Intervertebral disc degeneration stage evaluation method and system based on T1rho-MR
By constructing a model based on T1-MR image, the disc degeneration stage prediction based on T1ρ-MR is realized, which solves the problem of long scanning time of T1ρ-MR in the prior art, and improves the efficiency and feasibility of disc degeneration stage evaluation.
Patent Information
- Application Number
- CN202411821869.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-11
- Publication Date
- 2025-04-11
AI Technical Summary
In the prior art, the T1ρ-MR image acquisition and post-processing methods are not uniform, and the scanning time is too long, which limits the application of intervertebral disc degeneration stage based on T1ρ-MR.
By using the lumbar spine T1-MR image samples of the labeled patients, a lumbar segment image extraction model, an eigenvalue prediction model and a stage evaluation model were constructed, and conventional T1-MR images were used for model training to achieve lumbar disc degeneration stage prediction based on T1ρ-MR.
The prediction of disc degeneration stage based on T1ρ-MR was achieved, reducing the time and money spent on clinically adding T1ρ-MR scan, and providing a feasible tool for predicting disc degeneration stage for medical institutions that do not have the conditions for T1ρ-MR scan.
Smart Images
Figure CN119811659A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical imaging detection, and in particular to a method and system for evaluating intervertebral disc degeneration staging based on T1ρ-MR. Background Art
[0002] my country has the largest elderly population in the world, and the aging of the population is further deepening. By 2040, it is estimated that the number of people over 60 years old in my country will increase to 402 million, accounting for about 28% of the total population. With the deepening of aging, the number of patients with degenerative diseases such as low back pain is also increasing day by day. Lumbar disc degeneration and a series of secondary diseases are important causes of low back pain, which seriously affect the quality of life of patients and even cause disability, bringing a heavy burden to the patient's family and society. At present, surgical treatment is the main treatment for patients with advanced disc degeneration who are ineffective for conservative treatment and whose symptoms significantly affect their daily life. However, surgical treatment is expensive and has the risk of complications. Therefore, early diagnosis and early intervention of disc degeneration, early delay or reversal of disc degeneration are of great significance to improving the quality of life of patients and reducing medical expenses.
[0003] T1ρ-MR can reflect the interaction between water and macromolecules in tissues. It can sensitively and noninvasively quantify the changes in the content of proteoglycans in the nucleus pulposus in the early stage of degeneration in the intervertebral disc, which plays an important role in the early identification of intervertebral disc degeneration. In our previous study, T1ρ-MR found that there is a rapid degeneration period in the initial stage of intervertebral disc degeneration. During this period, the T1ρ value in the nucleus pulposus decreases rapidly in a short period of time, indicating that the proteoglycans in the intervertebral disc are rapidly lost and the intervertebral disc degeneration is accelerated. After the rapid degeneration period, the T1ρ value in the nucleus pulposus slows down and stabilizes at a lower level, indicating that it has entered the late stage of degeneration. Based on this, the process of intervertebral disc degeneration can be divided into three periods according to T1ρ-MR: early degeneration (Phase 1), rapid degeneration (Phase 1, T1ρ value), and late degeneration (Phase 1). Further studies have found that injecting hydrogel repair materials into the nucleus pulposus during the rapid degeneration period can effectively alleviate intervertebral disc degeneration. On the contrary, injection in the late stage of degeneration loses its therapeutic effect. This suggests that the staging of intervertebral disc degeneration based on T1ρ-MR has guiding significance for the selection of the timing of biomaterial treatment. However, due to the inconsistency of T1ρ-MR image acquisition and post-processing methods and the long scanning time, T1ρ-MR is not widely used at present, which limits the application of T1ρ-MR-based intervertebral disc degeneration staging. Summary of the invention
[0004] The purpose of the present invention is to provide a simple and feasible method and system for evaluating the staging of intervertebral disc degeneration based on T1ρ-MR in view of the defects in the above-mentioned prior art.
[0005] In an embodiment of the present invention, a method for evaluating intervertebral disc degeneration stages based on T1ρ-MR is provided, which comprises:
[0006] Using lumbar T1-MR image samples with each lumbar segment of the patient labeled for training, a lumbar segment image extraction model for extracting each lumbar segment image from the lumbar T1-MR image is obtained;
[0007] Using samples of each lumbar segment and its T1ρ eigenvalue in the patient's lumbar T1-MR image labeled for training, a eigenvalue prediction model for predicting the T1ρ eigenvalue of each lumbar segment image is obtained;
[0008] Using samples of the T1ρ characteristic values and intervertebral disc degeneration staging assessment results of each lumbar segment of the patient to mark out the patient's each lumbar segment, a staging assessment model for outputting the intervertebral disc degeneration staging assessment results of each lumbar segment of the patient is obtained;
[0009] Inputting the patient's lumbar T1-MR image into the lumbar segment image extraction model to obtain images of each lumbar segment of the patient;
[0010] Inputting the images of each lumbar segment of the patient into the eigenvalue prediction model to obtain the T1ρ eigenvalue of each lumbar segment of the patient;
[0011] The T1ρ characteristic values of each lumbar segment of the patient are input into the staging assessment model to obtain the patient's intervertebral disc degeneration staging assessment result.
[0012] In the embodiment of the present invention, the T1ρ characteristic value includes the minimum T1ρ value, the maximum T1ρ value and the average T1ρ value of the region of interest set at the center of the intervertebral disc nucleus pulposus of the T1ρ-MR image.
[0013] In the embodiment of the present invention, the lumbar segment image extraction model adopts a YOLOv7 model based on 314 layers.
[0014] In the embodiment of the present invention, the eigenvalue prediction model adopts a 22-layer based WRN model.
[0015] In the embodiment of the present invention, the staged evaluation model adopts the SVM model.
[0016] In an embodiment of the present invention, a system for evaluating intervertebral disc degeneration stages based on T1ρ-MR is also provided, which comprises:
[0017] A lumbar segment image extraction module is used to extract each lumbar segment image from the patient's lumbar T1-MR image using a pre-trained lumbar segment image extraction model;
[0018] A T1ρ eigenvalue prediction module is used to predict the T1ρ eigenvalues corresponding to each lumbar segment image using a pre-trained eigenvalue prediction model;
[0019] The staging assessment module is used to output the intervertebral disc degeneration staging assessment results of each lumbar segment according to the T1ρ eigenvalues corresponding to the images of each lumbar segment using a pre-trained staging assessment model.
[0020] In the embodiment of the present invention, the T1ρ characteristic value includes the minimum T1ρ value, the maximum T1ρ value and the average T1ρ value of the region of interest set at the center of the intervertebral disc nucleus pulposus of the T1ρ-MR image.
[0021] In the embodiment of the present invention, the lumbar segment image extraction model adopts a YOLOv7 model based on 314 layers.
[0022] In the embodiment of the present invention, the eigenvalue prediction model adopts a 22-layer based WRN model.
[0023] In the embodiment of the present invention, the staged evaluation model adopts the SVM model.
[0024] Compared with the prior art, the T1ρ-MR-based intervertebral disc degeneration staging assessment method and system of the present invention, by collecting the eigenvalues of conventional T1-MR images and corresponding T1ρ-MR images for model training, constructs a eigenvalue prediction model that can predict the T1ρ eigenvalues of each lumbar segment image, and adopts the T1-MR images routinely used in clinical lumbar examinations to achieve T1ρ-MR-based lumbar disc degeneration staging prediction, providing a new method for the convenient application of T1ρ-MR-based intervertebral disc degeneration staging, reducing the extra time and money spent on adding T1ρ-MR scanning in clinical practice, and providing a feasible tool for predicting T1ρ-MR-based intervertebral disc degeneration staging for medical institutions that do not have T1ρ-MR scanning conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 It is a flow chart of a method for evaluating intervertebral disc degeneration staging based on T1ρ-MR according to an embodiment of the present invention.
[0026] Figure 2 It is a schematic diagram of the structure of the intervertebral disc degeneration staging assessment system based on T1ρ-MR according to an embodiment of the present invention. DETAILED DESCRIPTION
[0027] like Figure 1As shown, in an embodiment of the present invention, a method for evaluating the staging of intervertebral disc degeneration based on T1ρ-MR is provided, using conventional T1-MR images, and through deep learning technology, a model is constructed and the trained model is used to predict the staging of lumbar disc degeneration based on T1ρ-MR. The method needs to construct a model and use the trained model to predict the staging of lumbar disc degeneration, and the method includes steps S1-S6. They are described below.
[0028] Step S1: training using labeled lumbar T1-MR image samples of patients to obtain a lumbar segment image extraction model for extracting images of each lumbar segment in the lumbar T1-MR images.
[0029] In step S1, the lumbar segment image extraction model adopts a YOLOv7 model based on 314 layers. LabelImg software can be used to outline the five lumbar intervertebral disc areas of L1 / L2-L5 / S1 of each patient from top to bottom in the T1-MR image in PNG format using a rectangular frame target selection tool, and mark the corresponding intervertebral disc segments as training samples to train the lumbar segment image extraction model. After the training is completed, the lumbar segment image extraction model automatically locates the lumbar intervertebral disc area in the T1-MR image in the form of a rectangular frame, and outputs each lumbar intervertebral disc area image and its segment according to the set image size.
[0030] Step S2: training using samples of T1ρ eigenvalues of each lumbar segment in the labeled patient's lumbar T1-MR images and the patient's lumbar T1ρ-MR images to obtain a eigenvalue prediction model for predicting the T1ρ eigenvalues of each lumbar segment image.
[0031] In step S2, the feature value prediction model adopts a 22-layer WRN model. The model consists of 1 input layer, 22 2D convolutional layers, 19 batch normalization layers, 18 activation layers, 9 Add fusion layers, 1 2D average pooling layer, 1 flattening layer, and 1 fully connected layer. After extracting features from the T1-MR intervertebral disc image, the model outputs the predicted maximum T1ρ value, minimum T1ρ value, and average T1ρ value.
[0032] It should be noted that, in the embodiment of the present invention, the T1ρ characteristic value includes the minimum T1ρ value, the maximum T1ρ value and the average T1ρ value of the region of interest (e.g., 1 cm2 of the center of the nucleus pulposus of the intervertebral disc) set in the T1ρ-MR image. When acquiring samples, not only the T1-MR images of the sampled population are scanned, but also the corresponding T1ρ-MR images. The corresponding characteristic values are obtained in the T1ρ-MR images. The corresponding lumbar segments in the T1-MR intervertebral disc images corresponding to these characteristic values are marked, and can be used as training samples to train the characteristic value prediction model. After the training is completed, the characteristic value prediction model can obtain the T1ρ characteristic values of each lumbar segment of the patient based on the images of each lumbar segment of the patient.
[0033] Step S3: training with samples of the labeled T1ρ characteristic values of each lumbar segment of the patient and the intervertebral disc degeneration staging assessment results to obtain a staging assessment model of the T1ρ characteristic values of each lumbar segment of the patient and the intervertebral disc degeneration staging assessment results.
[0034] Doctors can divide each patient's L1 / L2-L5 / S1 lumbar intervertebral disc into three stages according to the T1ρ feature value: early degeneration stage (Phase 1), rapid degeneration stage (Phase 2), and late degeneration stage (Phase 3), and then mark them in the training samples, which can be used as samples to train the staging assessment model.
[0035] It should be noted that the lumbar segment image extraction model, the eigenvalue prediction model and the staging assessment model can be trained separately or combined into a comprehensive model for comprehensive training. All required training parameters can be marked in the T1-MR image of the training sample, which can be used as the training sample of the comprehensive model for training.
[0036] Step S4: inputting the patient's lumbar T1-MR image into the lumbar segment image extraction model to obtain images of each lumbar segment of the patient.
[0037] Step S5: inputting the images of each lumbar segment of the patient into the eigenvalue prediction model to obtain the T1ρ eigenvalue of each lumbar segment of the patient.
[0038] Step S6: inputting the T1ρ characteristic value of each lumbar segment of the patient into the staging assessment model to obtain the patient's intervertebral disc degeneration staging assessment result.
[0039] After the model training is completed, the patient's lumbar T1-MR image can be sequentially passed through the lumbar segment image extraction model, the eigenvalue prediction model and the staging assessment model to finally obtain the patient's intervertebral disc degeneration staging assessment result.
[0040] like Figure 2 As shown, corresponding to the above-mentioned T1ρ-MR-based intervertebral disc degeneration staging assessment method, in an embodiment of the present invention, a T1ρ-MR-based intervertebral disc degeneration staging assessment system is also provided, which includes a lumbar segment image extraction module 1, a T1ρ eigenvalue prediction module 2 and a staging assessment module 3.
[0041] The lumbar segment image extraction module 1 is used to extract each lumbar segment image in the patient's lumbar T1-MR image using a pre-trained lumbar segment image extraction model. In an embodiment of the present invention, the lumbar segment image extraction model uses a 314-layer YOLOv7 model.
[0042] The T1ρ eigenvalue prediction module 2 is used to predict the T1ρ eigenvalues corresponding to each lumbar segment image using a pre-trained eigenvalue prediction model. In an embodiment of the present invention, the T1ρ eigenvalues include the minimum T1ρ value, the maximum T1ρ value, and the average T1ρ value of the region of interest set in the center of the intervertebral disc nucleus pulposus of the T1ρ-MR image. The eigenvalue prediction model adopts a 22-layer WRN model.
[0043] The staging assessment module 3 is used to output the staging assessment results of intervertebral disc degeneration of each lumbar segment according to the T1ρ eigenvalues corresponding to each lumbar segment image using a pre-trained staging assessment model. In the embodiment of the present invention, the staging assessment model adopts an SVM model.
[0044] In summary, the T1ρ-MR-based intervertebral disc degeneration staging assessment method and system of the present invention, by collecting the eigenvalues of conventional T1-MR images and corresponding T1ρ-MR images for model training, constructs a eigenvalue prediction model that can predict the T1ρ eigenvalues of each lumbar segment image, and adopts the T1-MR images routinely used in clinical lumbar examinations to achieve T1ρ-MR-based lumbar disc degeneration staging prediction, providing a new method for the convenient application of T1ρ-MR-based intervertebral disc degeneration staging, reducing the extra time and money spent on adding T1ρ-MR scanning in clinical practice, and providing a feasible tool for predicting T1ρ-MR-based intervertebral disc degeneration staging for medical institutions that do not have T1ρ-MR scanning conditions.
[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for evaluating the staging of intervertebral disc degeneration based on T1ρ-MR, characterized in that: include: Using lumbar T1-MR image samples with each lumbar segment of the patient labeled for training, a lumbar segment image extraction model for extracting each lumbar segment image from the lumbar T1-MR image is obtained; Using samples of each lumbar segment and its T1ρ eigenvalue in the patient's lumbar T1-MR image labeled for training, a eigenvalue prediction model for predicting the T1ρ eigenvalue of each lumbar segment image is obtained; Using samples of the T1ρ characteristic values and intervertebral disc degeneration staging assessment results of each lumbar segment of the patient to mark out the patient's each lumbar segment, a staging assessment model for outputting the intervertebral disc degeneration staging assessment results of each lumbar segment of the patient is obtained; Inputting the patient's lumbar T1-MR image into the lumbar segment image extraction model to obtain images of each lumbar segment of the patient; Inputting the images of each lumbar segment of the patient into the eigenvalue prediction model to obtain the T1ρ eigenvalue of each lumbar segment of the patient; The T1ρ characteristic values of each lumbar segment of the patient are input into the staging assessment model to obtain the patient's intervertebral disc degeneration staging assessment result.
2. The method for evaluating intervertebral disc degeneration staging based on T1ρ-MR according to claim 1, characterized in that: The T1ρ characteristic values include a minimum T1ρ value, a maximum T1ρ value and an average T1ρ value of a region of interest set at the center of the intervertebral disc nucleus pulposus of the T1ρ-MR image.
3. The method for evaluating intervertebral disc degeneration staging based on T1ρ-MR according to claim 1, characterized in that: The lumbar segment image extraction model adopts a YOLOv7 model based on 314 layers.
4. The method for evaluating intervertebral disc degeneration staging based on T1ρ-MR according to claim 1, characterized in that: The eigenvalue prediction model adopts a 22-layer based WRN model.
5. The method for evaluating intervertebral disc degeneration staging based on T1ρ-MR according to claim 1, characterized in that: The staging assessment model adopts the SVM model.
6. A T1ρ-MR-based intervertebral disc degeneration staging assessment system, characterized in that: include: A lumbar segment image extraction module is used to extract each lumbar segment image from the patient's lumbar T1-MR image using a pre-trained lumbar segment image extraction model; A T1ρ eigenvalue prediction module is used to predict the T1ρ eigenvalues corresponding to each lumbar segment image using a pre-trained eigenvalue prediction model; The staging assessment module is used to output the intervertebral disc degeneration staging assessment results of each lumbar segment according to the T1ρ eigenvalues corresponding to the images of each lumbar segment using a pre-trained staging assessment model.
7. The intervertebral disc degeneration staging assessment system based on T1ρ-MR as claimed in claim 6, characterized in that: The T1ρ characteristic values include a minimum T1ρ value, a maximum T1ρ value and an average T1ρ value of a region of interest set at the center of the intervertebral disc nucleus pulposus of the T1ρ-MR image.
8. The intervertebral disc degeneration staging assessment system based on T1ρ-MR as claimed in claim 1, characterized in that: The lumbar segment image extraction model adopts a YOLOv7 model based on 314 layers.
9. The intervertebral disc degeneration staging assessment system based on T1ρ-MR according to claim 1, characterized in that: The eigenvalue prediction model adopts a 22-layer based WRN model.
10. The intervertebral disc degeneration staging assessment system based on T1ρ-MR according to claim 1, characterized in that: The staging assessment model adopts the SVM model.