Crohn's disease intestinal fibrosis condition early warning method based on large model driving

Abstract

This invention relates to the field of intelligent medical technology, and more particularly to a method for early warning of intestinal fibrosis in Crohn's disease based on a large model. The method includes: acquiring intestinal MRI image sequences of Crohn's disease patients; inputting the intestinal MRI image sequences into a pre-trained large model for intestinal structure analysis to locate the initial intestinal fibrosis region; determining whether the degree of intestinal luminal stenosis in Crohn's disease patients meets the standard based on the intestinal stenosis progression index, and adjusting the feature contrast threshold of the large model based on the difference; determining whether the acute inflammation masking effect of Crohn's disease patients is adequate based on intestinal wall edema characterization parameters, and adjusting the T2-weighted imaging modal weights of the large model based on the relative difference; and determining whether the patency of the fibrotic permeation channel in Crohn's disease patients is adequate based on the fibrotic permeation resistivity, and adjusting the boundary loss weights of the large model based on the ratio. This invention improves the accuracy of early warning of intestinal fibrosis in Crohn's disease.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical technology, and in particular to a method for early warning of Crohn's disease intestinal fibrosis based on a large model. Background Technology

[0002] Crohn's disease is a chronic, relapsing inflammatory bowel disease, and intestinal fibrosis is a common complication. It is mainly characterized by excessive deposition of fibrotic materials such as collagen in the intestinal wall, which can lead to intestinal stenosis, obstruction, and even perforation, severely impacting patients' quality of life and increasing surgical risks. Early and accurate prediction of the progression of intestinal fibrosis is of significant clinical importance for adjusting treatment plans and avoiding irreversible damage. Currently, clinical assessment of the degree of fibrosis mainly relies on intestinal MRI imaging; however, accuracy is affected by various factors. Fibrosis often coexists with acute inflammation, which may obscure fibrotic features, leading to misdiagnosis. Fibrosis is a dynamic process, and static images cannot reflect its temporal trends. Blurred boundaries of fibrotic areas and altered interstitial flow characteristics also affect the accuracy of traditional imaging analysis methods.

[0003] Chinese Patent Application Publication No. CN114067154A discloses a method and related equipment for grading Crohn's disease fibrosis based on multi-sequence MRI. The method includes obtaining the first radiomics features of intestinal fibrous tissue in each of several abdominal MRI images of the same target object by controlling the radiomics feature module, fusing the first radiomics features to obtain fused radiomics features, and controlling the classification module to determine the Crohn's disease fibrosis category of the target object based on the fused radiomics features.

[0004] However, the existing technology has the following problems: feature extraction and classification are based only on static multi-sequence images and rely on morphological features, which leads to insufficient comprehensiveness in the identification of Crohn's disease intestinal fibrosis, resulting in low accuracy in the early warning of Crohn's disease intestinal fibrosis. Summary of the Invention

[0005] To address this, the present invention provides a large model-driven early warning method for Crohn's disease intestinal fibrosis, which overcomes the problem that existing technologies rely solely on static multi-sequence images for feature extraction and classification, depending on morphological features, resulting in insufficient comprehensiveness in identifying Crohn's disease intestinal fibrosis and thus low accuracy in early warning of Crohn's disease intestinal fibrosis.

[0006] To achieve the above objectives, this invention provides a method for early warning of Crohn's disease intestinal fibrosis based on a large model, comprising:

[0007] Obtain intestinal MRI image sequences from Crohn's disease patients;

[0008] The intestinal MRI image sequence is input into a pre-trained large intestinal structure analysis model to locate the initial intestinal fibrosis region.

[0009] The intestinal stenosis progression index is determined based on intestinal MRI image sequences to determine whether the degree of intestinal stenosis in Crohn's disease patients meets the target, and the feature contrast threshold of the large model is adjusted based on the difference between the dynamic progression index of intestinal stenosis and the preset stenosis threshold.

[0010] The parameters for characterizing intestinal wall edema were determined by T2-weighted imaging in intestinal MRI sequences to determine whether the acute inflammation masking effect in Crohn's disease patients was adequate. The T2-weighted imaging modal weights of the large model were adjusted based on the relative difference between the parameters for characterizing intestinal wall edema and the preset parameters for characterizing intestinal wall edema.

[0011] The fibrotic flow resistance was determined based on intestinal MRI image sequences to determine whether the patency of the fibrotic flow channels in Crohn's disease patients was acceptable, and the boundary loss weights of the large model were adjusted based on the ratio of the preset fibrotic flow resistance to the fibrotic flow resistance.

[0012] Furthermore, the determination of the degree of intestinal stenosis in the Crohn's disease patients not meeting the standard is based on the comparison results of the intestinal stenosis progression index being greater than the preset intestinal stenosis progression index.

[0013] Furthermore, the process of determining the intestinal stenosis progression index includes:

[0014] The time when Crohn's disease patients first underwent intestinal MRI scans was recorded as the baseline period. Subsequent MRI scans were performed according to the clinical follow-up plan, and a total of n effective images were acquired to form a time series.

[0015] In baseline MRI images, the minimum lumen diameter of the stenotic segment is measured along the central axis of the diseased intestinal segment using the U-Net model and recorded as the baseline diameter.

[0016] For each subsequent MRI image, the minimum lumen diameter of the stenotic segment at the corresponding time point was repeatedly measured and recorded as the temporal diameter; the difference between the temporal diameter and the baseline diameter at each time point was divided by the baseline diameter and recorded as the relative diameter change rate.

[0017] The relative changes at each time point are weighted, and the time weight is linearly positively correlated with the time interval. The time weight is the ratio of the time interval from the baseline period to the total monitoring duration.

[0018] The intestinal stenosis progression index is obtained by multiplying the relative diameter change rate at each time point by the corresponding time weight, summing the results, and then dividing by the sum of the time weights.

[0019] Furthermore, the process of adjusting the feature comparison threshold of the large model includes:

[0020] Calculate the difference between the intestinal stenosis progression index and the preset intestinal stenosis progression index under the condition that the degree of intestinal stenosis in Crohn's disease patients does not meet the target;

[0021] Based on the comparison results where the difference is less than or equal to a preset difference, a first preset feature adjustment coefficient is determined to reduce the feature comparison threshold.

[0022] Based on the comparison results where the difference is greater than a preset difference, a second preset feature adjustment coefficient is determined to reduce the feature comparison threshold.

[0023] Furthermore, the failure of the acute inflammation masking effect in the Crohn's disease patients is determined based on the comparison results of the intestinal wall edema characterization parameter being greater than the preset intestinal wall edema characterization parameter.

[0024] Furthermore, the process of determining the parameters characterizing intestinal wall edema includes:

[0025] The T2WI sequence is denoised using a nonlocal mean denoising algorithm. The normalized pixel gray value is the ratio of the denoised image gray value minus the global minimum gray value in the current image to the global maximum gray value minus the global minimum gray value in the current image.

[0026] The grayscale threshold of the high signal region of edema is determined by the Otsu adaptive thresholding method. By maximizing the inter-class variance, the region with a normalized pixel grayscale value greater than or equal to the grayscale threshold is recorded as the edema region, and the region with a normalized pixel grayscale value less than the grayscale threshold is recorded as the normal region.

[0027] Extract the grayscale values ​​of all pixels in the edema area, and record the ratio of the number of pixels corresponding to each grayscale level to the total number of pixels in the edema area as the grayscale probability of the edema area.

[0028] Calculate the information entropy of the edema region based on the gray-level probability of the edema region;

[0029] The ratio of the information entropy of the edema region to the theoretical maximum value of the information entropy of the edema region is the characteristic parameter of intestinal wall edema.

[0030] Furthermore, the process of adjusting the T2-weighted imaging modal weights of the large model includes:

[0031] Calculate the relative difference between the intestinal wall edema characterization parameter and the preset intestinal wall edema characterization parameter under the condition that the acute inflammation masking effect is not qualified in Crohn's disease patients;

[0032] Based on the comparison results where the relative difference is less than or equal to the preset relative difference, the first preset model weight adjustment coefficient is determined to reduce the T2 weighted imaging modality weight;

[0033] Based on the comparison results where the relative difference is greater than the preset relative difference, the second preset model weight adjustment coefficient is determined to reduce the T2 weighted imaging modality weight.

[0034] Furthermore, the patency of the fibrotic permeation channel in the Crohn's disease patient is deemed unqualified based on a comparison result where the fibrotic permeation resistance is greater than a preset fibrotic permeation resistance.

[0035] Further, the fibrosis permeation resistance is the result of multiplying the reciprocal of the product of the permeation rate constant and the interstitial space volume fraction by the ratio of the mean of the permeation rate constant to the standard deviation of the permeation rate constant by the ratio of the mean of the interstitial space volume fraction to the standard deviation of the interstitial space volume fraction, wherein the permeation rate constant is the diffusion rate of the contrast agent from plasma to the interstitial space, and the interstitial space volume fraction is the proportion of the extracellular space in the fibrotic region to the total volume.

[0036] Furthermore, the process of adjusting the boundary loss weights of the large model includes:

[0037] Calculate the ratio of the preset fibrotic flow resistance to the fibrotic flow resistance under the condition that the patency of the fibrotic flow channel is not up to standard in Crohn's disease patients.

[0038] Based on the comparison results where the ratio is less than or equal to a preset ratio, a first preset boundary loss adjustment coefficient is determined to increase the boundary loss weight of the large model;

[0039] Based on the comparison results where the ratio is greater than a preset ratio, a second preset boundary loss adjustment coefficient is determined to increase the boundary loss weight of the large model.

[0040] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention acquires intestinal MRI image sequences from Crohn's disease patients, inputs them into a pre-trained large-scale model to locate the initial fibrotic region, and combines three indicators—intestinal stenosis progression index, intestinal wall edema characterization parameter, and fibrotic permeability—to evaluate the achievement of intestinal stenosis targets, the qualification of acute inflammation masking effect, and the patency of permeability channels, respectively. Furthermore, it dynamically adjusts the feature contrast threshold, T2-weighted imaging modality weight, and boundary loss weight of the large-scale model to achieve precise early warning of the disease, improving the accuracy of fibrotic region localization. This is achieved by adjusting the feature contrast threshold and the T2-weighted imaging modality. The addition of weights and boundary loss weights improves the adaptability of the large model to the pathological states of different patients, enhances the robustness of the early warning results, and indicates that an excessively high intestinal stenosis progression index suggests a risk of fibrosis progression. Excessive edema parameters may mask the true extent of fibrosis, and increased osmotic resistance reflects impaired blood supply to the fibrotic area. This allows for a more comprehensive assessment of disease severity, reduces the rate of missed diagnoses in clinical validation, and links clinical indicators with large model parameters, shortening the time from image acquisition to clinical decision-making. This improves the comprehensiveness and real-time nature of early warning for Crohn's disease intestinal fibrosis, thereby enhancing the accuracy of early warning for Crohn's disease intestinal fibrosis.

[0041] Furthermore, this invention uses an intestinal stenosis progression index to determine whether the stenosis has reached the target level. If it does not, the feature comparison threshold of the large model is reduced, and multi-time point fusion is used to match the pathological mechanism, which improves the accuracy of the assessment, strengthens the model's ability to identify rapidly progressing fibrosis, avoids model instability caused by drastic parameter fluctuations, improves the model's diagnostic accuracy for patients with different progression rates, reduces the rate of fibrosis progression, and improves the accuracy of assessing intestinal stenosis in Crohn's disease, thereby improving the accuracy of early warning of intestinal fibrosis in Crohn's disease.

[0042] Furthermore, this invention uses intestinal wall edema characterization parameters to compare with preset intestinal wall edema characterization parameters to determine whether the acute inflammation masking effect is qualified. If it is not qualified, the T2-weighted imaging modal weight is reduced to achieve dynamic suppression of edema interference by the model. This enables the identification of fibrosis under the acute inflammation masking effect of Crohn's disease. The information entropy of the edema area reflects the disorder of the pixel gray-level distribution in the edema area. The ratio to the theoretical maximum value quantifies the interference intensity of edema. Edema interference is suppressed as needed to avoid excessively weakening the fibrosis identification ability and improve the robustness of the large model. T1CE enhanced scanning is sensitive to abnormal blood supply and reduces the misjudgment of fibrosis by high signal in the edema area. Delayed scanning reflects the integrity of the tissue barrier. The increased weight eliminates edema interference and improves the accuracy of identifying fibrosis masked by edema.

[0043] Furthermore, this invention determines the patency of fibrotic seepage channels by assessing the fibrotic seepage resistance. If the channels are obstructed, the boundary loss weight of the large model is increased to enable the model to accurately identify the boundaries of the seepage channels. A slight increase in boundary loss can guide the model to focus on subtle changes in the seepage channels, while a significant increase in boundary loss can force the model to focus on the edge contour of the obstructed area, thereby improving the contour accuracy of the boundary and enhancing the accuracy of the large model in identifying different degrees of obstruction. This improves the accuracy of assessing the patency of fibrotic seepage channels in Crohn's disease, and thus enhances the accuracy of early warning of intestinal fibrosis in Crohn's disease. Attached Figure Description

[0044] Figure 1 This is a flowchart of a method for early warning of Crohn's disease intestinal fibrosis based on a large model, according to an embodiment of the present invention.

[0045] Figure 2 This is a flowchart illustrating how to determine whether the degree of intestinal luminal stenosis in patients with Crohn's disease meets the target criteria, as described in this invention.

[0046] Figure 3 This is a flowchart illustrating the process of determining whether the acute inflammation shielding effect in Crohn's disease patients is satisfactory, as part of an embodiment of the present invention.

[0047] Figure 4 This is a flowchart illustrating how to determine the patency of fibrotic drainage channels in Crohn's disease patients according to an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0049] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0050] It should be noted that the data in this embodiment are all derived from a comprehensive analysis and evaluation of historical test data and corresponding historical test results from the three months prior to this test. Those skilled in the art will understand that the determination of the above-mentioned parameters for any single item in this invention can be achieved by selecting the value with the highest percentage based on the data distribution as the preset standard parameter, using weighted summation to obtain the value as the preset standard parameter, substituting each historical data point into a specific formula and using the value obtained from that formula as the preset standard parameter, or other selection methods, as long as the invention can clearly define different specific situations in the single-item judgment process through the obtained values.

[0051] Please see Figure 1 The flowchart shown is a process for early warning of Crohn's disease intestinal fibrosis based on a large model driven method according to an embodiment of the present invention.

[0052] This invention provides a large-model-driven early warning method for Crohn's disease intestinal fibrosis, comprising:

[0053] Step S1: Obtain intestinal MRI image sequences from Crohn's disease patients;

[0054] Step S2: Input the intestinal MRI image sequence into the pre-trained large intestinal structure analysis model to locate the initial intestinal fibrosis region.

[0055] Step S3: Determine the intestinal stenosis progression index based on the intestinal MRI image sequence to determine whether the degree of intestinal stenosis in Crohn's disease patients meets the target, and determine the feature contrast threshold of the large model based on the difference between the dynamic progression index of intestinal stenosis and the preset stenosis threshold.

[0056] Step S4: Determine the intestinal wall edema characterization parameters based on T2-weighted imaging in intestinal MRI image sequences to determine whether the acute inflammation masking effect of Crohn's disease patients is qualified, and determine the adjustment of the T2-weighted imaging modal weights of the large model based on the relative difference between the intestinal wall edema characterization parameters and the preset intestinal wall edema characterization parameters.

[0057] Step S5: Determine the fibrotic permeation resistance based on the intestinal MRI image sequence to determine whether the patency of the fibrotic permeation channel in Crohn's disease patients is acceptable, and determine the boundary loss weight of the large model based on the ratio of the preset fibrotic permeation resistance to the fibrotic permeation resistance.

[0058] Specifically, this invention acquires intestinal MRI image sequences from Crohn's disease patients, inputs them into a pre-trained large-scale model to locate the initial fibrotic region, and combines three indicators—intestinal stenosis progression index, intestinal wall edema characterization parameter, and fibrosis osmotic resistance—to evaluate the achievement of intestinal stenosis targets, the qualification of acute inflammation masking effect, and the patency of osmotic channels, respectively. Furthermore, it dynamically adjusts the feature contrast threshold, T2-weighted imaging modal weights, and boundary loss weights of the large-scale model to achieve precise early warning of the disease and improve the accuracy of fibrotic region localization. Weighting improves the adaptability of the large model to the pathological states of different patients, enhances the robustness of early warning results, and indicates that an excessively high intestinal stenosis progression index suggests a risk of fibrosis progression. Excessive edema characterization parameters may mask the true extent of fibrosis, and increased osmotic resistance reflects impaired blood supply to the fibrotic area. This allows for a more comprehensive assessment of disease severity, reduces the rate of missed diagnoses in clinical validation, and links clinical indicators with large model parameters. This shortens the time from image acquisition to clinical decision-making, improves the comprehensiveness and real-time nature of early warning for Crohn's disease intestinal fibrosis, and thus improves the accuracy of early warning for Crohn's disease intestinal fibrosis.

[0059] In this embodiment of the invention, the intestinal MRI imaging sequence includes T2-weighted imaging (T2WI), T1-weighted contrast-enhanced scanning (T1CE), and delayed phase scanning sequence (LATE).

[0060] Specifically, the intestinal MRI image sequence is input into a pre-trained large-scale intestinal structure analysis model. The large-scale model outputs a segmentation mask of the fibrotic region by comparing the image features of the diseased intestinal segment and the normal intestinal segment, and marks the spatial distribution range of fibrosis.

[0061] Specifically, the large model registers T2WI, T1CE, and LATE sequences to the same spatial coordinate system through affine transformation, eliminating spatial offset caused by intestinal peristalsis; each image sequence is individually Z-score normalized. The encoder consists of 5 layers, each containing 2 3D convolutional blocks; layer 1 has 3 input channels, a 3×3×3 convolutional kernel, and an output feature map size half that of the original image, with 64 channels; layers 2-5 double the number of output channels with each layer, and the spatial resolution is halved at each level; a max-pooling layer is added after each convolutional block. The decoder is symmetrical to the encoder, restoring spatial resolution through transposed convolution upsampling, while fusing the skip connection features of the encoder. Each layer uses 3D transposed convolution, halving the number of channels and doubling the spatial resolution at each level; the feature maps of the corresponding layers of the encoder are concatenated with the upsampled feature maps of the decoder. After skip connections in each layer of the decoder, a CBAM module is introduced to enhance the feature response of the lesion area. Channel weights are generated by global average pooling and max pooling via MLP to suppress irrelevant modalities. Based on the feature map output by channel attention, spatial weights are generated through 7×7×7 convolution to focus on areas of intestinal wall thickening or abnormal signal. The decoder finally outputs a 1×H×W×D feature map, which generates a probability map through a sigmoid activation function.

[0062] Specifically, the loss function parameters of the large model include Dice Loss weight, with a value of 0.7, FocalLoss weight, with a value of 0.3, boundary loss weight, with a value of 0.1, feature comparison threshold, with a value of 0.25, and T2WI modality weight, T1CE modality weight, and LATE modality weight, all equally divided and 0.33.

[0063] Please see Figure 2 The flowchart shown is a process for determining whether the degree of stenosis of the intestinal lumen in a Crohn's disease patient meets the standard according to an embodiment of the present invention.

[0064] Specifically, in this embodiment of the invention, the degree of intestinal stenosis in Crohn's disease patients is determined by comparing the intestinal stenosis progression index determined by the MRI image sequence with the preset intestinal stenosis progression index.

[0065] When the intestinal stenosis progression index is less than or equal to the preset intestinal stenosis progression index, the degree of intestinal stenosis in Crohn's disease patients is determined to be within the target range.

[0066] When the intestinal stenosis progression index is greater than the preset intestinal stenosis progression index, it is determined that the degree of intestinal stenosis in Crohn's disease patients does not meet the standard.

[0067] In this embodiment of the invention, the preset range of the intestinal stenosis progression index is [0.1, 0.2], preferably 0.15, but the above value is not limited to this, and those skilled in the art can adjust the value according to actual needs.

[0068] In this embodiment of the invention, the process of obtaining the intestinal stenosis progression index is as follows: the time when the patient first undergoes an intestinal MRI scan is recorded as the baseline period. Subsequent MRI scans are performed according to the clinical follow-up plan, acquiring a total of n effective images to form a time series. In the baseline MRI images, the minimum lumen diameter of the stenotic segment is measured along the central axis of the diseased intestinal segment using a U-Net model and recorded as the baseline diameter. For each subsequent MRI image, the minimum lumen diameter of the stenotic segment at the corresponding time point is repeatedly measured and recorded as the temporal diameter. The difference between the temporal diameter and the baseline diameter at each time point is divided by the baseline diameter and recorded as the relative diameter change rate. The relative changes at each time point are weighted, with the time weight being linearly positively correlated with the time interval. The time weight is the ratio of the time interval from the baseline period to the total monitoring duration. The relative diameter change rate at each time point is multiplied by the corresponding time weight, summed, and then divided by the sum of the time weights to obtain the intestinal stenosis progression index.

[0069] Specifically, in this embodiment of the invention, when it is determined that the degree of intestinal stenosis in a Crohn's disease patient does not meet the standard, the feature comparison threshold of the large model is determined based on the comparison result of the difference between the intestinal stenosis progression index and the preset intestinal stenosis progression index and the preset difference.

[0070] When the difference is less than or equal to the preset difference, it is determined that the feature comparison threshold will be reduced to the corresponding value by the first preset feature adjustment coefficient of 0.97.

[0071] When the difference is greater than the preset difference, it is determined that the feature comparison threshold will be reduced to the corresponding value by the second preset feature adjustment coefficient of 0.93;

[0072] The difference is the difference between the intestinal stenosis progression index and the preset intestinal stenosis progression index.

[0073] In this embodiment of the invention, the preset difference value range is [0.05, 1.05], preferably 1.02, but the above value is not limited to this, and those skilled in the art can adjust the value according to actual needs.

[0074] In this embodiment of the invention, the reduced feature comparison threshold is the product of the feature comparison threshold and the preset feature adjustment coefficient. The preset feature adjustment coefficient includes a first preset feature adjustment coefficient with a value of 0.97 and a second preset feature adjustment coefficient with a value of 0.93. In order to ensure that the adjusted feature comparison threshold meets the actual needs, the adjustment range should not be too large. Therefore, an adjustment coefficient is set to control the adjustment range.

[0075] Specifically, this invention uses an intestinal stenosis progression index to determine whether stenosis has reached a certain level. If it does not, the feature comparison threshold of the large model is reduced, and multi-time point fusion is used to match the pathological mechanism, thereby improving the accuracy of the assessment, strengthening the model's ability to identify rapidly progressing fibrosis, avoiding model instability caused by drastic parameter fluctuations, improving the model's diagnostic accuracy for patients with different progression rates, reducing the rate of fibrosis progression, and improving the accuracy of assessing intestinal stenosis in Crohn's disease, thus improving the accuracy of early warning of intestinal fibrosis in Crohn's disease.

[0076] Please see Figure 3 As shown, it is a flowchart for determining whether the acute inflammation shielding effect of Crohn's disease patients is qualified according to an embodiment of the present invention.

[0077] Specifically, in this embodiment of the invention, the acute inflammation masking effect of Crohn's disease patients is determined based on the comparison between the intestinal wall edema characterization parameters determined by the MRI image sequence and the preset intestinal wall edema characterization parameters.

[0078] When the intestinal wall edema characterization parameter is less than or equal to the preset intestinal wall edema characterization parameter, the acute inflammation shielding effect of Crohn's disease patients is deemed qualified.

[0079] When the intestinal wall edema characterization parameter is greater than the preset intestinal wall edema characterization parameter, the acute inflammation masking effect of Crohn's disease patients is determined to be unqualified.

[0080] In this embodiment of the invention, the preset range of the intestinal wall edema characterization parameter is [0.75, 0.85], preferably 0.8, but the above value is not limited to this, and those skilled in the art can also adjust the value according to actual needs.

[0081] In this embodiment of the invention, the process of obtaining the intestinal wall edema characterization parameter is as follows: The T2WI sequence is denoised using a nonlocal mean denoising algorithm. The ratio of the denoised image gray value minus the current image's global minimum gray value to the current image's global maximum gray value minus the current image's global minimum gray value is recorded as the normalized pixel gray value. The Otsu adaptive thresholding method is used to determine the gray value threshold of the edema high-signal region. By maximizing the inter-class variance, regions with normalized pixel gray values ​​greater than or equal to the gray value threshold are recorded as edema regions, and regions with normalized pixel gray values ​​less than the gray value threshold are recorded as normal regions. All pixel gray values ​​within the edema region are extracted, and the ratio of the number of pixels corresponding to each gray level to the total number of pixels in the edema region is recorded as the edema region gray probability. Based on the edema region gray probability, the edema region information entropy is calculated. The ratio of the edema region information entropy to the theoretical maximum value of the edema region information entropy is the intestinal wall edema characterization parameter.

[0082] Specifically, in this embodiment of the invention, under the condition that the acute inflammation masking effect of Crohn's disease patients is unqualified, the T2-weighted imaging modal weights of the large model are determined to be adjusted based on the comparison results of the relative difference between the intestinal wall edema characterization parameter and the preset intestinal wall edema characterization parameter and the preset relative difference.

[0083] When the relative difference is less than or equal to the preset relative difference, it is determined that the T2 weighted imaging modal weight will be reduced to the corresponding value by the first preset model weight adjustment coefficient of 0.95.

[0084] When the relative difference is greater than the preset relative difference, it is determined that the T2 weighted imaging modal weight will be reduced to the corresponding value by the second preset model weight adjustment coefficient of 0.91.

[0085] The relative difference is the relative difference between the intestinal wall edema characterization parameter and the preset intestinal wall edema characterization parameter.

[0086] In this embodiment of the invention, the preset relative difference range is [0.05, 0.10], preferably 0.07, but the above value is not limited to this, and those skilled in the art can adjust the value according to actual needs.

[0087] In this embodiment of the invention, the reduced T2-weighted imaging modal weight is the product of the T2-weighted imaging modal weight and the preset model weight adjustment coefficient. The preset model weight adjustment coefficient includes a first preset model weight adjustment coefficient with a value of 0.95 and a second preset model weight adjustment coefficient with a value of 0.91. To ensure that the adjusted T2-weighted imaging modal weight meets the actual needs, the adjustment range should not be too large. Therefore, an adjustment coefficient is set to control the adjustment range. After reducing the T2-weighted imaging modal weight, the T1CE modal weight and the LATE modal weight are increased equally.

[0088] Specifically, this invention uses intestinal wall edema characterization parameters to compare with preset intestinal wall edema characterization parameters to determine whether the acute inflammation masking effect is qualified. If it is not qualified, the T2-weighted imaging modal weight is reduced to achieve dynamic suppression of edema interference by the model. This enables the identification of fibrosis under the acute inflammation masking effect of Crohn's disease. The information entropy of the edema area reflects the disorder of the pixel gray-level distribution in the edema area. The ratio of the entropy to the theoretical maximum value quantifies the interference intensity of edema. Edema interference is suppressed as needed to avoid excessively weakening the fibrosis identification ability and improve the robustness of the large model. T1CE enhanced scanning is sensitive to abnormal blood supply and reduces the misjudgment of fibrosis by high signal in the edema area. Delayed scanning reflects the integrity of the tissue barrier. The increased weight eliminates edema interference and improves the accuracy of identifying fibrosis masked by edema.

[0089] Please see Figure 4The flowchart shown is a process for determining whether the patency of the fibrotic drainage channels in Crohn's disease patients is up to standard according to an embodiment of the present invention.

[0090] Specifically, in this embodiment of the invention, the patency of the fibrotic permeation channel in Crohn's disease patients is determined based on the comparison between the fibrotic permeation resistance and the preset fibrotic permeation resistance.

[0091] When the fibrotic permeation resistance is less than or equal to the preset fibrotic permeation resistance, the patency of the fibrotic permeation channel in Crohn's disease patients is determined to be qualified.

[0092] When the fibrotic permeation resistance is greater than the preset fibrotic permeation resistance, the fibrotic permeation channel patency of the Crohn's disease patient is determined to be unqualified.

[0093] In this embodiment of the invention, the preset range of the fiberized seepage resistance is [0.78, 0.89], preferably 0.86, but the above value is not limited to this, and those skilled in the art can adjust the value according to actual needs.

[0094] In this embodiment of the invention, the fibrosis permeation resistance is the result of multiplying the reciprocal of the product of the permeation rate constant and the interstitial space volume fraction by the ratio of the mean of the permeation rate constant to the standard deviation of the permeation rate constant by the ratio of the mean of the interstitial space volume fraction to the standard deviation of the interstitial space volume fraction. The permeation rate constant is the diffusion rate of the contrast agent from plasma to the interstitial space, and the interstitial space volume fraction is the proportion of the extracellular space in the fibrotic region to the total volume.

[0095] Specifically, in this embodiment of the invention, under the condition that the patency of the fibrotic permeation channel in Crohn's disease patients is unqualified, the boundary loss weight of the large model is determined by comparing the preset fibrotic permeation resistance with the preset ratio.

[0096] When the ratio is less than or equal to the preset ratio, it is determined that the boundary loss weight will be increased to the corresponding value by the first preset boundary loss adjustment coefficient of 1.05.

[0097] When the ratio is greater than the preset ratio, it is determined that the boundary loss weight will be increased to the corresponding value by the second preset boundary loss adjustment coefficient of 1.09.

[0098] The ratio is the ratio of the preset fibrous seepage resistance to the fibrous seepage resistance.

[0099] In this embodiment of the invention, the preset ratio range is [0.54, 0.65], preferably 0.59, but the above value is not limited to this, and those skilled in the art can adjust the value according to actual needs.

[0100] In this embodiment of the invention, the increased boundary loss weight is the product of the boundary loss weight and the preset boundary loss adjustment coefficient. The preset boundary loss adjustment coefficient includes a first preset boundary loss adjustment coefficient with a value of 1.05 and a second preset boundary loss adjustment coefficient with a value of 1.09. In order to ensure that the adjusted boundary loss weight meets the actual needs, the adjustment range should not be too large. Therefore, an adjustment coefficient is set to control the adjustment range, and the Dice Loss weight and Focal Loss weight are reduced equally.

[0101] Specifically, this invention determines the patency of fibrotic seepage channels by assessing the fibrotic seepage resistance. If the channels are obstructed, the boundary loss weight of the large model is increased to enable the model to accurately identify the boundaries of the seepage channels. A slight increase in the boundary loss can guide the model to focus on subtle changes in the seepage channels, while a significant increase in the boundary loss can force the model to focus on the edge contour of the obstructed area, thereby improving the contour accuracy of the boundary and the accuracy of the large model in identifying different degrees of obstruction. This improves the accuracy of assessing the patency of fibrotic seepage channels in Crohn's disease, and thus enhances the accuracy of early warning of intestinal fibrosis in Crohn's disease.

[0102] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for early warning of Crohn's disease intestinal fibrosis based on a large model, characterized in that, include: Obtain intestinal MRI image sequences from Crohn's disease patients; The intestinal MRI image sequence is input into a pre-trained large intestinal structure analysis model to locate the initial intestinal fibrosis region. The intestinal stenosis progression index is determined based on intestinal MRI image sequences to determine whether the degree of intestinal stenosis in Crohn's disease patients meets the target. The feature contrast threshold of the large model is adjusted based on the difference between the dynamic progression index of intestinal stenosis and the preset stenosis threshold. The intestinal stenosis progression index is determined based on the baseline diameter of the diseased intestinal segment in the baseline MRI image and the temporal diameter at each time point in the subsequent MRI images. The parameters for characterizing intestinal wall edema were determined by T2-weighted imaging in intestinal MRI sequences to determine whether the acute inflammation masking effect in Crohn's disease patients was adequate. The T2-weighted imaging modal weights of the large model were adjusted based on the relative difference between the parameters for characterizing intestinal wall edema and the preset parameters for characterizing intestinal wall edema. The fibrotic flow resistance was determined based on intestinal MRI image sequences to determine whether the patency of the fibrotic flow channels in Crohn's disease patients was acceptable, and the boundary loss weights of the large model were adjusted based on the ratio of the preset fibrotic flow resistance to the fibrotic flow resistance.

2. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 1, characterized in that, The determination of whether the degree of intestinal stenosis in Crohn's disease patients meets the standard is based on the comparison results of intestinal stenosis progression index being greater than the preset intestinal stenosis progression index.

3. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 2, characterized in that, The process of determining the intestinal stenosis progression index includes: The time when Crohn's disease patients first underwent intestinal MRI scans was recorded as the baseline period. Subsequent MRI scans were performed according to the clinical follow-up plan, and a total of n effective images were acquired to form a time series. In baseline MRI images, the minimum lumen diameter of the stenotic segment is measured along the central axis of the diseased intestinal segment using the U-Net model and recorded as the baseline diameter. For each subsequent MRI image, the minimum lumen diameter of the stenotic segment at the corresponding time point was repeatedly measured and recorded as the temporal diameter; the difference between the temporal diameter and the baseline diameter at each time point was divided by the baseline diameter and recorded as the relative diameter change rate. The relative changes at each time point are weighted, and the time weight is linearly positively correlated with the time interval. The time weight is the ratio of the time interval from the baseline period to the total monitoring duration. The intestinal stenosis progression index is obtained by multiplying the relative diameter change rate at each time point by the corresponding time weight, summing the results, and then dividing by the sum of the time weights.

4. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 3, characterized in that, The process of adjusting the feature comparison threshold of the large model includes: Calculate the difference between the intestinal stenosis progression index and the preset intestinal stenosis progression index under the condition that the degree of intestinal stenosis in Crohn's disease patients does not meet the target; Based on the comparison results where the difference is less than or equal to a preset difference, a first preset feature adjustment coefficient is determined to reduce the feature comparison threshold. Based on the comparison results where the difference is greater than a preset difference, a second preset feature adjustment coefficient is determined to reduce the feature comparison threshold.

5. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 4, characterized in that, The failure of the acute inflammation masking effect in Crohn's disease patients was determined based on the comparison results of the intestinal wall edema characterization parameter being greater than the preset intestinal wall edema characterization parameter.

6. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 5, characterized in that, The process of determining the parameters characterizing intestinal wall edema includes: The T2WI sequence is denoised using a nonlocal mean denoising algorithm. The normalized pixel gray value is the ratio of the denoised image gray value minus the global minimum gray value in the current image to the global maximum gray value minus the global minimum gray value in the current image. The grayscale threshold of the high signal region of edema is determined by the Otsu adaptive thresholding method. By maximizing the inter-class variance, the region with a normalized pixel grayscale value greater than or equal to the grayscale threshold is recorded as the edema region, and the region with a normalized pixel grayscale value less than the grayscale threshold is recorded as the normal region. Extract the grayscale values ​​of all pixels in the edema area, and record the ratio of the number of pixels corresponding to each grayscale level to the total number of pixels in the edema area as the grayscale probability of the edema area. Calculate the information entropy of the edema region based on the gray-level probability of the edema region; The ratio of the information entropy of the edema region to the theoretical maximum value of the information entropy of the edema region is the characteristic parameter of intestinal wall edema.

7. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 6, characterized in that, The process of adjusting the T2-weighted imaging modal weights of the large model includes: Calculate the relative difference between the intestinal wall edema characterization parameter and the preset intestinal wall edema characterization parameter under the condition that the acute inflammation masking effect is not qualified in Crohn's disease patients; Based on the comparison results where the relative difference is less than or equal to the preset relative difference, the first preset model weight adjustment coefficient is determined to reduce the T2 weighted imaging modality weight; Based on the comparison results where the relative difference is greater than the preset relative difference, the second preset model weight adjustment coefficient is determined to reduce the T2 weighted imaging modality weight.

8. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 7, characterized in that, The patency of the fibrotic permeation channel in Crohn's disease patients was deemed unqualified based on the comparison results of fibrotic permeation resistance being greater than the preset fibrotic permeation resistance.

9. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 8, characterized in that, The fibrosis permeation resistance is the result of multiplying the reciprocal of the product of the permeation rate constant and the interstitial space volume fraction by the ratio of the mean of the permeation rate constant to the standard deviation of the permeation rate constant by the ratio of the mean of the interstitial space volume fraction to the standard deviation of the interstitial space volume fraction, wherein the permeation rate constant is the diffusion rate of the contrast agent from plasma to the interstitial space, and the interstitial space volume fraction is the proportion of the extracellular space in the fibrotic region to the total volume.

10. The method for early warning of Crohn's disease intestinal fibrosis based on a large model as described in claim 9, characterized in that, The process of adjusting the boundary loss weights of the large model includes: Calculate the ratio of the preset fibrotic flow resistance to the fibrotic flow resistance under the condition that the patency of the fibrotic flow channel is not up to standard in Crohn's disease patients. Based on the comparison results where the ratio is less than or equal to a preset ratio, a first preset boundary loss adjustment coefficient is determined to increase the boundary loss weight of the large model; Based on the comparison results where the ratio is greater than a preset ratio, a second preset boundary loss adjustment coefficient is determined to increase the boundary loss weight of the large model.

Images