of bladder cancer grading in the existing technology.

[0034] Based on this, the present application discloses an imaging genomics bladder cancer grade assessment method, device and storage medium for improving the accuracy and reliability of bladder cancer grading assessment.

[0035] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0036] The method of the present application can be applied to a server, device, terminal or other device with logic processing capability, and the present application does not limit this. For the convenience of description, the following description is based on the example of the execution subject being a terminal.

[0037] See also Figure 1 The present application provides an embodiment of a radiomics-based bladder cancer grade assessment method, comprising: 101. Obtain multi-parametric MRI images of bladder cancer, including T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging; In this embodiment, the terminal acquires a multi-parameter MRI image of bladder cancer, wherein the multi-parameter MRI image includes T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging, which are most relevant to bladder cancer.

[0038] Please refer to Figure 11 , Figure 11This is a schematic diagram of T2-weighted imaging for bladder cancer used in this application. T2-weighted imaging primarily reflects the attenuation characteristics of the transverse magnetization vector (T2 relaxation) in bladder tissue. In T2WI (T2-weighted imaging), image contrast is primarily determined by the T2 value of the bladder tissue. Bladder tissue with long T2 values ​​(such as free water, cysts, and mucus) appears as a high signal (bright) on the image, while tissue with short T2 values ​​(such as calcification, fibrous tissue, and acute hemorrhage) appears as a low signal (dark).

[0039] Please refer to Figure 12 , Figure 12 This is a schematic diagram of the diffusion-weighted imaging (DWI) technique for bladder cancer used in this application. DWI is based on the Brownian motion (diffusion motion) of water molecules. In DWI, phase changes caused by the diffusion motion of water molecules are converted into signal intensity changes by applying a diffusion-sensitive gradient field. Bladder tissue with restricted diffusion (such as tumors and infarctions) appears as a high signal on DWI, while bladder tissue with free diffusion (such as cerebrospinal fluid) appears as a low signal.

[0040] Please refer to Figure 13 , Figure 13 This is a schematic diagram of the structure of dynamic contrast-enhanced imaging for bladder cancer in this application. Dynamic contrast-enhanced imaging for bladder cancer involves continuously acquiring a series of MRI images after contrast agent injection to observe the dynamic distribution of the contrast agent in bladder tissue. Due to active angiogenesis and increased vascular permeability in bladder cancer tumor tissue, the rate of contrast agent entry and washout differs from that of normal tissue, resulting in a specific time-signal intensity curve characteristic on DCE-MRI.

[0041] The above three images can capture the specific characteristics of bladder cancer tumor tissue, and combining multiple complementary modalities can provide more comprehensive characteristics. In this embodiment, the potential correlation between multi-parameter MRI of bladder cancer is explored as a means to improve the accuracy and reliability of bladder cancer grading assessment.

[0042] 102. Delineate tumor lesions on multi-parameter MRI images; In this embodiment, the preprocessing process of the multi-parameter MRI image can be performed by manually outlining the bladder cancer lesion area, or it can be automatically outlining using an artificial intelligence model. There is no limitation here, and the specific outlining method will be described later.

[0043] 103. Use the radiomics package to extract features from the delineated T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images to generate a bladder cancer feature set; In this embodiment, the terminal uses an imaging omics package to extract features from the delineated T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging images, generating a bladder cancer feature set. Specifically, during the bladder cancer feature extraction and screening process, key features such as shape, texture, and signal intensity distribution are first extracted from the delineated bladder cancer ROI region. Specifically, the shape characteristics of the ROI region on each image are analyzed, such as area, perimeter, and shape complexity. Texture analysis techniques, such as gray-level co-occurrence matrix (GLCM) and run-length matrix (RLM), are then used to evaluate the internal texture characteristics of the ROI. The distribution of the ROI at different signal intensities is evaluated, such as the mean, variance, skewness, and kurtosis of the signal intensity.

[0044] The terminal uses an imaging omics package to extract several different features from multi-parameter MRI (including T2WI, DCE, and DWI) of bladder cancer patients. In this embodiment, these features include first-order statistical features, second-order features, and higher-order features. Specifically, these bladder cancer features include first-order statistical features describing image signal intensity (such as mean and standard deviation), second-order features describing the relationship between pixels in the image (such as gray-level co-occurrence matrix and gray-level dependence matrix), and more complex higher-order features (such as shape descriptors and texture features).

[0045] 104. Screening for feature differences in a bladder cancer feature set based on clinical diagnostic labels; 105. Perform dimensionality reduction processing on the screened bladder cancer feature set; To reduce redundant features in radiomics, statistical and machine learning methods are further used to screen and reduce the dimensionality of features. In this embodiment, the terminal needs to screen the bladder cancer feature set based on clinical diagnostic labels for feature differences, and then perform dimensionality reduction on the screened bladder cancer feature set.

[0046] In this example, the clinical diagnostic label is a bladder cancer lesion diagnostic label derived from a multi-faceted feature analysis of bladder cancer in the same set of multi-parameter MRI images. The clinical diagnostic label is used to perform feature difference screening for the bladder cancer feature set. Each bladder cancer feature in the bladder cancer feature set is screened for feature differences based on its clinical diagnostic label and other bladder cancer features. Bladder cancer features with sufficient differences to construct a classification model are selected. The details of this screening and dimensionality reduction will be described later.

[0047] 106. Construct a single-parameter MRI classification model and a multi-parameter joint classification model based on the bladder cancer feature set after dimension reduction; 107. Use a single-parameter MRI classification model and a multi-parameter joint classification model to jointly evaluate the bladder cancer grade of the test data.

[0048] In this embodiment, the terminal constructs a single-parameter MRI classification model based on the three features in the bladder cancer feature set after dimensionality reduction. A multi-parameter joint classification model is then constructed based on the three features. Independent models (single-parameter MRI classification models) and a joint model (multi-parameter joint classification model) are established for the features extracted from the multi-parameter MRI. During training, SMOTE oversampling technology is used to address sample imbalance, increasing the number of samples in the minority class and thus preventing the model from favoring the majority class. The specific construction method will be described in detail later.

[0049] Finally, the single-parameter MRI classification model and the multi-parameter joint classification model are used to evaluate the bladder cancer grade of the test data. Figure 10 , Figure 10 FIG. 4 is a flow chart of the bladder cancer grade assessment method of this embodiment.

[0050] In this embodiment, a multi-parameter MRI image of bladder cancer is first obtained, wherein the multi-parameter MRI image includes T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images, and each set of multi-parameter MRI images includes a T2-weighted imaging, a diffusion-weighted imaging, and a dynamic contrast-enhanced image. Next, the tumor lesion portion is delineated for the T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images respectively to reduce the workload of subsequent feature extraction and increase the quality of subsequent feature extraction. The imaging omics package is used to extract features for the delineated T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images respectively to generate a bladder cancer feature set. The bladder cancer features extracted from these three types of images can capture the specific features of bladder cancer and can provide comprehensive bladder cancer lesion features in combination with multiple complementary modalities. Next, the bladder cancer feature set is screened for feature differences based on clinical diagnostic labels. The screened bladder cancer feature set is subjected to dimensionality reduction processing. Next, a single-parameter MRI classification model and a multi-parameter joint classification model were constructed based on the reduced bladder cancer feature set. Through screening and dimensionality reduction, the corresponding bladder cancer features of T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were obtained. These features were then combined to form a multi-parameter joint classification model. Finally, the single-parameter MRI classification model and the multi-parameter joint classification model were used to jointly assess the bladder cancer grade of the test data.

[0051] The clinical diagnostic label is a diagnostic label for bladder cancer lesions obtained after a multi-faceted feature analysis of bladder cancer in the same set of multi-parameter MRI images. The clinical diagnostic label is used to screen the feature differences of the bladder cancer feature set, so that each bladder cancer feature in the bladder cancer feature set can be screened for feature differences with other bladder cancer features based on its clinical diagnostic label, and the bladder cancer features whose differences meet the requirements for building a classification model are screened. Next, three independent classification models based on a single imaging sequence were constructed, and a joint classification model was further developed that integrates three parameter features that are strongly correlated with bladder cancer. The aim is to improve the accuracy and reliability of bladder cancer grading assessments by integrating multimodal imaging data, and to provide more accurate decision support for clinicians.

[0052] Secondly, by making full use of multi-parameter MRI features and merging feature information, a joint classification model was constructed. The construction of this joint model can improve the accuracy and stability of bladder cancer classification and provide more reliable auxiliary information for medical imaging diagnosis.

[0053] See also Figure 2 The present application provides an embodiment of a method for constructing a bladder cancer MRI classification model, comprising: 201. Single-parameter MRI classification models are constructed based on the T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features in the bladder cancer feature set. The single-parameter MRI classification models include T2WI classification model, DWI classification model, and DCE classification model. 202. Perform strong correlation feature analysis on T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features to generate grayscale dependency matrix features, grayscale size region features, and first-order features. The first-order features include bladder tumor heterogeneity features and bladder tumor necrosis region features. 203. A multi-parameter joint classification model is constructed through grayscale dependency matrix features, grayscale size region features and first-order features.

[0054] In this example, an RF classifier was first used to construct initial bladder cancer grading models for different parameter sequences of DWI, T2WI, and DCE. The single-parameter MRI classification models included T2WI, DWI, and DCE classification models. Strong correlation feature analysis was then performed on T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features to generate grayscale dependency matrix features, grayscale size region features, and first-order features. The first-order features included bladder tumor heterogeneity features and bladder tumor necrotic region features. Specifically, the filtered and dimensionality-reduced bladder cancer features were further filtered to identify the grayscale dependency matrix features, grayscale size region features, and first-order features with the highest weights among the three features (DWI, T2WI, and DCE). This analysis was conducted from the perspectives of grayscale dependency matrix, grayscale size region, and bladder tumor heterogeneity and necrotic region.

[0055] For example: after completing the screening and dimensionality reduction, in the process of constructing the joint classification model of multi-parameter MRI, the second round of T-test and LASSO feature screening was first performed on 30 imaging genomics features (10 bladder cancer features under T2-weighted imaging images, 10 bladder cancer features under diffusion-weighted imaging, and 10 bladder cancer features under dynamic contrast-enhanced images), and the 10 most correlated features were obtained to establish the joint classification model of multi-parameter MRI. Figure 14 As shown, Figure 14The 10 features and their corresponding weights are displayed. The weight directly reflects the importance of the feature in the multi-parameter joint classification model. Features with high weights have a greater impact on the prediction results of the multi-parameter joint classification model, while features with low weights have a relatively small impact. This weight distribution provides insight into the inner workings of the multi-parameter joint classification model and helps understand which radiomic features play a decisive role in disease classification. Next, we conducted a strong correlation feature analysis and identified 10 strong correlation features. The top 10 features are: Squareroot_gldm_DependenceNonUniformityNormalized, wavelet-LLL_gldm_Larg-eDependenceHighGrayLevelEmphasis, wavelet-HLL_glszm_SizeZoneNonUniform-ityNormalized, square_gldm_DependenceVariance, wavelet-LLL_glszm_SmallArea-Emphasis, wavelet-LLH_glszm_SmallAreaEmphasis, Squareroot_firstorder_Entropy, wavelet-HLL_glszm_HighGrayLevelZoneEmphasis, logarithm_firstorder_Entropy, and wavelet-HLL_firstorder_Minimum.

[0056] As can be seen, the above 10 features include three gray level dependence matrix (gldm), three first-order features, and four gray level size zone (glszm) features. These features reflect characteristics such as texture, structure, and grayscale distribution. Features such as dependency non-uniformity and grayscale non-uniformity are related to the heterogeneity of the included subjects. Features such as total energy and energy reflect the statistical grayscale properties of the image. Shape features such as surface area may be related to the morphology of the subjects. The two features with the largest weight coefficients reveal a strong correlation between bladder cancer and high heterogeneity and localized necrosis. The most important first-order features for constructing a multi-parameter joint classification model are logarithm_firstorder_entropy, which enhances the influence of high grayscale values ​​through logarithmic transformation and measures the randomness of the image grayscale distribution. Higher entropy values ​​indicate greater internal heterogeneity of bladder tumors. Wavelet-HLL_Firstorder_Minimum reflects low signal intensity in localized regions of the image. Low grayscale values ​​correspond to necrotic areas within bladder cancer tumors. These two first-order features are unique to bladder cancer. A multi-parameter joint classification model constructed using these strongly correlated features is more accurate than conventional models using grayscale region features and grayscale dependency matrix features, resulting in more accurate classification results. Once the ten most correlated bladder cancer features are identified, a multi-parameter joint classification model for multi-parameter MRI can be established.

[0057] In this embodiment, the terminal uses a RF classifier to construct a bladder cancer grading model (a single-parameter MRI classification model) based on different parameter sequences (bladder cancer features) of DWI, T2WI, and DCE, and generates classification results. During the training process, SMOTE oversampling technology was used to address sample imbalance. This method preserves the intra-class relationships of the original samples while increasing sample diversity and avoiding the risk of overfitting caused by simple replication. After multiple experimental verifications and adjustments to the classifier parameters, the number of decision trees in the RF was ultimately set to 20. A larger number of decision trees improves model stability and accuracy. To fully utilize the more contributing multi-parameter MRI features, a joint model was constructed using feature fusion to produce a more robust feature representation. Information from different features was combined to create a richer and more comprehensive feature set. This multi-parameter joint classification model can be combined with the single-parameter MRI classification model to analyze bladder cancer features, improving the accuracy and stability of bladder cancer classification and providing more reliable auxiliary information for medical imaging diagnosis.

[0058] The experimental environment of this embodiment includes a processor with 16 GB of memory and a Pycharm 2023 with radiomics.

[0059] According to the experimental steps, the test was conducted on 110 patients with pathologically confirmed bladder cancer, including 70 cases of high-grade bladder cancer and 40 cases of low-grade bladder cancer. In order to comprehensively evaluate the model, the accuracy, F1 value and AUC value were used. The specific results are shown in Table 1. The classification accuracy of the present invention is 0.9091, the F1 value is 0.9231, and the AUC value is 0.9060. The ROC curve of this experiment is shown in Table 1. Figure 15 shown.

[0060]

[0061] Table 1 Experimental results As can be seen, when using single-parameter MRI input, while T2WI and DWI alone demonstrate high diagnostic performance, demonstrating their importance in bladder cancer classification, the combined use of T2WI, DWI, and DCE can further improve diagnostic accuracy, F1 value, and AUC, thereby providing more comprehensive diagnostic information in medical image analysis. Therefore, the single-parameter MRI classification model and multi-parameter combined classification model in this embodiment have considerable clinical significance for bladder cancer detection.

[0062] The specific embodiments of the present invention are described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in this field without departing from the scope of the present invention.

[0063] See also Figure 3 , the present application provides an embodiment of a method for feature difference screening, comprising: 301. Normalizing the bladder cancer features in the bladder cancer feature set; 302. Grouping and feature difference analysis are performed based on clinical diagnostic labels, and bladder cancer features are screened based on the difference analysis results.

[0064] In order to reduce redundant features in the bladder cancer feature set and determine the most significant bladder cancer features related to bladder cancer in the bladder cancer feature set, in this embodiment, multiple feature selection methods are used. Specifically, clinical diagnostic labels and T tests are used to analyze each bladder cancer feature in the extracted bladder cancer feature set to screen out bladder cancer features with significant differences.

[0065] First, in conventional methods, before screening bladder cancer features, the extracted T2WI, DWI, and DCE imaging feature values ​​(feature values ​​of bladder cancer features) are directly normalized using Z-score to avoid the influence of feature scale differences on T-test results. The method is as follows:

[0066] Where x is the original eigenvalue, is the mean of all eigenvalues, is the standard deviation of all eigenvalues.

[0067] Next, the terminal performs grouping and feature difference analysis based on the clinical diagnostic labels, and then screens for bladder cancer features based on the difference analysis results. Specifically, the terminal can use a conventional T-test to analyze the extracted bladder cancer features (T2WI, DWI, and DCE) to screen for features with significant differences.

[0068] For example, for bladder cancer features extracted from T2-weighted imaging, the samples are directly divided into two groups according to the clinical diagnosis label (high-grade bladder cancer and low-grade bladder cancer). For each feature, the mean of the two groups is calculated. and , standard deviation and , sample size and . Use the T test formula:

[0069] Calculate the T statistic. Set the significance level , and then use the ttest_ind function in the Python library to obtain the degrees of freedom When the calculated t value is greater than the critical value or the P value is less than 0.05, it is judged that there is a significant difference between the two groups and the bladder cancer feature is retained. Finally, several bladder cancer features are screened for T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging.

[0070] The above method can quickly screen out features with significant grade differences based on established clinical diagnostic labels. It can also directly group bladder cancer grades to quickly screen out features with significant grade differences for subsequent model construction.

[0071] See also Figure 4 , the present application provides an embodiment of a method for feature difference screening, comprising: 401. Generate a strong correlation coefficient for each bladder cancer feature in the bladder cancer feature set according to the bladder cancer grade label, the epithelial cell differentiation degree label, and the differentiation morphology label in the clinical diagnosis label; In this embodiment, the conventional Z-score method directly normalizes the extracted three radiomics feature values ​​(feature values ​​of bladder cancer features) of T2WI, DWI, and DCE, which have certain defects. It does not take into account the severity of bladder cancer within each bladder cancer feature, and different severity levels are manifested differently on different multi-parameter MRI images. Direct normalization will lose information about the severity of the disease. To address the above issues, in this embodiment, clinical diagnostic labels are further obtained when the multi-parameter MRI images of bladder cancer are acquired. The clinical diagnostic labels include three labels: a bladder cancer grade label, which is the growth degree of bladder cancer diagnosed by humans or equipment; a second epithelial cell differentiation degree label, which is the differentiation level label of bladder cancer tumor cells diagnosed by humans or equipment; and a third differentiation morphology label, which is a label that characterizes the degree of differentiation morphology of epithelial tumor cells of bladder cancer.

[0072] The terminal determines the level weight parameter corresponding to the bladder cancer level label, the differentiation level weight parameter of the epithelial cell differentiation level label, and the differentiation morphology weight parameter corresponding to the differentiation morphology label based on the bladder cancer level label, epithelial cell differentiation level label, and differentiation morphology label corresponding to each bladder cancer feature. These weights are obtained according to the training target. The training target can set the corresponding importance of the specific bladder cancer level, epithelial cell differentiation level, and differentiation morphology, so that the characteristic value of the corresponding bladder cancer feature can be enhanced subsequently. The three weight parameters are all values ​​greater than 0 and less than 1. Next, the display degree coefficient of the disease degree (bladder cancer level label, epithelial cell differentiation level label, and differentiation morphology label) on different multi-parameter MRI images (T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images) is obtained. These three weight parameters and the display degree coefficient are calculated for each bladder cancer feature to generate a strong correlation coefficient. Taking a bladder cancer feature on a T2-weighted image as an example, the strong correlation coefficient is calculated. , the formula is as follows:

[0073] in, is the level weight parameter corresponding to the bladder cancer level label of this T2-weighted image, is the differentiation degree weight parameter of the epithelial cell differentiation degree label of this T2-weighted image, is the differentiation morphology weight parameter corresponding to the differentiation morphology label of this T2-weighted image, is the degree coefficient of bladder cancer grade on T2-weighted imaging, is the coefficient of epithelial cell differentiation displayed on T2-weighted imaging, It is the degree coefficient of differentiation morphology on T2-weighted imaging.

[0074] By analogy, a strong correlation coefficient of a diffusion-weighted image , the formula is as follows:

[0075] in, The level weight parameter corresponding to the bladder cancer level label of this diffusion-weighted imaging is: The differentiation degree weight parameter of the epithelial cell differentiation degree label of this diffusion-weighted imaging, is the differentiation morphology weight parameter of the differentiation morphology label of this diffusion-weighted imaging, is the degree coefficient of bladder cancer grade on diffusion-weighted imaging, is the degree coefficient of epithelial cell differentiation displayed on diffusion-weighted imaging, is the degree of differentiation displayed on diffusion-weighted imaging.

[0076] Strong correlation coefficient of a dynamic contrast enhanced image , the formula is as follows:

[0077] in, is the level weight parameter corresponding to the bladder cancer grade label of this dynamic contrast enhanced image, is the differentiation degree weight parameter of the epithelial cell differentiation degree label of this dynamic contrast enhanced image, is the differentiation morphological weight parameter of the differentiation morphological label of this dynamic contrast enhanced image, is the display degree coefficient of bladder cancer grade on dynamic contrast-enhanced images, is the degree coefficient of epithelial cell differentiation displayed on dynamic contrast-enhanced images, It is the display degree coefficient of differentiated morphology on dynamic contrast enhanced images.

[0078] 402. Perform feature normalization processing based on strong correlation weights on the bladder cancer features in the bladder cancer feature set according to the strong correlation coefficients; Next, the terminal performs feature normalization based on the strong correlation weight on the bladder cancer features in the bladder cancer feature set according to the strong correlation coefficient. The formula is as follows:

[0079] 403. The normalized bladder cancer features are grouped according to the bladder cancer grade label, epithelial cell differentiation degree label, and differentiation morphology label; Next, the terminal first groups the normalized bladder cancer features according to the bladder cancer level labels to generate a first group of bladder cancer features and a second group of bladder cancer features.

[0080] Next, the terminal groups the normalized bladder cancer features according to the epithelial cell differentiation degree labels to generate the third group of bladder cancer features and the fourth group of bladder cancer features.

[0081] Finally, the terminal groups the normalized bladder cancer features according to the differentiation morphology labels to generate the fifth group of bladder cancer features and the sixth group of bladder cancer features.

[0082] 404. Perform feature difference analysis on the bladder cancer features after grouping, and assign a differential value to each bladder cancer feature based on the feature difference analysis result; The terminal first performs a T-test on the first and second groups of bladder cancer features, assigning differential values ​​to the selected bladder cancer features and 0 to all others. Next, the terminal performs a T-test on the third and fourth groups of bladder cancer features, again assigning differential values ​​to the selected bladder cancer features and 0 to all others. The same process is repeated for the fifth and sixth groups of bladder cancer features, resulting in three differential values ​​for each feature.

[0083] 405. Filter out bladder cancer features from the bladder cancer feature set based on the differential assignment of each bladder cancer feature.

[0084] The terminal selects bladder cancer features from the bladder cancer feature set based on the differentially assigned values ​​for each bladder cancer feature. Specifically, the differentially assigned values ​​for each bladder cancer feature are summed up, and the bladder cancer features with a ranking higher than a preset value on T2-weighted imaging are selected, while the remaining features are discarded. Similarly, diffusion-weighted imaging and dynamic contrast-enhanced images are screened, retaining those that meet the criteria.

[0085] See also Figure 5 The present application provides an embodiment of a method for outlining a tumor lesion, comprising: 501. Use the automatic delineation tool to perform preliminary delineation on T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images; In this embodiment, an automatic delineation tool is first used to perform preliminary delineation on the T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images, respectively, to reduce the time for manual delineation and increase efficiency.

[0086] 502. Manually adjust the preliminary outlined T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced images.

[0087] Next, the outlined parts are manually adjusted, and several attending radiologists with rich experience in imaging diagnosis are responsible for the data set outline, ensuring the accuracy and reliability of the outline work. During the outline process, doctors need to identify and select the tumor area on the image layer by layer, which involves careful inspection of each layer to ensure that the complete outline of the tumor is accurately captured. After the automatic outline tool generates the preliminary tumor area, the doctors will perform manual corrections to correct any inaccurate outlines. This method greatly reduces the time of purely manual outline and increases efficiency.

[0088] See also Figure 6 The present application provides an embodiment of a method for enhancing multi-parameter MRI features, comprising: 601. Obtain clinical diagnostic labels for T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging, respectively. The clinical diagnostic labels include epithelial cell differentiation degree labels and differentiation morphology labels. In this embodiment, the terminal needs to obtain clinical diagnostic labels of T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images. The clinical diagnostic labels include epithelial cell differentiation degree labels and differentiation morphology labels, that is, to determine the area, epithelial cell differentiation degree, and differentiation morphology of bladder cancer in multi-parameter MRI.

[0089] 602. Obtain bladder cancer lesion features of T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging from a database; According to the degree of differentiation and differentiation morphology of epithelial cells in each image, a threshold-matched bladder cancer standard image is determined, and corresponding bladder cancer lesion features are extracted from the bladder cancer standard image.

[0090] 603. Generate bladder cancer enhancement features based on T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images according to the epithelial cell differentiation degree label, differentiation morphology label, and bladder cancer lesion characteristics; The terminal adjusts the bladder cancer lesion characteristics into data formats that conform to T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging. If the bladder cancer lesion characteristics in the database are already in data formats corresponding to T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging, no adjustment is required.

[0091] For example: determine the epithelial cell differentiation degree label and differentiation morphology label of a T2-weighted image, determine a reference T2-weighted image corresponding to the epithelial cell differentiation degree label and differentiation morphology label from the database, extract bladder cancer features from the reference T2-weighted image, and generate corresponding bladder cancer lesion features. In this case, the bladder cancer lesion features are bladder cancer enhancement features.

[0092] 604. Perform lesion area enhancement processing on the bladder area of ​​T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging using bladder cancer enhancement features.

[0093] After acquiring the bladder cancer enhancement feature, the terminal uses it to enhance the bladder lesion region on T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images. In other words, a bladder cancer enhancement feature only enhances the corresponding multi-parameter MRI image. This enhancement facilitates delineation and increases the accuracy of subsequent feature extraction.

[0094] See also Figure 7 The present application provides an embodiment of a method for bladder cancer feature dimensionality reduction, comprising: 701. The dimension of the screened bladder cancer feature set was reduced using the LASSO regression analysis method.

[0095] In this embodiment, the terminal uses LASSO regression analysis to further narrow the feature range and focus on key features related to bladder cancer. To further narrow the feature range and focus on key features related to bladder cancer, LASSO regression analysis is employed. The LASSO regression analysis method in this embodiment incorporates an L1 regularization term into the objective function. This compresses the coefficients of features with minimal impact on the dependent variable to zero during model training, thereby achieving feature selection and dimensionality reduction. For each trained LASSO model (T2WI, DWI, DCE), its coefficients are examined. Features with coefficients compressed to zero are eliminated, and features with non-zero coefficients are identified as key features related to bladder cancer. After LASSO regression dimensionality reduction screening, a sufficient number of bladder cancer features of the required type are screened from T2WI, DWI, and DCE.

[0096] See also Figure 8 The present application provides an embodiment of a radiomics-based bladder cancer grade assessment device, comprising: A first acquisition unit 801 is configured to acquire a multi-parameter MRI image of bladder cancer, wherein the multi-parameter MRI image includes T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging; A second acquisition unit 802 is configured to respectively acquire clinical diagnostic labels of T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging, wherein the clinical diagnostic labels include epithelial cell differentiation degree labels and differentiation morphology labels; The third acquisition unit 803 is used to respectively acquire bladder cancer lesion features of T2-weighted imaging, diffusion-weighted imaging and dynamic contrast-enhanced imaging from the database; The second generating unit 804 is configured to generate bladder cancer enhancement features based on the epithelial cell differentiation degree label, the differentiation morphology label, and the bladder cancer lesion feature for T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced image, respectively; The enhancement unit 805 is configured to perform lesion area enhancement processing on the bladder area of ​​the T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced image using bladder cancer enhancement features; A delineation unit 806 is configured to delineate the tumor lesion portion of the multi-parameter MRI image; Optionally, the delineation unit 806 includes: T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were preliminarily delineated using an automatic delineation tool; The preliminary outlined T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced images were manually adjusted.

[0097] A first generating unit 807 is configured to extract features from the delineated T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging using an imaging omics package to generate a bladder cancer feature set; a screening unit 808 for performing feature difference screening for a bladder cancer feature set based on clinical diagnostic labels; Optionally, the screening unit 808 includes: Normalizing the bladder cancer features in the bladder cancer feature set; Grouping and feature difference analysis were performed according to clinical diagnostic labels, and bladder cancer features were screened based on the difference analysis results.

[0098] Optionally, the clinical diagnostic label includes a bladder cancer grade label, an epithelial cell differentiation degree label, and a differentiation morphology label; The screening unit 808 includes: generating a strong correlation coefficient for each bladder cancer feature in the bladder cancer feature set according to the bladder cancer grade label, the epithelial cell differentiation degree label, and the differentiation morphology label in the clinical diagnosis label; performing feature normalization processing based on strong correlation weights on the bladder cancer features in the bladder cancer feature set according to the strong correlation coefficients; The normalized bladder cancer features were grouped according to the bladder cancer grade label, epithelial cell differentiation degree label, and differentiation morphology label; Perform feature difference analysis on the bladder cancer features after grouping, and assign differential values ​​to each bladder cancer feature based on the feature difference analysis results; Bladder cancer features are screened out from the bladder cancer feature set according to the differential assignment of each bladder cancer feature.

[0099] A dimension reduction unit 809 is used to perform dimension reduction processing on the screened bladder cancer feature set; Optionally, the dimension reduction unit 809 is configured to include: The LASSO regression analysis method was used to reduce the dimension of the screened bladder cancer feature set.

[0100] A construction unit 810 is configured to construct a single-parameter MRI classification model and a multi-parameter joint classification model based on the bladder cancer feature set after dimension reduction; Optionally, the bladder cancer feature set includes T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features; The construction unit 810 includes: Single-parameter MRI classification models were constructed based on the T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features in the bladder cancer feature set. The single-parameter MRI classification models included T2WI classification model, DWI classification model, and DCE classification model. Strong correlation feature analysis is performed on T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features to generate grayscale dependency matrix features, grayscale size region features, and first-order features. The first-order features include bladder tumor heterogeneity features and bladder tumor necrosis region features. A multi-parameter joint classification model is constructed through grayscale dependency matrix features, grayscale size region features and first-order features.

[0101] The classification unit 811 is configured to jointly evaluate the bladder cancer grade of the test data using a single-parameter MRI classification model and a multi-parameter joint classification model.

[0102] See also Figure 9 The present application provides a radiomics-based bladder cancer grade assessment device, comprising: Processor 901 , memory 902 , input / output unit 903 , and bus 904 .

[0103] The processor 901 is connected to the memory 902 , the input / output unit 903 , and the bus 904 .

[0104] The memory 902 stores a program, and the processor 901 calls the program to execute the following Figure 1 、 Figure 2and Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 and Figure 7 Methods for assessing the grade of bladder cancer in .

[0105] The present application provides a computer-readable storage medium, wherein a program is stored on the computer-readable storage medium, and when the program is executed on a computer, the program performs the following operations: Figure 1 、 Figure 2 and Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 and Figure 7 Methods for assessing the grade of bladder cancer in .

[0106] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0107] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0108] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0109] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0110] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk, and other media that can store program code.

Claims

1. A radiomics-based method for assessing the grade of bladder cancer, characterized in that: include: Acquiring a multi-parameter MRI image of bladder cancer, wherein the multi-parameter MRI image includes T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging; Delineating a tumor lesion portion on the multi-parameter MRI image; Using an imaging omics package, feature extraction is performed on the delineated T2-weighted imaging, the diffusion-weighted imaging, and the dynamic contrast-enhanced imaging to generate a bladder cancer feature set; performing feature difference screening for the bladder cancer feature set according to the clinical diagnostic label; performing dimensionality reduction processing on the screened bladder cancer feature set; Constructing a single-parameter MRI classification model and a multi-parameter joint classification model based on the bladder cancer feature set after dimension reduction; The single-parameter MRI classification model and the multi-parameter joint classification model are used to jointly evaluate the bladder cancer grade of the test data.

2. The method for assessing the grade of bladder cancer according to claim 1, wherein: The bladder cancer feature set includes T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features; The step of constructing a single-parameter MRI classification model and a multi-parameter joint classification model based on the bladder cancer feature set after dimension reduction includes: Constructing single-parameter MRI classification models according to the T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features in the bladder cancer feature set, respectively, wherein the single-parameter MRI classification models include a T2WI classification model, a DWI classification model, and a DCE classification model; Performing strong correlation feature analysis on T2-weighted features, diffusion-weighted features, and dynamic contrast enhancement features to generate grayscale dependency matrix features, grayscale size region features, and first-order features, wherein the first-order features include bladder tumor heterogeneity features and bladder tumor necrosis region features; A multi-parameter joint classification model is constructed through grayscale dependency matrix features, grayscale size region features and first-order features.

3. The method for assessing the grade of bladder cancer according to claim 2, wherein: The step of performing feature difference screening for the bladder cancer feature set based on the clinical diagnostic label comprises: performing normalization processing on the bladder cancer features in the bladder cancer feature set; Grouping processing and feature difference analysis are performed according to the clinical diagnostic labels, and bladder cancer features are screened according to the difference analysis results.

4. The method for assessing the grade of bladder cancer according to claim 2, wherein: The clinical diagnostic labels include bladder cancer grade label, epithelial cell differentiation degree label, and differentiation morphology label; The step of performing feature difference screening for the bladder cancer feature set based on the clinical diagnostic label comprises: generating a strong correlation coefficient for each bladder cancer feature in the bladder cancer feature set according to the bladder cancer grade label, the epithelial cell differentiation degree label, and the differentiation morphology label in the clinical diagnosis label; performing feature normalization processing based on strong correlation weights on the bladder cancer features in the bladder cancer feature set according to the strong correlation coefficients; grouping the normalized bladder cancer features according to the bladder cancer grade label, the epithelial cell differentiation degree label, and the differentiation morphology label; Perform feature difference analysis on the bladder cancer features after grouping, and assign differential values ​​to each bladder cancer feature based on the feature difference analysis results; Bladder cancer features are screened out from the bladder cancer feature set according to the differential assignment of each bladder cancer feature.

5. The method for evaluating the grade of bladder cancer according to any one of claims 1 to 4, characterized in that: The step of outlining the tumor lesion portion of the multi-parameter MRI image comprises: Using an automatic delineation tool to preliminarily delineate the T2-weighted imaging, the diffusion-weighted imaging, and the dynamic contrast-enhanced imaging, respectively; The T2-weighted imaging, the diffusion-weighted imaging, and the dynamic contrast-enhanced imaging that were initially outlined were manually adjusted.

6. The method for assessing the grade of bladder cancer according to claim 5, wherein: After the step of acquiring a multi-parameter MRI image of bladder cancer and the step of delineating the tumor lesion portion of the multi-parameter MRI image, the bladder cancer grade assessment method further includes: respectively acquiring clinical diagnostic labels of the T2-weighted imaging, the diffusion-weighted imaging, and the dynamic contrast-enhanced image, wherein the clinical diagnostic labels include an epithelial cell differentiation degree label and a differentiation morphology label; respectively acquiring bladder cancer lesion features of the T2-weighted imaging, the diffusion-weighted imaging, and the dynamic contrast-enhanced imaging from a database; generating bladder cancer enhancement features for the T2-weighted imaging, the diffusion-weighted imaging, and the dynamic contrast-enhanced imaging according to the epithelial cell differentiation degree label, the differentiation morphology label, and the bladder cancer lesion feature; The bladder region of the T2-weighted imaging, the diffusion-weighted imaging and the dynamic contrast-enhanced image is enhanced by using bladder cancer enhancement features.

7. The method for evaluating the grade of bladder cancer according to any one of claims 1 to 4, characterized in that: The step of performing dimensionality reduction processing on the screened bladder cancer feature set comprises: The bladder cancer feature set after screening was subjected to dimension reduction processing by LASSO regression analysis method.

8. A radiomics-based bladder cancer grade assessment device, characterized in that: include: a first acquisition unit, configured to acquire a multi-parameter MRI image of bladder cancer, wherein the multi-parameter MRI image includes T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging; a delineation unit, configured to delineate a tumor lesion portion of the multi-parameter MRI image; A first generating unit is configured to extract features from the delineated T2-weighted imaging, the diffusion-weighted imaging, and the dynamic contrast-enhanced imaging using a radiomics package to generate a bladder cancer feature set; a screening unit, configured to perform feature difference screening for the bladder cancer feature set based on the clinical diagnostic label; a dimensionality reduction unit, configured to perform dimensionality reduction processing on the screened bladder cancer feature set; A construction unit, configured to construct a single-parameter MRI classification model and a multi-parameter joint classification model based on the bladder cancer feature set after dimension reduction; The classification unit is configured to jointly evaluate the bladder cancer grade of the test data using the single-parameter MRI classification model and the multi-parameter joint classification model.

9. A radiomics-based bladder cancer grade assessment device, characterized in that: include: A processor, a memory, an input-output unit, and a bus, wherein the processor is connected to the memory, the input-output unit, and the bus, the memory stores a program, and the processor calls the program to execute the bladder cancer grade assessment method according to any one of claims 1 to 7. 10 . A computer-readable storage medium having a program stored thereon, wherein when the program is executed on a computer, the method for evaluating the grade of bladder cancer according to claim 1 is executed.

Images