CN-121121240-B - Crohn's disease intestinal fibrosis disease early warning method based on large model driving

CN121121240BCN 121121240 BCN121121240 BCN 121121240BCN-121121240-B

Abstract

The invention relates to the technical field of intelligent medicine, in particular to a method for early warning the intestinal fibrosis disease state of a Crohn disease based on large model driving, which comprises the steps of obtaining an intestinal MRI image sequence of the Crohn disease patient, inputting the intestinal MRI image sequence into a pre-trained intestinal structure analysis large model, positioning the initial intestinal fibrosis area position, determining whether the intestinal lumen stenosis degree of the Crohn disease patient meets the standard based on an intestinal lumen stenosis progress index, determining a characteristic contrast threshold value of the large model according to a difference value, determining whether the acute inflammation shielding effect of the Crohn disease patient is qualified based on an intestinal wall edema characterization parameter, determining a T2 weighted imaging mode weight of the large model according to a relative difference, determining whether the smoothness of a fibrosis seepage channel of the Crohn disease is qualified based on a fibrosis seepage rate, and determining a boundary loss weight of the large model according to a ratio. The invention improves the accuracy of early warning of the Crohn disease intestinal fibrosis.

Inventors

LI XUEHUA
HUANG LI
Fan Cien
LIN SHAOCHUN
Peng Zhenpeng
Feng Shiting
ZHANG RUONAN
Li Zhoulei
Wang Yangdi
SHEN XIAODI
ZHENG QINGZHU
HE WEITAO
LIN JINJIANG

Assignees

中山大学附属第一医院

Dates

Publication Date: 20260505
Application Date: 20250831

Claims (10)

1. The method for early warning the intestinal fibrosis of the Crohn disease based on large model driving is characterized by comprising the following steps of: Acquiring an intestinal MRI image sequence of a patient suffering from Crohn's disease; Inputting the intestinal MRI image sequence into a pre-trained intestinal structure analysis large model, and positioning the initial intestinal fibrosis region; Determining an intestinal lumen stenosis progress index based on an intestinal MRI image sequence to determine whether the intestinal lumen stenosis degree of a crohn patient meets the standard, and determining a feature contrast threshold of the adjustment large model according to a difference value between the intestinal lumen stenosis dynamic progress index and a preset stenosis threshold, wherein the intestinal lumen stenosis progress index is determined based on a baseline diameter of a diseased intestinal segment in an MRI image of a baseline period and a time sequence diameter of each time point in a subsequent MRI image; Determining an intestinal wall edema characterization parameter based on T2 weighted imaging in an intestinal MRI image sequence to determine whether an acute inflammatory shielding effect of a Crohn patient is qualified or not, and determining a T2 weighted imaging modal weight of the large model according to a relative difference between the intestinal wall edema characterization parameter and a preset intestinal wall edema characterization parameter; And determining the fibrosis seepage flow rate based on the intestinal MRI image sequence so as to determine whether the smoothness of a fibrosis seepage flow channel of a Crohn patient is qualified, and determining and adjusting the boundary loss weight of the large model according to the ratio of the preset fibrosis seepage flow rate to the fibrosis seepage flow rate.
2. The method for early warning of a large model driven intestinal fibrosis condition of crohn's disease according to claim 1, wherein the extent of intestinal lumen stenosis in the crohn's disease patient is determined based on a comparison result that the intestinal lumen stenosis progression index is greater than a preset intestinal lumen stenosis progression index.
3. The method for early warning of a large model driven intestinal fibrosis condition of crohn's disease according to claim 2, wherein the determining the enteroluminal stenosis progression index includes: recording the time of the first time of the patient suffering from the Crohn disease to receive the MRI scanning of the intestinal tract as a baseline period, and subsequently carrying out MRI scanning for a plurality of times according to a clinical follow-up plan, and collecting n effective images to form a time sequence; In an MRI image of a baseline period, measuring the minimum lumen diameter of a narrow section along the central axis of a diseased intestinal section through a U-Net model, and marking the minimum lumen diameter as the baseline diameter; repeatedly measuring the minimum lumen diameter of the narrow section at the corresponding time point for the MRI image at each subsequent time, and recording the minimum lumen diameter as a time sequence diameter; Weighting the relative change of each time point, wherein the time weight and the time interval are in linear positive correlation, and the time weight is the ratio of the time interval from the baseline period to the total monitoring duration; multiplying the relative diameter change rate of each time point by the corresponding time weight, summing, and dividing by the time weight sum to obtain the intestinal lumen stenosis progression index.
4. The method of claim 3, wherein the adjusting the feature contrast threshold of the large model comprises: calculating the difference value between the intestinal lumen stenosis progress index of the Crohn patient under the condition that the intestinal lumen stenosis degree does not reach the standard and the preset intestinal lumen stenosis progress index; Determining a first preset feature adjustment coefficient reduction feature comparison threshold based on a comparison result that the difference is less than or equal to a preset difference; and determining a second preset characteristic adjustment coefficient to reduce the characteristic contrast threshold value based on the contrast result that the difference value is larger than the preset difference value.
5. The method for early warning of intestinal fibrosis of crohn's disease based on large model driving according to claim 4, wherein the disqualification of acute inflammatory shadowing effect of crohn's disease is determined based on a comparison result that intestinal wall edema characterization parameters are greater than preset intestinal wall edema characterization parameters.
6. The method for early warning of a large model driven crohn's disease intestinal fibrosis condition according to claim 5, wherein the determining the intestinal wall edema characterization parameter includes: denoising the T2WI sequence by adopting a non-local mean denoising algorithm, and marking 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 as a normalized pixel gray value; Determining a gray level threshold value of an edema high signal area by using an Otsu self-adaptive threshold value method, marking an area with a normalized pixel gray level value larger than or equal to the gray level threshold value as an edema area and marking an area with a normalized pixel gray level value smaller than the gray level threshold value as a normal area by maximizing an inter-class variance; extracting all pixel gray values in the edema area, and marking the ratio of the number of pixels corresponding to the gray level to the total number of pixels in the edema area as the gray probability of the edema area; calculating information entropy of the edema area according to the gray scale probability of the edema area; the ratio of the edema area information entropy to the theoretical maximum value of the edema area information entropy is the characteristic parameter of the intestinal wall edema.
7. The method of claim 6, wherein the adjusting the T2-weighted imaging modality weights of the large model comprises: calculating the relative difference between the intestinal wall edema characterization parameter and the preset intestinal wall edema characterization parameter of the patient with Crohn's disease under the condition of disqualification of acute inflammatory shielding effect; Determining a first preset model weight adjustment coefficient to reduce the T2 weighted imaging modality weight based on a comparison result that the relative difference is less than or equal to a preset relative difference; and determining a second preset model weight adjustment coefficient to reduce the T2 weighted imaging modality weight based on the comparison result that the relative difference is larger than the preset relative difference.
8. The method for early warning of a large model driven intestinal fibrosis disease based on crohn's disease according to claim 7, wherein the poor smoothness of the fibrotic percolation path of the crohn's disease is determined based on a comparison result that the fibrotic percolation ratio is greater than a preset fibrotic percolation ratio.
9. The method of claim 8, wherein the fibrotic permeability is a result of multiplying the inverse of the product of the permeability constant and the interstitial volume fraction by the mean of the permeability constant and the standard deviation of the permeability constant times the mean of the interstitial volume fraction and the standard deviation of the interstitial volume fraction, wherein the permeability constant is the diffusion rate of the contrast agent from the plasma to the interstitial volume, and the interstitial volume fraction is the ratio of the extracellular space to the total volume in the fibrotic region.
10. The method of claim 9, wherein the step of adjusting the boundary loss weight of the large model comprises: calculating the ratio of the preset fibrosis seepage resistivity to the fibrosis seepage resistivity of the Crohn's disease patient under the condition of unqualified smoothness of the fibrosis seepage channel; Determining a first preset boundary loss adjustment coefficient to increase the boundary loss weight of the large model based on the comparison result that the ratio is smaller than or equal to the preset ratio; and determining that the second preset boundary loss adjustment coefficient increases the boundary loss weight of the large model based on the comparison result that the ratio is larger than the preset ratio.

Description

Crohn's disease intestinal fibrosis disease early warning method based on large model driving Technical Field The invention relates to the technical field of intelligent medicine, in particular to a Crohn disease intestinal fibrosis disease early warning method based on large model driving. Background Crohn's disease is a chronic recurrent intestinal inflammatory disease, and intestinal fibrosis is a common complication thereof, and is mainly represented by excessive deposition of collagen and other fibrous substances in intestinal walls, which can cause intestinal lumen stenosis, obstruction and even perforation, seriously affect the life quality of patients and increase the operation risk. The early accurate early warning of the progress of intestinal fibrosis has important clinical significance in adjusting a treatment scheme and avoiding irreversible damage, at present, the clinical evaluation of fibrosis degree mainly depends on an intestinal MRI image, but the accuracy is influenced by various factors, fibrosis often coexists with acute inflammation, the inflammation can shield fibrosis characteristics, erroneous judgment is caused, fibrosis is a dynamic progress process, a static image is difficult to reflect the change trend in the time dimension of the static image, the boundary of a fibrosis area is fuzzy, the tissue gap seepage characteristic is changed and the like, and the accuracy of a traditional image analysis method is also influenced. CN114067154a discloses a method for grading crohn's disease fibrosis based on multi-sequence MRI and related equipment, the method includes obtaining first image histology characteristics of intestinal fibrous tissue in each of a plurality of preset abdominal MRI images of the same target object by controlling an image histology characteristic module, and fusing each first image histology characteristic to obtain fused image histology characteristics, and determining the type of crohn's disease fibrosis of the target object based on the fused image histology characteristics by controlling a classification module. However, the prior art has the following problems that the characteristic extraction and classification are only based on static multi-sequence images, and the characteristic extraction and classification depend on morphological characteristics, so that the recognition of the intestinal fibrosis disease condition of the Crohn disease is insufficient comprehensively, and the early warning of the intestinal fibrosis disease condition of the Crohn disease is low in accuracy. Disclosure of Invention Therefore, the invention provides a large-model-driven early warning method for the intestinal fibrosis of the Crohn's disease, which is used for solving the problems that in the prior art, feature extraction and classification are only based on static multi-sequence images, and the recognition of the intestinal fibrosis of the Crohn's disease is comprehensively insufficient due to morphological features, so that the early warning accuracy of the intestinal fibrosis of the Crohn's disease is low. In order to achieve the above purpose, the present invention provides a method for early warning the disease state of Crohn's disease intestinal fibrosis based on large model driving, comprising: Acquiring an intestinal MRI image sequence of a patient suffering from Crohn's disease; Inputting the intestinal MRI image sequence into a pre-trained intestinal structure analysis large model, and positioning the initial intestinal fibrosis region; Determining an intestinal lumen stenosis progress index based on the intestinal MRI image sequence to determine whether the intestinal lumen stenosis degree of a Crohn patient meets the standard or not, and determining a characteristic contrast threshold of the large model according to the difference value between the intestinal lumen stenosis dynamic progress index and a preset stenosis threshold; Determining an intestinal wall edema characterization parameter based on T2 weighted imaging in an intestinal MRI image sequence to determine whether an acute inflammatory shielding effect of a Crohn patient is qualified or not, and determining a T2 weighted imaging modal weight of the large model according to a relative difference between the intestinal wall edema characterization parameter and a preset intestinal wall edema characterization parameter; And determining the fibrosis seepage flow rate based on the intestinal MRI image sequence so as to determine whether the smoothness of a fibrosis seepage flow channel of a Crohn patient is qualified, and determining and adjusting the boundary loss weight of the large model according to the ratio of the preset fibrosis seepage flow rate to the fibrosis seepage flow rate. Further, the extent of the intestinal lumen stenosis of the crohn's disease patient is determined based on a comparison result that the intestinal lumen stenosis progress index is greater than a preset intestinal lumen s