A morphology-driven two-dimensional cine-MRI small intestine motion evaluation method

Abstract

The present application is suitable for the technical field of medical signal processing, and provides a shape-driven two-dimensional Cine-MRI small intestine movement evaluation method. The present application shows significant effects in the aspects of small intestine movement quantitative evaluation and clinical application. The present application takes the geometric continuity of the intestinal cavity as a constraint, automatically segments, space-time filters, and reconstructs a closed and hole-free space-time shell through a weighted level set, extracts a time-consistent center line and optimizes the processing, and then converts the cavity wall movement through radiation encoding and constructs four types of quantitative indexes. The method can distinguish between healthy and disease groups, the parameter difference is significant and the consistency with expert annotation is high, can provide a reviewable quantitative baseline for efficacy follow-up, and can also complement the MRE index to improve the stability of small intestine comprehensive evaluation. Without changing the acquisition protocol, the device sensitivity is low, the single GPU near real-time processing is easy to integrate, a multi-module system can be constructed to adapt to the DICOM process, and the reliability of the results can be ensured by distinguishing real deformation from interference factors.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical signal processing, and particularly relates to a shape-driven two-dimensional Cine-MRI small intestine movement evaluation method. BACKGROUND

[0002] In recent years, the role of Cine-MRI in small intestine movement evaluation has been continuously improved. Its advantages lie in the absence of ionizing radiation, a large coverage range, high time resolution, and compatibility with conventional abdominal examination procedures. Clinical studies have shown that under the conditions of two-dimensional sequences with exhalation breath holding, a time span of at least 15-20 seconds and a time resolution of no less than 1 frame / s can stably observe one or several peristaltic cycles. Under these acquisition conditions, Cine-MRI has good visualization ability for movement patterns such as propulsion, segmentation, retardation, or disorder. In addition to Cine-MRI, current techniques for small intestine movement evaluation include pressure measurement, radionuclide imaging, and capsule endoscopy. However, these techniques all have obvious limitations: pressure measurement is limited to local assessment and cannot achieve large-scale evaluation; radionuclide imaging is limited by time resolution and cannot accurately capture dynamic changes; and capsule endoscopy is limited by morphological details and cannot clearly present relevant morphological features, making it difficult for all three to balance "range-resolution-explainability".

[0003] Although two-dimensional Cine-MRI has advantages, its quantification has long been subject to two types of physical constraints: first, due to insufficient local intestinal lumen expansion, there may be discontinuous segments in the time sequence, forming "holes" in the space-time domain, leading to distortion in registration or lumen diameter estimation; second, the displacement of the intestinal tract in the cross-layer direction can be manifested as "false contraction" on a single slice, which is easily misjudged as "contraction-dilation", leading to ambiguity in intensity optical flow methods after brightness compensation. In recent years, although there have been productized attempts based on "motion heat maps" or intensity change fields for compensation, when factors such as contrast agent dynamics, respiratory motion, cross-device differences, and brightness drift are superimposed, the pixel intensity consistency assumption is easily invalidated, and the correspondence of the same intestinal loop on the time axis may also be destroyed, leading to subsequent statistical analysis only being able to stay at the global average level, lacking local reliability. In addition, existing registration methods based on intensity consistency cannot distinguish between true tissue movement and brightness changes, further introducing systematic errors.

[0004] In view of the above problems of existing methods, it is necessary to establish a "shape-anchored" space-time geometric representation and index system around two-dimensional Cine-MRI, so that it remains repeatable under cross-center and cross-device conditions, becoming a more practical approach. Therefore, the present application proposes a shape-driven two-dimensional Cine-MRI small intestine movement evaluation method. SUMMARY

[0005] The present application aims to provide a morphology-driven two-dimensional Cine-MRI small intestine movement evaluation method, aiming to solve the problems raised in the background art.

[0006] The purpose of the present application is achieved by the following technical solutions:

[0007] A morphology-driven two-dimensional Cine-MRI small intestine movement evaluation method, comprising the following steps:

[0008] Automatic segmentation and spatio-temporal screening: a small intestine segmentation network based on JointSwinUNETR is used to automatically segment the two-dimensional Cine-MRI images of the complete time sequence to generate a binary three-dimensional mask; the connected regions obtained by segmentation are labeled in the three-dimensional space, and the connected domain with the largest number of voxels is retained as the quantitative target region;

[0009] Spatio-temporal shell reconstruction and centerline extraction: the frame-by-frame segmentation results are used as data items to minimize the weighted minimum surface energy in the spatio-temporal domain, and a stable conservative difference is used to realize numerical calculation, and the calculation is stopped when the energy relatively decreases to a set threshold or the iteration upper limit, at which time the zero level set is defined as a closed spatio-temporal shell; the lumen centerline consistent in time sequence is extracted from the spatio-temporal shell by shortest path optimization and inter-frame smoothing constraint, and the extracted centerline is B-spline fitted and then equally spaced resampled to obtain a smooth and structurally stable centerline sequence;

[0010] Radiation encoding and core index construction: each node in the centerline sequence is used as a source point to emit equal-angle rays outward, the intersection points of the rays and the spatio-temporal shell boundary are recorded, the average value of the intersection point distance and direction is calculated to obtain the radial length and phase, which together constitute the spatio-temporal matrix representing the small intestine movement state; based on the radial length and phase, four types of indexes including amplitude dominance, synchronicity, spatial heterogeneity and rhythm stability are extracted to construct a quantitative index system for evaluating the small intestine movement ability.

[0011] Further, the small intestine segmentation network based on JointSwinUNETR is composed of a local sequential encoder, a global multi-scale encoder and a spatial sequential hybrid decoder.

[0012] Further, the calculation of the weighted minimum surface energy follows the following formula:

[0013] ;

[0014] wherein, is the weighted minimum surface energy; is a 2-dimensional spatial domain; is a time domain; is a smooth Heaviside function; is difference; Spatial coordinates; For a specific moment; The weights are differential proportions; To increase the confidence weight of reliable region fit; For segmentation mask.

[0015] Furthermore, the radial length and phase are calculated using the following formulas:

[0016] ;

[0017] ;

[0018] in, Radial length; For phase; The total number of rays; For the first ray; Center Online point t Time of the first Length of the ray; Center Online point t Time of the first The direction angle of the ray.

[0019] Furthermore, the amplitude dominance is constructed by calculating cumulative energy, which measures the kinetic energy generated from the previous... K The proportions explained by each main model depict the dominance and energy concentration of the main model.

[0020] Furthermore, the synchronicity is calculated using Kuramoto order, and the time mean of Kuramoto order is taken as the synchronicity index to measure the degree of phase synchronization of different spatial locations of the small intestine throughout the entire time series. The larger the index value, the stronger the synchronicity.

[0021] Furthermore, the spatial heterogeneity is calculated using the Laplacian of the chain graph to determine the spatial roughness of each frame. After normalizing the spatial roughness, the time mean is taken to obtain the spatial heterogeneity index. The larger the index value, the stronger the local inconsistency and the lower the compliance.

[0022] Furthermore, the rhythm stability is obtained by performing Vietoris-Rips filtering on the point cloud and then performing persistent cohomology analysis to obtain the birth and death pair thresholds of one-dimensional loop features. The lifetime of all one-dimensional loop features is aggregated to obtain a rhythm stability index. The larger the index value, the stronger the topological stability of the small intestinal motility rhythm under multi-scale distance thresholds.

[0023] A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method as described above.

[0024] An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method as described above when executing the computer program.

[0025] Compared with the prior art, the present application has the following advantages:

[0026] The present application proposes a morphology-driven two-dimensional Cine-MRI small intestine movement evaluation method, which exhibits multidimensional significant effects in small intestine movement quantitative evaluation and clinical application.

[0027] At the technical core level, it takes the geometric continuity of the intestinal cavity as the key constraint, and after generating a three-dimensional mask through automatic segmentation, it retains the coherent intestinal loop through spatiotemporal screening, and then reconstructs a closed and hole-free spatiotemporal shell with the help of a weighted level set to repair the original mask fracture problem, extracts the time-consistent centerline and optimizes the processing, laying a precise geometric foundation for subsequent analysis; then it converts the cavity wall movement into a spatiotemporal matrix of "radial length" and "phase" through radiation encoding, and then constructs four types of complementary indicators of amplitude dominance, synchronicity, spatial heterogeneity and rhythm stability, realizing the comprehensive quantification of small intestine movement ability, and the index definition and physical meaning correspond clearly, facilitating report and decision support.

[0028] In terms of clinical application value, this method performs excellently in distinguishing between healthy and diseased groups, with significant differences in parameter distribution and linear separability, and high consistency with image expert annotation, which can be used as an important basis for clinically distinguishing between healthy and diseased small intestines, providing a reviewable quantitative baseline for efficacy follow-up, and complementing MRE indexes (such as MaRIA and Clermont) to improve the stability of small intestine comprehensive evaluation.

[0029] In terms of technology landing and popularization, it only relies on two-dimensional Cine-MRI routine acquisition without changing the acquisition protocol, and the key parameter range is clear, which is not sensitive to equipment and center, and is conducive to multi-center reuse; it adopts lightweight segmentation backbone and analytical geometric calculation, and single GPU can realize near real-time processing, the interface is unified and easy to integrate into a scientific research station, and a system containing multiple modules (segmentation and screening module - level set shell module - centerline consistency module - radiation encoding and index module - report module) can be built, forming an independent quantitative evaluation terminal or hospital server, adapting to standard DICOM process, and having hardware and software integrated landing ability.

[0030] In addition, the method takes the geometric shell and the center line as anchor points, effectively distinguishes real deformation and brightness change, layer loss, avoids the influence of contrast agent, respiration and cross-device conditions on the intensity model, eliminates ambiguity, and guarantees the reliability of the evaluation results. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 A technical flowchart of the present application.

[0032] Figure 2 A motion description index for clinical classification effect diagram; wherein a is a box plot, and b is a confusion matrix diagram. DETAILED DESCRIPTION

[0033] In order to have a clearer understanding of the technical features, objectives and beneficial effects of the present application, the technical solutions of the present application will be described in detail below, but it cannot be understood as limiting the scope of the present application.

[0034] The specific implementation of the present application will be described in detail below in combination with specific embodiments.

[0035] The present application provides a shape-driven two-dimensional Cine-MRI small intestine motion evaluation method, and its flowchart is shown in Figure 1 The method starts from frame-by-frame segmentation to obtain a small intestine mask; maximum connected components (3D-LCC) screening is performed in the spatial-temporal voxel domain, only the coherent loop is retained in the entire sequence, and the stable analysis region is cut according to this. Then, based on the weighted level set, the closed and hole-free space-time shell is reconstructed in the (x, y, t) domain, so that the cavity wall naturally transitions in time; the center line consistent in time is extracted inside the shell and is equally arc length resampled. With the center line as the reference, equally angular radiation rays are emitted, the intersection with the shell boundary is recorded, and the space-time matrix of "radial length" and "phase" is obtained; on this basis, indexes such as amplitude dominance (PMC), synchronicity (MSI), spatial heterogeneity (LS) and rhythm stability (PHI) are constructed, and finally a lightweight model is used to complete the small intestine health or disease discrimination based on motion ability. The specific process is as follows:

[0036] I. Automatic segmentation and space-time screening

[0037] Small intestine automatic segmentation: a small intestine segmentation network based on JointSwinUNETR is used to automatically segment the two-dimensional Cine-MRI images of the complete time sequence to generate a binary three-dimensional mask. The network is composed of a local sequential encoder, a global multi-scale encoder and a spatial sequential hybrid decoder, which can effectively balance between fine-grained and large-scale features, thereby realizing high-precision segmentation.

[0038] Space-time screening: the connected regions obtained by segmentation are labeled in three-dimensional space, and the connected domain with the largest number of voxels is retained as the quantitative target region.

[0039] II. Spatiotemporal shell reconstruction and centerline extraction;

[0040] Spatiotemporal shell reconstruction: the result of frame-by-frame segmentation is taken as data item, and the weighted minimal surface energy is minimized in spatiotemporal domain (x, y, t). The energy calculation follows the formula:

[0041] ;

[0042] wherein, is the weighted minimal surface energy; is the 2-dimensional spatial domain; is the time domain; is the smooth Heaviside function; is the difference; is the spatial coordinate; is the time instant; is the difference proportion weight; is the confidence weight, used to improve the fitting degree of reliable areas and suppress noise and holes; is the segmentation mask. Numerical implementation uses stable conservative difference. When the energy is relatively reduced to the set threshold or the iteration upper limit, the calculation is stopped. The zero level set at this time defines a closed "spatiotemporal shell", which can make the small intestine cavity wall naturally transition in the time dimension, with no holes and complete structure.

[0043] Centerline extraction: the spatiotemporal shell is extracted from the inside to extract the cavity centerline consistent in time sequence through shortest path optimization and inter-frame smoothing constraint; the extracted centerline is B-spline fitted and then equally-arc-length resampled to finally obtain a smooth and structurally stable centerline sequence, which provides an accurate reference benchmark for subsequent radiation encoding.

[0044] III. Radiation encoding and core index construction;

[0045] Radiation encoding: taking each node in the centerline sequence as a source point, a M strip of isoclines is emitted outward, the intersection of the isoclines and the spatiotemporal shell boundary is recorded, and the average value of the intersection distance and direction is calculated to obtain the radial length and phase , the calculation formulas are respectively:

[0046] ;

[0047] ;

[0048] wherein is used to describe the spatiotemporal variation of the cavity diameter, The phase coordination characteristics used to characterize small intestinal motility, together with the phase coordination characteristics, constitute a spatiotemporal matrix characterizing the state of small intestinal motility. The total number of rays; For the first ray; Center Online point t Time of the first Length of the ray; Center Online point t Time of the first The direction angle of the ray.

[0049] Core metric construction: based on radial length With phase Four complementary indices were extracted: amplitude dominance (PMC), synchronicity (MSI), spatial heterogeneity (LS), and rhythm stability (PHI). These indices quantify small intestinal motility from different dimensions. The calculation logic and correlation formulas for each index are as follows:

[0050] ① Amplitude Dominance (PMC): This index is constructed by calculating cumulative energy. Used to measure the energy of motion from the previous K The proportions explained by each principal model characterize the dominance and energy concentration of the principal model; among them The total number of energy modes; This is the kth pattern; Amplitude; This represents the total number of sampling points along the center line. This represents the total number of frames.

[0051] ② Synchronicity (MSI): with For each spatial location i At any moment t The motion phase is determined using Kuramoto order. Its time average It is used to measure the degree of phase synchronization of different spatial locations of the small intestine throughout the entire time series. (The larger the value, the more synchronized the data). This represents the total number of sampling points along the center line. Spatial coordinates; It is the direction angle; These are equivalent phases; For equivalent modulus; Total number of frames; For a moment.

[0052] ③ Spatial Heterogeneity (LS): Laplace of a chain diagram L Calculate the spatial roughness of each frame After normalizing to energy, the time average is obtained. LS This is used to characterize the spatial heterogeneity of radial variation (a larger value indicates stronger local inconsistency and lower compliance); among which Let be the ray length combination vector at time t; for The transpose of .

[0053] ④ Rhythm Stability (PHI): For point clouds { , Vietoris-Rips filtering was performed on the ,t / T)}, followed by persistent cohomology analysis to obtain the birth and death pair thresholds of one-dimensional loop features ( , );in The birth threshold; The death threshold is defined as the aggregated lifetime of all one-dimensional loop features to obtain the index. This is used to quantify the topological stability of small intestinal motility rhythms under multi-scale distance thresholds (the larger the sum, the more stable the rhythm).

[0054] Using four complementary indices—PMC, MSI, LS, and PHI—a lightweight model was employed to differentiate between healthy and disease groups based on small intestinal motility. The results, as seen in the box plot, demonstrate... Figure 2 (a) The parameter distributions obtained using this method show significant differences between the disease group and the healthy group. The parameters in the disease group are generally lower than those in the healthy group, and the parameters of the two groups are not overlapping and are linearly separable, indicating that the motion quantification method has good discriminative power. The confusion matrix can characterize the consistency between the health and disease discrimination results based on motion parameters and the annotations by imaging experts. The results show that (a) Figure 2 (b) In the figure, there are 25 cases with a true label of 0 (corresponding to the healthy group) and a predicted label of 0, and 30 cases with a true label of 1 (corresponding to the disease group) and a predicted label of 1. This shows that the method is in good agreement with the expert score, proving that it has high clinical application value.

[0055] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility, characterized in that, Includes the following steps: Automatic segmentation and spatiotemporal filtering: The small intestine segmentation network based on JointSwinUNETR is used to automatically segment the complete time series of two-dimensional Cine-MRI images and generate a binarized three-dimensional mask; the connected regions obtained by segmentation are marked in three-dimensional space, and the connected region with the largest number of voxels is retained as the quantization target region; Spatiotemporal shell reconstruction and centerline extraction: Using frame-by-frame segmentation results as data items, the weighted minimum surface energy is minimized in the spatiotemporal domain. Stable conserved difference is used to achieve numerical calculation. The calculation stops when the energy decreases to a set threshold or the upper limit of iteration. At this time, the zero level set is defined as a closed spatiotemporal shell. Through shortest path optimization and inter-frame smoothing constraints, temporally consistent cavity centerlines are extracted from the spatiotemporal shell. After B-spline fitting of the extracted centerlines, equal arc length resampling is performed to obtain a smooth and structurally stable centerline sequence. Radiation coding and core index construction: Taking each node in the centerline sequence as the source point, equiangular rays are emitted outward, and the intersection points of the rays with the spatiotemporal shell boundary are recorded. The average value of the distance and direction of the intersection points is calculated to obtain the radial length and phase, which together constitute the spatiotemporal matrix characterizing the small intestinal motility state. Based on the radial length and phase, four types of indicators are extracted: amplitude dominance, synchronicity, spatial heterogeneity, and rhythm stability, and a quantitative index system for evaluating small intestinal motility is constructed.

2. The morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility according to claim 1, characterized in that, The small intestine segmentation network based on JointSwinUNETR consists of a local sequential encoder, a global multi-scale encoder, and a spatial sequential hybrid decoder.

3. The morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility according to claim 1, characterized in that, The calculation of the weighted minimum surface energy follows the formula below: ； in, The weighted minimum surface energy; It is a 2-dimensional spatial domain; For the time domain; For smoothing the Heaviside function; for difference; Spatial coordinates; For a specific moment; The weights are differential proportions; To increase the confidence weight of reliable region fit; For segmentation mask.

4. The morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility according to claim 1, characterized in that, The radial length and phase are calculated using the following formulas: ； ； in, Radial length; For phase; The total number of rays; For the first ray; Center Online point t Time of the first Length of the ray; Center Online point t Time of the first The direction angle of the ray.

5. The morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility according to claim 1, characterized in that, The amplitude dominance is constructed by calculating cumulative energy, which measures the kinetic energy generated from the previous... K The proportions explained by each main model depict the dominance and energy concentration of the main model.

6. The morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility according to claim 1, characterized in that, The synchronicity is calculated using Kuramoto order, and the time mean of Kuramoto order is taken as the synchronicity index to measure the degree of phase synchronization of different spatial locations of the small intestine in the entire time series. The larger the index value, the stronger the synchronicity.

7. The morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility according to claim 1, characterized in that, The spatial heterogeneity is calculated using the Laplacian of the chain graph to determine the spatial roughness of each frame. After normalizing the spatial roughness, the time mean is taken to obtain the spatial heterogeneity index. The larger the index value, the stronger the local inconsistency and the lower the compliance.

8. The morphology-driven two-dimensional Cine-MRI method for assessing small bowel motility according to claim 1, characterized in that, The rhythm stability is obtained by performing Vietoris-Rips filtering on the point cloud and then performing persistent cohomology analysis to obtain the birth and death pair thresholds of one-dimensional loop features. The lifetime of all one-dimensional loop features is aggregated to obtain the rhythm stability index. The larger the index value, the stronger the topological stability of the small intestinal motility rhythm under multi-scale distance thresholds.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-8.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-8.

Images