A magnetic resonance intelligent image diagnosis system for dysphagia

Abstract

The application discloses a kind of about dysphagia magnetic resonance intelligent image diagnosis system, it is related to medical image processing and diagnostic technique field, including: swallowing state functional magnetic resonance acquisition module;Swallowing movement timing decomposition module;Virtual swallowing contrast parameter quantification module;Aspiration pneumonia risk prediction module;Diagnostic report generation module.The application will be obtained by traditional dependence X ray swallowing contrast leakage-Aspiration degree, pharyngeal residual state and pharyngeal transport time such as key quantitative index, by the brain activation timing and functional connection characteristics of functional magnetic resonance imaging Equivalent mapping, so that patients do not need to bear ionizing radiation can obtain and contrast gold standard highly comparable quantitative evaluation parameters, the application not only can realize no radiation equivalent quantitative evaluation, also can create new "functional survival brain area" evaluation and rehabilitation target positioning ability, establish the model for preventing aspiration pneumonia to occur, provide reliable quantitative standard for clinical rehabilitation.

Description

Technical Field

[0001] This invention relates to the field of medical image processing and diagnostic technology, specifically to a magnetic resonance imaging intelligent diagnostic system for swallowing disorders. Background Technology

[0002] Dysphagia is a common complication after stroke, which can lead to serious consequences such as aspiration pneumonia and malnutrition. Currently, the "gold standard" for clinical diagnosis of dysphagia is video fluoroscopy swallowing (VFSS). This method uses X-ray imaging to observe the transport process of the food bolus in the oral cavity, pharynx, and esophagus in real time, and uses quantitative parameters such as the Rosenbek leak-aspiration scale, pharyngeal residue grading, and pharyngeal transport time to assess swallowing function and aspiration risk.

[0003] However, existing VFSS technology, as the gold standard, has two major flaws:

[0004] First, there is the risk of ionizing radiation: it relies on X-ray imaging, which poses a cumulative radiation risk to stroke rehabilitation patients who require multiple follow-up examinations, limiting its repeated use in children, pregnant women, and populations requiring long-term dynamic monitoring.

[0005] Second, the blind spot in functional assessment: VFSS can only provide indirect images of pharyngeal structure movement and cannot assess the functional integrity of the central nervous system. It also cannot diagnose whether there are still surviving neurons with compensatory potential in the brain regions that regulate the swallowing process (such as the sensorimotor cortex and the brainstem swallowing center). This makes it impossible for clinicians to develop targeted neurorehabilitation plans for patients (such as transcranial magnetic stimulation and electroacupuncture), and misses the best opportunity to rebuild swallowing function using plastic brain tissue.

[0006] In response to this, this application proposes a magnetic resonance imaging intelligent diagnostic system for dysphagia to solve the above-mentioned problems. Summary of the Invention

[0007] The purpose of this invention is to provide a magnetic resonance imaging intelligent diagnostic system for dysphagia, addressing the shortcomings of existing technologies such as VFSS (Vibration-Free Swallowing Surgery) which relies on X-ray ionizing radiation, posing a cumulative radiation risk to stroke rehabilitation patients requiring multiple follow-up examinations. This limits its repeated application in children, pregnant women, and populations requiring long-term dynamic monitoring. Furthermore, VFSS only provides indirect imaging of pharyngeal structure movement and cannot assess the functional integrity of the central nervous system. This invention not only achieves radiation-free equivalent quantitative assessment but also pioneers new capabilities for assessing "functionally viable brain regions" and locating rehabilitation targets, establishing a model for preventing aspiration pneumonia, and providing reliable quantitative standards for clinical rehabilitation.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A magnetic resonance imaging intelligent diagnostic system for dysphagia includes:

[0010] The functional magnetic resonance imaging (fMRI) acquisition module during swallowing is used to acquire time-series images of whole-brain oxygen level-dependent (BOLD) signals in real time when the subject performs standard swallowing actions, and obtain brain region temporal activation data volume reflecting the functional viability and activity intensity of neurons in each brain region;

[0011] The swallowing motor temporal decomposition module is used to segment and label the brain region temporal activation data volume based on the physiological phases of swallowing, namely the oral phase, pharyngeal phase, and esophageal phase, and quantify the motor intensity, generating a topological map that can intuitively display the range and intensity of brain region functional activation in each phase—wherein, the significantly activated areas directly represent surviving brain tissue with blood perfusion and neural function.

[0012] The virtual swallowing imaging parameter quantification module is used to compare the similarity of the phased brain region activation topology map with the preset swallowing imaging pathology pattern library and map it into a set of virtual swallowing imaging parameters including leakage-aspiration simulation rating, pharyngeal residue index and pharyngeal transport time equivalent.

[0013] The aspiration pneumonia risk prediction module is used to input the set of virtual swallowing imaging parameters into a risk stratification model validated by multi-center data, and calculate the probability value and risk level of aspiration pneumonia.

[0014] The diagnostic report generation module is used to generate a visual diagnostic report containing a three-dimensional brain function activation map and virtual imaging quantitative indicators based on the risk level and the virtual swallowing imaging parameter set.

[0015] Furthermore, the functional magnetic resonance imaging (fMRI) acquisition module during swallowing uses synchronous trigger pulses and swallowing motion sensors to timestamp each swallowing motion, and aligns the timestamps with the whole-brain fMRI image sequence frame by frame to obtain brain region temporal activation data with swallowing event time-locked markers.

[0016] Furthermore, the swallowing movement temporal decomposition module includes a pharyngeal movement segmentation unit, which is used to identify the pharyngeal start frame and end frame based on the zero-crossing point of the first derivative of the brain region activation intensity change curve corresponding to the hyoid-larynx complex, and calculate the average activation intensity of each brain region during the pharyngeal period as the quantitative value of the pharyngeal movement intensity, thereby constituting the phased brain region activation topology map.

[0017] Furthermore, the virtual swallowing angiography parameter quantification module is used for:

[0018] The functional connectivity strength matrix between the brain region activation topology map of the time-phased brain regions and the pharyngeal sensory-motor cortex and the brainstem swallowing center is calculated. The functional connectivity strength matrix is ​​then registered with the typical abnormal connectivity patterns in the preset swallowing imaging pathology pattern library. Based on the registration error value, the leakage-aspiration simulation rating and the pharyngeal residue index are derived.

[0019] Furthermore, the virtual swallowing angiography parameter quantification module is also used for:

[0020] The temporal dynamic change characteristics of the functional connectivity strength matrix are regressed and mapped to the standard duration distribution of pharyngeal transport time in pharyngography. The pharyngeal transport time equivalent is generated by calculating the time delay for functional connectivity to reach its peak.

[0021] Furthermore, the aspiration pneumonia risk prediction module also includes a dynamic correction unit for acquiring the subject's clinical characteristic factors, including age, oral cleanliness score, tracheotomy status, and active cough ability grade. The clinical characteristic factors and the virtual swallowing imaging parameter set are input into the risk stratification model to obtain the corrected probability value and risk level of aspiration pneumonia.

[0022] Furthermore, the diagnostic report generation module is also used to bind the leakage-aspiration simulation rating and pharyngeal residue index in the virtual swallowing imaging parameter set with a preset color scale to generate a pseudo-color risk heat map overlaid on a standard brain template, and to fuse the risk heat map with the three-dimensional brain function activation map for display.

[0023] Furthermore, the preset swallowing imaging pathology model library is a quantitative comparative database established by extracting and annotating features from a large number of fMRI data of patients with swallowing disorders who have undergone swallowing imaging examinations, and using canonical correlation analysis to correlate brain region functional connectivity features with PAS rating, pharyngeal residue grading and pharyngeal transport time threshold.

[0024] Furthermore, the system also includes:

[0025] The model self-updating module is used to incrementally learn and update the risk stratification model by using the brain region temporal activation data, the virtual swallowing imaging parameter set, and the clinical outcome feedback data of the case as new training samples after each diagnostic report is generated.

[0026] Furthermore, the scanning parameters of the functional magnetic resonance imaging (fMRI) acquisition module during swallowing are set as follows: repetition time 2000ms, echo time 30ms, slice thickness 4mm, interslice spacing 0mm, and voxel size 3.5×3.5×4.0mm. 3 To match the hemodynamic response function of standard swallowing movements.

[0027] Compared with the prior art, the beneficial effects of the present invention are reflected in the following three aspects:

[0028] First, it achieves radiation-free equivalent quantitative assessment: This invention uses functional magnetic resonance imaging (fMRI) to map key quantitative indicators such as leakage-aspiration degree, pharyngeal residual state, and pharyngeal transport time, which are traditionally obtained by X-ray swallowing imaging, to the brain activation sequence and functional connectivity characteristics. This allows patients (especially those in the recovery period who need repeated follow-ups, children, and pregnant women) to obtain quantitative assessment parameters that are highly comparable to the gold standard of imaging without being exposed to ionizing radiation.

[0029] Second, it pioneers a new capability for assessing and locating rehabilitation targets in "functionally viable brain regions": this is the core innovation that distinguishes this invention from all existing technologies. Through blood oxygen level dependent (BOLD) signals from fMRI during swallowing, the system can directly identify brain regions that remain significantly activated during swallowing tasks. Even if these regions show structural damage on imaging, their BOLD activation proves that local neurons still have blood oxygen supply and functional viability, possessing compensatory potential. Clinicians can use the three-dimensional brain functional activation map output by this system to precisely locate treatment targets for transcranial magnetic stimulation, electroacupuncture, or neurofeedback training, actively inducing neural plasticity and reconstructing swallowing function—upgrading a diagnostic tool to a treatment decision-making tool.

[0030] Third, it provides clinically practical risk prediction: The system-integrated aspiration pneumonia risk prediction module, after dynamic correction of multi-center clinical factors (age, oral cleanliness, cough ability, etc.), can provide patients with a quantitative probability and risk level of pneumonia, directly guiding the choice of feeding method and preventive anti-infection treatment.

[0031] Fourth, this invention not only enables radiation-free equivalent quantitative assessment, but also pioneers new capabilities for assessing "functionally viable brain regions" and locating rehabilitation targets, establishes a model for preventing aspiration pneumonia, and provides reliable quantitative standards for clinical rehabilitation. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0033] Figure 1 This is a block diagram of a magnetic resonance imaging intelligent diagnostic system for swallowing disorders, provided as an embodiment of the present invention. Detailed Implementation

[0034] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0035] As attached Figure 1 As shown:

[0036] Example 1:

[0037] A magnetic resonance imaging intelligent diagnostic system for dysphagia includes:

[0038] The functional magnetic resonance imaging (fMRI) acquisition module during swallowing is used to acquire time-series images of whole-brain oxygen level-dependent (BOLD) signals in real time when the subject performs standard swallowing actions, and obtain brain region temporal activation data volume reflecting the functional viability and activity intensity of neurons in each brain region;

[0039] Specifically, the functional magnetic resonance acquisition module in the swallowing state uses synchronous trigger pulses and swallowing action sensors to timestamp each swallowing action, and aligns the timestamps with the whole brain functional magnetic resonance image sequence frame by frame to obtain brain region temporal activation data volume with swallowing event time-locked markers;

[0040] The scanning parameters for the functional magnetic resonance imaging (fMRI) acquisition module during swallowing were set as follows: repetition time 2000 ms, echo time 30 ms, slice thickness 4 mm, interslice spacing 0 mm, and voxel size 3.5 × 3.5 × 4.0 mm. 3 To match the hemodynamic response function of standard swallowing movements;

[0041] Furthermore, the standard swallowing task refers to the subject swallowing a fixed volume (e.g., 5 ml) of saline or thickened liquid orally at a pre-set rhythm (e.g., every 20 seconds), with the entire task lasting approximately 6 minutes.

[0042] Whole-brain functional magnetic resonance imaging sequence refers to time-series images of blood oxygen level dependent (BOLD) signals covering the entire brain, acquired using gradient echo planar imaging sequences, containing 180 to 240 time points.

[0043] Brain region temporal activation data volume refers to the four-dimensional data (three-dimensional space + temporal dimension) formed after the original DICOM image is corrected by head movement and registered to a standard space (such as MNI space). Each voxel corresponds to a BOLD signal change curve.

[0044] Before the data collection, the subjects were trained in swallowing movements, and a synchronous triggering device was used to record the actual time of each swallow.

[0045] Preferably, each swallowing action is time-stamped by synchronously triggering pulses and swallowing motion sensors (such as laryngeal piezoelectric sensors or neck electromyography electrodes), and the calibrated timestamps are aligned frame by frame with the fMRI image sequence, thereby eliminating the difference in response delay when the subject performs the task.

[0046] The specific alignment method is as follows: based on the swallowing start time recorded by the sensor, find the closest time point (at an integer multiple of TR) in the image sequence, and mark this time point as the swallowing event lock mark.

[0047] After this processing, each time point in the obtained brain region temporal activation data volume is clearly marked as being in the pre-swallowing, during-swallowing, or post-swallowing stage;

[0048] To ensure image quality matches the hemodynamic response during swallowing, the recommended scanning parameters for the functional magnetic resonance imaging (fMRI) acquisition module during swallowing are: repetition time 2000 ms, echo time 30 ms, slice thickness 4 mm, slice spacing 0 mm, flip angle 90°, and field of view 240 × 240 mm. 2 The matrix is ​​64×64, and the voxel size is 3.5×3.5×4.0mm. 3 Each cycle collects 180 time points. This parameter can effectively cover the BOLD signal peak in swallowing-related brain regions (usually occurring 4-6 seconds after swallowing).

[0049] The swallowing motor temporal decomposition module is used to segment and label the brain region temporal activation data volume based on the physiological phases of swallowing, namely the oral phase, pharyngeal phase, and esophageal phase, and quantify the motor intensity, generating a topological map that can intuitively display the range and intensity of brain region functional activation in each phase—wherein, the significantly activated areas directly represent surviving brain tissue with blood perfusion and neural function.

[0050] Specifically, the swallowing movement temporal decomposition module includes a pharyngeal movement segmentation unit, which is used to identify the pharyngeal start frame and end frame based on the zero-crossing point of the first derivative of the brain region activation intensity change curve corresponding to the hyoid-larynx complex, and calculate the average activation intensity of each brain region during the pharyngeal period as the quantitative value of the pharyngeal movement intensity, thereby constituting the phased brain region activation topology map.

[0051] Furthermore, the physiological phases of swallowing are divided into the oral phase (the tongue pushes the food bolus into the pharynx), the pharyngeal phase (the larynx rises, the glottis closes, and the upper esophageal sphincter opens), and the esophageal phase (the food bolus moves into the stomach via esophageal peristalsis).

[0052] The brain functional areas corresponding to each stage are different: the oral stage mainly activates the primary motor cortex (tongue area) and the supplementary motor area; the pharyngeal stage mainly activates the brainstem swallowing center, insula, anterior cingulate cortex, etc.; and the esophageal stage involves the dorsal nucleus of the vagus nerve, etc.

[0053] Segmentation labeling refers to determining the start and end frames of each period in an image sequence based on temporal activation patterns. Motion intensity quantization refers to calculating the average amplitude or peak value of activation signals in a specific brain region within that period.

[0054] The final generated brain region activation topology map is a three-dimensional brain map collection. Each brain map corresponds to a physiological period, and the voxel color in the map represents the average activation intensity (T value) of that period.

[0055] For the segmentation of the pharyngeal phase, the zero-crossing point of the first derivative of the activation intensity change curve of the brain regions corresponding to the hyoid-laryngostomy complex (such as the lower part of the precentral gyrus and the insula) is used to identify the start and end frames of the pharyngeal phase.

[0056] First, the average signal time-series curve of the bilateral sensorimotor cortex ROL (region of interest) is extracted; after smoothing and filtering the curve, its first derivative is calculated; the zero-crossing point where the derivative changes from positive to negative corresponds to the activation peak, that is, the moment when the pharyngeal phase function activity is most significant; one TR before and one after the peak is used as the pharyngeal phase window.

[0057] Subsequently, the average activation intensity of all voxels within the window was calculated as the pharyngeal motor intensity.

[0058] For the oral and esophageal phases, similar segmentation can be performed using signal changes in the tongue representation area and the dorsal nucleus of the vagus nerve, respectively.

[0059] Finally, by superimposing the activation intensity maps of the three periods onto the same standard brain template, a phased brain region activation topology map is obtained.

[0060] The virtual swallowing imaging parameter quantification module is used to compare the similarity of the phased brain region activation topology map with the preset swallowing imaging pathology pattern library and map it into a set of virtual swallowing imaging parameters including leakage-aspiration simulation rating, pharyngeal residue index and pharyngeal transport time equivalent.

[0061] Specifically, the virtual swallowing angiography parameter quantification module is used for:

[0062] The functional connectivity strength matrix between the brain region activation topology map of the time-phased brain regions and the pharyngeal sensory-motor cortex and the brainstem swallowing center is calculated. The functional connectivity strength matrix is ​​then registered with the typical abnormal connectivity patterns in the preset swallowing imaging pathology pattern library. Based on the registration error value, the leakage-aspiration simulation rating and the pharyngeal residue index are derived.

[0063] The virtual swallowing angiography parameter quantification module is also used for:

[0064] The temporal dynamic change characteristics of the functional connectivity strength matrix are regressed and mapped to the standard duration distribution of pharyngeal transport time in pharyngography. The pharyngeal transport time equivalent is generated by calculating the time delay for functional connectivity to reach its peak.

[0065] The preset swallowing imaging pathological pattern library is a quantitative control database established by extracting and annotating features from a large number of fMRI data of patients with swallowing disorders who have undergone swallowing imaging examinations, and using canonical correlation analysis to correlate brain region functional connectivity features with PAS rating, pharyngeal residue grading and pharyngeal transport time threshold.

[0066] Furthermore, the swallowing imaging pathology pattern library is a pre-built dataset that stores a large number of fMRI activation maps of patients diagnosed with dysphagia and corresponding radiographic swallowing imaging (VFSS) quantitative indicators (such as Rosenbek leak-aspiration scale PAS rating, pharyngeal residual volume percentage, and pharyngeal transport time in seconds).

[0067] Similarity comparison refers to matching the current subject's stage-specific brain region activation topology map with typical pathological patterns in the database to obtain the VFSS parameters of the most similar case, or predicting parameter values ​​through a regression model. Leakage-aspiration simulation rating refers to PAS levels 1-8 in the simulated VFSS; the pharyngeal residue index refers to a continuous value (0-1) simulating the proportion of residue in the vallecula epiglottis or pyriform fossa; the pharyngeal transport time equivalent refers to the time (in seconds) it takes for the simulated food bolus to pass from the pharynx to the upper esophageal sphincter.

[0068] Specifically, the functional connectivity strength matrix is ​​used as the comparison feature. The functional connectivity strength matrix between multiple key brain regions (including the pharyngeal sensory-motor cortex, brainstem swallowing center, insula, and anterior cingulate cortex) in the phased brain region activation topology map is calculated, that is, the Pearson correlation coefficient of the BOLD signal time-series curves between each pair of brain regions.

[0069] The matrix has a dimension of N×N (N represents 10-15 key regions). Then, the matrix is ​​matched with pre-calculated typical connectivity abnormality patterns in the swallowing contrast pathology pattern library—using spatial cosine similarity or a kernel-based matching algorithm to find the k closest cases in the library.

[0070] Based on the actual VFSS parameter values ​​of these k cases, the leakage-aspiration simulation rating and pharyngeal residue index of the current subject are derived by weighted averaging (weights are proportional to similarity). Furthermore, for the generation of pharyngeal transport time equivalents, the temporal dynamics of the functional connectivity strength matrix (i.e., the curves of each connectivity strength changing over time) are regressed and mapped to the standard duration distribution of pharyngeal transport time measured under VFSS.

[0071] Specifically, a support vector regression model is selected, with the input feature being the time delay of the peak functional connectivity strength within the pharyngeal window (relative to the start of the task), and the model output being the pharyngeal transport time equivalent (unit: seconds).

[0072] The greater the delay, the worse the pharyngeal nerve coordination, and the longer the simulated pharyngeal transport time.

[0073] The pre-defined swallowing imaging pathology model database is constructed as follows: by collecting a large amount of data from patients with swallowing disorders who have undergone both VFSS and fMRI examinations, canonical correlation analysis (CCA) is used to establish a quantitative comparison database between brain region functional connectivity characteristics and PAS ratings, pharyngeal residue grades, and pharyngeal transport time thresholds. This database needs to be updated regularly to ensure the accuracy of the matching.

[0074] The aspiration pneumonia risk prediction module is used to input the set of virtual swallowing imaging parameters into a risk stratification model validated by multi-center data, and calculate the probability value and risk level of aspiration pneumonia.

[0075] Specifically, the aspiration pneumonia risk prediction module also includes a dynamic correction unit, which is used to acquire the subject's clinical characteristic factors, including age, oral cleanliness score, tracheotomy status and active cough ability grade, and input the clinical characteristic factors and the virtual swallowing imaging parameter set into the risk stratification model to obtain the corrected probability value and risk level of aspiration pneumonia.

[0076] The phased brain region activation topology map itself can serve as a key basis for clinical rehabilitation decisions. Brain regions exhibiting significant BOLD signal activation, even if they appear as damaged areas on structural imaging (such as conventional MRI or CT), can still be identified as viable tissue with potential for blood oxygenation and functional compensation. Clinicians can use this to locate interventional targets for transcranial magnetic stimulation or electroacupuncture, implement precise neuromodulation, and promote the reconstruction of swallowing function.

[0077] Furthermore, the risk stratification model is a predictive model trained and validated with multicenter clinical data. The input variables are a set of virtual swallowing imaging parameters (including leakage-aspiration simulation rating, pharyngeal residue index, and pharyngeal transport time equivalent) and optional clinical characteristic factors. The output is the predicted probability of aspiration pneumonia (a continuous value between 0 and 1) and the risk level (e.g., low risk <0.2, medium risk 0.2~0.5, high risk >0.5).

[0078] This module also includes a dynamic correction unit. This unit acquires the subject's clinical characteristics, including age (continuous value), oral cleanliness score (0-IV), tracheostomy status (yes / no), and active cough ability grade (1-3). These factors, along with the virtual swallowing imaging parameter set, are input into the risk stratification model.

[0079] The model can employ an ensemble learning framework (such as random forest or XGBoost) to integrate the aforementioned clinical factors and virtual imaging parameters during the modeling phase. For new subjects, the model automatically outputs a corrected probability value for aspiration pneumonia.

[0080] The risk levels are classified as follows: a probability value <0.1 indicates extremely low risk, 0.1–0.3 indicates low risk, 0.3–0.6 indicates medium risk, and >0.6 indicates high risk. Clinicians can determine whether prophylactic antibiotics or changes in feeding methods are necessary based on the risk level.

[0081] The diagnostic report generation module is used to generate a visual diagnostic report containing a three-dimensional brain function activation map and virtual imaging quantitative indicators based on the risk level and the virtual swallowing imaging parameter set.

[0082] Specifically, the diagnostic report generation module is also used to bind the leakage-aspiration simulation rating and pharyngeal residue index in the virtual swallowing imaging parameter set with a preset color scale to generate a pseudo-color risk heat map overlaid on a standard brain template, and to fuse the risk heat map with the three-dimensional brain function activation map for display.

[0083] Furthermore, a three-dimensional brain functional activation map refers to overlaying a topological map of brain region activation at different time periods onto a standard brain structural image or a rendered three-dimensional brain model, displaying the activation intensity of each brain region during a swallowing task in pseudo-color or heat map form.

[0084] Virtual imaging quantitative indicators include leakage-aspiration simulation rating, pharyngeal residue index, pharyngeal transport time equivalent, and their corresponding normal reference ranges. Reports should be presented in a structured, visually friendly format for easy interpretation by clinicians.

[0085] The diagnostic report generation module first binds the leakage-aspiration simulation rating (levels 1-8) from the virtual swallowing imaging parameter set to a preset color scale: level 1-2 is represented by green, level 3-4 by yellow, level 5-6 by orange, and level 7-8 by red. The pharyngeal residue index (0-1) is mapped to varying shades of blue.

[0086] Then, these pseudo-colors are overlaid on the corresponding functional areas of the standard brain template—for example, color patches representing leakage-aspiration risk are overlaid on the brainstem swallowing center and insula region; color patches representing pharyngeal residue are overlaid on the sensorimotor cortex.

[0087] The final output is a composite visualization report that combines a three-dimensional brain functional activation map (showing the original activation intensity of each brain region) and a risk heatmap (showing the risk of virtual imaging parameter mapping).

[0088] The report also lists specific values, normal reference values, and clinical recommendations in tabular form. This report can be directly exported as a PDF or DICOM structured report and uploaded to the hospital's PACS system.

[0089] Specifically, the system also includes:

[0090] The model self-updating module is used to incrementally learn and update the risk stratification model by using the brain region temporal activation data, the virtual swallowing imaging parameter set, and the clinical outcome feedback data of the case as new training samples after each diagnostic report is generated.

[0091] Specifically, after each diagnostic report is generated, if the clinical outcome of the case (such as whether aspiration pneumonia subsequently occurred) is confirmed through follow-up, the system uses the brain region temporal activation data, the virtual swallowing imaging parameter set, and the actual clinical outcome as new training samples to incrementally learn and update the risk stratification model. Online random forests or incremental support vector machines can be used to achieve continuous model optimization and avoid data obsolescence.

[0092] As can be seen from the above, the beneficial effects of the present invention are reflected in the following three aspects:

[0093] First, it achieves radiation-free equivalent quantitative assessment: This invention uses functional magnetic resonance imaging (fMRI) to map key quantitative indicators such as leakage-aspiration degree, pharyngeal residual state, and pharyngeal transport time, which are traditionally obtained by X-ray swallowing imaging, to the brain activation sequence and functional connectivity characteristics. This allows patients (especially those in the recovery period who need repeated follow-ups, children, and pregnant women) to obtain quantitative assessment parameters that are highly comparable to the gold standard of imaging without being exposed to ionizing radiation.

[0094] Second, it pioneers a new capability for assessing and locating rehabilitation targets in "functionally viable brain regions": this is the core innovation that distinguishes this invention from all existing technologies. Through blood oxygen level dependent (BOLD) signals from fMRI during swallowing, the system can directly identify brain regions that remain significantly activated during swallowing tasks. Even if these regions show structural damage on imaging, their BOLD activation proves that local neurons still have blood oxygen supply and functional viability, possessing compensatory potential. Clinicians can use the three-dimensional brain functional activation map output by this system to precisely locate treatment targets for transcranial magnetic stimulation, electroacupuncture, or neurofeedback training, actively inducing neural plasticity and reconstructing swallowing function—upgrading a diagnostic tool to a treatment decision-making tool.

[0095] Third, it provides clinically practical risk prediction: The system-integrated aspiration pneumonia risk prediction module, after dynamic correction based on multi-center clinical factors (age, oral cleanliness, cough ability, etc.), can provide patients with a quantitative probability and risk level of pneumonia, directly guiding the choice of feeding method and preventive anti-infection treatment.

[0096] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A magnetic resonance imaging intelligent diagnostic system for dysphagia, characterized in that, include: The functional magnetic resonance imaging (fMRI) acquisition module during swallowing is used to acquire time-series images of whole-brain blood oxygenation level-dependent signals in real time when the subject performs standard swallowing actions, and obtain brain region temporal activation data volume reflecting the functional viability and activity intensity of neurons in each brain region; The swallowing motor temporal decomposition module is used to segment and label the brain region temporal activation data volume based on the physiological phases of swallowing, namely the oral phase, pharyngeal phase, and esophageal phase, and quantify the motor intensity, generating a topological map that can intuitively display the range and intensity of brain region functional activation in each phase—wherein, the significantly activated areas directly represent surviving brain tissue with blood perfusion and neural function. The virtual swallowing imaging parameter quantification module is used to compare the similarity of the phased brain region activation topology map with the preset swallowing imaging pathology pattern library and map it into a set of virtual swallowing imaging parameters including leakage-aspiration simulation rating, pharyngeal residue index and pharyngeal transport time equivalent. The aspiration pneumonia risk prediction module is used to input the set of virtual swallowing imaging parameters into a risk stratification model validated by multi-center data, and calculate the probability value and risk level of aspiration pneumonia. The diagnostic report generation module is used to generate a visual diagnostic report containing a three-dimensional brain function activation map and virtual imaging quantitative indicators based on the risk level and the virtual swallowing imaging parameter set.

2. The intelligent magnetic resonance imaging diagnostic system for dysphagia according to claim 1, characterized in that, The functional magnetic resonance imaging (fMRI) acquisition module during swallowing uses synchronous trigger pulses and swallowing motion sensors to timestamp each swallowing action, and aligns the timestamps with the whole-brain fMRI image sequence frame by frame to obtain brain region temporal activation data with swallowing event time-locked markers.

3. The intelligent magnetic resonance imaging diagnostic system for dysphagia according to claim 1, characterized in that, The swallowing movement temporal decomposition module includes a pharyngeal movement segmentation unit, which is used to identify the pharyngeal start frame and end frame based on the zero-crossing point of the first derivative of the brain region activation intensity change curve corresponding to the hyoid-larynx complex, and calculate the average activation intensity of each brain region during the pharyngeal period as the quantitative value of the pharyngeal movement intensity, thereby constructing the phased brain region activation topology map.

4. The intelligent magnetic resonance imaging diagnostic system for dysphagia according to claim 1, characterized in that, The virtual swallowing angiography parameter quantification module is used for: The functional connectivity strength matrix between the brain region activation topology map of the time-phased brain regions and the pharyngeal sensory-motor cortex and the brainstem swallowing center is calculated. The functional connectivity strength matrix is ​​then registered with the typical abnormal connectivity patterns in the preset swallowing imaging pathology pattern library. Based on the registration error value, the leakage-aspiration simulation rating and the pharyngeal residue index are derived.

5. A magnetic resonance imaging intelligent imaging diagnostic system for dysphagia according to claim 4, characterized in that, The virtual swallowing angiography parameter quantification module is also used for: The temporal dynamic change characteristics of the functional connectivity strength matrix are regressed and mapped to the standard duration distribution of pharyngeal transport time in pharyngography. The pharyngeal transport time equivalent is generated by calculating the time delay for functional connectivity to reach its peak.

6. The intelligent magnetic resonance imaging diagnostic system for dysphagia according to claim 1, characterized in that, The aspiration pneumonia risk prediction module also includes a dynamic correction unit, which is used to acquire the subject's clinical characteristic factors, including age, oral cleanliness score, tracheotomy status and active cough ability grade, and input the clinical characteristic factors and the virtual swallowing imaging parameter set into the risk stratification model to obtain the corrected probability value and risk level of aspiration pneumonia.

7. A magnetic resonance imaging intelligent imaging diagnostic system for dysphagia according to claim 1, characterized in that, The diagnostic report generation module is also used to bind the leakage-aspiration simulation rating and pharyngeal residue index in the virtual swallowing imaging parameter set with a preset color scale to generate a pseudo-color risk heat map overlaid on a standard brain template, and to fuse the risk heat map with the three-dimensional brain function activation map for display.

8. A magnetic resonance imaging intelligent imaging diagnostic system for dysphagia according to claim 1, characterized in that, The pre-defined swallowing imaging pathology model library is a quantitative comparative database established by extracting and annotating features from a large number of fMRI data of patients with swallowing disorders who have undergone swallowing imaging examinations, and using canonical correlation analysis to correlate brain region functional connectivity features with PAS rating, pharyngeal residue grading, and pharyngeal transport time threshold.

9. A magnetic resonance imaging intelligent imaging diagnostic system for dysphagia according to claim 1, characterized in that, The system also includes: The model self-updating module is used to incrementally learn and update the risk stratification model by using the brain region temporal activation data, the virtual swallowing imaging parameter set, and the clinical outcome feedback data of the case as new training samples after each diagnostic report is generated.

10. A magnetic resonance imaging intelligent imaging diagnostic system for dysphagia according to claim 1, characterized in that, The scanning parameters for the functional magnetic resonance imaging (fMRI) acquisition module during swallowing were set as follows: repetition time 2000 ms, echo time 30 ms, slice thickness 4 mm, interslice spacing 0 mm, and voxel size 3.5 × 3.5 × 4.0 mm. 3 To match the hemodynamic response function of standard swallowing movements.

Images