Swallowing process temperature change analysis and detection method based on thermal imaging technology

Abstract

The invention relates to the technical field of medical imaging and detection, in particular to a swallowing process temperature change analysis and detection method based on a thermal imaging technology, which comprises the following steps: S1, continuously collecting a thermal imaging sequence of an anterior cervical region of a subject through an infrared thermal imager; s2, generating a space-time fusion feature tensor; s3, extracting temperature gradient features of different anatomical partitions in the region of interest; s4, establishing a three-dimensional layered heat conduction model, and solving real-time heat flux distribution of each anatomical layer through inverse operation; s5, constructing a swallowing function abnormity probability model, and calculating a swallowing function abnormity probability value; s6, whether swallowing dysfunction exists or not is judged, and an abnormal stage is positioned. According to the method, collaborative analysis of heat change of superficial and deep tissues in the swallowing process is realized by constructing the space-time fusion feature and layered heat conduction model, so that intelligent judgment of swallowing dysfunction and accurate positioning of an abnormal stage are completed.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging and detection technology, and in particular to a method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology. Background Art

[0002] Swallowing is a core process for the human body to complete basic physiological functions such as feeding, vocalization, and respiratory protection. It involves the coordinated movement of multiple anatomical structures, including the oral cavity, pharynx, larynx, and esophagus. In recent years, with the increasing aging of the population, the incidence of swallowing disorders caused by factors such as stroke, neurological diseases, and tumor surgery has increased significantly, becoming a major issue that seriously affects the quality of life and safety of patients.

[0003] Functional assessment is often performed clinically through video fluoroscopy or fiberoptic endoscopy. However, these methods are highly invasive, complex, and rely on specialized equipment and personnel, limiting their widespread application in primary care or home rehabilitation settings. Therefore, developing a convenient, non-contact, real-time dynamic method for swallowing function assessment has become a key technical issue that needs to be addressed. Summary of the Invention

[0004] Based on the above objectives, the present invention provides a method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology.

[0005] The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology includes the following steps:

[0006] S1: Continuously capture thermal imaging sequences of the subject's anterior neck area using an infrared thermal imager, perform motion distortion correction on the thermal imaging sequences, and output standardized thermal imaging sequences;

[0007] S2: Locate the laryngeal anatomical landmarks in the standardized thermal imaging sequence, delineate the annular region of interest with the Adam's apple as the core, and perform multi-dimensional fusion of the temperature value of each pixel with its spatial coordinates and timestamp to generate a spatiotemporal fusion feature tensor;

[0008] S3: Detect temperature change rate mutation points along the time axis in the spatiotemporal fusion feature tensor and extract temperature gradient characteristics of different anatomical regions within the region of interest, including the maximum temperature difference between the anterior and posterior pharyngeal regions, the integral value of the laryngeal temperature rise period, and the temperature drop rate during the esophageal period;

[0009] S4: Based on the spatiotemporal fusion feature tensor, a three-dimensional layered heat conduction model including the mucosa, muscle layer, and cartilage layer is established. The real-time heat flux distribution of each anatomical layer is solved through inverse calculation, and the deep tissue heat conduction feature vector is output;

[0010] S5: Construct a swallowing dysfunction probability model, use the maximum temperature difference, temperature integral value, temperature drop rate, and deep tissue heat conduction characteristic vector as input features, and calculate the swallowing dysfunction probability value;

[0011] S6: Determine whether swallowing dysfunction exists based on the swallowing dysfunction probability value, and locate the abnormal stage based on the contribution of each input feature.

[0012] Optionally, the S1 specifically includes:

[0013] S11: An infrared thermal imager fixed in front of the subject's anterior neck was used to continuously capture the entire swallowing process at an angle perpendicular to the skin surface of the anterior neck, and a multi-frame raw thermal imaging sequence was acquired at a constant sampling frequency.

[0014] S12: Using the anatomical landmarks of the anterior neck of the subject during swallowing as a reference, the rigid and non-rigid motion components of the anterior neck in multiple frames of the original thermal imaging sequence are tracked using a feature point matching algorithm to obtain the spatial motion vector of each frame of thermal imaging.

[0015] S13: Apply the spatial motion vector to the original thermal imaging sequence for motion compensation, use a distortion correction method based on spatial transformation to eliminate image dislocation and deformation caused by swallowing, and output a standardized thermal imaging sequence.

[0016] Optionally, the S2 specifically includes:

[0017] S21: In the initial frame of the standardized thermal imaging sequence, the spatial position of the Adam's apple is determined using an automatic image segmentation algorithm based on the grayscale gradient change of the thermal imaging, and this position is used as the anatomical landmark of the laryngeal region;

[0018] S22: With the laryngeal anatomical landmark as the center, set inner and outer concentric circles with radii R1 and R2 respectively to form a circular region of interest with the Adam's apple as the core;

[0019] S23: In subsequent frames of the standardized thermal imaging sequence, the positions of laryngeal anatomical landmarks are tracked based on the inter-frame image registration algorithm, and the position of the annular region of interest is dynamically adjusted to ensure that the anatomic position of the annular region of interest is consistent in each frame of the sequence;

[0020] S24: Extract the temperature values ​​of all pixels in the annular region of interest, and fuse them with the two-dimensional spatial coordinates of each pixel and the timestamp of the sampling moment to form a spatiotemporal fusion feature tensor T with a three-dimensional matrix structure.

[0021] Optionally, the S3 specifically includes:

[0022] S31: Taking the spatiotemporal fusion feature tensor T as input, the first-order derivative of the temperature change of each pixel in the annular region of interest is calculated along the time axis to obtain the temperature change rate sequence of each pixel;

[0023] S32: Set a fixed threshold in the temperature change rate sequence. When the temperature change rate of a certain pixel exceeds the fixed threshold for the first time, it is determined as a temperature change rate mutation point, and the corresponding time t is marked. p , and use this to determine the time dividing point between the laryngeal ascending phase and the esophageal phase during swallowing;

[0024] S33: Taking the mutation point time t p As the boundary, the annular region of interest is divided into two anatomical partitions, the anterior pharyngeal region and the posterior pharyngeal region, and the pixel points in the two anatomical partitions are calculated at t p The average temperature difference at the time is used to determine the maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area D max ;

[0025] S34: During the throat rising period, the temperatures of all pixels in the annular region of interest are accumulated and integrated using a time integration method to obtain a temperature integral value characteristic Q during the throat rising period;

[0026] S35: During the esophageal period, a linear regression method is used to fit the temperature decrease trend of all pixels in the annular region of interest over time to determine the temperature decrease rate characteristic K during the esophageal period.

[0027] Optionally, the S4 specifically includes:

[0028] S41: Based on the two-dimensional spatial information of the annular region of interest in the spatiotemporal fusion feature tensor and the anatomical structure of the subject's anterior neck, a three-dimensional hierarchical tissue structure with the mucosal layer, muscle layer, and cartilage layer from shallow to deep was constructed;

[0029] S42: Based on the thermal conductivity, tissue thickness, and physiological characteristic parameters of the mucosa, muscularis, and cartilage layers, a three-dimensional layered heat conduction model is established based on the annular region of interest. The model boundary conditions use the real-time temperature measured at each pixel in the spatiotemporal fusion feature tensor.

[0030] S43: In a three-dimensional layered heat conduction model, the real-time surface temperature of the mucosal layer is used as the input condition for heat conduction. Based on the spatial topological relationship between the mucosal layer, muscular layer, and cartilage layer, the heat conduction process is discretized into a spatial grid using the finite element method to form a numerical calculation matrix for heat conduction. Based on the calculation matrix, the inverse operation method is used to iteratively solve the real-time heat flux density distribution from the shallow to the deep tissue layer, and the real-time heat flux distribution of the mucosal layer, muscular layer, and cartilage layer is output;

[0031] S44: The real-time heat flux distribution of the mucosal layer, muscular layer, and cartilage layer is feature extracted along the spatial dimension to form a deep tissue heat conduction feature vector for swallowing function analysis.

[0032] Optionally, the S42 specifically includes:

[0033] S421: Based on the two-dimensional spatial information of the annular region of interest, a three-dimensional coordinate system is established along the depth direction perpendicular to the skin surface of the anterior neck region, and the mucosal layer, muscle layer, and cartilage layer are mapped into three-dimensional layered geometric models at different depth intervals.

[0034] S422: setting material characteristic parameters of each layer in the three-dimensional layered heat conduction model according to the thickness, density, heat capacity, and thermal conductivity of each anatomical layer;

[0035] S423: Based on the Pennes bioheat transfer equation, the heat conduction control equation of the three-dimensional layered heat conduction model is established, which is expressed as: Where T(x, y, z, t) is the tissue temperature at the three-dimensional spatial coordinate (x, y, z) at time t; ρ represents the tissue density of the corresponding anatomical layer; c represents the tissue specific heat capacity of the corresponding anatomical layer; k represents the tissue thermal conductivity of the corresponding anatomical layer; q(x, y, z, t) is the biological tissue heat production term at the spatial coordinate (x, y, z) inside the tissue; Represents the three-dimensional Laplace operator in space;

[0036] S424: In the three-dimensional layered heat conduction model, initial conditions and boundary conditions are defined, where the initial condition uses the steady-state temperature distribution of the tissue before swallowing, and the boundary condition uses the real-time temperature value measured at each pixel point on the surface of the mucosal layer in the spatiotemporal fusion feature tensor.

[0037] Optionally, the S43 specifically includes:

[0038] S431: Divide the spatial region of the three-dimensional layered heat conduction model into a finite number of spatial units, and perform mesh discretization on the mucosal layer, muscle layer, and cartilage layer using the finite element method to obtain a spatial discretization model consisting of multiple mesh nodes;

[0039] S432: Based on the spatially discretized grid nodes, the weighted residual method is used to transform the three-dimensional heat conduction governing equations into a finite element discretized equation system.

[0040] S433: Matrix assembly is performed based on all spatial unit equations to form the numerical calculation matrix of the overall heat conduction;

[0041] S434: Using the real-time temperature of the mucosal surface as the boundary condition, the above matrix equation is iteratively solved using the inverse operation method to obtain the node temperature distribution within each time step, and the real-time heat flux density of each anatomical layer is calculated using Fourier's law;

[0042] S435: Through the inverse operation iterative process, the real-time heat flux distribution of each anatomical layer of the mucosa layer, muscle layer and cartilage layer is obtained layer by layer from shallow to deep, forming a real-time heat flux distribution field.

[0043] Optionally, the S5 specifically includes:

[0044] S51: The maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area, the integral value of the laryngeal temperature rise period, and the temperature drop rate during the esophageal period extracted in S3 are used as temperature gradient characteristic parameters, and the deep tissue heat conduction characteristic vector calculated in S4 is used as the heat conduction characteristic parameter to jointly construct the sample characteristic vector;

[0045] S52: Collect sample data of subjects with swallowing function assessment standards, establish a training sample set containing feature vectors and corresponding clinical assessment labels, and classify the assessment labels as normal or abnormal;

[0046] S53: Logistic regression model was used as the classification algorithm to establish a probability model for abnormal swallowing function, and each parameter in the feature vector was used as the input variable;

[0047] S54: Train the model using the training sample set and optimize the regression coefficient to improve the model's discriminative ability;

[0048] S55: Inputting the temperature gradient characteristics of the individual to be tested and the deep tissue heat conduction characteristic vector into the trained swallowing dysfunction probability model, and calculating and outputting the corresponding swallowing dysfunction probability value.

[0049] Optionally, the swallowing dysfunction probability model is expressed as:

[0050] Where, P represents the probability value of abnormal swallowing function; D max represents the maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area; Q represents the integral value of the temperature during the laryngeal ascending phase; K represents the temperature drop rate during the esophageal phase; H ... i represents the i-th characteristic component in the deep tissue heat conduction characteristic vector; β0 represents the intercept term of the model; β1,β2,β3,γ i They represent the weight coefficients corresponding to the input features in the model respectively; n1 represents the number of characteristic components in the deep tissue heat conduction characteristic vector.

[0051] Optionally, the S6 specifically includes:

[0052] S61: Compare the swallowing function abnormality probability value obtained in S5 with the preset abnormality determination threshold to determine whether the swallowing function is abnormal. The determination expression is:

[0053] Where, P is the calculated probability value of abnormal swallowing function; P th The preset probability threshold for determining abnormal swallowing function;

[0054] S62: Based on the expression of the swallowing function abnormality probability model, calculate the contribution C of each input feature to the abnormality probability j ;

[0055] S63: Contribution C to all input features j Sort by numerical value, determine the input feature with the greatest contribution, and map it to the corresponding swallowing anatomical stage; specifically,

[0056] The maximum temperature difference between the current and posterior pharyngeal regions corresponds to the prepharyngeal stage;

[0057] The integral value of the temperature during the throat rising period corresponds to the throat contraction stage;

[0058] The rate of temperature decrease during the esophageal phase corresponds to the esophageal propulsion phase;

[0059] The deep tissue heat conduction eigenvector corresponds to the overall heat conduction consistency stage;

[0060] S64: Based on the determined spatial position and occurrence time corresponding to the abnormal stage, the specific spatial position where the swallowing dysfunction occurs and the specific time when the abnormality occurs are marked and output in the standardized thermal imaging sequence.

[0061] Beneficial effects of the present invention:

[0062] The present invention corrects motion distortion of the thermal imaging sequence of the anterior cervical region and constructs a spatiotemporal fusion feature tensor based on the position of the Adam's apple to accurately extract the temperature gradient characteristics that reflect the physiological activity state of each anatomical region. At the same time, combined with the three-dimensional layered heat conduction model and the inverse operation method, it realizes the quantitative inversion of the heat flux of deep tissues such as the mucosal layer, muscle layer and cartilage layer, effectively breaking through the technical bottleneck of traditional thermal imaging technology that cannot obtain dynamic information of deep tissues.

[0063] The present invention, by constructing a probability model for abnormal swallowing function and integrating the maximum temperature difference, temperature integral value, temperature drop rate and deep heat conduction characteristic vector, realizes intelligent judgment of swallowing dysfunction and abnormal stage positioning, solving the problem of lack of objective quantitative indicators and stage discrimination ability in the existing technology; this method does not rely on complex equipment or invasive operations, is suitable for rapid assessment and long-term monitoring of swallowing function in various scenarios, and has good prospects for clinical promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Figure 1 Schematic diagram of a method for analyzing and detecting temperature changes during swallowing according to an embodiment of the present invention;

[0066] Figure 2 Schematic diagram of the process of outputting deep tissue heat conduction characteristic vectors according to an embodiment of the present invention. DETAILED DESCRIPTION

[0067] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0068] It should be noted that references in the specification to "one embodiment," "an embodiment," "an exemplary embodiment," "some embodiments," etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not every embodiment necessarily includes such specific features, structures, or characteristics. In addition, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is within the knowledge of persons skilled in the relevant art to implement such features, structures, or characteristics in conjunction with other embodiments (whether or not explicitly described).

[0069] In general, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending at least in part on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0070] like Figure 1-Figure 2 As shown, the swallowing process temperature change analysis and detection method based on thermal imaging technology includes the following steps:

[0071] S1: Continuously capture thermal imaging sequences of the subject's anterior neck area using an infrared thermal imager, perform motion distortion correction on the thermal imaging sequences, and output standardized thermal imaging sequences;

[0072] S2: Locate the laryngeal anatomical landmarks in the standardized thermal imaging sequence, delineate the annular region of interest with the Adam's apple as the core, and perform multi-dimensional fusion of the temperature value of each pixel with its spatial coordinates and timestamp to generate a spatiotemporal fusion feature tensor;

[0073] S3: Detect temperature change rate mutation points along the time axis in the spatiotemporal fusion feature tensor and extract temperature gradient characteristics of different anatomical regions within the region of interest, including the maximum temperature difference between the anterior and posterior pharyngeal regions, the integral value of the laryngeal temperature rise period, and the temperature drop rate during the esophageal period;

[0074] S4: Based on the spatiotemporal fusion feature tensor, a three-dimensional layered heat conduction model including the mucosa, muscle layer, and cartilage layer is established. The real-time heat flux distribution of each anatomical layer is solved through inverse calculation, and the deep tissue heat conduction feature vector is output;

[0075] S5: Construct a swallowing dysfunction probability model, use the maximum temperature difference, temperature integral value, temperature drop rate, and deep tissue heat conduction characteristic vector as input features, and calculate the swallowing dysfunction probability value;

[0076] S6: Determine whether swallowing dysfunction exists based on the swallowing dysfunction probability value, and locate the abnormal stage based on the contribution of each input feature.

[0077] S1 specifically includes:

[0078] S11: An infrared thermal imager fixed in front of the subject's anterior neck was used to continuously capture the entire swallowing process at an angle perpendicular to the skin surface of the anterior neck, and a multi-frame raw thermal imaging sequence was acquired at a constant sampling frequency.

[0079] S12: Using the anatomical landmarks of the subject's anterior neck during swallowing as a reference, the rigid and non-rigid motion components of the anterior neck region in multiple consecutive frames of the original thermal imaging sequence are tracked using a feature point matching algorithm to obtain the spatial motion vector of each frame of thermal imaging. The spatial motion vector is determined by the following formula:

[0080] Where V t is the spatial motion vector of the t-th frame relative to the t-1-th frame; V is the spatial displacement to be optimized; I t Represents the grayscale value of the original thermal imaging image of the tth frame; I t-1 represents the gray value of the original thermal imaging image of frame t-1; p i is the coordinate of the i-th anatomical landmark point in the image; M is the total number of anatomical landmark points involved in the matching; t is the frame number in the thermal imaging sequence;

[0081] S13: Apply the spatial motion vector to the original thermal imaging sequence for motion compensation, use a distortion correction method based on spatial transformation to eliminate image dislocation and deformation caused by swallowing, and output a standardized thermal imaging sequence; the distortion correction method is implemented by the following formula: t ′(x, y)=I t (x+u t , y+w t ), where I t ′(x, y) represents the grayscale value of the t-th frame standardized thermal image at the pixel coordinate (x, y) after correction; I t represents the grayscale value of the original thermal imaging image of frame t before correction; u t Represents the spatial motion vector V of the thermal imaging of the tth frame t The displacement component in the horizontal direction (x-axis) of the image; w t Represents the spatial motion vector V of the thermal imaging of the tth frame t The displacement component in the longitudinal direction (y-axis) of the image; (x, y) represents the two-dimensional coordinate position of a single pixel point in the thermal imaging; through the above steps, precise spatial alignment is used to eliminate the motion artifacts caused by the subject's swallowing action, so that the subsequent analysis steps are based on spatially consistent and accurate thermal imaging data, effectively ensuring the reliability and accuracy of the analysis results.

[0082] S2 specifically includes:

[0083] S21: In the initial frame of the standardized thermal imaging sequence, the spatial position of the Adam's apple is determined using an automatic image segmentation algorithm based on the grayscale gradient change of the thermal imaging, and this position is used as the anatomical landmark of the laryngeal region;

[0084] The specific steps are as follows:

[0085] S211: performing Gaussian filtering denoising on the initial frame normalized thermal image, and outputting a smoothed thermal image;

[0086] S212: The grayscale gradient of the smoothed thermal imaging image is calculated using the Sobel operator, and a thermal imaging gradient amplitude image is output. The calculation formula is: Where G(x, y) is the gradient amplitude at the coordinate point (x, y); G x (x, y) and G y (x, y) are the horizontal and vertical grayscale gradient components calculated by the Sobel operator respectively;

[0087] S213: performing binarization processing on the gradient amplitude image to obtain a binary mask image of the Adam's apple region, and determining a preliminary outline of the Adam's apple edge region;

[0088] S214: performing contour closing on the binary mask image, and calculating the centroid coordinates of the closed area to determine the precise spatial position coordinates of the Adam's apple.

[0089] S22: With the laryngeal anatomical landmark as the center, set inner and outer concentric circles with radii R1 and R2 respectively to form a circular region of interest with the Adam's apple as the core. The radii R1 and R2 are determined according to the spatial dimensions of the subject's anterior cervical anatomical structure.

[0090] S23: In subsequent frames of the standardized thermal imaging sequence, the positions of laryngeal anatomical landmarks are tracked based on the inter-frame image registration algorithm, and the position of the annular region of interest is dynamically adjusted to ensure that the anatomic position of the annular region of interest is consistent in each frame of the sequence;

[0091] The specific steps are as follows:

[0092] S231: Based on the standardized thermal images of the t-th frame and the t-1-th frame, the normalized cross-correlation coefficient registration method is used to perform inter-frame position registration. The calculation formula is:

[0093]

[0094] Where C(u,v) is the correlation coefficient when the spatial displacement of the t-th frame relative to the t-1-th frame is (u,v); I t ′(x,y) and I t-1 ′(x, y) is the temperature value of the corresponding position (x, y) of the normalized thermal imaging of the t-th frame and the t-1-th frame respectively; and are the mean temperatures of the annular regions of interest in the tth and t-1th frames respectively;

[0095] S232: Using the displacement (u, v) corresponding to the maximum correlation coefficient as the movement of the laryngeal anatomical landmark in the current frame, the position of the region of interest is adjusted to achieve consistency in the position of the annular region of interest in the standardized thermal imaging sequence.

[0096] S24: Extract the temperature values ​​of all pixels in the annular region of interest and fuse them with the two-dimensional spatial coordinates of each pixel and the timestamp of the sampling moment to form a spatiotemporal fusion feature tensor T with a three-dimensional matrix structure. The expression is: T(x, y, t) = [I t ′(x, y), (x, y), t], where T(x, y, t) represents the feature vector of the t-th frame at the spatial coordinate (x, y) in the spatiotemporal fusion feature tensor T; I t′(x, y) represents the temperature value of the t-th frame of standardized thermal imaging at the pixel coordinate (x, y). Through the above steps, the precise anatomical positioning and spatiotemporal data fusion of the standardized thermal imaging sequence are achieved, providing an accurate and stable data basis for the subsequent detailed analysis of temperature changes during swallowing, effectively improving the accuracy of anomaly detection.

[0097] S3 specifically includes:

[0098] S31: Taking the spatiotemporal fusion feature tensor T as input, calculate the first-order derivative of the temperature of each pixel in the annular region of interest along the time axis to obtain the temperature change rate sequence of each pixel. The calculation formula is: Where R(x, y, t) represents the temperature change rate of the pixel at the spatial coordinate (x, y) at time t; I t ′(x, y) is the temperature value of the normalized thermal image at the coordinate (x, y) of the t-th frame; Δt is the sampling time interval between two consecutive frames;

[0099] S32: Set a fixed threshold R in the temperature change rate sequence th , when the temperature change rate of a pixel exceeds the fixed threshold R for the first time th When the temperature change rate is determined to be a sudden change point, the corresponding time t is marked. p , and use this to determine the time dividing point between the laryngeal ascending phase and the esophageal phase during swallowing;

[0100] S33: Taking the mutation point time t p As the boundary, the annular region of interest is divided into two anatomical partitions, the anterior pharyngeal region and the posterior pharyngeal region, and the pixel points in the two anatomical partitions are calculated at t p The average temperature difference at the time is used to determine the maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area D max , the calculation formula is: Where D max Indicates the difference between the anterior pharyngeal area and the posterior pharyngeal area at the mutation time t p Maximum temperature difference; Indicates that all pixels in the anterior pharyngeal area at time t p average temperature; Indicates that all pixels in the posterior pharyngeal area at time t p average temperature;

[0101] S34: During the throat rising period, the temperature of all pixels in the annular region of interest is accumulated and integrated using the time integration method to obtain the throat rising period temperature integral value characteristic Q. The calculation formula is:

[0102] Where Q represents the integral value of the throat temperature during the rising period; t s Indicates the starting time of the laryngeal rising phase; t prepresents the moment of temperature change rate mutation; S represents the set of all pixels in the annular region of interest;

[0103] S35: During the esophageal phase, the temperature decrease trend of all pixels in the annular region of interest is fitted with a linear regression method to determine the temperature decrease rate characteristic K during the esophageal phase. The calculation formula is:

[0104] Where K represents the temperature drop rate during the esophageal phase; t p Indicates the starting time of the esophageal phase; t e Indicates the end of the esophageal phase; I t ′ represents the average temperature of all pixels in the annular region of interest during the esophageal phase at time t; represents the average value of all sampling moments during the esophageal period; Represents the average temperature of all pixel points during the esophageal phase; through the above steps, the precise extraction of temperature gradient characteristics during swallowing is achieved, providing an objective, accurate and quantitative analysis indicator for abnormal swallowing function analysis, which helps to accurately identify abnormal stages and locations during swallowing.

[0105] S4 specifically includes:

[0106] S41: Based on the two-dimensional spatial information of the annular region of interest in the spatiotemporal fusion feature tensor and the anatomical structure of the subject's anterior neck, a three-dimensional hierarchical tissue structure with the mucosal layer, muscle layer, and cartilage layer from shallow to deep was constructed;

[0107] S42: Based on the thermal conductivity, tissue thickness, and physiological characteristic parameters of the mucosa, muscularis, and cartilage layers, a three-dimensional layered heat conduction model is established based on the annular region of interest. The model boundary conditions use the real-time temperature measured at each pixel in the spatiotemporal fusion feature tensor.

[0108] S43: In a three-dimensional layered heat conduction model, the real-time surface temperature of the mucosal layer is used as the input condition for heat conduction. Based on the spatial topological relationship between the mucosal layer, muscular layer, and cartilage layer, the heat conduction process is discretized into a spatial grid using the finite element method to form a numerical calculation matrix for heat conduction. Based on the calculation matrix, the inverse operation method is used to iteratively solve the real-time heat flux density distribution from the shallow to the deep tissue layer, and the real-time heat flux distribution of the mucosal layer, muscular layer, and cartilage layer is output;

[0109] S44: The real-time heat flux distribution of the mucosal layer, muscle layer, and cartilage layer is feature extracted along the spatial dimension to form a deep tissue heat conduction feature vector for swallowing function analysis; through the above steps, the real-time heat conduction feature information of each anatomical tissue layer during the swallowing process can be accurately obtained, providing a more comprehensive basis for deep tissue temperature changes for swallowing function abnormality analysis, and effectively improving the accuracy and sensitivity of swallowing dysfunction judgment.

[0110] The establishment of a three-dimensional layered heat conduction model in S42 specifically includes:

[0111] S421: Based on the two-dimensional spatial information of the annular region of interest, a three-dimensional coordinate system is established along the depth direction perpendicular to the skin surface of the anterior neck region, and the mucosal layer, muscle layer, and cartilage layer are mapped into three-dimensional layered geometric models at different depth intervals.

[0112] S422: setting material characteristic parameters of each layer in the three-dimensional layered heat conduction model according to the thickness, density, heat capacity, and thermal conductivity of each anatomical layer;

[0113] S423: Based on the Pennes bioheat transfer equation, the heat conduction control equation of the three-dimensional layered heat conduction model is established, which is expressed as: Where T(x, y, z, t) is the tissue temperature at the three-dimensional spatial coordinate (x, y, z) at time t; ρ represents the tissue density of the corresponding anatomical layer; c represents the tissue specific heat capacity of the corresponding anatomical layer; k represents the tissue thermal conductivity of the corresponding anatomical layer; q(x, y, z, t) is the biological tissue heat production term at the spatial coordinate (x, y, z) inside the tissue; Represents the three-dimensional Laplace operator in space;

[0114] S424: In the three-dimensional layered heat conduction model, define initial conditions and boundary conditions. The initial condition uses the steady-state temperature distribution of the tissue before swallowing, and the boundary condition uses the real-time temperature value measured at each pixel on the mucosal surface in the spatiotemporal fusion feature tensor. The boundary condition expression is:

[0115] T(x, y, z=0, t)=I t ′(x, y), where T(x, y, z = 0, t) represents the boundary temperature at the mucosal surface (z = 0); I t ′(x, y) represents the temperature value of the standardized thermal image at the pixel coordinate (x, y) at time t; through the above steps, the dynamic change characteristics of heat conduction of anatomical tissue structures during swallowing can be accurately described, providing a reliable model basis for the subsequent inverse operation to solve the real-time heat flux distribution, effectively improving the accuracy and quantification of swallowing dysfunction assessment.

[0116] S43 specifically includes:

[0117] S431: Divide the spatial region of the three-dimensional layered heat conduction model into a finite number of spatial units, and perform mesh discretization on the mucosal layer, muscle layer, and cartilage layer using the finite element method to obtain a spatial discretization model consisting of multiple mesh nodes;

[0118] S432: Based on the spatially discretized grid nodes, the weighted residual method is used to transform the three-dimensional heat conduction control equation into a finite element discrete equation system. The numerical calculation matrix expression of heat conduction is established as follows:

[0119] In the formula, [C e ] is the unit heat capacity matrix, which represents the heat storage capacity of the unit tissue; [K e ] is the unit thermal conduction stiffness matrix, which describes the spatial conduction characteristics of heat within the unit; {T e} is the unit node temperature vector, which represents the real-time temperature of the nodes in the unit; is the rate of change of the unit node temperature over time; {F e} is the unit heat source load vector, representing biological heat production and other external heat input;

[0120] S433: Matrix assembly is performed based on all spatial unit equations to form the numerical calculation matrix of the overall heat conduction, which is expressed as: Where [C] is the global heat capacity matrix, through each unit heat capacity matrix [C e ] is assembled; [K] is the global thermal conduction stiffness matrix, which is obtained by the stiffness matrix of each unit [K e ] is assembled; {T} is the global node temperature vector; is the rate of change of global node temperature over time; {F} is the global heat source load vector, which is calculated by each unit load vector {F e}Assembled;

[0121] S434: Using the real-time temperature of the mucosal surface as the boundary condition, the above matrix equation is iteratively solved using the inverse operation method to obtain the node temperature distribution within each time step. The real-time heat flux density of each anatomical layer is calculated using Fourier's law. The real-time heat flux density expression is: Where q n represents the real-time heat flux density of the nth layer of tissue node; k represents the tissue thermal conductivity of the corresponding anatomical layer; is the temperature gradient of the node temperature in the depth direction;

[0122] S435: Through the inverse operation iterative process, the real-time heat flux distribution of each anatomical layer of the mucosa, muscle layer and cartilage layer is solved layer by layer from shallow to deep, forming a real-time heat flux distribution field; through the above steps, the accurate numerical solution of the real-time heat flux of each anatomical tissue layer can be effectively achieved, providing quantitative deep tissue heat conduction characteristics for the evaluation of swallowing dysfunction, and effectively improving the sensitivity and accuracy of swallowing dysfunction detection.

[0123] S5 specifically includes:

[0124] S51: The maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area, the integral value of the laryngeal temperature rise period, and the temperature drop rate during the esophageal period extracted in S3 are used as temperature gradient characteristic parameters, and the deep tissue heat conduction characteristic vector calculated in S4 is used as the heat conduction characteristic parameter to jointly construct the sample characteristic vector;

[0125] S52: Collect sample data of subjects with swallowing function assessment standards, establish a training sample set containing feature vectors and corresponding clinical assessment labels, and classify the assessment labels as normal or abnormal;

[0126] S53: Logistic regression model was used as the classification algorithm to establish a probability model for abnormal swallowing function, and each parameter in the feature vector was used as the input variable;

[0127] S54: Train the model using the training sample set and optimize the regression coefficient to improve the model's discriminative ability;

[0128] S55: Input the temperature gradient characteristics of the individual to be tested and the deep tissue heat conduction feature vector into the trained swallowing function abnormality probability model, calculate and output the corresponding swallowing function abnormality probability value, and use it for subsequent abnormality judgment; through the above steps, a probability model that can integrate surface temperature changes and deep heat conduction information is constructed, which effectively improves the recognition accuracy of swallowing function abnormality and realizes the quantitative mapping between multi-dimensional thermal physiological feature vectors and clinical abnormal conditions.

[0129] The expression of the swallowing dysfunction probability model is:

[0130] Where, P represents the probability value of abnormal swallowing function; D max represents the maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area; Q represents the integral value of the temperature during the laryngeal ascending phase; K represents the temperature drop rate during the esophageal phase; H ... i represents the i-th characteristic component in the deep tissue heat conduction characteristic vector; β0 represents the intercept term of the model; β1,β2,β3,γ i They represent the weight coefficients corresponding to the input features in the model respectively; n1 represents the number of characteristic components in the deep tissue heat conduction characteristic vector.

[0131] S6 specifically includes:

[0132] S61: Compare the swallowing function abnormality probability value obtained in S5 with the preset abnormality determination threshold to determine whether the swallowing function is abnormal. The determination expression is:

[0133] Where, P is the calculated probability value of abnormal swallowing function; P th The preset probability threshold for determining abnormal swallowing function;

[0134] S62: Based on the expression of the swallowing function abnormality probability model, calculate the contribution C of each input feature to the abnormality probability j , the calculation formula is: C j =|β j X j |, where C j is the contribution of the j-th input feature to the probability of abnormal swallowing function; β j is the model weight coefficient corresponding to the j-th input feature; X j is the normalized value of the corresponding input feature; j corresponds to the maximum temperature difference feature D in the input feature. max , temperature integral value characteristic Q, temperature drop rate characteristic K and deep tissue heat conduction characteristic vector H i Each characteristic component in ;

[0135] S63: Contribution C to all input features j Sort by numerical value, determine the input feature with the greatest contribution, and map it to the corresponding swallowing anatomical stage; specifically,

[0136] The maximum temperature difference between the current and posterior pharyngeal regions corresponds to the prepharyngeal stage;

[0137] The integral value of the temperature during the throat rising period corresponds to the throat contraction stage;

[0138] The rate of temperature decrease during the esophageal phase corresponds to the esophageal propulsion phase;

[0139] The deep tissue heat conduction eigenvector corresponds to the overall heat conduction consistency stage;

[0140] S64: Based on the spatial position and occurrence time corresponding to the determined abnormal stage, the specific spatial position of the swallowing dysfunction and the specific time of the abnormality are marked and output in the standardized thermal imaging sequence; through the above steps, the quantitative determination of the swallowing dysfunction can be accurately achieved, and the key stages, spatial positions and moments of the swallowing abnormality can be precisely located, providing a clear, objective and effective analysis basis for the diagnosis and intervention of the swallowing dysfunction.

[0141] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0142] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology, characterized in that: The following steps are involved: S1: Continuously capture thermal imaging sequences of the subject's anterior neck area using an infrared thermal imager, perform motion distortion correction on the thermal imaging sequences, and output standardized thermal imaging sequences; S2: Locate the laryngeal anatomical landmarks in the standardized thermal imaging sequence, delineate the annular region of interest with the Adam's apple as the core, and perform multi-dimensional fusion of the temperature value of each pixel with its spatial coordinates and timestamp to generate a spatiotemporal fusion feature tensor; S3: Detect temperature change rate mutation points along the time axis in the spatiotemporal fusion feature tensor and extract temperature gradient characteristics of different anatomical regions within the region of interest, including the maximum temperature difference between the anterior and posterior pharyngeal regions, the integral value of the laryngeal temperature rise period, and the temperature drop rate during the esophageal period; S4: Based on the spatiotemporal fusion feature tensor, a three-dimensional layered heat conduction model including the mucosa, muscle layer, and cartilage layer is established. The real-time heat flux distribution of each anatomical layer is solved through inverse calculation, and the deep tissue heat conduction feature vector is output; S5: Construct a swallowing dysfunction probability model, use the maximum temperature difference, temperature integral value, temperature drop rate, and deep tissue heat conduction characteristic vector as input features, and calculate the swallowing dysfunction probability value; S6: Determine whether swallowing dysfunction exists based on the swallowing dysfunction probability value, and locate the abnormal stage based on the contribution of each input feature.

2. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 1, characterized in that: Said S1 specifically includes: S11: An infrared thermal imager fixed in front of the subject's anterior neck was used to continuously capture the entire swallowing process at an angle perpendicular to the skin surface of the anterior neck, and a multi-frame raw thermal imaging sequence was acquired at a constant sampling frequency. S12: Using the anatomical landmarks of the anterior neck of the subject during swallowing as a reference, the rigid and non-rigid motion components of the anterior neck in multiple frames of the original thermal imaging sequence are tracked using a feature point matching algorithm to obtain the spatial motion vector of each frame of thermal imaging. S13: Apply the spatial motion vector to the original thermal imaging sequence for motion compensation, use a distortion correction method based on spatial transformation to eliminate image dislocation and deformation caused by swallowing, and output a standardized thermal imaging sequence.

3. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 1, characterized in that: The S2 specifically includes: S21: In the initial frame of the standardized thermal imaging sequence, the spatial position of the Adam's apple is determined using an automatic image segmentation algorithm based on the grayscale gradient change of the thermal imaging, and this position is used as the anatomical landmark of the laryngeal region; S22: With the laryngeal anatomical landmark as the center, set inner and outer concentric circles with radii R1 and R2 respectively to form a circular region of interest with the Adam's apple as the core; S23: In subsequent frames of the standardized thermal imaging sequence, the positions of laryngeal anatomical landmarks are tracked based on the inter-frame image registration algorithm, and the position of the annular region of interest is dynamically adjusted to ensure that the anatomic position of the annular region of interest is consistent in each frame of the sequence; S24: Extract the temperature values ​​of all pixels in the annular region of interest, and fuse them with the two-dimensional spatial coordinates of each pixel and the timestamp of the sampling moment to form a spatiotemporal fusion feature tensor T with a three-dimensional matrix structure.

4. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 3, characterized in that: The S3 specifically includes: S31: Taking the spatiotemporal fusion feature tensor T as input, the first-order derivative of the temperature change of each pixel in the annular region of interest is calculated along the time axis to obtain the temperature change rate sequence of each pixel; S32: Set a fixed threshold in the temperature change rate sequence. When the temperature change rate of a certain pixel exceeds the fixed threshold for the first time, it is determined as a temperature change rate mutation point, and the corresponding time t is marked. p , and use this to determine the time dividing point between the laryngeal ascending phase and the esophageal phase during swallowing; S33: Taking the mutation point time t p As the boundary, the annular region of interest is divided into two anatomical partitions, the anterior pharyngeal region and the posterior pharyngeal region, and the pixel points in the two anatomical partitions are calculated at t p The average temperature difference at the time is used to determine the maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area D max ; S34: During the throat rising period, the temperatures of all pixels in the annular region of interest are accumulated and integrated using a time integration method to obtain a temperature integral value characteristic Q during the throat rising period; S35: During the esophageal period, a linear regression method is used to fit the temperature decrease trend of all pixels in the annular region of interest over time to determine the temperature decrease rate characteristic K during the esophageal period.

5. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 1, characterized in that: The S4 specifically includes: S41: Based on the two-dimensional spatial information of the annular region of interest in the spatiotemporal fusion feature tensor and the anatomical structure of the subject's anterior neck, a three-dimensional hierarchical tissue structure with the mucosal layer, muscle layer, and cartilage layer from shallow to deep was constructed; S42: Based on the thermal conductivity, tissue thickness, and physiological characteristic parameters of the mucosa, muscularis, and cartilage layers, a three-dimensional layered heat conduction model is established based on the annular region of interest. The model boundary conditions use the real-time temperature measured at each pixel in the spatiotemporal fusion feature tensor. S43: In a three-dimensional layered heat conduction model, the real-time surface temperature of the mucosal layer is used as the input condition for heat conduction. Based on the spatial topological relationship between the mucosal layer, muscular layer, and cartilage layer, the heat conduction process is discretized into a spatial grid using the finite element method to form a numerical calculation matrix for heat conduction. Based on the calculation matrix, the inverse operation method is used to iteratively solve the real-time heat flux density distribution from the shallow to the deep tissue layer, and the real-time heat flux distribution of the mucosal layer, muscular layer, and cartilage layer is output; S44: The real-time heat flux distribution of the mucosal layer, muscular layer, and cartilage layer is feature extracted along the spatial dimension to form a deep tissue heat conduction feature vector for swallowing function analysis.

6. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 5, characterized in that: The S42 specifically includes: S421: Based on the two-dimensional spatial information of the annular region of interest, a three-dimensional coordinate system is established along the depth direction perpendicular to the skin surface of the anterior neck region, and the mucosal layer, muscle layer, and cartilage layer are mapped into three-dimensional layered geometric models at different depth intervals. S422: setting material characteristic parameters of each layer in the three-dimensional layered heat conduction model according to the thickness, density, heat capacity, and thermal conductivity of each anatomical layer; S423: Based on the Pennes bioheat transfer equation, the heat conduction control equation of the three-dimensional layered heat conduction model is established, which is expressed as: Where T(x, y, z, t) is the tissue temperature at the three-dimensional spatial coordinate (x, y, z) at time t; ρ represents the tissue density of the corresponding anatomical layer; c represents the tissue specific heat capacity of the corresponding anatomical layer; k represents the tissue thermal conductivity of the corresponding anatomical layer; q(x, y, z, t) is the biological tissue heat production term at the spatial coordinate (x, y, z) inside the tissue; Represents the three-dimensional Laplace operator in space; S424: In the three-dimensional layered heat conduction model, initial conditions and boundary conditions are defined, where the initial condition uses the steady-state temperature distribution of the tissue before swallowing, and the boundary condition uses the real-time temperature value measured at each pixel point on the surface of the mucosal layer in the spatiotemporal fusion feature tensor.

7. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 6, characterized in that: The S43 specifically includes: S431: Divide the spatial region of the three-dimensional layered heat conduction model into a finite number of spatial units, and perform mesh discretization on the mucosal layer, muscle layer, and cartilage layer using the finite element method to obtain a spatial discretization model consisting of multiple mesh nodes; S432: Based on the spatially discretized grid nodes, the weighted residual method is used to transform the three-dimensional heat conduction governing equations into a finite element discretized equation system. S433: Matrix assembly is performed based on all spatial unit equations to form the numerical calculation matrix of the overall heat conduction; S434: Using the real-time temperature of the mucosal surface as the boundary condition, the above matrix equation is iteratively solved using the inverse operation method to obtain the node temperature distribution within each time step, and the real-time heat flux density of each anatomical layer is calculated using Fourier's law; S435: Through the inverse operation iterative process, the real-time heat flux distribution of each anatomical layer of the mucosa layer, muscle layer and cartilage layer is obtained layer by layer from shallow to deep, forming a real-time heat flux distribution field.

8. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 1, characterized in that: The S5 specifically includes: S51: The maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area, the integral value of the laryngeal temperature rise period, and the temperature drop rate during the esophageal period extracted in S3 are used as temperature gradient characteristic parameters, and the deep tissue heat conduction characteristic vector calculated in S4 is used as the heat conduction characteristic parameter to jointly construct the sample characteristic vector; S52: Collect sample data of subjects with swallowing function assessment standards, establish a training sample set containing feature vectors and corresponding clinical assessment labels, and classify the assessment labels as normal or abnormal; S53: Logistic regression model was used as the classification algorithm to establish a probability model for abnormal swallowing function, and each parameter in the feature vector was used as the input variable; S54: Train the model using the training sample set and optimize the regression coefficient to improve the model's discriminative ability; S55: Inputting the temperature gradient characteristics of the individual to be tested and the deep tissue heat conduction characteristic vector into the trained swallowing dysfunction probability model, and calculating and outputting the corresponding swallowing dysfunction probability value.

9. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 8, characterized in that: The expression of the swallowing dysfunction probability model is: Where, P represents the probability value of abnormal swallowing function; D max represents the maximum temperature difference between the anterior pharyngeal area and the posterior pharyngeal area; Q represents the integral value of the temperature during the laryngeal ascending phase; K represents the temperature drop rate during the esophageal phase; H ... i represents the i-th characteristic component in the deep tissue heat conduction characteristic vector; β0 represents the intercept term of the model; β1, β2, β3, γ i They represent the weight coefficients corresponding to the input features in the model respectively; n1 represents the number of characteristic components in the deep tissue heat conduction characteristic vector.

10. The method for analyzing and detecting temperature changes during swallowing based on thermal imaging technology according to claim 9, characterized in that: The S6 specifically includes: S61: Compare the swallowing function abnormality probability value obtained in S5 with the preset abnormality determination threshold to determine whether the swallowing function is abnormal. The determination expression is: Where, P is the calculated probability value of abnormal swallowing function; P th The preset probability threshold for determining abnormal swallowing function; S62: Based on the expression of the swallowing function abnormality probability model, calculate the contribution C of each input feature to the abnormality probability j ; S63: Contribution C to all input features j Sort by numerical value, determine the input feature with the greatest contribution, and map it to the corresponding swallowing anatomical stage; specifically, The maximum temperature difference between the current and posterior pharyngeal regions corresponds to the prepharyngeal stage; The integral value of the temperature during the throat rising period corresponds to the throat contraction stage; The rate of temperature decrease during the esophageal phase corresponds to the esophageal propulsion phase; The deep tissue heat conduction eigenvector corresponds to the overall heat conduction consistency stage; S64: Based on the determined spatial position and occurrence time corresponding to the abnormal stage, the specific spatial position where the swallowing dysfunction occurs and the specific time when the abnormality occurs are marked and output in the standardized thermal imaging sequence.

Images