Body position auxiliary detection method and system based on DR equipment

By aligning the patient's pressure distribution and three-dimensional depth information in spatial coordinates, combining physiological parameters and medical history data, dynamic correction marks are generated, which solves the problem of time-consuming and susceptible to human factors in traditional position placement methods, and achieves efficient and accurate adjustment of DR equipment position and improves image quality.

CN120267321AInactive Publication Date: 2025-07-08老桂红
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202510343676.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-22
Publication Date
2025-07-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The traditional DR equipment positioning method depends on the experience of doctors, is time-consuming and susceptible to human factors, resulting in inaccurate positioning, affecting image quality and diagnostic results.

Method used

By aligning the patient's pressure distribution data and three-dimensional depth information, an initial position three-dimensional point cloud model is generated, and combining the patient's physiological parameters and medical history data, the reference bone-muscle model is matched from the pre-trained three-dimensional anatomical database, non-rigid registration is performed, the projection angle deviation vector is calculated, dynamic correction marks are generated, and the laser projection device is driven for real-time adjustment.

Benefits of technology

Real-time feedback and closed-loop control of position adjustment are realized, the accuracy and correction effect of position detection are improved, clearer and more standardized medical images are obtained, and the risk of missed diagnosis or misdiagnosis is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure HDA0005323905260000011
    Figure HDA0005323905260000011
  • Figure HDA0005323905260000021
    Figure HDA0005323905260000021
Patent Text Reader

Abstract

The invention relates to a body position auxiliary detection method and system based on DR equipment. The method comprises the following steps: performing coordinate alignment on pressure distribution data and three-dimensional depth information of a patient to generate an initial body position three-dimensional point cloud model; matching a corresponding reference skeleton-muscle model based on the registration information, generating a registered body position three-dimensional model through non-rigid registration, and extracting coordinates of key anatomical mark points; a target structure theoretical projection area is calculated by combining a DR equipment C arm projection angle, an angle deviation vector is acquired by comparing an actual projection boundary, and a dynamic correction mark is generated. According to the method, through space coordinate alignment, personalized model registration and real-time projection angle deviation correction, the accuracy and efficiency of body position placement are remarkably improved, radiation exposure caused by repeated shooting is reduced, and therefore the medical image quality and diagnosis reliability are remarkably improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and particularly relates to a body position assisted detection method and system based on a DR device. Background Art

[0002] With the development of medical imaging technology, digital radiography (DR) devices are increasingly widely used in clinical diagnosis. DR devices have the characteristics of high-resolution imaging and low radiation dose, and can provide doctors with clearer and safer image data, thus significantly improving the accuracy and efficiency of diagnosis. In the actual operation of DR devices, the positioning of the patient's body position is crucial for obtaining high-quality images. However, traditional body position placement methods mainly rely on doctors' experience and manual adjustment. This method not only takes time, but is also easily affected by human factors, resulting in inaccurate body positions, which in turn affects the image quality and diagnostic results. Summary of the Invention

[0003] Based on this, in view of the above technical problems, it is necessary to provide a body position assisted detection method based on a DR device to assist in adjusting the patient's body position, improve the efficiency and accuracy of body position placement, thereby obtaining clearer and more standardized medical images, and reducing the risk of missed diagnosis or misdiagnosis caused by body position deviation.

[0004] In a first aspect, the present application provides a body position assisted detection method based on a DR device, the method comprising:

[0005] Performing spatial coordinate alignment processing on the pressure distribution data and three-dimensional depth information of the patient to generate an initial body position three-dimensional point cloud model;

[0006] According to the patient registration information, matching the corresponding reference bone-muscle model from a pre-trained three-dimensional anatomy database, and performing non-rigid registration on the reference bone-muscle model and the initial body position three-dimensional point cloud model to generate a registered body position three-dimensional model, and outputting the coordinates of key anatomical landmark points, where the patient registration information includes physiological parameters and medical history data;

[0007] According to the current projection angle of the C-arm of the DR device and the coordinates of the key anatomical landmark points, calculating the theoretical projection area of the target anatomical structure, and outputting a projection angle deviation vector by comparing the theoretical projection area with the actual projection boundary of the registered body position three-dimensional model, where the projection angle deviation vector includes a rotation compensation angle and a translation compensation amount;

[0008] According to the projection angle deviation vector, performing body position correction guidance marker generation processing to generate a dynamic correction marker, where the dynamic correction marker is used to drive a laser projection device to display a real-time adjustment direction.

[0009] In one embodiment, according to the patient registration information, a corresponding reference bone-muscle model is matched from a pre-trained three-dimensional anatomical database, and the reference bone-muscle model is non-rigidly registered with the initial body position three-dimensional point cloud model to generate a registered body position three-dimensional model, and the coordinates of key anatomical landmark points are output, including:

[0010] According to the physiological parameters, a reference bone-muscle model is selected from the pre-trained three-dimensional anatomical database, and the physiological parameters include age, height, and weight data;

[0011] Through the iterative closest point algorithm, the bone surface point cloud of the reference bone-muscle model is roughly matched with the initial body position point cloud of the initial body position three-dimensional point cloud model, and the initial rigid body transformation matrix is calculated;

[0012] The initial rigid body transformation matrix is applied to the bone surface point cloud of the reference bone-muscle model to obtain the roughly matched reference model;

[0013] Based on the thin plate spline deformation field model, non-rigid registration is performed on the roughly matched reference model to optimize the deformation field parameters and generate a registered body position three-dimensional model;

[0014] The coordinates of key anatomical landmark points are extracted from the registered body position three-dimensional model, and the coordinates of key anatomical landmark points include three-dimensional space coordinates and confidence scores.

[0015] In one embodiment, according to the current projection angle of the C-arm of the DR device and the coordinates of key anatomical landmark points, the theoretical projection area of the target anatomical structure is calculated, and by comparing the theoretical projection area with the actual projection boundary of the registered body position three-dimensional model, a projection angle deviation vector is output, including:

[0016] According to the current projection angle and the focus-detector geometric relationship, the theoretical projection boundary of the target anatomical structure is calculated;

[0017] The bone edge point cloud in the registered body position three-dimensional model is projected onto the detector plane to extract the actual projection boundary;

[0018] The maximum deviation distance between the theoretical projection boundary and the actual projection boundary is compared through the Hausdorff distance algorithm. If the maximum deviation distance is greater than the preset threshold, a projection angle deviation vector is generated.

[0019] In one embodiment, according to the projection angle deviation vector, a body position correction guidance marker generation process is performed to generate a dynamic correction marker, including:

[0020] According to the rotation compensation angle in the projection angle deviation vector, an angle scale line in the joint flexion direction is generated;

[0021] Generate an arrow mark in the direction of the torso displacement according to the translation compensation amount in the projection angle deviation vector;

[0022] Superimpose the angle scale line and the arrow mark, and adjust the color of the dynamic correction mark according to the deviation level to obtain the dynamic correction mark.

[0023] In one embodiment, the method further includes:

[0024] Calculate the displacement vector of the patient's body surface based on the continuously acquired real-time frame body surface point cloud data, and generate a dynamic change signal, where the dynamic change signal includes a displacement direction change signal and an amplitude direction change signal;

[0025] Perform motion type classification processing on the dynamic change signal and the contact area change rate through the Kalman filter algorithm to obtain a classification result, where the classification result includes physiological micro-movement and body position offset, and the contact area change rate is used to characterize the rate of change of the area of the contact area between the patient's body surface and the examination bed over time;

[0026] When the classification result is body position offset, update the thin plate spline deformation field parameters according to the displacement vector to generate a dynamically corrected deformation field;

[0027] Apply the dynamically corrected deformation field to the registered three-dimensional body position model, recalculate the coordinates of the key anatomical landmark points, and output the corrected coordinates of the key anatomical landmark points, where the corrected coordinates of the key anatomical landmark points are used to update and represent the coordinates of the key anatomical landmark points.

[0028] In one embodiment, the method further includes:

[0029] Calculate the local pressure gradient change rate of different body regions according to the adjusted body position pressure distribution data. If the local pressure gradient change rate is greater than the preset pressure limit value, generate an acoustic and optical alarm instruction;

[0030] Based on the gray-scale distribution characteristics of the adjusted low-dose preview image, calculate the maximum equivalent tissue thickness of the target anatomical region through the pre-stored gray-scale-thickness calibration curve to generate a tube voltage recommendation value;

[0031] Predict through a random forest model according to the historical image quality data to obtain the optimal milliamperage-seconds parameter, where the historical image quality data includes projection angle, tissue thickness, and detector type parameters;

[0032] Output the tube voltage recommendation value and the optimal milliamperage-seconds parameter to the DR device exposure control system to generate a device parameter optimization instruction, where the device parameter optimization instruction is used to adjust the exposure conditions.

[0033] In a second aspect, the present application also provides a body position auxiliary detection system based on a DR device, and the system includes:

[0034] An initial body position generation module, configured to perform spatial coordinate alignment processing on the pressure distribution data and three-dimensional depth information of a patient to generate an initial body position three-dimensional point cloud model;

[0035] A reference model matching module, configured to match a corresponding reference bone-muscle model from a pre-trained three-dimensional anatomy database according to patient registration information, perform non-rigid registration on the reference bone-muscle model and the initial body position three-dimensional point cloud model to generate a registered body position three-dimensional model, and output the coordinates of key anatomical landmark points, where the patient registration information includes physiological parameters and medical history data;

[0036] A deviation vector calculation module, configured to calculate the theoretical projection area of a target anatomical structure according to the current projection angle of the C-arm of the DR device and the coordinates of key anatomical landmark points, and output a projection angle deviation vector by comparing the theoretical projection area with the actual projection boundary of the registered body position three-dimensional model, where the projection angle deviation vector includes a rotation compensation angle and a translation compensation amount;

[0037] A guiding mark generation module, configured to perform body position correction guiding mark generation processing according to the projection angle deviation vector to generate dynamic correction marks, where the dynamic correction marks are used to drive a laser projection device to display a real-time adjustment direction.

[0038] In a third aspect, the present application further provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps in the first aspect are implemented.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the first aspect are implemented.

[0040] The above-mentioned body position assisted detection method and system based on DR equipment generate an initial three-dimensional point cloud model of the body position through the coordinate alignment of pressure distribution data and three-dimensional depth information, realizing the three-dimensional modeling of the spatial position of the patient's body position, and providing an accurate geometric basis for subsequent anatomical structure analysis. Moreover, by combining the patient's physiological parameters and medical history data, a reference bone-muscle model is matched from a pre-trained three-dimensional anatomical database, and personalized anatomical structure fusion is carried out through non-rigid registration technology, effectively solving the problem of insufficient adaptation of the model to individual anatomical differences in traditional methods, and further improving the positioning accuracy of key anatomical landmark points. Secondly, based on the real-time projection angle of the C-arm and the spatial coordinates of the registration model, a theoretical projection mathematical model of the target anatomical structure is established, and through the dynamic comparison with the actual projection boundary, the rotation compensation angle and translation compensation amount are accurately calculated, removing the subjectivity of visually judging the angle deviation, and realizing the quantitative correction of the projection parameters. Finally, the projection angle deviation vector is converted into a real-time adjustment instruction of the laser projection device, and the body position correction is guided through visual dynamic marking, replacing the cumbersome operations of traditional verbal instructions or mechanical positioners, and realizing the real-time feedback and closed-loop control of the body position adjustment process.

[0041] Compared with the traditional body position detection and correction method, this method improves the accuracy and correction effect of body position detection through multi-dimensional data fusion, personalized modeling, quantitative deviation analysis, and real-time dynamic correction guidance, so as to obtain clearer and more standardized medical images, which helps to improve the accuracy of diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0043] Figure 1 It is a flowchart of a body position assisted detection method based on a DR device provided by an exemplary embodiment of the present invention;

[0044] Figure 2 It is a schematic structural diagram of a body position assisted detection system based on a DR device provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] In order to make the purpose, technical solutions and advantages of the present application clearer, the following further details the present application in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0046] In one embodiment, as Figure 1 shown, a posture assistance detection method based on a DR device is provided. In this embodiment, taking the application of this method to a terminal as an example, it can be understood that this method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0047] S101: Perform spatial coordinate alignment processing on the pressure distribution data and three-dimensional depth information of the patient to generate an initial posture three-dimensional point cloud model.

[0048] Specifically, the pressure distribution data of each part of the patient's body can be collected by a pressure sensor array. The array consists of multiple high-sensitivity pressure sensors and can be arranged on the surface of the examination bed, so as to be able to obtain the pressure distribution data of the contact area between the patient and the bed surface in real time. And the three-dimensional depth information of the patient can be obtained by using a depth camera. Schematically, the depth camera can be based on the principle of structured light or time of flight. By emitting infrared light or laser of a specific pattern and receiving the reflected light to calculate the depth value, a depth image of the patient's body surface can be generated. Align the obtained pressure distribution data and three-dimensional depth information in spatial coordinates. Through a coordinate transformation algorithm, map the data in different coordinate systems to the same spatial coordinate system to ensure accurate matching of the two in spatial position. Finally, combining the aligned pressure distribution data and three-dimensional depth information, a point cloud generation algorithm can be used to construct an initial posture three-dimensional point cloud model. The initial posture three-dimensional point cloud model can present the three-dimensional shape and pressure distribution characteristics of the patient's body in the form of a point cloud, providing a basis for subsequent posture analysis.

[0049] S102: According to the patient registration information, match the corresponding reference bone-muscle model from the pre-trained three-dimensional anatomy database, and perform non-rigid registration on the reference bone-muscle model and the initial posture three-dimensional point cloud model to generate a registered posture three-dimensional model, and output the coordinates of key anatomical landmark points. The patient registration information includes physiological parameters and medical history data.

[0050] Specifically, the patient registration information includes physiological parameters such as height, weight, age, and gender, as well as medical history data such as past medical history, surgical history, and trauma history. Based on this registration information, a reference bone-muscle model that best matches the patient's anatomical structure can be selected from a pre-trained three-dimensional anatomical database. Schematically, this database contains a large number of bone-muscle models of different individuals and can be pre-trained through machine learning algorithms, enabling a suitable model to be quickly and accurately matched according to the input physiological parameters and medical history data. The matched reference bone-muscle model is non-rigidly registered with the three-dimensional point cloud model of the initial body position. Among them, the non-rigid registration algorithm can consider the deformation and flexibility of the patient's body and adjust the shape and position of the reference model through an optimization algorithm to achieve the best match with the three-dimensional point cloud model of the initial body position in terms of shape and spatial position. After the registration is completed, a three-dimensional model of the registered body position can be generated. This three-dimensional model of the registered body position combines anatomical structure and actual body position information and can more accurately reflect the patient's current body position state. And through feature extraction algorithms, etc., key anatomical landmark coordinates, such as joint positions and spinal key points, can be identified and output from the three-dimensional model of the registered body position, providing reference data for subsequent medical imaging and body position correction.

[0051] S103: According to the current projection angle of the C-arm of the DR device and the key anatomical landmark coordinates, calculate the theoretical projection area of the target anatomical structure, and by comparing the theoretical projection area with the actual projection boundary of the three-dimensional model of the registered body position, output a projection angle deviation vector, where the projection angle deviation vector includes a rotation compensation angle and a translation compensation amount.

[0052] Specifically, the current projection angle information of the C-arm of the DR device can be obtained through an angle sensor built into the device. And this angle can include multi-dimensional angle parameters such as the rotation angle and pitch angle of the C-arm. Secondly, in combination with the key anatomical landmark coordinates, the theoretical projection area of the target anatomical structure at the current projection angle can be calculated through a projection geometric model. Schematically, this projection geometric model can accurately predict the ideal projection range of the target structure on the imaging plane based on the X-ray imaging principle and the spatial geometric relationship of the anatomical structure. And by projecting the three-dimensional model of the registered body position at the current projection angle, the actual projection boundary can be obtained. By comparing the theoretical projection area and the actual projection boundary, calculate the difference between the two to obtain the projection angle deviation vector. Among them, the rotation compensation angle can represent the angle that the C-arm needs to adjust in the rotation direction, and the translation compensation amount can represent the distance that it needs to adjust in the translation direction to make the actual projection match the theoretical projection.

[0053] S104: According to the projection angle deviation vector, perform processing for generating a body position correction guidance mark to generate a dynamic correction mark, where the dynamic correction mark is used to drive the laser projection device to display the real-time adjustment direction.

[0054] Specifically, the projection angle deviation vector can be used as the input, and the corresponding body position correction guidance mark can be generated through the body position correction algorithm. This body position correction algorithm can determine the way and degree of adjustment required for the patient's body position according to the magnitude and direction of the deviation vector, such as rotating to the left or translating to the right. Among them, the dynamic correction mark is a visual guidance information, which can be displayed through graphic elements such as arrows, lines, and text, and can then intuitively indicate how the patient should adjust the body position. The dynamic correction mark is sent to the laser projection device to drive it to display the adjustment direction in real time near the imaging area, and the patient or medical staff can adjust the body position according to the displayed mark until the deviation vector is zero, reaching the ideal body position state, thereby improving the imaging quality and diagnostic accuracy of the DR device.

[0055] In the above method, by performing spatial coordinate alignment processing on the patient's pressure distribution data and three-dimensional depth information to generate the initial body position three-dimensional point cloud model, the patient's body position information can be obtained more comprehensively and accurately, providing a basis for subsequent precise body position detection and correction. And according to the patient registration information, the corresponding reference bone-muscle model is matched from the pre-trained three-dimensional anatomy database, and non-rigid registration is performed with the initial body position three-dimensional point cloud model to generate the registered body position three-dimensional model and output the coordinates of the key anatomical landmark points. This not only realizes personalized modeling but also can accurately locate the key anatomical parts, greatly improving the efficiency and accuracy compared with the traditional manual identification of key points, providing an accurate position reference for subsequent body position correction. Secondly, calculating the theoretical projection area of the target anatomical structure according to the current projection angle of the DR device C-arm and the coordinates of the key anatomical landmark points, and comparing the theoretical projection area with the actual projection boundary of the registered body position three-dimensional model to output the projection angle deviation vector can accurately quantify the body position deviation, improving the pertinence and scientificity of body position correction. In addition, generating the body position correction guidance mark according to the projection angle deviation vector, generating the dynamic correction mark, and using it to drive the laser projection device to display the real-time adjustment direction realizes real-time dynamic body position correction guidance, avoiding the situation in traditional correction methods where the patient may have deviations in understanding the doctor's instructions and cannot provide real-time feedback on the adjustment effect, and improving the patient's comfort and correction efficiency.

[0056] In one of the embodiments, according to the patient registration information, the corresponding reference bone-muscle model is matched from the pre-trained three-dimensional anatomy database, and non-rigid registration is performed between the reference bone-muscle model and the initial body position three-dimensional point cloud model to generate the registered body position three-dimensional model and output the coordinates of the key anatomical landmark points, including:

[0057] According to the physiological parameters, the reference bone-muscle model is screened from the pre-trained three-dimensional anatomy database, and the physiological parameters include age, height, and weight data;

[0058] Through the iterative closest point algorithm, the bone surface point cloud of the reference bone-muscle model is roughly matched with the initial body position point cloud of the initial body position three-dimensional point cloud model, and the initial rigid body transformation matrix is calculated;

[0059] Apply the initial rigid body transformation matrix to the bone surface point cloud of the reference bone-muscle model to obtain the roughly matched reference model;

[0060] Based on the thin plate spline deformation field model, perform non-rigid registration on the roughly matched reference model, optimize the deformation field parameters, and generate the registered body position three-dimensional model;

[0061] Extract the coordinates of key anatomical landmark points from the registered body position three-dimensional model. The coordinates of key anatomical landmark points include three-dimensional space coordinates and confidence scores.

[0062] Specifically, in the pre-trained three-dimensional anatomical database, pre-training can be carried out through machine learning algorithms so that it can quickly and accurately match the model most similar to the patient's anatomical structure, that is, the reference bone-muscle model, according to the input physiological parameters. Among them, the matching process can comprehensively consider the influence of age, height, and weight on the bone-muscle structure. For example, for pediatric patients, models with incompletely developed bones and corresponding cartilage structures can be preferentially selected. The iterative closest point algorithm can find the optimal match by minimizing the distance between two point clouds. Initially, the bone surface point cloud of the reference bone-muscle model can be initially aligned with the initial body position point cloud of the initial body position three-dimensional point cloud model based on a rough alignment method. Then, the closest point pairs of the two point clouds can be iteratively calculated, and the transformation matrix can be calculated by methods such as the least squares method according to the closest point pairs to update the point cloud positions until the convergence condition is reached, such as the upper limit of the number of iterations or the error threshold, and the initial rigid body transformation matrix can be obtained. This matrix contains translation and rotation parameters, which can describe the rough position and pose relationship of the reference bone-muscle model relative to the initial body position point cloud, and can roughly align the bone surface point cloud of the reference bone-muscle model in the space coordinate system of the initial body position three-dimensional point cloud model.

[0063] Schematically, through matrix transformation operations, the coordinates of the surface points of each bone in the reference bone-muscle model can be substituted into the initial rigid body transformation matrix to calculate their new coordinates in the spatial coordinate system of the initial body position three-dimensional point cloud model, thereby realizing the translation and rotation of the reference bone-muscle model to align it with the initial body position point cloud in terms of approximate position and orientation, providing an initial alignment state for subsequent non-rigid registration. The thin plate spline deformation field model is a mathematical model that can describe complex non-linear deformations and is suitable for describing the deformations of non-rigid structures such as human soft tissues. In the non-rigid registration of the coarsely matched reference model based on the thin plate spline deformation field model, the coarsely matched reference model and the initial body position three-dimensional point cloud model can be used as inputs, and the parameters of the deformation field can be adjusted through an optimization algorithm. Under the action of the deformation field, the coarsely matched reference model not only matches the overall position and orientation of the initial body position point cloud, but also better fits the actual body position shape in terms of local details, fully considering the non-rigid characteristics of human soft tissues such as elasticity and compressibility, thereby generating a more accurate registered body position three-dimensional model. In addition, through model matching and feature extraction algorithms, key anatomical landmark points can be identified and located in the registered body position three-dimensional model, and their accurate coordinates in the three-dimensional space can be calculated. And in order to evaluate the reliability of the positioning of each landmark point, a confidence score mechanism can be introduced, and the confidence score can be comprehensively evaluated based on multiple factors such as the stability of feature matching and the similarity of surrounding structures, providing a more reliable position reference for subsequent body position correction.

[0064] In one embodiment, according to the current projection angle of the C-arm of the DR device and the coordinates of the key anatomical landmark points, the theoretical projection area of the target anatomical structure is calculated, and by comparing the theoretical projection area with the actual projection boundary of the registered body position three-dimensional model, a projection angle deviation vector is output, including:

[0065] According to the current projection angle and the focal point-detector geometric relationship, the theoretical projection boundary of the target anatomical structure is calculated;

[0066] The bone edge point cloud in the registered body position three-dimensional model is projected onto the detector plane to extract the actual projection boundary;

[0067] The maximum deviation distance between the theoretical projection boundary and the actual projection boundary is compared through the Hausdorff distance algorithm. If the maximum deviation distance is greater than the preset threshold, a projection angle deviation vector is generated.

[0068] Specifically, the focus-detector geometric relationship includes parameters such as the distance to the X-ray tube focus, the detector position, and the relative distance between the X-ray tube focus and the detector position. These parameters jointly determine the geometric path and imaging scale relationship of the X-ray emitted from the focus and projected onto the detector after passing through the patient's body. Based on the current projection angle and the focus-detector geometric relationship, the projection direction of the X-ray and the imaging area of the detector can be determined; then, the three-dimensional coordinate points of the target anatomical structure are converted into two-dimensional coordinate points on the detector plane through the projection transformation formula, and then the extreme value range of these two-dimensional coordinate points can be collected to form the contour of the theoretical projection boundary.

[0069] Schematically, the bone edge point cloud can represent the contour information of the patient's bone in three-dimensional space. Based on the current projection angle and the focus-detector geometric relationship, the bone edge point cloud can be projected onto the detector plane one by one, and the corresponding position of each edge point on the detector plane can be calculated through coordinate transformation. Then, all the projected points can be subjected to clustering analysis and boundary fitting to extract the actual projection boundary. The Hausdorff distance algorithm can measure the similarity between two point sets. Through this algorithm, the maximum distance difference between the two boundaries can be found, so as to effectively evaluate the matching degree between the theoretical projection boundary and the actual projection boundary. Schematically, the theoretical projection boundary and the actual projection boundary can be represented as two point sets, and then the Hausdorff distance algorithm is used to calculate the maximum deviation distance between the two point sets. The calculated maximum deviation distance is compared with a preset threshold. If the maximum deviation distance exceeds the preset threshold, it indicates that there is a large deviation between the current projection angle and the ideal angle, and adjustment is required to generate the corresponding projection angle deviation vector.

[0070] In one embodiment, according to the projection angle deviation vector, a body position correction guidance mark generation process is performed to generate a dynamic correction mark, including:

[0071] According to the rotation compensation angle in the projection angle deviation vector, an angle scale line in the joint flexion direction is generated;

[0072] According to the translation compensation amount in the projection angle deviation vector, an arrow mark in the trunk displacement direction is generated;

[0073] The angle scale line and the arrow mark are superimposed, and the color of the dynamic correction mark is adjusted according to the deviation level to obtain the dynamic correction mark.

[0074] Specifically, the rotation compensation angle in the projection angle deviation vector is a key parameter for guiding the patient to adjust joint flexion. The rotation compensation angle can be used as the input, and combined with the anatomical characteristics of the human joints, the starting position, direction, and scale interval of the angle scale line can be determined to generate the angle scale line in the joint flexion direction. This angle scale line can accurately indicate the joint direction and angle size that the patient needs to flex or extend, facilitating the patient's understanding and operation. The translation compensation amount can be used to determine the direction and distance that the patient's torso needs to be adjusted. The shape, length, and pointing direction of the arrow can be designed according to the magnitude and direction of the translation compensation amount. This arrow mark can intuitively indicate the direction and approximate distance that the patient needs to move, facilitating the patient to quickly and accurately adjust the body position. During the process of image overlay of the generated angle scale line and the arrow mark, it is necessary to ensure that the two marks do not interfere with each other visually, and the different deviation levels such as slight, medium, and severe can be determined through quantitative evaluation of the projection angle deviation vector. According to the deviation level, the color of the dynamic correction mark is adjusted by using the color mapping algorithm, which can more effectively guide the patient to adjust the body position.

[0075] In one embodiment, the method further includes:

[0076] According to the continuously acquired real-time frame body surface point cloud data, calculate the displacement vector of the patient's body surface, generate a dynamic change signal, and the dynamic change signal includes a displacement direction change signal and an amplitude direction change signal;

[0077] According to the dynamic change signal and the contact area change rate, perform motion type classification processing through the Kalman filter algorithm to obtain a classification result, and the classification result includes physiological micro-movement and body position offset. The contact area change rate is used to characterize the rate of change of the area of the contact region between the patient's body surface and the examination bed over time;

[0078] When the classification result is body position offset, update the thin plate spline deformation field parameters according to the displacement vector to generate a dynamically corrected deformation field;

[0079] Apply the dynamically corrected deformation field to the registered three-dimensional body position model, recalculate the coordinates of the key anatomical landmark points, and output the corrected coordinates of the key anatomical landmark points. The corrected coordinates of the key anatomical landmark points are used to update the representation of the coordinates of the key anatomical landmark points.

[0080] Specifically, when the patient's body position moves, based on the continuously acquired real-time frame body surface point cloud data, through the Iterative Closest Point (ICP) algorithm, the body surface point cloud data of consecutive frames can be matched and aligned, and the coordinate changes of each point in adjacent frames can be compared to calculate the displacement vector of the patient's body surface. And by analyzing the direction and magnitude of the displacement vector, a dynamic change signal can be generated. The contact area change rate can be obtained by using a pressure sensor array to continuously acquire the area information of the contact area between the patient's body surface and the examination bed and calculating its change rate over time. This contact area change rate reflects the dynamic changes in the contact situation between the patient's body position adjustment and the examination bed.

[0081] The Kalman filter algorithm is a recursive least squares algorithm that can effectively process measurement data containing noise and estimate the state of the system. The dynamic change signal and the contact area change rate can be used as inputs, and the Kalman filter algorithm can be applied for filtering and estimation to remove noise interference, extract relevant features of the motion type, and obtain classification results, including physiological micro-movements and body position offsets. Schematically, physiological micro-movements refer to the small body surface changes caused by physiological activities such as breathing and heartbeat during the examination of the patient, while body position offsets are the relatively large-scale body surface changes caused by the patient actively changing the body position due to discomfort or other reasons.

[0082] When the classification result is a body position offset, it indicates that the patient's body position has changed significantly, and the three-dimensional model of the registered body position needs to be adjusted again. Schematically, the displacement vector provides the displacement information of the patient's body surface at different positions. By integrating the displacement vector into the deformation field model, the parameters of the deformation field can be adjusted to better reflect the actual situation after the patient's body position changes. The updated thin plate spline deformation field parameters are used to generate a dynamically corrected deformation field through the deformation field generation algorithm. This deformation field can accurately describe the deformation of the body surface and internal anatomical structures after the patient's body position changes. Based on this dynamically corrected deformation field, through the three-dimensional model deformation algorithm, coordinate transformation and deformation processing can be performed on each point of the registered three-dimensional model of the body position according to the parameters of the deformation field, and the coordinates of the key anatomical landmark points can be recalculated to obtain the corrected coordinates of the key anatomical landmark points. These coordinates can more accurately reflect the anatomical structure position of the patient in the current body position, providing a reliable basis for subsequent body position correction and medical imaging, and ensuring the accuracy and effectiveness of diagnosis and treatment.

[0083] In one embodiment, the method further includes:

[0084] Calculating the local pressure gradient change rate of different body regions according to the adjusted body position pressure distribution data. If the local pressure gradient change rate is greater than a preset pressure limit value, an audible and visual alarm instruction is generated;

[0085] Based on the gray-scale distribution characteristics of the adjusted low-dose preview image, calculate the maximum equivalent tissue thickness of the target anatomical region through the pre-stored gray-scale-thickness calibration curve, and generate a tube voltage recommendation value;

[0086] According to the historical image quality data, predict through a random forest model to obtain the optimal milliampere-second parameter. The historical image quality data includes projection angle, tissue thickness, and detector type parameters;

[0087] Output the tube voltage recommendation value and the optimal milliampere-second parameter to the DR device exposure control system to generate a device parameter optimization instruction, which is used to adjust the exposure conditions.

[0088] Specifically, the adjusted body position pressure distribution data can be preprocessed to remove noise and outliers, and then methods such as the difference method or gradient operator are used to calculate the pressure gradient of each region, and the change rate of the pressure gradient over time is analyzed to obtain the local pressure gradient change rate. The preset pressure limit is used to judge whether the pressure distribution may cause discomfort to the patient or affect the imaging quality. If the local pressure gradient change rate exceeds the preset pressure limit, it indicates that the pressure change in a certain area of the patient's body is too fast or too large, generate an audible and visual alarm instruction, trigger the audible and visual alarm device of the device, and remind medical staff or patients to pay attention to the possible problems of the current body position.

[0089] Before the DR device performs a formal exposure, a low-dose preview image can be obtained first. This preview image can provide the basic gray-scale distribution information of the target anatomical area on the premise of ensuring patient safety. Analyze the gray-scale distribution characteristics of the low-dose preview image, and extract key parameters such as the gray-scale range, gray-scale mean, and gray-scale variance of the target anatomical area. Among them, the gray-scale distribution characteristics reflect information such as the density and thickness of the tissues in the target anatomical area. By analyzing these key parameters, the organizational structure of the target anatomical area can be initially understood. The pre-stored gray-scale-thickness calibration curve describes the relationship between different tissue thicknesses and the corresponding gray-scale values. The corresponding tissue thickness can be found on the gray-scale-thickness calibration curve according to the gray-scale value of the preview image, so as to obtain the maximum equivalent tissue thickness of the target anatomical area. According to the calculated maximum equivalent tissue thickness, combined with the X-ray imaging principle, a tube voltage recommendation value can be generated. Schematically, the tube voltage is a key parameter that affects the X-ray penetration ability and image contrast. Different tissue thicknesses and densities require corresponding tube voltages to ensure the best imaging effect. In addition, the historical image quality data can include information in multiple dimensions such as the projection angle, tissue thickness, and detector type parameters. Train the historical image quality data through a random forest model, use the projection angle, tissue thickness, and detector type parameters in the historical data as input features, and use the corresponding milliampere-second value as the output label, so that the model learns the complex non-linear relationship between the input features and the output label. Therefore, by inputting the current projection angle, tissue thickness, and detector type parameters into the trained random forest model, the optimal milliampere-second parameter can be obtained. This parameter can further reduce the radiation dose of X-rays and improve the imaging effect and safety of the DR device on the premise of ensuring image quality. Generate a device parameter optimization instruction based on the tube voltage recommendation value and the optimal milliampere-second parameter, and through the communication interface of the device, transmit the instruction to the control unit of the DR device to adjust its exposure conditions.

[0090] Based on the same inventive concept, as Figure 2 shown, the present application also provides a body position assisted detection system based on a DR device. The system includes:

[0091] An initial body position generation module 201, configured to perform spatial coordinate alignment processing on the pressure distribution data and three-dimensional depth information of the patient to generate an initial body position three-dimensional point cloud model;

[0092] A reference model matching module 202, configured to match the corresponding reference bone-muscle model from the pre-trained three-dimensional anatomical database according to the patient registration information, and perform non-rigid registration on the reference bone-muscle model and the initial body position three-dimensional point cloud model to generate a registered body position three-dimensional model, and output the coordinates of key anatomical landmark points. The patient registration information includes physiological parameters and medical history data;

[0093] The deviation vector calculation module 203 is configured to calculate the theoretical projection area of the target anatomical structure according to the current projection angle of the C-arm of the DR device and the coordinates of the key anatomical landmark points, and output a projection angle deviation vector by comparing the theoretical projection area with the actual projection boundary of the registered three-dimensional body position model. The projection angle deviation vector includes a rotation compensation angle and a translation compensation amount.

[0094] The guidance mark generation module 204 is configured to perform processing for generating body position correction guidance marks according to the projection angle deviation vector, and generate dynamic correction marks, which are used to drive the laser projection device to display the real-time adjustment direction.

[0095] In the above body position assistance detection system based on the DR device, the initial body position generation module 201 generates an initial body position three-dimensional point cloud model by performing spatial coordinate alignment processing on the pressure distribution data and three-dimensional depth information of the patient, which can comprehensively and accurately reflect the current body position state of the patient and provide basic data for subsequent model matching and analysis. The reference model matching module 202 can match the corresponding reference bone-muscle model from the pre-trained three-dimensional anatomy database according to the patient registration information, and perform non-rigid registration on it and the initial body position three-dimensional point cloud model, which can consider the elasticity and deformability of human tissues and make the reference model better fit the actual body position of the patient. The deviation vector calculation module 203 can calculate the theoretical projection area of the target anatomical structure according to the current projection angle of the C-arm of the DR device and the coordinates of the key anatomical landmark points, and by comparing the theoretical projection area with the actual projection boundary of the registered three-dimensional body position model, it can quantify the difference between the theoretical projection and the actual projection of the target anatomical structure at the current projection angle, obtain the projection angle deviation vector, and provide a clear direction and specific adjustment parameters for body position correction. The guidance mark generation module 204 can perform processing for generating body position correction guidance marks according to the projection angle deviation vector, and generate dynamic correction marks, enabling the patient to adjust the body position according to the visual indication, improving the efficiency and accuracy of body position correction, and reducing the time and number of times for the patient to adjust the body position.

[0096] Furthermore, the reference model matching module 202 includes:

[0097] A reference matching subunit, configured to screen out the reference bone-muscle model from the pre-trained three-dimensional anatomy database according to physiological parameters, where the physiological parameters include age, height, and weight data;

[0098] A personal dynamic feature subunit, configured to:

[0099] Through the iterative closest point algorithm, perform rough matching on the bone surface point cloud of the reference bone-muscle model and the initial body position point cloud of the initial body position three-dimensional point cloud model, and calculate the initial rigid body transformation matrix;

[0100] Apply the initial rigid body transformation matrix to the bone surface point cloud of the reference bone-muscle model to obtain the reference model after rough matching;

[0101] Perform non-rigid registration on the reference model after rough matching based on the thin plate spline deformation field model, optimize the deformation field parameters, and generate the three-dimensional model of the registered body position;

[0102] Extract the coordinates of key anatomical landmark points from the three-dimensional model of the registered body position. The coordinates of key anatomical landmark points include three-dimensional space coordinates and confidence scores.

[0103] Furthermore, the deviation vector calculation module 203 includes:

[0104] A projection subunit for:

[0105] Calculate the theoretical projection boundary of the target anatomical structure according to the current projection angle and the focal-detector geometric relationship;

[0106] Project the bone edge point cloud in the three-dimensional model of the registered body position onto the detector plane and extract the actual projection boundary;

[0107] A projection deviation quantification subunit for comparing the maximum deviation distance between the theoretical projection boundary and the actual projection boundary through the Hausdorff distance algorithm. If the maximum deviation distance is greater than the preset threshold, generate a projection angle deviation vector.

[0108] Furthermore, the guidance marker generation module 204 includes:

[0109] A rotation subunit for generating an angular scale line in the joint flexion direction according to the rotation compensation angle in the projection angle deviation vector;

[0110] A translation subunit for generating an arrow mark in the trunk displacement direction according to the translation compensation amount in the projection angle deviation vector;

[0111] A marker generation subunit for superimposing the angular scale line and the arrow mark and adjusting the color of the dynamic correction marker according to the deviation level to obtain the dynamic correction marker.

[0112] Furthermore, the system further includes a dynamic body position adjustment module for:

[0113] Calculate the displacement vector of the patient's body surface based on the continuously acquired real-time frame body surface point cloud data, generate a dynamic change signal, and the dynamic change signal includes a displacement direction change signal and an amplitude direction change signal;

[0114] According to the dynamic change signal and the contact area change rate, motion type classification processing is performed through the Kalman filtering algorithm to obtain a classification result, where the classification result includes physiological micro-motion and body position offset, and the contact area change rate is used to characterize the rate of change of the area of the contact region between the patient's body surface and the examination table over time;

[0115] When the classification result is body position offset, the thin plate spline deformation field parameters are updated according to the displacement vector to generate a dynamically corrected deformation field;

[0116] The dynamically corrected deformation field is applied to the registered three-dimensional body position model, the coordinates of the key anatomical landmark points are recalculated, and the corrected coordinates of the key anatomical landmark points are output, where the corrected coordinates of the key anatomical landmark points are used to update the representation of the coordinates of the key anatomical landmark points.

[0117] Furthermore, the system further includes an exposure adjustment module for:

[0118] Calculating the local pressure gradient change rate of different body regions according to the adjusted body position pressure distribution data, and if the local pressure gradient change rate is greater than a preset pressure limit value, generating an audible and visual alarm instruction;

[0119] Based on the gray-scale distribution characteristics of the adjusted low-dose preview image, calculating the maximum equivalent tissue thickness of the target anatomical region through a pre-stored gray-scale - thickness calibration curve to generate a tube voltage recommended value;

[0120] Predicting through a random forest model according to the historical image quality data to obtain the optimal milliampere-second parameter, where the historical image quality data includes the projection angle, tissue thickness, and detector type parameters;

[0121] Outputting the tube voltage recommended value and the optimal milliampere-second parameter to the DR device exposure control system to generate a device parameter optimization instruction, where the device parameter optimization instruction is used to adjust the exposure conditions.

[0122] In an exemplary embodiment, the present invention further provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of a method for position-assisted detection based on a DR device in this application are implemented. Preferably, a multi-core processor is used to improve the parallel processing ability of the system. Memory: Provide sufficient temporary storage space to support the operation of the program and the processing of data. The memory capacity should be large enough to accommodate a large amount of supply information and computing tasks.

[0123] In an exemplary embodiment, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a posture-assisted detection method based on a DR device according to the present application are implemented. The computer-readable storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), solid state drives (SSD, Solid State Drives), or optical discs, etc. Among them, the random access memory may include resistive random access memory (ReRAM, Resistance Random Access Memory) and dynamic random access memory (DRAM, Dynamic Random Access Memory).

[0124] The above-described embodiments merely represent several implementation manners of the embodiments of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A body position assisted detection method based on a DR device, characterized in that The method includes: Performing spatial coordinate alignment processing on the pressure distribution data and three-dimensional depth information of the patient to generate an initial body position three-dimensional point cloud model; According to the patient registration information, matching the corresponding reference bone-muscle model from the pre-trained three-dimensional anatomical database, and performing non-rigid registration on the reference bone-muscle model and the initial body position three-dimensional point cloud model to generate a registered body position three-dimensional model, and outputting the coordinates of key anatomical landmark points, where the patient registration information includes physiological parameters and medical history data; According to the current projection angle of the C-arm of the DR device and the coordinates of the key anatomical landmark points, calculating the theoretical projection area of the target anatomical structure, and outputting a projection angle deviation vector by comparing the theoretical projection area with the actual projection boundary of the registered body position three-dimensional model, where the projection angle deviation vector includes a rotation compensation angle and a translation compensation amount; Performing body position correction guidance marker generation processing according to the projection angle deviation vector to generate a dynamic correction marker, where the dynamic correction marker is used to drive the laser projection device to display the real-time adjustment direction.

2. The method according to claim 1, wherein The step of, according to the patient registration information, matching the corresponding reference bone-muscle model from the pre-trained three-dimensional anatomical database, and performing non-rigid registration on the reference bone-muscle model and the initial body position three-dimensional point cloud model to generate a registered body position three-dimensional model, and outputting the coordinates of key anatomical landmark points, includes: Screening out the reference bone-muscle model from the pre-trained three-dimensional anatomical database according to the physiological parameters, where the physiological parameters include age, height, and weight data; Performing rough matching on the bone surface point cloud of the reference bone-muscle model and the initial body position point cloud of the initial body position three-dimensional point cloud model through the iterative closest point algorithm, and calculating an initial rigid body transformation matrix; Applying the initial rigid body transformation matrix to the bone surface point cloud of the reference bone-muscle model to obtain a roughly matched reference model; Performing non-rigid registration on the roughly matched reference model based on the thin plate spline deformation field model, optimizing the deformation field parameters, and generating the registered body position three-dimensional model; Extracting the coordinates of the key anatomical landmark points from the registered body position three-dimensional model, where the coordinates of the key anatomical landmark points include three-dimensional space coordinates and confidence scores.

3. The method according to claim 2, characterized in that, The step of, according to the current projection angle of the C-arm of the DR device and the coordinates of the key anatomical landmark points, calculating the theoretical projection area of the target anatomical structure, and outputting a projection angle deviation vector by comparing the theoretical projection area with the actual projection boundary of the registered body position three-dimensional model, includes: Calculating the theoretical projection boundary of the target anatomical structure according to the current projection angle and the focus-detector geometric relationship; Projecting the bone edge point cloud in the registered body position three-dimensional model onto the detector plane, and extracting the actual projection boundary; Comparing the maximum deviation distance between the theoretical projection boundary and the actual projection boundary through the Hausdorff distance algorithm, and if the maximum deviation distance is greater than a preset threshold, generating the projection angle deviation vector.

4. The method according to claim 1, wherein Performing a body position correction guidance mark generation process based on the projection angle deviation vector to generate a dynamic correction mark, including: Generating an angle scale line in the joint flexion direction according to the rotation compensation angle in the projection angle deviation vector; Generating an arrow mark in the trunk displacement direction according to the translation compensation amount in the projection angle deviation vector; Superimposing the angle scale line and the arrow mark, and adjusting the color of the dynamic correction mark according to the deviation level to obtain the dynamic correction mark.

5. The method according to claim 2, characterized in that, The method further includes: Calculating a displacement vector on the patient's body surface based on continuously acquired real-time frame body surface point cloud data, and generating a dynamic change signal, where the dynamic change signal includes a displacement direction change signal and an amplitude direction change signal; Performing a motion type classification process on the dynamic change signal and the contact area change rate through a Kalman filter algorithm to obtain a classification result, where the classification result includes physiological micro-movement and body position offset, and the contact area change rate is used to characterize the rate of change of the area of the contact area between the patient's body surface and the examination bed over time; When the classification result is the body position offset, updating the thin plate spline deformation field parameters according to the displacement vector to generate a dynamically corrected deformation field; Applying the dynamically corrected deformation field to the registered body position three-dimensional model, recalculating the coordinates of key anatomical landmark points, and outputting the corrected coordinates of key anatomical landmark points, where the corrected coordinates of key anatomical landmark points are used to update and represent the coordinates of the key anatomical landmark points.

6. The method according to claim 1, characterized in that, The method further includes: Calculating the local pressure gradient change rate of different body regions according to the adjusted body position pressure distribution data, and generating an acoustic-optic alarm instruction if the local pressure gradient change rate is greater than a preset pressure limit value; Calculating the maximum equivalent tissue thickness of the target anatomical region based on the gray distribution characteristics of the adjusted low-dose preview image through a pre-stored gray-thickness calibration curve to generate a tube voltage recommendation value; Predicting through a random forest model according to historical image quality data to obtain the optimal milliamperage-second parameter, where the historical image quality data includes projection angle, tissue thickness, and detector type parameters; Outputting the tube voltage recommendation value and the optimal milliamperage-second parameter to the DR device exposure control system to generate a device parameter optimization instruction, where the device parameter optimization instruction is used to adjust the exposure conditions.

7. A body position assistance detection system based on a DR device, characterized in that The system includes: An initial body position generation module, configured to perform spatial coordinate alignment processing on the patient's pressure distribution data and three-dimensional depth information to generate an initial body position three-dimensional point cloud model; A reference model matching module, configured to match a corresponding reference bone-muscle model from a pre-trained three-dimensional anatomy database according to the patient registration information, and perform non-rigid registration on the reference bone-muscle model and the initial body position three-dimensional point cloud model to generate a registered body position three-dimensional model, and output the coordinates of key anatomical landmark points, where the patient registration information includes physiological parameters and medical history data; A deviation vector calculation module, configured to calculate a theoretical projection area of a target anatomical structure according to a current projection angle of the C-arm of the DR device and the coordinates of the key anatomical landmark points, and output a projection angle deviation vector by comparing the theoretical projection area with an actual projection boundary of the registered three-dimensional body position model, where the projection angle deviation vector includes a rotation compensation angle and a translation compensation amount; A guiding mark generation module, configured to perform a body position correction guiding mark generation process according to the projection angle deviation vector to generate a dynamic correction mark, where the dynamic correction mark is used to drive a laser projection device to display a real-time adjustment direction.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Citation Information

Cited By

  • Whole center dose verification method based on no-wash film EBT3

    CN120762083A

  • Curing shrinkage compensation and forming precision control method and system for 3D printing

    CN121671004A