A dysphagia screening system based on intelligent visual recognition

Abstract

This invention relates to the field of visual recognition technology, and more particularly to a swallowing disorder screening system based on intelligent visual recognition, comprising: an initial detection and analysis module for determining an initial swallowing detection area and acquiring motion state parameters of the initial swallowing detection area during a target object's execution of a preset action; a target detection and analysis module for determining the detection confidence level of the initial swallowing detection area to determine the target swallowing detection area; a visual recognition and analysis module for acquiring a region image sequence of the target swallowing detection area during the target object's execution of the preset action and determining the motion coordination rate of each swallowing-related point; a preliminary judgment module for determining the preliminary swallowing type of the target object; and a secondary judgment module for determining the secondary swallowing type of the target object if the preliminary swallowing type is suspected to be a swallowing disorder. This invention can improve the accuracy of swallowing disorder screening.

Description

Technical Field

[0001] This invention relates to the field of visual recognition technology, and in particular to a swallowing disorder screening system based on intelligent visual recognition. Background Technology

[0002] Swallowing, an indispensable physiological action in the human body during eating and drinking, relies on the coordinated action of multiple soft tissues and skeletal structures, including the hyoid bone in the neck, larynx, pharyngeal wall, and upper esophageal sphincter. Abnormalities in the coordination, range of motion, and timing of these movements can easily lead to dysphagia. This condition is prevalent among the elderly, stroke survivors, those recovering from neck injuries, and individuals with impaired neurological function. It not only causes difficulty eating and insufficient nutrient intake but can also lead to food leakage and aspiration, potentially causing aspiration pneumonia and endangering personal health. Therefore, efficient and accurate early screening for dysphagia is of significant clinical value for disease intervention, rehabilitation, and risk control.

[0003] Currently, commonly used clinical methods for screening dysphagia are mainly divided into two categories: manual assessment and invasive instrument testing. Manual assessment relies on medical staff's clinical experience to observe the patient's swallowing performance and combine this with medical history to determine the level of dysphagia. The assessment results are highly subjective, with different medical staff using different criteria, leading to potential misjudgments and missed diagnoses. Furthermore, the assessment efficiency is low, making it difficult to meet the needs of rapid screening for large populations. Invasive testing often involves inserting instruments such as endoscopes and pressure testing devices into the pharyngeal cavity to collect data. The testing process can cause significant discomfort to the subject due to a foreign body sensation.

[0004] As visual recognition technology is gradually applied to the field of physiological behavior detection, some existing swallowing detection devices have begun to rely on image acquisition to capture neck movement images. However, existing visual detection solutions generally have obvious technical defects. For example, the detection area adopts a fixed delineation mode and does not adjust the range according to individual physiological differences such as neck fat distribution, muscle morphology, and bone position of different subjects. This easily leads to problems such as the omission of effective movement features and interference from irrelevant surface movement signals. The initial delineation area cannot be dynamically corrected according to the actual swallowing movement response, which limits the accuracy of subsequent feature recognition and the accuracy of screening results. The use of single-point sensors to analyze and record the movement trajectory of the larynx during swallowing results in a single analysis dimension.

[0005] In summary, existing swallowing disorder screening technologies suffer from numerous drawbacks, such as overly coarse or fixed detection area positioning, lack of dynamic verification mechanisms for the effectiveness of detection areas, and limited analytical dimensions of swallowing movements. Therefore, developing an intelligent swallowing disorder screening system that can adaptively define and dynamically correct detection areas and analyze swallowing movements from multiple dimensions has become an urgent technical problem to be solved in this field. Summary of the Invention

[0006] To address this issue, the present invention provides a swallowing disorder screening system based on intelligent visual recognition, which overcomes the problem of insufficient accuracy in swallowing disorder screening caused by fixed detection areas and neglect of individual physiological differences in the prior art.

[0007] To achieve the above objectives, the present invention provides a swallowing disorder screening system based on intelligent visual recognition, comprising: The initial detection and analysis module is used to determine the initial swallowing detection area based on the target physiological state parameters of the target object's neck preset area, and to obtain the motion state parameters of the initial swallowing detection area during the target object's execution of preset actions; The target detection and analysis module is used to determine the detection confidence of the initial swallowing detection area based on the motion state parameters of the initial swallowing detection area during the target object's execution of a preset action, and to determine the target swallowing detection area based on the detection confidence. The visual recognition and analysis module is used to acquire a sequence of regional images of the target swallowing detection area during the target object's execution of a preset action, and to determine the motion trajectory of several swallowing-related points based on the regional image sequence, so as to determine the motion coordination rate of each swallowing-related point. The preliminary judgment module is used to determine the preliminary swallowing type of the target object based on the motor coordination rate, wherein the preliminary swallowing type includes swallowing disorder type, suspected swallowing disorder type, and no swallowing disorder type; The secondary determination module is used to determine the maximum swallowing characterization value based on the regional image sequence when the initial determination of the swallowing type of the target object is a suspected swallowing disorder type, so as to determine the secondary determination of the swallowing type of the target object. The secondary determination of the swallowing type includes a swallowing disorder type and a no-swallowing disorder type.

[0008] Furthermore, the initial detection and analysis module includes: An initial state analysis unit is used to determine the target adjustment coefficient based on the comparison results between the target physiological state parameters and the standard physiological state parameters; An initial adjustment analysis unit is used to determine the initial swallowing detection area based on the target adjustment coefficient and the standard detection area.

[0009] Furthermore, the target detection and analysis module includes: The parameter analysis unit is used to divide the initial swallowing detection area into several detection sub-regions, and construct corresponding parameter change curves based on the motion state parameters of each detection sub-region during the target object's execution of a preset action. The confidence analysis unit is used to determine several key change feature points based on the change curves of each parameter, and to determine the detection confidence of the initial swallowing detection area based on the correlation of each key change feature point.

[0010] Furthermore, the target detection and analysis module also includes: The target adjustment analysis unit is used to determine whether to adjust the initial swallowing detection area based on the comparison result between the detection confidence and the preset confidence, so as to determine the target swallowing detection area.

[0011] Furthermore, the visual recognition and analysis module includes: A visual recognition unit is used to acquire a sequence of regional images of the target swallowing detection area during the target object's performance of a preset action; A visual analysis unit, connected to a visual recognition unit, is used to determine the motion trajectories of several swallowing-related points based on the image sequence of the region, and to determine the overlap of the trajectories of any two swallowing-related points based on the motion trajectories of each swallowing-related point, so as to determine the motion coordination rate of each swallowing-related point.

[0012] Furthermore, the preliminary determination module determines the preliminary swallowing type of the target object based on the comparison results of the motor coordination rate with the first preset coordination rate and the second preset coordination rate, wherein, If the motor coordination rate is less than the first preset coordination rate, the preliminary judgment of the target object's swallowing type is swallowing disorder. If the motor coordination rate is greater than or equal to the first preset coordination rate and less than the second preset coordination rate, then the preliminary judgment of the target object's swallowing type is determined to be a suspected swallowing disorder. If the motor coordination rate is greater than or equal to the second preset coordination rate, the preliminary judgment of the target object's swallowing type is determined to be the type without swallowing disorder.

[0013] Furthermore, the secondary determination module determines the maximum movement distance of several detection points within the target swallowing detection area during the target object's execution of a preset action based on the regional image sequence, and determines the maximum swallowing characterization value based on the maximum movement distance of each detection point.

[0014] Furthermore, the secondary determination module determines the secondary swallowing type of the target object based on the comparison result between the maximum swallowing characteristic value and the preset swallowing characteristic value.

[0015] Furthermore, the target adjustment analysis unit determines, based on the first determination condition, to increase or adjust the initial swallowing detection area to determine the target swallowing detection area; The first determination condition is that the detection confidence level is less than the preset confidence level.

[0016] Furthermore, the target adjustment analysis unit determines, based on the second determination condition, that the initial swallowing detection area will not be adjusted, and determines the initial swallowing detection area as the target swallowing detection area; The second determination condition is that the detection confidence level is greater than or equal to the preset confidence level.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention adaptively determines the initial swallowing detection area by combining the physiological state parameters of the target object, and dynamically corrects the detection area according to the running state during the execution of the preset action. This accurately locks the core response area of ​​the swallowing action, achieving adaptive contraction and calibration of the detection range, avoiding interference from irrelevant factors, and improving the accuracy of subsequent visual recognition. By determining the motion trajectory of several swallowing-related points through the regional image sequence of the target swallowing detection area during the target object's execution of the preset action, the motion coordination rate of each swallowing-related point can be adaptively determined, quantifying the degree of multi-point structural linkage and coordination of the target object, objectively reflecting the coordination of swallowing movements. Using the motion coordination rate as the criterion, three preliminary swallowing types are classified, and stratified screening and initial identification are completed, improving screening efficiency. Furthermore, through secondary judgment, a second in-depth analysis is performed on suspected swallowing disorder types, extracting the maximum swallowing characteristic value to complete the verification judgment, thereby improving the accuracy and reliability of swallowing disorder screening.

[0018] Furthermore, the initial detection and analysis module compares the collected individual target physiological state parameters with the standard physiological state parameters to quantify the target adjustment coefficient that adapts to the individual's physical characteristics. This objectively quantifies the differences in the neck physiological structure of different subjects. By using the target adjustment coefficient to adaptively modify and adjust the standard detection area, the initial swallowing detection area is determined. This reduces the deviation in area delineation and ensures that the area adjustment is data-supported and reasonable.

[0019] Furthermore, the target detection and analysis module further refines the initial swallowing detection area into sub-regions, simultaneously collects motion state parameters for each region, and constructs parameter change curves. This allows for a clear visualization of the differences in swallowing action responses in different local areas, facilitating accurate identification of effective motion areas and interference areas. By extracting key change feature points from the parameter change curves, the module determines the overall detection confidence level of the region based on the correlation between these feature points. Combined with the comparison results with the preset confidence level, the module optimizes and adjusts the initial swallowing detection area, ultimately obtaining the target swallowing detection area. This achieves dynamic calibration of the detection range, enhances the ability to capture subsequent effective motion features, and further improves the accuracy of swallowing disorder screening.

[0020] Furthermore, the visual recognition and analysis module continuously acquires dynamic image sequences of the target swallowing detection area, accurately tracks the motion trajectories of multiple swallowing-related points based on the regional image sequences, and quantitatively calculates the motion coordination rate by calculating the overlap of trajectories between points, reflecting the overall degree of coordination of the swallowing structure. Attached Figure Description

[0021] Figure 1 This is a structural block diagram of the swallowing disorder screening system based on intelligent visual recognition, according to an embodiment of the present invention. Figure 2 This is a structural block diagram of the initial detection and analysis module in an embodiment of the present invention; Figure 3 This is a structural block diagram of the target detection and analysis module according to an embodiment of the present invention; Figure 4 This is a structural block diagram of the visual recognition and analysis module in an embodiment of the present invention. Detailed Implementation

[0022] 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.

[0023] 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.

[0024] Please see Figure 1 The diagram shown is a structural block diagram of a swallowing disorder screening system based on intelligent visual recognition according to an embodiment of the present invention. The swallowing disorder screening system based on intelligent visual recognition provided in this embodiment of the present invention includes: The initial detection and analysis module is used to determine the initial swallowing detection area based on the target physiological state parameters of the target object's neck preset area, and to obtain the motion state parameters of the initial swallowing detection area during the target object's execution of preset actions; Please see Figure 2 The diagram shown is a structural block diagram of the initial detection and analysis module according to an embodiment of the present invention; specifically, the initial detection and analysis module includes: Physiological state detection unit, which is used to acquire target physiological state parameters of a preset area of ​​the neck of the target object; An initial state analysis unit, connected to a physiological state detection unit, is used to determine a target adjustment coefficient based on the comparison results between the target physiological state parameters and standard physiological state parameters. An initial adjustment analysis unit, connected to the initial state analysis unit, is used to determine the initial swallowing detection area based on the target adjustment coefficient and the standard detection area; The motion detection unit, which is connected to the initial adjustment analysis unit, is used to obtain the motion state parameters of the initial swallowing detection area during the target object's execution of a preset action.

[0025] In this embodiment, the preset neck region is a pre-defined neck swallowing-related region, with the thyroid cartilage as the core positioning landmark. It can be delineated based on an expert system. Preferably, the boundaries of the preset neck region are marked by anatomical landmarks. The upper boundary is a horizontal line 10mm above the upper edge of the hyoid bone, the lower boundary is a horizontal line 10mm below the lower edge of the cricoid cartilage, and the left and right boundaries are the range of the line connecting the anterior edges of the two sternocleidomastoid muscles. Physiological parameters include, but are not limited to, fat layer thickness, cross-sectional area of ​​swallowing muscles, height of laryngeal bone protrusion, and elastic modulus of soft tissue. Target physiological parameters are used to characterize individual physiological differences in the target subject, while standard physiological parameters are used to characterize the standard physiological state of the standard subject. Practitioners can build and store a database to pre-store standard physiological parameter datasets adapted to different age groups and body types. Standard physiological parameters include standard fat layer thickness, standard swallowing muscle cross-sectional area, standard laryngeal bone protrusion height, and standard soft tissue elastic modulus in a preset area of ​​the neck corresponding to the standard body type. Each standard physiological parameter has a corresponding standard detection area, which can be obtained through a limited number of recognition tests or expert system marking. The standard detection area is the optimal swallowing disorder observation area distributed on the outer surface of the neck corresponding to the standard physiological parameter.

[0026] Understandably, the target adjustment coefficient is used to characterize the degree of deviation between the target physiological state parameters of the target object and its corresponding standard physiological state parameters. A larger target adjustment coefficient indicates a greater degree of deviation between the target physiological state parameters and their corresponding standard physiological state parameters, and thus a greater degree of adjustment of the initial standard detection region relative to the standard detection region. This can be determined based on the similarity between the target physiological state parameters and the standard physiological state parameters. The sum of the similarity between the target physiological state parameters and the standard physiological state parameters and the target adjustment coefficient is 1. In practical applications, the similarity between the target physiological state parameters and the standard physiological state parameters can be calculated based on the Pearson similarity coefficient algorithm, which is existing technology and will not be elaborated further. The target adjustment coefficient is used as a scaling factor to process the standard detection region to obtain the initial swallowing detection region.

[0027] Understandably, the preset action could be the voluntary swallowing of 5ml of room temperature purified water. The motion state parameters include, but are not limited to, the longitudinal lifting distance and lateral offset of the tissue in the area. The corresponding motion state parameters can be obtained based on an inertial measurement unit or a high-speed industrial camera.

[0028] Specifically, the initial detection and analysis module compares the collected individual target physiological state parameters with standard physiological state parameters to quantify the target adjustment coefficient that adapts to individual physical signs. This objectively quantifies the differences in the neck physiological structure of different subjects. By using the target adjustment coefficient to adaptively modify and adjust the standard detection area, the initial swallowing detection area is determined. This reduces the deviation in area delineation and ensures that the area adjustment is data-supported and reasonable.

[0029] The target detection and analysis module, which is connected to the initial detection and analysis module, is used to determine the detection confidence of the initial swallowing detection area based on the motion state parameters of the initial swallowing detection area during the target object's execution of a preset action, and to determine the target swallowing detection area based on the detection confidence. Please see Figure 3 The diagram shown is a structural block diagram of the target detection and analysis module according to an embodiment of the present invention; specifically, the target detection and analysis module includes: The parameter analysis unit is used to divide the initial swallowing detection area into several detection sub-regions, and construct corresponding parameter change curves based on the motion state parameters of each detection sub-region during the target object's execution of a preset action. A confidence analysis unit, connected to the parameter analysis unit, is used to determine several key change feature points based on the parameter change curves, and to determine the detection confidence of the initial swallowing detection area based on the correlation of the key change feature points. The target adjustment analysis unit, which is connected to the confidence analysis unit, is used to determine whether to adjust the initial swallowing detection area based on the comparison result between the detection confidence and the preset confidence, so as to determine the target swallowing detection area.

[0030] In this embodiment, the initial swallowing detection area is uniformly divided into several detection sub-regions. If the size of the detection sub-region is too large, local motion details will be lost; if the compensation is too small, computational redundancy will increase. Preferably, the size of the detection sub-region can be set to 5mm × 5mm. In practical applications, the motion state parameters of the center point of the detection sub-region are used as the motion state parameters corresponding to the detection sub-region, or the average value of the motion state parameters at each position of the detection sub-region is used as the motion state parameters corresponding to the detection sub-region. With time as the independent variable and the motion state parameters as the dependent variable, parameter change curves are constructed for the corresponding parameters. For example, the parameter change curve corresponding to the longitudinal lifting distance is constructed with time as the independent variable and the longitudinal lifting distance as the dependent variable.

[0031] Understandably, key change feature points are used to characterize points where the swallowing state differs from that of normal individuals during the swallowing process. For any detection sub-region, during the target object's execution of a preset action, the slope of the curve at each position is calculated based on the parameter change curve corresponding to the longitudinal lifting distance. The time point corresponding to the peak value is determined as the key time point. If the absolute value of the maximum slope before the key time point is less than the absolute value of the maximum slope after the key time point, then the center point of the detection sub-region is determined as the key change feature point. Alternatively, based on the parameter change curve corresponding to the lateral offset, if the curve has multiple peaks, then the center point of the detection sub-region is determined as the key change feature point. Detection confidence is used to characterize the credibility of judging the target object as having a swallowing disorder. It can be determined based on the clustering state and number of each key change feature point. If the clustering state of each key change feature point is relatively scattered and the number is small, then the credibility of judging the target object as having a swallowing disorder is relatively low, and in this case, the detection confidence is low. Preferably, the key change feature points can be clustered to obtain several cluster sets. Each cluster set includes at least one key change feature point, and the detection sub-regions corresponding to each key change feature point in each cluster set are at least pairwise adjacent. Cluster sets with more than a preset number of key change feature points are defined as key cluster sets. The ratio of the total area of ​​the regions corresponding to the key change feature points in the key cluster sets to the area of ​​the preset neck region is defined as the detection confidence level. The larger the preset number, the higher the required degree of aggregation of the key change feature points in the cluster sets. Preferably, the preset number is set to 1 / 3 to 1 / 4 of the total number of detection sub-regions.

[0032] Specifically, the target adjustment analysis unit determines to increase the size of the initial swallowing detection area based on a first determination condition to determine the target swallowing detection area; wherein, the first determination condition is that the detection confidence level is less than a preset confidence level.

[0033] Specifically, the target adjustment analysis unit determines that the initial swallowing detection area will not be adjusted based on the second determination condition, and determines the initial swallowing detection area as the target swallowing detection area; wherein, the second determination condition is that the detection confidence level is greater than or equal to the preset confidence level.

[0034] Understandably, a higher preset reliability level requires a higher level of explicitness in the initial swallowing detection area. If the detection confidence level is lower than the preset reliability level, it indicates that the current initial swallowing detection area has insufficient coverage or that the target subject may not have a swallowing disorder. To further determine the target subject's condition, the initial swallowing detection area can be enlarged by expanding its boundaries to include surrounding tissues within the detection range, compensating for the initial inadequacy in localization, and the detection can be re-evaluated. The implementer can adaptively set the enlargement percentage; for example, the ratio of the preset reliability level to the detection confidence level can be used to determine the enlargement ratio and adjust the initial swallowing detection area to obtain the target swallowing detection area. If the detection confidence level is greater than or equal to the preset reliability level, it indicates that the current initial swallowing detection area boundary matches the target subject's neck physiological signs and the target subject is highly likely to have a swallowing disorder; therefore, no adjustment to the initial swallowing detection area is necessary. In practical applications, the detection confidence level ranges from 0 to 1. To ensure the reliability of the target swallowing detection area, the preset confidence level can be set to 0.6 or higher. Alternatively, during the generalization validation phase before system deployment, the detection confidence level calculation results from no fewer than 500 patients with swallowing disorders from no fewer than 3 clinical centers under the initial swallowing detection area are collected to construct the overall distribution of detection confidence level in order to determine the preset confidence level.

[0035] Specifically, the target detection and analysis module performs refined sub-regional segmentation of the initial swallowing detection area, synchronously collects motion state parameters in each region, and constructs parameter change curves. This allows for a visual representation of the differences in swallowing action responses in different local areas, facilitating accurate identification of effective motion areas and interference areas. By extracting key change feature points from the parameter change curves, the overall detection confidence of the region is determined based on the correlation between these feature points. Combined with the comparison results with the preset confidence level, the initial swallowing detection area is optimized and adjusted, ultimately yielding the target swallowing detection area. This achieves dynamic calibration of the detection range, enhances the ability to capture subsequent effective motion features, and further improves the accuracy of swallowing disorder screening.

[0036] The visual recognition and analysis module, which is connected to the target detection and analysis module, is used to obtain the regional image sequence of the target swallowing detection area during the target object's performance of a preset action, and to determine the motion trajectory of several swallowing-related points based on the regional image sequence, so as to determine the motion coordination rate of each swallowing-related point. Please see Figure 4 The diagram shown is a structural block diagram of the visual recognition analysis module according to an embodiment of the present invention; specifically, the visual recognition analysis module includes: A visual recognition unit is used to acquire a sequence of regional images of the target swallowing detection area during the target object's performance of a preset action; A visual analysis unit, connected to a visual recognition unit, is used to determine the motion trajectories of several swallowing-related points based on the image sequence of the region, and to determine the overlap of the trajectories of any two swallowing-related points based on the motion trajectories of each swallowing-related point, so as to determine the motion coordination rate of each swallowing-related point.

[0037] In this embodiment, swallowing association points are feature points within the target swallowing detection area that have clear anatomical functions and motor coordination relationships. The swallowing association points of healthy individuals without swallowing disorders exhibit extremely high similarity in their movement states during swallowing, such as synchronized start and stop, and consistent movement speed and trajectory. Practitioners can mark the target swallowing detection area based on visual images of several healthy individuals without swallowing disorders performing a preset action to obtain several swallowing association points and their corresponding movement trajectories. The region image sequence includes several frames of region images of the target swallowing detection area during the target object's performance of the preset action, arranged in chronological order. For any swallowing association point, time is used as the independent variable, the position of the swallowing association point in the first frame of the region image is taken as the origin, and the position coordinates in each region image in the region image sequence are taken as the dependent variable to generate the corresponding movement trajectory. The degree of overlap between the trajectories of any two swallowing association points can characterize the morphological similarity of their motion trajectories. The greater the degree of overlap between the trajectories of two swallowing association points, the greater the morphological similarity of their motion trajectories. Preferably, the degree of overlap between the trajectories of any two swallowing association points can be calculated using algorithms such as the Fraser distance algorithm, Euclidean distance algorithm, and Pearson correlation coefficient. It is understandable that the motor coordination rate is used to characterize the consistency of the overall motor trend of each swallowing association point. The higher the motor coordination rate, the greater the consistency of the overall motor trend of each swallowing association point. The motor coordination rate can be determined based on the trajectory overlap of each pair of swallowing association points. Preferably, the motor coordination rate can be determined as the ratio of the average trajectory overlap of each pair of swallowing association points to the maximum trajectory overlap of each pair of swallowing association points.

[0038] Specifically, the visual recognition and analysis module continuously acquires dynamic image sequences of the target swallowing detection area, accurately tracks the motion trajectories of multiple swallowing-related points based on the regional image sequences, and quantitatively calculates the motion coordination rate by calculating the overlap of trajectories between points, reflecting the overall degree of coordination of the swallowing structure.

[0039] The preliminary judgment module, which is connected to the visual recognition and analysis module, is used to determine the preliminary swallowing type of the target object based on the motor coordination rate. The preliminary swallowing type includes swallowing disorder type, suspected swallowing disorder type, and no swallowing disorder type. Specifically, the preliminary determination module determines the preliminary swallowing type of the target object based on the comparison results of the motor coordination rate with the first preset coordination rate and the second preset coordination rate, wherein, If the motor coordination rate is less than the first preset coordination rate, the preliminary judgment of the target object's swallowing type is swallowing disorder. If the motor coordination rate is greater than or equal to the first preset coordination rate and less than the second preset coordination rate, then the preliminary judgment of the target object's swallowing type is determined to be a suspected swallowing disorder. If the motor coordination rate is greater than or equal to the second preset coordination rate, the preliminary judgment of the target object's swallowing type is determined to be the type without swallowing disorder.

[0040] In this embodiment, due to factors such as varying swallowing effort, slight neck posture deviation, and minor differences in individual swallowing movements, the motor coordination rate may fluctuate slightly within a normal range. The dual threshold setting provides an intermediate buffer zone for potential dysphagia, preventing direct misjudgment of swallowing disorders based on minor numerical fluctuations. Only when the motor coordination rate is significantly low and the trajectory matching of multiple associated points deteriorates significantly is a direct determination of dysphagia made, thus reducing the probability of false positives. In practical applications, implementers can determine the second preset coordination rate based on the average motor coordination rate of each swallowing associated point during a limited number of times healthy individuals without swallowing disorders perform the preset actions, and determine the first preset coordination rate based on the minimum motor coordination rate of each swallowing associated point during a limited number of times individuals with swallowing disorders perform the preset actions.

[0041] The secondary determination module, which is connected to the preliminary determination module and the visual recognition analysis module, is used to determine the maximum swallowing characterization value based on the regional image sequence when the preliminary determination of the swallowing type of the target object is a suspected swallowing disorder type, so as to determine the secondary determination of the swallowing type of the target object. The secondary determination of the swallowing type includes swallowing disorder type and no swallowing disorder type.

[0042] Specifically, the secondary determination module determines the maximum movement distance of several detection points within the target swallowing detection area during the target object's execution of a preset action based on the regional image sequence, and determines the maximum swallowing characterization value based on the maximum movement distance of each detection point.

[0043] Specifically, the secondary judgment module determines the secondary judgment swallowing type of the target object based on the comparison result between the maximum swallowing characterization value and the preset swallowing characterization value; If the maximum swallowing characteristic value is less than the preset swallowing characteristic value, the secondary swallowing type of the target object is determined to be a swallowing disorder type; if the maximum swallowing characteristic value is greater than or equal to the preset swallowing characteristic value, the secondary swallowing type of the target object is determined to be a non-swallowing disorder type.

[0044] In this embodiment, the maximum swallowing characterization value is used to characterize the peak motion displacement of each detection point during the execution of a preset action by the target object. The larger the maximum swallowing characterization value, the larger the peak motion displacement of the target object during the execution of the preset action. Under normal circumstances, the peak motion displacement of each detection point in healthy individuals without swallowing disorders during the execution of the preset action will be greater than that in individuals with swallowing disorders. Preferably, the detection points can be set as the thyroid cartilage and any random points around it. The maximum value of the maximum motion distance among each detection point is determined as the key motion distance, the detection point corresponding to the key motion distance is determined as the key detection point, and the ratio of the key motion distance to the preset motion distance is determined as the maximum swallowing characterization value. In practical applications, the implementer can set the preset motion distance based on the average of the maximum motion distances at various positions within the target swallowing detection area during a limited number of times healthy individuals without swallowing disorders perform the preset action. The larger the preset swallowing characterization value, the higher the requirement for the matching degree of the peak motion displacement of the target object during the execution of the preset action. Generally, it can be determined according to the ratio of the first motion distance to the second motion distance. The first motion distance can be determined according to the maximum value of the maximum motion distance at each position in the target swallowing detection area during the execution of the preset action by a limited number of people with swallowing disorders. The second motion distance can be determined according to the minimum value of the maximum motion distance at each position in the target swallowing detection area during the execution of the preset action by a limited number of healthy people without swallowing disorders.

[0045] This invention adaptively determines the initial swallowing detection area by combining the physiological state parameters of the target object, and dynamically corrects the detection area according to the running state during the execution of the preset action, thereby accurately locking the core response area of ​​the swallowing action. This achieves adaptive contraction and calibration of the detection range, avoiding interference from irrelevant factors and improving the accuracy of subsequent visual recognition. By determining the motion trajectory of several swallowing-related points through the regional image sequence of the target swallowing detection area during the target object's execution of the preset action, the motion coordination rate of each swallowing-related point can be adaptively determined, quantifying the degree of multi-point structural coordination of the target object and objectively reflecting the coordination of swallowing movements. Based on the motion coordination rate, three preliminary swallowing types are classified, and stratified screening and initial identification are completed, improving screening efficiency. Through secondary judgment, a second in-depth analysis is performed on suspected swallowing disorder types, extracting the maximum swallowing characteristic value for verification, which improves the accuracy and reliability of swallowing disorder screening.

[0046] 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 swallowing disorder screening system based on intelligent visual recognition, characterized in that, include: The initial detection and analysis module is used to determine the initial swallowing detection area based on the target physiological state parameters of the target object's neck preset area, and to obtain the motion state parameters of the initial swallowing detection area during the target object's execution of preset actions; The target detection and analysis module is used to determine the detection confidence of the initial swallowing detection area based on the motion state parameters of the initial swallowing detection area during the target object's execution of a preset action, and to determine the target swallowing detection area based on the detection confidence. The visual recognition and analysis module is used to acquire a sequence of regional images of the target swallowing detection area during the target object's execution of a preset action, and to determine the motion trajectory of several swallowing-related points based on the regional image sequence, so as to determine the motion coordination rate of each swallowing-related point. The preliminary judgment module is used to determine the preliminary swallowing type of the target object based on the motor coordination rate, wherein the preliminary swallowing type includes swallowing disorder type, suspected swallowing disorder type, and no swallowing disorder type; The secondary determination module is used to determine the maximum swallowing characterization value based on the regional image sequence when the initial determination of the swallowing type of the target object is a suspected swallowing disorder type, so as to determine the secondary determination of the swallowing type of the target object. The secondary determination of the swallowing type includes a swallowing disorder type and a no-swallowing disorder type.

2. The swallowing disorder screening system based on intelligent visual recognition according to claim 1, characterized in that, The initial detection and analysis module includes: An initial state analysis unit is used to determine the target adjustment coefficient based on the comparison results between the target physiological state parameters and the standard physiological state parameters; An initial adjustment analysis unit is used to determine the initial swallowing detection area based on the target adjustment coefficient and the standard detection area.

3. The swallowing disorder screening system based on intelligent visual recognition according to claim 2, characterized in that, The target detection and analysis module includes: The parameter analysis unit is used to divide the initial swallowing detection area into several detection sub-regions, and construct corresponding parameter change curves based on the motion state parameters of each detection sub-region during the target object's execution of a preset action. The confidence analysis unit is used to determine several key change feature points based on the change curves of each parameter, and to determine the detection confidence of the initial swallowing detection area based on the correlation of each key change feature point.

4. The swallowing disorder screening system based on intelligent visual recognition according to claim 3, characterized in that, The target detection and analysis module further includes: The target adjustment analysis unit is used to determine whether to adjust the initial swallowing detection area based on the comparison result between the detection confidence and the preset confidence, so as to determine the target swallowing detection area.

5. The swallowing disorder screening system based on intelligent visual recognition according to claim 4, characterized in that, The visual recognition and analysis module includes: A visual recognition unit is used to acquire a sequence of regional images of the target swallowing detection area during the target object's performance of a preset action; A visual analysis unit, connected to a visual recognition unit, is used to determine the motion trajectories of several swallowing-related points based on the image sequence of the region, and to determine the overlap degree of the trajectories of any two swallowing-related points based on the motion trajectories of each swallowing-related point, so as to determine the motion coordination rate of each swallowing-related point.

6. The swallowing disorder screening system based on intelligent visual recognition according to claim 1 or 5, characterized in that, The preliminary judgment module determines the preliminary swallowing type of the target object based on the comparison results of the motor coordination rate with the first preset coordination rate and the second preset coordination rate. If the motor coordination rate is less than the first preset coordination rate, the preliminary judgment of the target object's swallowing type is swallowing disorder. If the motor coordination rate is greater than or equal to the first preset coordination rate and less than the second preset coordination rate, then the preliminary judgment of the target object's swallowing type is determined to be a suspected swallowing disorder. If the motor coordination rate is greater than or equal to the second preset coordination rate, the preliminary judgment of the target object's swallowing type is determined to be the type without swallowing disorder.

7. The swallowing disorder screening system based on intelligent visual recognition according to claim 1, characterized in that, The secondary determination module determines the maximum movement distance of several detection points within the target swallowing detection area during the target object's execution of a preset action based on the regional image sequence, and determines the maximum swallowing characterization value based on the maximum movement distance of each detection point.

8. The swallowing disorder screening system based on intelligent visual recognition according to claim 1 or 7, characterized in that, The secondary judgment module determines the secondary judgment swallowing type of the target object based on the comparison result between the maximum swallowing characterization value and the preset swallowing characterization value.

9. The swallowing disorder screening system based on intelligent visual recognition according to claim 4, characterized in that, The target adjustment analysis unit determines, based on the first determination condition, to increase the size of the initial swallowing detection area in order to determine the target swallowing detection area; The first determination condition is that the detection confidence level is less than the preset confidence level.

10. The swallowing disorder screening system based on intelligent visual recognition according to claim 4 or 9, characterized in that, The target adjustment analysis unit determines, based on the second determination condition, that the initial swallowing detection area will not be adjusted, and determines the initial swallowing detection area as the target swallowing detection area; The second determination condition is that the detection confidence level is greater than or equal to the preset confidence level.

Images