Performance evaluation method and device of magnetic resonance cytometry imaging in distinguishing high-grade bladder cancer from low-grade bladder cancer

Abstract

The invention discloses a performance evaluation method and device of magnetic resonance cytometry imaging in distinguishing high-level and low-level bladder cancer. The method comprises the following steps: respectively carrying out imaging parameter calculation on a PGSE sequence image group with an ROI (Region of Interest), a 20Hz OGSE sequence image group with the ROI and a 40Hz OGSE sequence image group with the ROI so as to obtain imaging parameter information; and performing performance analysis according to the imaging parameter information to obtain an analysis result. According to the method, PGSE and two OGSE sequences are combined, specific microscopic parameters are extracted by using an IMPULSED technology, the limitation that the specificity of a traditional ADC is insufficient is broken through, and the image quality is guaranteed through 3T MRI and parameter optimization; the bladder cancer can be differentiated and graded noninvasively, the joint parameter AUC reaches 0.92, and reliable support is provided for personalized preoperative treatment.

Description

Technical Field

[0001] This application relates to the field of image processing technology, specifically to a method and apparatus for performance evaluation of magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer. Background Technology

[0002] Bladder cancer is the most common malignant tumor of the urinary system, with more than 500,000 new cases and 200,000 deaths worldwide each year. Tumor grading is crucial for treatment and prognosis. High-grade tumors have a higher risk of recurrence, invasion, and metastasis, and usually require aggressive treatment, such as intravesical BCG instillation or radical cystectomy. Conversely, low-grade tumors are often treated with more conservative bladder-preserving strategies. Accurate preoperative non-invasive grading is essential for personalized treatment and optimized prognosis.

[0003] Diffusion-weighted magnetic resonance imaging (dMRI), particularly time-dependent dMRI, is indispensable in bladder imaging, revealing tumor biology by characterizing the movement of water molecules within tissues. The apparent diffusion coefficient (ADC) reflects the overall diffusion capacity of tissue, aiding in tumor detection and characterization. However, as a composite indicator, ADC lacks specificity for cell density, membrane permeability, or extracellular structures, and is affected by inflammation, edema, and partial volume effects, limiting its accuracy in tumor grading. These challenges highlight the need for advanced imaging methods capable of probing the microstructure of tumors more specifically.

[0004] Magnetic resonance cytometry has emerged as a promising technique to overcome these limitations. Unlike traditional dMRI metrics, which reflect mixed signals from multiple microstructural factors, magnetic resonance cytometry employs multiple diffusion times (t_diff) to probe tissue microstructures with greater specificity [13-16]. This technique can extract specific quantitative microstructural parameters such as cell diameter (d), intracellular volume fraction (v_in), cell density (ρ), and extracellular diffusion coefficient (D_ex). Imaging microstructural parameterization using finite-spectral edited diffusion (IMPULSED) [17, 18] is the most widely used magnetic resonance cytometry technique, which integrates pulsed gradient spin echo (PGSE) and oscillating gradient spin echo (OGSE) sequences to enhance sensitivity to a wide range of microstructural features. Studies in breast, head and neck, endometrial, and glioma cancers have demonstrated its clinical potential in biologically relevant tumor characterization.

[0005] Despite its promising prospects, the feasibility and diagnostic utility of magnetic resonance cytometry in bladder cancer remain largely unexplored. Unlike solid organs such as the breast, the bladder presents unique technical challenges, including signal interference from residual urine and the difficulty in accurately delineating tumors close to or attached to the bladder wall and detrusor muscle. These factors can affect image quality and diagnostic accuracy, thus necessitating disease-specific assessments. Summary of the Invention

[0006] The purpose of this invention is to provide a method for detecting and evaluating arterial plaques to at least solve one of the above-mentioned technical problems.

[0007] One aspect of the present invention provides a method for evaluating the performance of magnetic resonance cytometry in distinguishing between high-grade and low-grade bladder cancer, the method comprising:

[0008] Acquire PGSE sequence image groups, 20Hz OGSE sequence image groups, and 40Hz OGSE sequence image groups;

[0009] Imaging preprocessing was performed on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group respectively to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group.

[0010] Regions of interest (ROIs) were identified in the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group, respectively, to obtain PGSE sequence image group with ROIs, 20Hz OGSE sequence image group with ROIs, and 40Hz OGSE sequence image group with ROIs.

[0011] Imaging parameters were calculated for the PGSE image sequence group with ROI region, the 20Hz OGSE image sequence group with ROI region, and the 40Hz OGSE image sequence group with ROI region, respectively, to obtain imaging parameter information.

[0012] Performance analysis is performed based on imaging parameter information to obtain analysis results.

[0013] Optionally, the step of performing imaging preprocessing on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group respectively to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group includes:

[0014] Using the 20Hz OGSE sequence image group as a reference standard, the imregtform and imwarp functions are used to align the PGSE sequence image group and the 40Hz OGSE sequence image group with the reference standard, thereby obtaining the aligned PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group.

[0015] Four-dimensional denoising was performed on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group using Marchenko-Pastur principal component analysis to obtain the denoised PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group.

[0016] The denoised PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group were corrected by IVIM to obtain the corrected PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group.

[0017] Log-linear fitting was performed on the corrected PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group.

[0018] Optionally, the step of calculating imaging parameters for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, respectively, to obtain imaging parameter information includes:

[0019] Microstructure parameters were calculated for PGSE sequence image groups with ROI regions, 20Hz OGSE sequence image groups with ROI regions, and 40Hz OGSE sequence image groups with ROI regions to obtain the average value data of MPULSED derived microstructure parameters.

[0020] Time-dependent ADC metrics were calculated for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, thereby obtaining the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data.

[0021] Optionally, the step of calculating microstructure parameters for the PGSE sequence image group with ROI regions, the 20Hz OGSE sequence image group with ROI regions, and the 40Hz OGSE sequence image group with ROI regions to obtain the average value data of MPULSED-derived microstructure parameters includes:

[0022] Normalization processing is performed on the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region to obtain the PGSE sequence ROI normalized signal set, the 20Hz OGSE sequence ROI normalized signal set, and the 40Hz OGSE sequence ROI normalized signal set. The PGSE sequence ROI normalized signal set, the 20Hz OGSE sequence ROI normalized signal set, and the 40Hz OGSE sequence ROI normalized signal set constitute a joint normalized signal dataset.

[0023] The IMPULSED model was constructed and its parameters were fitted on the joint normalized signal dataset to obtain the average value data of the microstructure parameters derived from IMPULSED.

[0024] Optionally, the step of calculating time-dependent ADC metrics for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region to obtain ADC PGSE average data, ADC 20Hz average data, and ADC 40Hz average data includes:

[0025] Normalize the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region to obtain the normalized PGSE sequence image group, the normalized 20Hz OGSE sequence image group, and the normalized 40Hz OGSE sequence image group.

[0026] The normalized PGSE sequence image group was fitted with a single exponential model and ADC was calculated to obtain the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data.

[0027] Optionally, the step of performing performance analysis based on imaging parameter information to obtain analysis results includes:

[0028] Obtain a preset database, which includes multiple preset pathological parameter data;

[0029] Performance analysis is performed based on the pathological parameters and imaging parameters to obtain the analysis results.

[0030] Optionally, obtaining the preset database includes:

[0031] Surgical specimens from patients corresponding to the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group were obtained.

[0032] The surgical specimen was stained with H&E to obtain H&E-stained tumor tissue sections;

[0033] Bladder cancer grading was performed on each H&E-stained tumor tissue section to obtain pathological grading results;

[0034] Pathological parameters are calculated based on tumor tissue sections stained with H&E to obtain standard parameter information; wherein, the standard parameter information and pathological grading results of a patient constitute the preset pathological parameter data of a patient.

[0035] Optionally, the step of calculating pathological parameters based on H&E-stained tumor tissue sections to obtain standard parameter information includes:

[0036] Image recognition was performed on H&E stained tumor tissue sections to obtain cells and cell nuclei in the sections, thereby obtaining the nuclear diameter of each identified cell and the total number of identified cells;

[0037] The volume-weighted average nuclear diameter is obtained based on the nuclear diameter of the identified cells and the total number of identified cells;

[0038] The pathological cell density is obtained based on the total number of cells identified.

[0039] Optionally, the step of performing performance analysis based on the pathological parameter data and imaging parameter information to obtain analysis results includes:

[0040] The Mann–Whitney U test was performed on the mean data of each IMPULSED-derived microstructural parameter to obtain statistical differences between the low- and high-level groups.

[0041] This application also provides a performance evaluation device for magnetic resonance cytometry in distinguishing between high-grade and low-grade bladder cancer, the performance evaluation device for magnetic resonance cytometry in distinguishing between high-grade and low-grade bladder cancer includes:

[0042] An image group acquisition module is used to acquire PGSE sequence image groups, 20Hz OGSE sequence image groups, and 40Hz OGSE sequence image groups.

[0043] The preprocessing module is used to perform imaging preprocessing on the PGSE sequence image group, the 20Hz OGSE sequence image group and the 40Hz OGSE sequence image group respectively, so as to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group and the preprocessed 40Hz OGSE sequence image group.

[0044] The region of interest (ROI) identification module is used to identify the ROI of the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group, respectively, so as to obtain the PGSE sequence image group with ROI, the 20Hz OGSE sequence image group with ROI, and the 40Hz OGSE sequence image group with ROI.

[0045] An imaging parameter information acquisition module is used to calculate imaging parameters for PGSE sequence image groups with ROI regions, 20Hz OGSE sequence image groups with ROI regions, and 40Hz OGSE sequence image groups with ROI regions, respectively, thereby acquiring imaging parameter information.

[0046] The performance analysis module is used to perform performance analysis based on imaging parameter information, thereby obtaining analysis results.

[0047] The performance evaluation method of magnetic resonance cytometry proposed in this application for differentiating high-grade and low-grade bladder cancer combines PGSE and two OGSE sequences, and uses IMPULSED technology to extract specific microscopic parameters (v). in (e.g., p, etc.), overcoming the limitations of traditional ADCs in terms of specificity; 3T MRI and optimized parameters ensure image quality; comprehensive preprocessing, including registration and denoising, improves data accuracy; thorough pathological validation, with imaging parameters strongly correlated with pathological indicators (r = 0.74-0.78); scientific statistical methods prevent overfitting and comprehensively evaluate diagnostic efficacy; can non-invasively differentiate bladder cancer grades, v in The combined parameter AUC reached 0.92, providing reliable support for personalized preoperative treatment. Attached Figure Description

[0048] Figure 1 This is a schematic flowchart illustrating a method for evaluating the performance of magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer according to an embodiment of this application.

[0049] Figure 2 This is a schematic diagram of a patient screening process according to an embodiment of this application.

[0050] Figure 3 This is a schematic diagram of ADC indices and IMPULSE-derived microstructural parameters of low-grade and high-grade bladder tumors according to an embodiment of this application.

[0051] Figure 4 This is a schematic diagram showing a comparative analysis of ADC indices and IMPULSE-derived microstructural parameters of low-grade and high-grade bladder cancer according to an embodiment of this application.

[0052] Figure 5 This is a schematic diagram of ROC curve analysis of ADC index and IMPULSE-derived microstructure parameters used for bladder cancer grading according to an embodiment of this application.

[0053] Figure 6 This is a schematic diagram illustrating the correlation between IMPULSED-derived microstructural parameters and corresponding histopathological features in one embodiment of this application. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all, embodiments of this application. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0055] like Figure 1 The performance evaluation methods for magnetic resonance cytometry in distinguishing between high-grade and low-grade bladder cancer, as shown, include:

[0056] Acquire PGSE sequence image groups, 20Hz OGSE sequence image groups, and 40Hz OGSE sequence image groups;

[0057] Imaging preprocessing was performed on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group respectively to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group.

[0058] Regions of interest (ROIs) were identified in the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group, respectively, to obtain PGSE sequence image group with ROIs, 20Hz OGSE sequence image group with ROIs, and 40Hz OGSE sequence image group with ROIs.

[0059] Imaging parameters were calculated for the PGSE image sequence group with ROI region, the 20Hz OGSE image sequence group with ROI region, and the 40Hz OGSE image sequence group with ROI region, respectively, to obtain imaging parameter information.

[0060] Performance analysis is performed based on imaging parameter information to obtain analysis results.

[0061] In this embodiment, the PGSE sequence image group includes multiple PGSE sequence images, and one PGSE sequence image belongs to one patient;

[0062] In this embodiment, the 20Hz OGSE sequence image group includes multiple 20Hz OGSE sequence images, and one 20Hz OGSE sequence image belongs to one patient;

[0063] In this embodiment, the 40Hz OGSE sequence image group includes multiple 40Hz OGSE sequence images, and one 40Hz OGSE sequence image belongs to one patient;

[0064] In this embodiment, each patient simultaneously possesses a set of PGSE sequence images, a set of 20Hz OGSE sequence images, and a set of 40Hz OGSE sequence images.

[0065] In this embodiment, patients can be selected in the following manner:

[0066] Between March 2024 and May 2025, 76 patients with pathologically confirmed bladder cancer who underwent MRI were initially included. The inclusion criteria were as follows: (1) a preoperative bladder MRI was performed within 2 weeks prior to surgical resection; and (2) no history of biopsy, radiotherapy, chemotherapy or other treatment interventions prior to MRI.

[0067] Exclusion criteria were as follows: (1) lack of clear pathological results (n=9); (2) tumor long diameter <10 mm; (3) insufficient bladder filling or presence of severe imaging artifacts (n=2). After applying these criteria, the final analysis included 60 patients with pathologically confirmed tumors, comprising 37 (61.7%) high-grade tumors and 23 (38.3%) low-grade tumors. The patient screening process is detailed below. Figure 2 Patient and tumor characteristics are summarized in Table 1:

[0068] Table 1. Clinical characteristics of patients with high-grade and low-grade bladder cancer

[0069]

[0070] In this embodiment, the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group are obtained in the following manner:

[0071] All bladder dMRI examinations were performed in the supine position on a 3T MRI system (MAGNETOM Vida, Siemens Healthineers AG, Forchheim, Germany), equipped with a phased array body coil and a whole-body gradient coil. A combined acquisition scheme using PGSE (t_diff=70ms), 20 Hz cosine-modulated OGSE (t_diff=12.5 ms), and 40 Hz OGSE (t_diff=6.25 ms) was employed to measure diffusion-weighted signals at different diffusion times. Detailed acquisition parameters are shown in Table 2.

[0072] Table 2. Acquisition parameters of dMRI sequences

[0073]

[0074] in, The gradient duration; For diffusion time (gradient separation); PGSE is the oscillation frequency; OGSE is the pulse gradient echo; OGSE is the oscillation gradient echo.

[0075] Imaging was performed using two-dimensional single-shot excitation spin-echo-planar echo imaging (SE-EPI) with three orthogonally diffused weighted gradient directions ([1,1,–0.5],[1,–0.5,1],[–0.5,1,1]). The following acquisition parameters were applied: TE / TR = 127 / 5100 ms; field of view = 300 × 224 mm²; in-plane resolution = 2.0 × 2.0 mm²; slice thickness = 4.0 mm; slice spacing = 0.8 mm; 20 slices; GRAPPA factor = 2; echo spacing = 0.62 ms; bandwidth = 1,852 Hz / pixel; partial Fourier factor = 6 / 8 (to shorten TE and improve signal-to-noise ratio). The geometric mean of the signals from the three diffusion directions was calculated to minimize directional noise.

[0076] In this embodiment, the imaging preprocessing of the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group includes:

[0077] Using the 20Hz OGSE sequence image group as a reference standard, the imregtform and imwarp functions are used to align the PGSE sequence image group and the 40Hz OGSE sequence image group with the reference standard, thereby obtaining the aligned PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group.

[0078] Specifically, the MATLAB R2022b software is opened, and the imregtform function is called to calculate the registration transformation matrix of the PGSE, 40Hz OGSE sequence images and the 20Hz OGSE sequence images. Then, the imwarp function is used to perform spatial transformation on the PGSE and 40Hz OGSE sequence images according to the matrix, so as to achieve the alignment of the three sequence images and eliminate image misalignment caused by slight movement during patient scanning.

[0079] Four-dimensional denoising was performed on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group using Marchenko-Pastur principal component analysis to obtain the denoised PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group.

[0080] Specifically, in MATLAB R2022b software, the three aligned image sequences are constructed as four-dimensional data with "time dimension (corresponding to different scan sequences) + spatial dimension (corresponding to image pixels)". The Marchenko-Pastur principal component analysis algorithm is applied to denoise the four-dimensional data to filter out random noise.

[0081] The denoised PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group were corrected by IVIM to obtain the corrected PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group.

[0082] Specifically, in MATLAB R2022b software, the non-diffusion weighted signals (b=0s / mm²) of the three sequences in the denoised image are extracted. By refitting the signal, the interference of non-diffusion factors such as blood perfusion on the diffusion signal is eliminated, ensuring that the signal only reflects the diffusion motion of water molecules.

[0083] Log-linear fitting was performed on the corrected PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group.

[0084] Specifically, in MATLAB R2022b software, signal data of b=200~1000 s / mm² in the image after IVIM correction are extracted, and then substituted into the preset logarithmic linear fitting formula for fitting calculation to eliminate residual interference factors and restore the true diffusion signal characteristics of water molecules.

[0085] In this embodiment, the step of calculating imaging parameters for the PGSE sequence image group with ROI regions, the 20Hz OGSE sequence image group with ROI regions, and the 40Hz OGSE sequence image group with ROI regions to obtain imaging parameter information includes:

[0086] Microstructure parameters were calculated for PGSE sequence image groups with ROI regions, 20Hz OGSE sequence image groups with ROI regions, and 40Hz OGSE sequence image groups with ROI regions to obtain the average value data of MPULSED derived microstructure parameters.

[0087] Time-dependent ADC metrics were calculated for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, thereby obtaining the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data.

[0088] In this embodiment, the calculation of microstructure parameters for the PGSE sequence image group with ROI regions, the 20Hz OGSE sequence image group with ROI regions, and the 40Hz OGSE sequence image group with ROI regions, thereby obtaining the average value data of MPULSED-derived microstructure parameters, includes:

[0089] Normalization processing is performed on the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region to obtain the PGSE sequence ROI normalized signal set, the 20Hz OGSE sequence ROI normalized signal set, and the 40Hz OGSE sequence ROI normalized signal set. The PGSE sequence ROI normalized signal set, the 20Hz OGSE sequence ROI normalized signal set, and the 40Hz OGSE sequence ROI normalized signal set constitute a joint normalized signal dataset.

[0090] The IMPULSED model was constructed and its parameters were fitted on the joint normalized signal dataset to obtain the average value data of the microstructure parameters derived from IMPULSED.

[0091] Specifically, the signal of each pixel in the PGSE image sequence group with ROI region, the 20Hz OGSE image sequence group with ROI region, and the 40Hz OGSE image sequence group with ROI region are normalized to obtain the normalized diffused signal S (calculated as: pixel signal value S at a certain b value). b The purpose of ÷ the signal value S0 of the pixel at b=0 s / mm² is to eliminate the signal strength deviation caused by the difference in scanning conditions of different sequences and ensure that the three types of sequence signals can be jointly modeled.

[0092] In this embodiment, the construction and parameter fitting of the IMPULSED model on the joint normalized signal dataset to obtain the average value data of the IMPULSED-derived microstructure parameters includes:

[0093] The normalized diffusion signal S is decomposed into intracellular signals (S0). in ) and extracellular signals (S ex The sum of contributions of ), i.e., S=v in ×Sin +(1-v in )×S ex (v) in (This represents the intracellular volume fraction, to be estimated), and it is assumed that "water exchange across the cell membrane is negligible";

[0094] Intracellular signaling modeling: S in =exp(-b.ADC r ), where: b is the diffusion weighting factor of the corresponding sequence (e.g., the b value of PGSE is 0, 250, etc., and the b value of 20Hz OGSE is 0, 250, etc.); ADC r For confined diffusion, ADC (diffusion coefficient of water molecules within cells) r The values ​​are determined to be cell diameter d (to be estimated) and D. in A function of (fixed at 1.56 μm² / ms);

[0095] Extracellular signal modeling: S ex =exp(-bD ex ), where D ex is the extracellular diffusion coefficient (to be estimated), and b is the diffusion weighting factor;

[0096] Model Fitting: Substituting the normalized diffuse signal S and corresponding b values ​​from the joint normalized signal dataset into the above model, the MATI platform was used for fitting to solve for the microstructural parameters of each pixel in the ROI region: cell diameter d, intracellular volume fraction v. in Extracellular diffusion coefficient D ex ;

[0097] Cell density ρ is calculated using the formula ρ = v in / d×100 calculates the ρ value (v) for each pixel. in (where d is the fitted pixel parameter).

[0098] Parameter averaging: Calculates the d and v values ​​of all pixels in the ROI region. in D ex The average values ​​of ρ were used as the IMPULSED-derived microstructural parameters of the patient's tumor.

[0099] In this embodiment, the calculation of time-dependent ADC metrics for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, thereby obtaining the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data, includes:

[0100] Normalize the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region to obtain the normalized PGSE sequence image group, the normalized 20Hz OGSE sequence image group, and the normalized 40Hz OGSE sequence image group.

[0101] The normalized PGSE sequence image group was fitted with a single exponential model and ADC was calculated to obtain the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data.

[0102] Specifically, based on the ROI pixel coordinates drawn on the 20Hz OGSE sequence image obtained above (denoted as ROI 20Hz coordinates, such as "row 10-50, column 20-60, layer 3-5");

[0103] The ROI 20Hz coordinates are directly mapped to the PGSE and 40Hz OGSE image sequences to obtain the pixel coordinate range that completely overlaps with the ROI 20Hz space. These are denoted as ROI PGSE coordinates and ROI 40Hz coordinates, respectively (for example, "row 10-50" of 20Hz OGSE corresponds to "row 10-50" of PGSE, and the coordinates are completely consistent).

[0104] Based on the ROI PGSE coordinates, ROI 20Hz coordinates, and ROI 40Hz coordinates, all pixel signals within the corresponding coordinate range are extracted from the corrected original signals of PGSE, 20Hz OGSE, and 40Hz OGSE, respectively. (This includes signal values ​​for different b values ​​in each sequence, such as signals corresponding to b=0, 250, and 500 for PGSE). This yields the GSE sequence ROI signal set, which includes all pixels within the ROI PGSE coordinate range and the corrected original signals corresponding to each b value (the format can be understood as "pixel coordinates + b value + signal value", such as "(10,20,3), b=0, signal value=1500; (10,20,3), b=250, signal value=1200").

[0105] 20Hz OGSE sequence ROI signal set: contains the original corrected signal (same format as above) of all pixels within the 20Hz coordinate range of the ROI and each b value.

[0106] 40Hz OGSE sequence ROI signal set: contains the original corrected signal (same format as above) for all pixels within the 40Hz coordinate range of the ROI and each b value.

[0107] For each pixel in the ROI signal set of each sequence, find the signal value of that pixel at b=0 s / mm² (denoted as S0);

[0108] For the signal value of this pixel at other b values ​​(such as 250, 500, etc.) (denoted as S) b According to the formula "normalized signal S=S b / S0” calculation eliminates the absolute deviation in signal strength caused by differences in scanning conditions of different sequences;

[0109] Normalization calculations were performed on the ROI signal sets of PGSE, 20Hz OGSE, and 40Hz OGSE respectively to obtain the following information:

[0110] PGSE sequence ROI normalized signal set: contains the normalized signal S corresponding to each b value of all pixels within the ROI_PGSE coordinate range (format: pixel coordinate + b value + normalized signal S);

[0111] 20Hz OGSE sequence ROI normalized signal set: contains the normalized signal S (same format as above) for all pixels within the ROI_20Hz coordinate range and each b value.

[0112] 40Hz OGSE sequence ROI normalized signal set: contains the normalized signal S (same format as above) for all pixels within the ROI_40Hz coordinate range and each b value.

[0113] The three are combined into a "Joint Normalized Signal Dataset" (containing normalized signals of the three types of sequences, used for subsequent IMPULSE model fitting).

[0114] Model construction: The normalized signal S of each pixel in the joint normalized signal dataset is modeled as an intracellular signal S. in +Extracellular signal S ex The weighted sum, formula: S=v in ×S in +(1-v in )×S ex (v) in =Intracellular volume fraction; assuming transmembrane water exchange is negligible).

[0115] Intracellular signaling modeling: S in =exp(-b×ADC r ), where ADC r =Confined diffusion ADC, calculated using the following formula:

[0116] ;

[0117] Among them, ADC rIt is calculated based on the analytical expression for confined diffusion in a spherical cavity, where the radius R of the sphere is defined as d / 2, and the intracellular diffusion coefficient D... in The diffusion time is fixed at 1.56 μm² / ms, and its equivalent diffusion time is denoted as Δeff.

[0118] Extracellular signal modeling: S ex =exp(-b×D ex (D) ex =Extracellular diffusion coefficient, to be estimated);

[0119] Model Fitting: Substitute the normalized signal S+ corresponding to the b value of each pixel into the above model, and use the MATI platform for fitting to solve for the d and v values ​​of that pixel. in D ex ;

[0120] Cell density ρ calculation: ρ = v (as per the document formula) in / d×100, substitute the v of each pixel in , d, calculate ρ for that pixel;

[0121] Parameter averaging: d, v values ​​for all pixels in the ROI region in D ex Calculate the average values ​​of ρ and ρ respectively to eliminate noise interference from individual pixels.

[0122] Through the above processing, the following content is obtained:

[0123] Mean values ​​of IMPULSED-derived microstructural parameters (calculated by combining three types of sequence signals, corresponding to the overall tumor level in the patient):

[0124] Average cell diameter d (unit: μm);

[0125] Intracellular volume fraction v in Average value (unitless, a percentage, such as 0.31);

[0126] extracellular diffusion coefficient D ex Average value (unit: μm² / ms);

[0127] Average cell density ρ (unit: 1 / μm).

[0128] For the PGSE sequence: Extract the multi-b value signals (b=0, 250, 500, 750, 1000, 1200, 1400s / mm²) of all pixels within the ROI PGSE coordinate range from the original signal after PGSE sequence correction; thereby obtaining the original signal set of multi-b values ​​of the ROI in the PGSE sequence (format: "pixel coordinates + b value + original signal value").

[0129] For the 20Hz OGSE sequence: Extract the multi-b value signals (b=0, 250, 500, 750, 1000 s / mm²) of all pixels within the 20Hz coordinate range of the ROI from the original signal after correction of the 20Hz OGSE sequence; thereby obtaining the original multi-b value signal set of the ROI of the 20Hz OGSE sequence (in the same format as above);

[0130] For the 40Hz OGSE sequence: Extract the multi-b value signals (b=0, 250, 500 s / mm²) of all pixels within the 40Hz coordinate range of the ROI from the original signal after correction of the 40Hz OGSE sequence. This yields the original multi-b value signal set of the ROI of the 40Hz OGSE sequence.

[0131] For each pixel in each sequence ROI signal set, according to the normalized signal S=S b / S0 calculation (same as the normalization method in the calculation of microstructure parameters);

[0132] Single exponential model fitting: Substitute the normalized signal S+ corresponding to the b value of each pixel into the model S=exp(-b×ADC), and fit the solution to obtain the ADC value of that pixel (corresponding to each sequence):

[0133] The ADC values ​​of PGSE sequence pixels are obtained by fitting the PGSE sequence pixels.

[0134] The ADC values ​​of the 20Hz OGSE sequence pixels were obtained by fitting the 20Hz OGSE sequence pixels.

[0135] The ADC values ​​of the 40Hz OGSE sequence pixels were obtained by fitting the 40Hz OGSE sequence pixels.

[0136] Average ADC value: The average ADC value of all pixels in the ROI region of each sequence is calculated to represent the ADC level of the tumor in that sequence.

[0137] This allows us to obtain the average time-dependent ADC index for each patient (corresponding to one group for each of the three sequence types):

[0138] ADCPGSE average value (PGSE sequence, unit: μm² / ms);

[0139] ADC 20Hz average value (20Hz OGSE sequence, unit: μm² / ms);

[0140] ADC 40Hz average value (40Hz OGSE sequence, unit: μm² / ms).

[0141] In this embodiment, the step of performing performance analysis based on imaging parameter information to obtain analysis results includes:

[0142] Obtain a preset database, which includes multiple preset pathological parameter data;

[0143] Performance analysis is performed based on the pathological parameters and imaging parameters to obtain the analysis results.

[0144] In this embodiment, obtaining the preset database includes:

[0145] Surgical specimens from patients corresponding to the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group were obtained.

[0146] The surgical specimen was stained with H&E to obtain H&E-stained tumor tissue sections;

[0147] Bladder cancer grading was performed on each H&E-stained tumor tissue section to obtain pathological grading results;

[0148] Pathological parameters are calculated based on tumor tissue sections stained with H&E to obtain standard parameter information; wherein, the standard parameter information and pathological grading results of a patient constitute the preset pathological parameter data of a patient.

[0149] In this embodiment, the step of calculating pathological parameters based on H&E-stained tumor tissue sections to obtain standard parameter information includes:

[0150] Image recognition was performed on H&E stained tumor tissue sections to obtain cells and cell nuclei in the sections, thereby obtaining the nuclear diameter of each identified cell and the total number of identified cells;

[0151] The volume-weighted average nuclear diameter is obtained based on the nuclear diameter of the identified cells and the total number of identified cells;

[0152] The pathological cell density is obtained based on the total number of cells identified.

[0153] Specifically, hematoxylin-eosin (H&E) stained surgical specimens underwent histopathological evaluation. Bladder cancer grading was determined according to the 2016 World Health Organization (WHO) classification criteria. Pathological diagnosis was based on specimens from 53 patients who underwent transurethral resection of bladder tumor (TURBT) and 7 patients who underwent radical cystectomy. Bladder cancer was classified as low-grade or high-grade based on histological characteristics.

[0154] All pathological assessments were performed by a pathologist with 6 years of experience (YW) and independently reviewed by another pathologist with 18 years of experience (ZL). For quantitative assessment of nuclear morphology, H&E-stained sections were analyzed using CellPose, a deep learning-based segmentation model pre-trained on various microscopic images. The volume-weighted average nuclear diameter was calculated using the following formula: d_mean = ∑d n 4 / ∑d n ³, where n is the number of cells recognized by CellPose. Pathological cell density is calculated as follows: ρ path =n / A, where A is the area of ​​the tissue slice.

[0155] In this embodiment, the step of performing performance analysis based on the pathological parameter data and imaging parameter information to obtain analysis results includes:

[0156] The Mann–Whitney U test was performed on the mean data of each IMPULSED-derived microstructural parameter to obtain statistical differences between the low- and high-level groups.

[0157] In this embodiment, the average values ​​of IMPULSED-derived microstructural parameters and ADC indices were calculated for each patient across the entire ROI. The Mann–Whitney U test was used to assess the statistical differences between low-grade and high-grade bladder cancer. Logistic regression was applied to evaluate the diagnostic performance of individual and combined imaging parameters for tumor grading. Receiver operating characteristic (ROC) curve analysis was performed, and the area under the curve (AUC) was calculated to quantify the diagnostic power of each model. To prevent overfitting, a hybrid feature selection method combining Lasso and Ridge regression was used, with a maximum of three features (corresponding to 1 / 20 of the sample size) in the multivariate model. The DeLong test was used to compare the AUC of different regression models. All statistical analyses were performed using SPSS software (IBM, Chicago, IL, USA).

[0158] Specifically, based on the ROI region imaging parameter data obtained for each patient above, including IMPULSE-derived microstructural parameters (d, v) in D ex , p), and time-dependent ADC metrics (ADCPGSE, ADC20Hz, ADC40Hz);

[0159] In SPSS software, for each patient, the average values ​​of the microstructure parameters and ADC indices of all pixels within the entire ROI region are calculated to eliminate the interference of individual pixel noise on subsequent analysis, thereby obtaining a dataset of average imaging parameters for each patient, containing the d and v values ​​for each of the 60 patients. in D ex ρ, ADCPGSE, ADC20Hz, ADC40Hz mean values, and corresponding pathological grading labels (low grade / high grade).

[0160] In SPSS software, using pathological grade as the grouping variable, the average value of each imaging parameter (d, v) is calculated. in D ex Mann-Whitney U tests were performed on the low- and high-level groups (ρ, ADCPGSE, ADC20Hz, and ADC40Hz) to analyze the statistical differences between them, thus obtaining a table of inter-group difference test results, including the test statistic (T-value) and p-value for each parameter, to determine whether there is a significant difference (P < 0.05 is considered significant). For example: v in The three ADC indices showed significant differences between groups (P < 0.001), d (P = 0.86), and D. ex (P=0.053) No significant difference.

[0161] For each imaging parameter, a logistic regression model was constructed, with the parameter as the independent variable and pathological grade as the dependent variable, to evaluate its predictive ability for tumor grade.

[0162] For each logistic regression model, plot the receiver operating characteristic (ROC) curve, calculate the area under the curve (AUC) and the 95% confidence interval (CI), and determine the optimal threshold.

[0163] Based on the optimal threshold, the diagnostic metrics for each parameter are calculated: Sensitivity, Specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), Accuracy, Positive Likelihood Ratio (+LR), and Negative Likelihood Ratio (-LR).

[0164] Finally, a single-parameter diagnostic performance result table is obtained, which includes indicators such as AUC (95% CI), threshold, sensitivity, and specificity for each parameter. For example, the AUC of v_in is 0.89 (95% CI: 0.80–0.97), and the AUC of ADCPGSE is 0.82 (95% CI: 0.71–0.94).

[0165] The hybrid feature selection method is applied from the pool of independent variables (mean values ​​of all imaging parameters: d, v). in Dex Features are selected from (p, ADCPGSE, ADC20Hz, ADC40Hz). A multivariate model can contain a maximum of three features (corresponding to 1 / 20 of the sample size, 60 cases × 1 / 20 = 3).

[0166] Based on the selected features (in this embodiment, v is ultimately selected) in D ex (ADCPGSE), construct a multivariate logistic regression model;

[0167] ROC curve analysis was performed on the multivariate model to calculate the AUC (95% CI) and the corresponding diagnostic indicators (sensitivity, specificity, etc.).

[0168] In SPSS software, the DeLong test is used to compare the AUC of different models (such as multivariate models and single-parameter ADCP-GSE models) to determine whether there is a statistically significant difference in their diagnostic efficacy.

[0169] Inter-group comparison of imaging parameters:

[0170] Figure 3 This paper presents representative feature maps of time-dependent ADC indices and IMPULSE-derived microstructural parameters for low-grade and high-grade bladder tumors. Representative feature maps of ADC indices and model-fitted microstructural parameters are superimposed on b=0 s / mm² averaged DWI images acquired using PGSE sequences in low-grade (top) and high-grade (bottom) bladder tumors. The averaged DWI is calculated as the geometric mean of images acquired in three orthogonal diffusion directions. Where ADC is the apparent diffusion coefficient; DWI is the diffusion-weighted image; v in d represents the intracellular volume fraction; d represents the cell diameter; D ex ρ is the extracellular diffusion coefficient; ρ is the cell density; PGSE is the pulse gradient spin echo.

[0171] Intergroup comparisons revealed significant differences between low-grade and high-grade bladder tumors. Figure 4 And Table 3, in which, Figure 4 (a) Comparison of ADC values ​​of low-grade and high-grade bladder tumors on PGSE, ADC20Hz and ADC40Hz sequences. Figure 4 (b) Comparison of IMPULSE-derived microstructural parameters between low-grade and high-grade tumors. P < 0.001. All three ADC parameters (ADCPGSE, ADC20Hz, and ADC40Hz) were significantly lower in high-grade lesions than in low-grade lesions (all p < 0.001). Conversely, the IMPULSE-derived parameter v in Both ρ and d were significantly higher in high-grade tumors (p < 0.001). Parameters d (p = 0.86) and D...ex (p = 0.053) No statistically significant difference was observed.

[0172] Table 3. Comparison of imaging parameters between high-grade and low-grade bladder tumors

[0173]

[0174] IQR is the interquartile range; IMPULSED is the average data of IMPULSED-derived microstructural parameters; d is the cell diameter; V in D represents the intracellular volume fraction. EX The extracellular diffusion coefficient; Cell density.

[0175] Diagnostic performance of imaging parameters in bladder cancer grading:

[0176] The diagnostic performance of single and combined imaging parameters is summarized in Figure 5 And Table 4. In Figure 5 In the table, (a) is the ROC curve for distinguishing between low-grade and high-grade bladder tumors using IMPULSED-derived microstructural parameters. (b) is the ROC curve for distinguishing between low-grade and high-grade bladder tumors using the ADC index. (c) is the ROC curve of a multivariate regression model combining the ADC index and microstructural parameters.

[0177] Among the ADC metrics, ADCPGSE exhibited the highest diagnostic performance, with an AUC of 0.82. A combined model integrating all three ADC metrics (Multi-ADC) produced a slightly higher AUC (0.84). Among the IMPULSE-derived parameters, v_in had the highest diagnostic accuracy (AUC = 0.89), followed by ρ (AUC = 0.85) and D_ex (AUC = 0.65). Integrating ADC and microstructure parameters further improved diagnostic performance. After hybrid feature selection, the three variables: v in D ex ADCPGSE was included in a multivariate regression model with an AUC of 0.92. However, the Delong test showed no statistically significant difference between the AUC of this combined model and the AUC (0.82) of the best-performing single ADC index, ADCPGSE (p=0.10).

[0178] Table 4. Diagnostic performance of ADC and model-fitted microstructural parameters in differentiating between high-grade and low-grade bladder cancer

[0179]

[0180] AUC values ​​are expressed as the mean (self-administered 95% confidence interval);

[0181] Correlation analysis:

[0182] Correlation analysis revealed a strong correlation between imaging-derived microstructural parameters and histopathological findings. Figure 6 ).exist Figure 6 In the images, (a) is a representative image of automatically segmented H&E-stained sections of low-grade bladder tumors. (b) is a representative image of automatically segmented H&E-stained sections of high-grade bladder tumors. (c) is a scatter plot showing a positive correlation between imaging-derived cell density (ρ) and corresponding pathological-derived cell density (r=0.74). (d) is a scatter plot showing a significant positive correlation between imaging-derived cell diameter (d) and the volume-weighted average nuclear diameter of pathologically derived cells (r=0.78).

[0183] The d-value derived from IMPULSED was strongly positively correlated with the volume-weighted average nuclear diameter obtained from H&E stained sections (r=0.74). Similarly, the ρ-value derived from imaging was positively correlated with the cell density derived from pathology (r=0.78).

[0184] In the above technical solutions, the intracellular diffusion coefficient (D) in Fixed at 1.56μm 2 This method, operating at / ms, does not consider the multi-parameter coupling regulation of tumor microstructure. Specific problems include:

[0185] Cell diameter (d) determines the spatial scale of intracellular water molecule diffusion and directly affects the degree of diffusion restriction;

[0186] Preliminary estimate of cell density (ρ) pre This reflects the degree of intercellular crowding; higher crowding will compress intracellular space and increase diffusion resistance.

[0187] Predicted intracellular volume fraction (v) in,pre The proportion of intracellular water molecules is related to the molecular collision probability; the higher the proportion, the greater the probability of molecular collisions and the lower the diffusion rate.

[0188] The time-dependent ADC ratio (ADC PGSE / ADC 40Hz) can indirectly characterize intracellular structural heterogeneity; stronger heterogeneity leads to greater fluctuations in diffusion rate. (Fixed D) in This will cause the confined diffusion ADC (ADC) r The calculations introduce systematic biases, especially in high-grade tumors with small cell counts, high crowding, and high v. in In the phenotype, the deviation will be amplified by v. in The estimation error of ρ reduces the diagnostic specificity.

[0189] Therefore, this application also provides a D in The method to obtain it is as follows:

[0190] ;

[0191] D in denoted as the personalized intracellular diffusion coefficient; α is the basic calibration coefficient, α∈[1.40,1.65]; d is the imaging-derived cell diameter; μ d μ serves as a standardized benchmark for cell diameter. d The diameter was 16.37 μm; β is the cell diameter regulation index, β∈[0.25,0.45], β∈[0.25,0.35] for high-grade tumors, and β∈[0.35,0.45] for low-grade tumors; γ is the cell density regulation index, γ∈[0.15,0.25]; δ is the intracellular volume fraction regulation coefficient, δ∈[0.6,0.8]; v in,pre To estimate the intracellular volume fraction, μ v As a standardized benchmark for intracellular volume fraction, μ v =0.255; ε is the intracellular volume fraction regulation index, ε∈[1.1,1.3]; ζ is the ADC ratio regulation coefficient, ζ∈[0.3,0.5]; ADC PGSE is the apparent diffusion coefficient of the PGSE sequence; ADC 40Hz is the apparent diffusion coefficient of the 40Hz OGSE sequence; η is the ADC ratio regulation index, η∈[0.7,0.9].

[0192] In this embodiment, ρ can be obtained when fitting the IMPULSED model. pre v in,pre d;

[0193] The obtained parameters can be used to calculate personalized D using the formula above. in Then use personalized D in Refit the IMPULSE model to finally output v in , d, D ex The precise value of ρ (replacing the original fixed D) in (Fitting results).

[0194] This application also provides a performance evaluation device for magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer. The device includes an image group acquisition module, a preprocessing module, a region of interest identification module, an imaging parameter information acquisition module, and a performance analysis module.

[0195] The image group acquisition module is used to acquire PGSE sequence image groups, 20Hz OGSE sequence image groups, and 40Hz OGSE sequence image groups.

[0196] The preprocessing module is used to perform imaging preprocessing on the PGSE sequence image group, the 20Hz OGSE sequence image group and the 40Hz OGSE sequence image group respectively, so as to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group and the preprocessed 40Hz OGSE sequence image group.

[0197] The region of interest (ROI) identification module is used to identify the ROI of the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group, respectively, so as to obtain the PGSE sequence image group with ROI, the 20Hz OGSE sequence image group with ROI, and the 40Hz OGSE sequence image group with ROI.

[0198] The imaging parameter information acquisition module is used to calculate imaging parameters for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, respectively, so as to obtain imaging parameter information.

[0199] The performance analysis module is used to perform performance analysis based on imaging parameter information, thereby obtaining analysis results.

[0200] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for evaluating the performance of magnetic resonance cytometry in distinguishing between high-grade and low-grade bladder cancer, characterized in that, The performance evaluation methods for magnetic resonance cytometry in distinguishing between high-grade and low-grade bladder cancer include: Acquire PGSE sequence image groups, 20Hz OGSE sequence image groups, and 40Hz OGSE sequence image groups; Imaging preprocessing was performed on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group respectively to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group. Regions of interest (ROIs) were identified in the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group, respectively, to obtain PGSE sequence image group with ROIs, 20Hz OGSE sequence image group with ROIs, and 40Hz OGSE sequence image group with ROIs. Imaging parameters were calculated for the PGSE image sequence group with ROI region, the 20Hz OGSE image sequence group with ROI region, and the 40Hz OGSE image sequence group with ROI region, respectively, to obtain imaging parameter information. Performance analysis is performed based on imaging parameter information to obtain analysis results.

2. The method for evaluating the performance of magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer as described in claim 1, characterized in that, The step of performing imaging preprocessing on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group respectively to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group includes: Using the 20Hz OGSE sequence image group as a reference standard, the imregtform and imwarp functions are used to align the PGSE sequence image group and the 40Hz OGSE sequence image group with the reference standard, thereby obtaining the aligned PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group. Four-dimensional denoising was performed on the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group using Marchenko-Pastur principal component analysis to obtain the denoised PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group. The denoised PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group were corrected by IVIM to obtain the corrected PGSE sequence image group, 20Hz OGSE sequence image group and 40Hz OGSE sequence image group. Log-linear fitting was performed on the corrected PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group.

3. The performance evaluation method for magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer as described in claim 2, characterized in that, The process of calculating imaging parameters for PGSE image sequences with ROI regions, 20Hz OGSE image sequences with ROI regions, and 40Hz OGSE image sequences with ROI regions to obtain imaging parameter information includes: Microstructure parameters were calculated for PGSE sequence image groups with ROI regions, 20Hz OGSE sequence image groups with ROI regions, and 40Hz OGSE sequence image groups with ROI regions to obtain average data of IMPULSE-derived microstructure parameters. Time-dependent ADC metrics were calculated for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, thereby obtaining the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data.

4. The performance evaluation method for magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer as described in claim 3, characterized in that, The calculation of microstructure parameters for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, to obtain the average value data of MPULSED derived microstructure parameters, includes: Normalization processing is performed on the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region to obtain the PGSE sequence ROI normalized signal set, the 20Hz OGSE sequence ROI normalized signal set, and the 40Hz OGSE sequence ROI normalized signal set. The PGSE sequence ROI normalized signal set, the 20Hz OGSE sequence ROI normalized signal set, and the 40Hz OGSE sequence ROI normalized signal set constitute a joint normalized signal dataset. An IMPULSE model was constructed and parameters were fitted on the joint normalized signal dataset to obtain the average value data of the microstructure parameters derived from IMPULSE.

5. The method for detecting and evaluating arterial plaques as described in claim 4, characterized in that, The calculation of time-dependent ADC metrics for the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region, to obtain the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data, includes: Normalize the PGSE sequence image group with ROI region, the 20Hz OGSE sequence image group with ROI region, and the 40Hz OGSE sequence image group with ROI region to obtain the normalized PGSE sequence image group, the normalized 20Hz OGSE sequence image group, and the normalized 40Hz OGSE sequence image group. The normalized PGSE sequence image group was fitted with a single exponential model and ADC was calculated to obtain the average ADC PGSE data, the average ADC 20Hz data, and the average ADC 40Hz data.

6. The method for evaluating the performance of magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer as described in claim 5, characterized in that, The process of performing performance analysis based on imaging parameter information to obtain analysis results includes: Obtain a preset database, which includes multiple preset pathological parameter data; Performance analysis is performed based on the pathological parameters and imaging parameters to obtain the analysis results.

7. The performance evaluation method for magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer as described in claim 6, characterized in that, The process of obtaining the preset database includes: Surgical specimens from patients corresponding to the PGSE sequence image group, the 20Hz OGSE sequence image group, and the 40Hz OGSE sequence image group were obtained. The surgical specimen was stained with H&E to obtain H&E-stained tumor tissue sections; Bladder cancer grading was performed on each H&E-stained tumor tissue section to obtain pathological grading results; Pathological parameters are calculated based on tumor tissue sections stained with H&E to obtain standard parameter information; wherein, the standard parameter information and pathological grading results of a patient constitute the preset pathological parameter data of a patient.

8. The method for evaluating the performance of magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer as described in claim 7, characterized in that, The calculation of pathological parameters based on H&E-stained tumor tissue sections to obtain standard parameter information includes: Image recognition was performed on H&E stained tumor tissue sections to obtain cells and cell nuclei in the sections, thereby obtaining the nuclear diameter of each identified cell and the total number of identified cells; The volume-weighted average nuclear diameter is obtained based on the nuclear diameter of the identified cells and the total number of identified cells; The pathological cell density is obtained based on the total number of cells identified.

9. The performance evaluation method for magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer as described in claim 8, characterized in that, The performance analysis based on the pathological parameter data and imaging parameter information, and the resulting analysis, includes: The Mann–Whitney U test was performed on the mean data of each IMPULSED-derived microstructure parameter to obtain statistical differences between the low- and high-level groups.

10. A performance evaluation device for magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer, characterized in that, The performance evaluation device for magnetic resonance cytometry imaging in distinguishing between high-grade and low-grade bladder cancer includes: An image group acquisition module is used to acquire PGSE sequence image groups, 20Hz OGSE sequence image groups, and 40Hz OGSE sequence image groups. The preprocessing module is used to perform imaging preprocessing on the PGSE sequence image group, the 20Hz OGSE sequence image group and the 40Hz OGSE sequence image group respectively, so as to obtain the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group and the preprocessed 40Hz OGSE sequence image group. The region of interest (ROI) identification module is used to identify the ROI of the preprocessed PGSE sequence image group, the preprocessed 20Hz OGSE sequence image group, and the preprocessed 40Hz OGSE sequence image group, respectively, so as to obtain the PGSE sequence image group with ROI, the 20Hz OGSE sequence image group with ROI, and the 40Hz OGSE sequence image group with ROI. An imaging parameter information acquisition module is used to calculate imaging parameters for PGSE sequence image groups with ROI regions, 20Hz OGSE sequence image groups with ROI regions, and 40Hz OGSE sequence image groups with ROI regions, respectively, thereby acquiring imaging parameter information. The performance analysis module is used to perform performance analysis based on imaging parameter information, thereby obtaining analysis results.

Images