Automatic matching method for X-ray machine working position and flat panel detector

Abstract

The invention discloses an automatic matching method for a working position of an X-ray machine and a flat panel detector, and relates to the technical field of radiographic imagines.The method comprises the steps that pet struggling features are recognized in real time through a motion state classification model, a compensation mechanism is triggered before displacement occurs, and matching delay caused by passive response in the prior art is avoided; the working position of the X-ray machine can synchronously track the position of the detector to ensure that an imaging area accurately covers a key anatomical structure; when the displacement is overtime, the system automatically stops exposure and reversely calibrates the position based on the anatomical mark point; an operator does not need to interrupt flow manual calibration, the repeated exposure risk of the pet is reduced, and the diagnosis and treatment efficiency is improved; the influence of equipment vibration and operation jitter is suppressed by gravity compensation and smooth filtering, and the displacement prediction model is compatible with motion characteristics of pets with different body types; in a narrow clinic or multi-pet assembly line scene, the matching stability is kept; an autonomous decision closed loop is formed from data acquisition, state judgment, real-time tracking to abnormal calibration.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of radiological imaging, in particular to an X-ray machine working position and flat panel detector automatic matching method. BACKGROUND

[0002] In the field of pet X-ray photography, the technology of automatically matching the working position of the X-ray machine and the flat panel detector has been gradually introduced. Most of these systems realize linkage between devices based on wireless communication protocols. For example, the position identifier is broadcasted by the detector, and the host system analyzes the coordinates to control the mechanical arm or gimbal to adjust the angle of the radiation source. Some solutions introduce infrared or Bluetooth sensor modules to assist positioning and establish a binding relationship between devices within a preset working distance.

[0003] In the live pet photography scenario, the core problem caused by sudden limb movement of the animal gradually emerges. When dogs and cats struggle out of fear, their torsos or limbs may shift position within milliseconds, causing the flat panel detector to shift position as well. At this time, the initial positioning parameters relied on by the device are invalid, and the matching process needs to be re-executed. The existing technology mainly uses two types of ways to deal with such dynamic changes: one is to forcibly interrupt the photography process and restart the matching sequence after the pet is fixed; the other is to add physical restraint devices to limit the animal's activity range. Both of these two ways require manual intervention to adjust and the photography process is forced to pause.

[0004] The limitations of existing solutions are mainly reflected in the real-time response link. The communication protocol polling mechanism has inherent delays, and re-matching needs to complete steps such as signal scanning, coordinate verification, and position calibration, which cannot be completed within a sub-second window for tracking compensation. The added motion sensors can only detect displacement results and lack the ability to predict sudden movements. When the pet is restless continuously, multiple matching failures may result in accumulated radiation dose or operation timeout. Some devices attempt to alleviate the problem by reducing the matching precision threshold, but this may cause misalignment of the imaging area or omission of key anatomical structures. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] The present application provides an X-ray machine working position and flat panel detector automatic matching method to solve the problem that existing matching technology cannot respond to pet displacement in real time, requires interruption of photography and manual adjustment, resulting in low efficiency and radiation risk.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] The present application provides an X-ray machine working position and flat panel detector automatic matching method, which includes the following steps:

[0009] Step S1: Real-time acquisition of motion state data of the flat panel detector, the motion state data including acceleration and angular velocity information;

[0010] Step S2: Based on the motion state data, determine whether the current state is in a dynamic displacement state using a motion state classification model;

[0011] Step S3: If the state is determined to be dynamic displacement, a working position coordinate adjustment command is generated and sent to the X-ray machine control system so that the X-ray machine working position tracks the position of the flat panel detector in real time.

[0012] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, step S2, the construction of the motion state classification model includes:

[0013] Extract time-domain waveform features from motion state data, including acceleration amplitude variance and angular velocity spectral entropy;

[0014] The time-domain waveform features are input into a lightweight neural network classifier, which outputs a probability threshold for the dynamic displacement state.

[0015] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, the training of the lightweight neural network classifier includes:

[0016] Collect a dataset of movement states under typical pet struggling behaviors;

[0017] The initial network is optimized using transfer learning based on time-domain waveform features;

[0018] Enhance the robustness of dynamic displacement state recognition by using adversarial examples.

[0019] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, the step of extracting the time-domain waveform features of the motion state data includes:

[0020] In length of Within the sliding time window, the acceleration and angular velocity sequences of the inertial measurement unit are synchronized, and the triaxial accelerations are synthesized into amplitudes:

[0021] ,

[0022] in, For the first Acceleration amplitude at each sampling point, in units , , , The first Each sampling point is at , , axial acceleration components, This is the sampling point index, and its value range is... ;

[0023] Calculate the variance of acceleration amplitude within this window:

[0024] ,

[0025] ,

[0026] in, The variance of the acceleration amplitude is dimensionless. This is the average acceleration amplitude, in units. , The window length is expressed in units of sample points.

[0027] An adjustable cosine window is applied to the angular velocity sequence to suppress spectral leakage:

[0028] ,

[0029] in, The window function weights are dimensionless. The shape factor is obtained through equipment calibration.

[0030] The angular velocity amplitude is obtained after applying the window function:

[0031] ,

[0032] in, For the first angular velocity amplitude at each sampling point, in units , , , These are the angular velocity components;

[0033] The power spectrum is obtained using the discrete Fourier transform:

[0034] ,

[0035] in, For the first The power spectral amplitude of each discrete frequency component, in units , For discrete frequency indexes, the range of values ​​is... The imaginary unit;

[0036] After normalizing the power spectrum, its probability components and spectral entropy are calculated as follows:

[0037] , ,

[0038] in, These are the probability components of the power spectrum, which are dimensionless. Angular velocity spectral entropy, unit: nat;

[0039] Constructing time-domain waveform eigenvectors:

[0040] ,

[0041] in, is the feature vector of the current window, with dimension 2 and dimensionless.

[0042] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, step S3, the generation of the working position coordinate adjustment command includes:

[0043] Calculate the displacement trend vector for future time periods based on motion state data;

[0044] The displacement trend vector is converted into the pitch angle and translation correction values ​​of the X-ray machine gimbal.

[0045] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, a collision protection mechanism is triggered when the continuous dynamic displacement state times out.

[0046] Suspend X-ray machine exposure operations;

[0047] Based on the location of the anatomical markers in the most recent valid imaging results, the actual coordinate offset of the flat panel detector is calculated in reverse.

[0048] The working position matching reference is calibrated using the actual coordinate offset.

[0049] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, the position of the anatomical marker point is obtained through the following steps:

[0050] The bone edge enhancement process for X-ray images includes multi-scale filtering to suppress soft tissue background noise, adaptive threshold segmentation to extract bone tissue contours, and curvature analysis to filter continuous edges.

[0051] Identify the geometric center point of the intervertebral space or joint cavity as a marker point.

[0052] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, the reverse calculation of the actual coordinate offset includes:

[0053] Non-rigid registration was performed between the anatomical markers and the standard position template.

[0054] Extract the spatial transformation matrix of the registration deformation field;

[0055] The translation components of the decomposed transformation matrix are used as coordinate offsets.

[0056] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, in the process of calculating the displacement trend vector within a future time period based on motion state data, the future displacement direction is estimated using the inertial measurement data of the last complete time window within the current exposure preparation cycle, as follows:

[0057] After gravity compensation, the net acceleration in the inertial coordinate system is obtained:

[0058] ,

[0059] in, Indicates the first Net acceleration vector at each sampling point, in units , For attitude quaternions The constructed rotation matrix is ​​dimensionless. For the first The pose quaternion of each sampling point The original acceleration vector, This is the Earth's gravity vector;

[0060] Estimating instantaneous velocity using recursive integral form:

[0061] ,

[0062] in, For the first The instantaneous velocity vector of each sampling point, in units. , The instantaneous velocity vector at the previous sampling point. The time interval between adjacent samplings, in seconds;

[0063] Applying exponential weighted smoothing to the velocity series yields the trend velocity:

[0064] ,

[0065] in, This represents the trend velocity vector of the current sampling point, in units. The smoothing coefficient, determined through offline measurement and calibration, is dimensionless. This represents the trend velocity vector of the previous sampling point;

[0066] In the prediction time domain Internally generated displacement trend vector:

[0067] ,

[0068] in, This represents the displacement trend vector for future time periods, in meters. The predicted duration is in seconds (s).

[0069] As a preferred embodiment of the automatic matching method between the X-ray machine working position and the flat panel detector described in this invention, the motion state data is collected by the inertial measurement unit built into the flat panel detector and transmitted to the X-ray machine control system through a low-latency wireless communication protocol, and the determination of dynamic displacement state and coordinate adjustment are completed within a single exposure preparation cycle.

[0070] The beneficial effects of this invention are as follows: This invention identifies pet struggling characteristics in real time through a motion state classification model, triggering a compensation mechanism before displacement occurs, avoiding matching delays caused by the passive response of existing technologies; the X-ray machine workstation can synchronously track the detector position, ensuring accurate coverage of key anatomical structures in the imaging area; when displacement exceeds the timeout period, the system automatically pauses exposure and reverses the position based on anatomical markers. Operators do not need to interrupt the process for manual calibration, reducing the risk of repeated pet exposure and improving diagnostic efficiency. Gravity compensation and smoothing filtering suppress the effects of equipment vibration and operational jitter, and the displacement prediction model is compatible with the motion characteristics of pets of different sizes. Matching stability is maintained in confined clinics or multi-pet assembly line scenarios.

[0071] This invention forms an autonomous decision-making closed loop, encompassing data acquisition, status assessment, real-time tracking, and anomaly calibration. All matching actions are completed within a single exposure cycle, shortening preparation time. Furthermore, it is based on the flat panel detector's built-in IMU and X-ray machine control system, requiring no hardware modifications. The motion classification model and image processing algorithm can be embedded in the device firmware, reducing upgrade costs. Attached Figure Description

[0072] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a flowchart illustrating the automatic matching method between the X-ray machine working position and the flat panel detector in Example 1. Detailed Implementation

[0074] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0075] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0076] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0077] Example 1, referring to Figure 1 This embodiment provides a method for automatically matching the working position of an X-ray machine with a flat panel detector, including the following steps:

[0078] Step S1: Real-time acquisition of motion state data of the flat panel detector, including acceleration and angular velocity information;

[0079] Step S2: Based on the motion state data, determine whether the current state is in a dynamic displacement state using a motion state classification model;

[0080] In step S2, the construction of the motion state classification model includes:

[0081] Extract the time-domain waveform features of the motion state data. The time-domain waveform features include the acceleration amplitude variance and the angular velocity spectrum entropy.

[0082] Input the time-domain waveform features into a lightweight neural network classifier and output the probability threshold of the dynamic displacement state;

[0083] Training a lightweight neural network classifier includes:

[0084] Collect a dataset of movement states under typical pet struggling behaviors;

[0085] The initial network is optimized using transfer learning based on time-domain waveform features;

[0086] Enhance the robustness of dynamic displacement state recognition through adversarial examples;

[0087] The steps for extracting time-domain waveform features from motion state data include:

[0088] In length of Within the sliding time window, the acceleration and angular velocity sequences of the inertial measurement unit are synchronized, and the triaxial accelerations are synthesized into amplitudes:

[0089] ,

[0090] in, For the first Acceleration amplitude at each sampling point, in units , , , The first Each sampling point is at , , axial acceleration components, This is the sampling point index, and its value range is... ;

[0091] Calculate the variance of acceleration amplitude within this window:

[0092] ,

[0093] ,

[0094] in, The variance of the acceleration amplitude is dimensionless. This is the average acceleration amplitude, in units. , The window length is expressed in units of sample points.

[0095] An adjustable cosine window is applied to the angular velocity sequence to suppress spectral leakage:

[0096] ,

[0097] in, The window function weights are dimensionless. This is the shape factor, ranging from 0.50 to 0.60, obtained through equipment calibration.

[0098] The angular velocity amplitude is obtained after applying the window function:

[0099] ,

[0100] in, For the first angular velocity amplitude at each sampling point, in units , , , These are the angular velocity components;

[0101] The power spectrum is obtained using the discrete Fourier transform:

[0102] ,

[0103] in, For the first The power spectral amplitude of each discrete frequency component, in units , For discrete frequency indexes, the range of values ​​is... The imaginary unit;

[0104] After normalizing the power spectrum, its probability components and spectral entropy are calculated as follows:

[0105] , ,

[0106] in, These are the probability components of the power spectrum, which are dimensionless. Angular velocity spectral entropy, unit: nat;

[0107] Constructing time-domain waveform eigenvectors:

[0108] ,

[0109] in, The feature vector of the current window, dimension 2, dimensionless;

[0110] Specifically, the sensor signal is processed in three layers: amplitudeization, statistical analysis, and information entropy reduction, so that translation and rotation dynamic information are comparable in the same feature space.

[0111] Acceleration amplitude variance is sensitive to the severity of displacement and can promptly reflect uncontrolled translational impacts. Angular velocity spectral entropy measures rotational complexity with uniform energy distribution and is particularly sensitive to frequent direction changes. Adjustable cosine windows suppress different noise spectrum shapes through shape factors to avoid window leakage, improving entropy stability while preserving... , As an adjustable hyperparameter, it provides room for subsequent transfer learning and model lightweighting;

[0112] Specifically, the flat panel detector incorporates a six-axis inertial measurement unit (IMU) to synchronously acquire raw triaxial acceleration and triaxial angular velocity data at a fixed sampling frequency (typically 200Hz). When constructing the angular velocity spectral entropy characteristics, an adjustable cosine window is used to suppress spectral leakage. Its shape factor λ is determined through the equipment's factory calibration process: a sinusoidal excitation of a known frequency is applied to a standard vibration table, and λ is adjusted to align the main lobe width of the window function with the excitation frequency, ultimately locking λ = 0.55. This value is compatible with most pet struggling frequency bands (1-5Hz), and the window function weight distribution satisfies:

[0113] ,

[0114] in, The window length is fixed at 50 sampling points (corresponding to a 0.25-second time interval) to balance real-time performance and feature stability.

[0115] Step S3: If it is determined to be a dynamic displacement state, a working position coordinate adjustment command is generated and sent to the X-ray machine control system so that the X-ray machine working position tracks the position of the flat panel detector in real time.

[0116] In step S3, the generation of the work position coordinate adjustment command includes:

[0117] Calculate the displacement trend vector for future time periods based on motion state data;

[0118] The displacement trend vector is converted into the pitch angle and translation correction values ​​of the X-ray machine gimbal.

[0119] When the continuous dynamic displacement state times out, the collision protection mechanism is triggered:

[0120] Suspend X-ray machine exposure operations;

[0121] Based on the location of the anatomical markers in the most recent valid imaging results, the actual coordinate offset of the flat panel detector is calculated in reverse.

[0122] The working position matching reference is calibrated using the actual coordinate offset.

[0123] The locations of anatomical markers are obtained through the following steps:

[0124] The bone edge enhancement process for X-ray images includes multi-scale filtering to suppress soft tissue background noise, adaptive threshold segmentation to extract bone tissue contours, and curvature analysis to filter continuous edges.

[0125] Identify the geometric center point of the intervertebral space or joint cavity as a marker point;

[0126] Specifically, the bone edge enhancement process involves three steps:

[0127] 1. Multi-scale filtering: Use a Gaussian difference filter (scale σ = 1.0 / 2.0 pixels) to suppress muscle tissue background;

[0128] 2. Adaptive thresholding: Calculate the local Otsu threshold using a sliding window (20×20 pixels) to extract the cortical bone region;

[0129] 3. Curvature Filtering: Calculate the radius of curvature of the contour points. ,reserve The concave points of the pixels are used as candidate sites for the joint cavity.

[0130] The reverse calculation of the actual coordinate offset includes:

[0131] Non-rigid registration was performed between anatomical markers and standard position templates;

[0132] Extract the spatial transformation matrix of the registration deformation field;

[0133] The translation components of the decomposed transformation matrix are used as coordinate offsets;

[0134] In calculating the displacement trend vector for future time periods based on motion state data, within the current exposure preparation cycle, the future displacement direction is estimated using the inertial measurement data of the last complete time window, as follows:

[0135] After gravity compensation, the net acceleration in the inertial coordinate system is obtained:

[0136] ,

[0137] in, Indicates the first Net acceleration vector at each sampling point, in units , For attitude quaternions The constructed rotation matrix is ​​dimensionless. For the first The pose quaternion of each sampling point The original acceleration vector, This is the Earth's gravity vector;

[0138] Estimating instantaneous velocity using recursive integral form:

[0139] ,

[0140] in, For the first The instantaneous velocity vector of each sampling point, in units. , The instantaneous velocity vector at the previous sampling point. The time interval between adjacent samplings, in seconds;

[0141] Applying exponential weighted smoothing to the velocity series yields the trend velocity:

[0142] ,

[0143] in, This represents the trend velocity vector of the current sampling point, in units. The smoothing coefficient, ranging from 0.6 to 0.8, was determined through offline measurement and calibration and is dimensionless. This represents the trend velocity vector of the previous sampling point;

[0144] In the prediction time domain Internally generated displacement trend vector:

[0145] ,

[0146] in, This represents the displacement trend vector for future time periods, in meters. For the predicted duration, a value of 0.10-0.15 seconds is typically used;

[0147] Specifically, this method utilizes attitude and acceleration information provided by the inertial measurement unit. First, the gravity component is eliminated under the action of the rotation matrix to obtain the net acceleration. Then, the velocity is recursively derived using discrete integration, and short-term fluctuations are suppressed by exponential weighted smoothing, so that the trend velocity maintains a moderate lag to sudden disturbances. The smoothing coefficient α can be calibrated offline according to the equipment noise spectrum and animal body size, thus finding a balance between response speed and anti-jitter. The predicted displacement adopts the classical uniform acceleration model, which incorporates both the trend velocity and the current net acceleration. Only one vector operation is needed to give the displacement direction at the level of 0.1s in the future, which meets the real-time requirements of the gimbal drive control loop.

[0148] Specifically, the exponentially weighted smoothing coefficient α is determined based on the pet's movement characteristics:

[0149] Behavioral statistics were performed on dogs and cats (weighing 5-40kg), and the rate of change of acceleration for sudden movements was concentrated in the range of 3-8m / s³. Through offline simulation, the tracking delay of step velocity was tested in IMU data with a noise level of ±0.2m / s². Finally, α=0.7 was selected to keep the trend velocity lag within 30ms, while filtering out more than 70% of high-frequency jitter.

[0150] Motion state data is collected by the inertial measurement unit built into the flat panel detector and transmitted to the X-ray machine control system through a low-latency wireless communication protocol. The determination of dynamic displacement state and coordinate adjustment are completed within a single exposure preparation cycle.

[0151] Specifically, the low-latency wireless communication protocol uses Bluetooth 5.0's LE Coded PHY mode (125kbps), with a transmission cycle of ≤10ms, ensuring that the end-to-end delay of motion state data from the detector to the control system is <15ms. The single exposure preparation cycle refers to the time from detector positioning to X-ray tube preheating (typically 0.8-1.2 seconds), and all matching actions are completed within this window.

[0152] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for automatically matching the working position of an X-ray machine with a flat panel detector, characterized in that, Includes the following steps: Step S1: Real-time acquisition of motion state data of the flat panel detector, the motion state data including acceleration and angular velocity information; Step S2: Based on the motion state data, determine whether the current state is in a dynamic displacement state using a motion state classification model; Step S3: If the state is determined to be dynamic displacement, a working position coordinate adjustment command is generated and sent to the X-ray machine control system so that the X-ray machine working position tracks the position of the flat panel detector in real time.

2. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 1, characterized in that, In step S2, the construction of the motion state classification model includes: Extract time-domain waveform features from motion state data, including acceleration amplitude variance and angular velocity spectral entropy; The time-domain waveform features are input into a lightweight neural network classifier, which outputs a probability threshold for the dynamic displacement state.

3. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 2, characterized in that, The training of the lightweight neural network classifier includes: Collect a dataset of movement states under typical pet struggling behaviors; The initial network is optimized using transfer learning based on time-domain waveform features; Enhance the robustness of dynamic displacement state recognition by using adversarial examples.

4. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 3, characterized in that, The step of extracting the time-domain waveform features of the motion state data includes: In length of Within the sliding time window, the acceleration and angular velocity sequences of the inertial measurement unit are synchronized, and the triaxial accelerations are synthesized into amplitudes: ， in, For the first Acceleration amplitude at each sampling point, in units , , , The first Each sampling point is at , , axial acceleration components, This is the sampling point index, and its value range is... ; Calculate the variance of acceleration amplitude within this window: ， ， in, The variance of the acceleration amplitude is dimensionless. This is the average acceleration amplitude, in units. , The window length is expressed in units of sample points. An adjustable cosine window is applied to the angular velocity sequence to suppress spectral leakage: ， in, The window function weights are dimensionless. The shape factor is obtained through equipment calibration. The angular velocity amplitude is obtained after applying the window function: ， in, For the first angular velocity amplitude at each sampling point, in units , , , These are the angular velocity components; The power spectrum is obtained using the discrete Fourier transform: ， in, For the first The power spectral amplitude of each discrete frequency component, in units , For discrete frequency indexes, the range of values ​​is... The imaginary unit; After normalizing the power spectrum, its probability components and spectral entropy are calculated as follows: ， ， in, These are the probability components of the power spectrum, which are dimensionless. Angular velocity spectral entropy, unit: nat; Constructing time-domain waveform eigenvectors: ， in, is the feature vector of the current window, with dimension 2 and dimensionless.

5. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 1, characterized in that, In step S3, the generation of the working position coordinate adjustment command includes: Calculate the displacement trend vector for future time periods based on motion state data; The displacement trend vector is converted into the pitch angle and translation correction values ​​of the X-ray machine gimbal.

6. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 5, characterized in that, When the continuous dynamic displacement state times out, the collision protection mechanism is triggered: Suspend X-ray machine exposure operations; Based on the location of the anatomical markers in the most recent valid imaging results, the actual coordinate offset of the flat panel detector is calculated in reverse. The working position matching reference is calibrated using the actual coordinate offset.

7. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 6, characterized in that, The location of the anatomical markers is obtained through the following steps: The bone edge enhancement process for X-ray images includes multi-scale filtering to suppress soft tissue background noise, adaptive threshold segmentation to extract bone tissue contours, and curvature analysis to filter continuous edges. Identify the geometric center point of the intervertebral space or joint cavity as a marker point.

8. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 7, characterized in that, The reverse calculation of the actual coordinate offset includes: Non-rigid registration was performed between anatomical markers and standard position templates; Extract the spatial transformation matrix of the registration deformation field; The translation components of the decomposed transformation matrix are used as coordinate offsets.

9. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 8, characterized in that, In the process of calculating the displacement trend vector for future time periods based on motion state data, the future displacement direction is estimated using the inertial measurement data of the last complete time window within the current exposure preparation cycle, as follows: After gravity compensation, the net acceleration in the inertial coordinate system is obtained: ， in, Indicates the first Net acceleration vector at each sampling point, in units , For attitude quaternions The constructed rotation matrix is ​​dimensionless. For the first The pose quaternion of each sampling point The original acceleration vector, This is the Earth's gravity vector; Estimating instantaneous velocity using recursive integral form: ， in, For the first The instantaneous velocity vector of each sampling point, in units. , The instantaneous velocity vector at the previous sampling point. The time interval between adjacent samplings, in seconds; Applying exponential weighted smoothing to the velocity series yields the trend velocity: ， in, This represents the trend velocity vector of the current sampling point, in units. The smoothing coefficient, determined through offline measurement and calibration, is dimensionless. This represents the trend velocity vector of the previous sampling point; In the prediction time domain Internally generated displacement trend vector: ， in, This represents the displacement trend vector for future time periods, in meters. The predicted duration is in seconds (s).

10. The method for automatic matching of X-ray machine working position and flat panel detector as described in claim 1, characterized in that, The motion state data is collected by the inertial measurement unit built into the flat panel detector and transmitted to the X-ray machine control system through a low-latency wireless communication protocol. The determination of dynamic displacement state and coordinate adjustment are completed within a single exposure preparation cycle.

Images