A ct scan assisting method and system based on image detection

Abstract

The present application relates to the technical field of CT scan auxiliary, in particular to a CT scan auxiliary method and system based on image detection, comprising: collecting reference image data of a standard safety posture of a target object before scanning, and establishing a reference posture model containing body surface contour and key anatomical structure position; collecting image data in real time during scanning, and comparing with the reference posture model to calculate posture change parameters; establishing a multi-layer virtual safety boundary according to CT device structure parameters and scanning protocol; dynamically comparing the posture change parameters with the virtual safety boundary, judging the safety state and generating a signal to trigger corresponding level warning. The method can accurately capture posture changes, realize graded safety warning, solve the problems of inaccurate posture monitoring and delayed warning in conventional technology, and improve the safety and efficiency of CT scanning.

Description

Technical Field

[0001] This invention relates to the field of CT scan assistance technology, and in particular to a CT scan assistance method and system based on image detection. Background Technology

[0002] CT scans are a commonly used imaging technique in clinical diagnosis. During the scan, the posture stability of the target subject directly affects the image quality and the safety of the examination. Currently, conventional CT scan assistance techniques mainly rely on medical staff to guide the target subject's posture before the scan to determine the approximate scanning position before starting the scan. During the scan, changes in the target subject's posture are monitored only through simple position sensors built into the equipment or by visual observation by medical staff.

[0003] Conventional techniques lack precise modeling of the target object's standard posture, making it impossible to capture subtle posture changes during the scanning process in real time and accurately, and difficult to quantify the degree of posture deviation. Furthermore, the absence of a safety boundary system adapted to CT scanning equipment means that only simple warnings can be issued when the target object's posture deviates significantly, without the ability to provide tiered warnings based on the degree of posture change. This can easily lead to artifacts in the scanned images, scan failures, and even damage to the equipment or injury to the target object if the target object's posture exceeds the equipment's safety limits. Therefore, how to accurately acquire the target object's posture changes during the scanning process, and how to establish scientific safety boundaries and implement tiered warnings, have become problems that need to be solved in current CT scan-assisted technologies. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a CT scan assistance method and system based on image detection.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a CT scan-assisted method based on image detection, comprising: Before the CT scan begins, baseline image data of the target object on the CT scan bed in a preset standard safe posture is acquired by an image acquisition device. The reference image data is processed to extract and establish a reference posture model that reflects the surface contour and key anatomical structure position of the target object in the standard safe posture. During a CT scan, the image acquisition device continuously and in real-time acquires real-time image data of the target object. The real-time image data is registered and compared with the reference pose model. The pose change parameters of the target object during the real-time scanning process are calculated by image registration and feature matching algorithms. Based on the structural parameters and scanning protocol of the CT scanning equipment, a multi-layered virtual security boundary is established inside and around the CT scanning aperture; The attitude change parameters are dynamically compared with the multi-layer virtual safety boundary to determine the real-time attitude of the target object relative to the multi-layer virtual safety boundary and generate a corresponding safety status signal. Based on the security status signal, a warning operation corresponding to different security status levels is triggered.

[0006] As a further aspect of the present invention, before the CT scan begins, reference image data of the target object located on the CT scan bed in a preset standard safe posture is acquired by an image acquisition device, including: After the CT scanning equipment has positioned the target object and the target object is in a preset standard safe posture, the image acquisition device installed on the CT scanning gantry or a fixed position in the scanning room is activated. The image acquisition device is a device for acquiring three-dimensional spatial coordinate information. The image acquisition device performs a complete scan on the target object in a stationary state to acquire a set of three-dimensional coordinate points covering the overall outline of the target object on the CT scanning bed. The set of three-dimensional coordinate points constitutes the reference image data.

[0007] As a further aspect of the present invention, the reference image data is processed to extract and establish a reference posture model reflecting the surface contour and key anatomical structure positions of the target object in the standard safe posture, including: The set of three-dimensional coordinate points in the reference image data is subjected to denoising filtering to remove isolated noise point coordinates; The denoised set of three-dimensional coordinate points is spatially rasterized to construct a dense three-dimensional voxel model representing the surface morphology of the target object. In the dense three-dimensional voxel model, a feature recognition algorithm is used to identify and mark the preset key anatomical structure regions, which include at least the head region, shoulder region, arm region, chest region, and lower limb region. Extract the spatial coordinate range and geometric center point coordinates of the key anatomical structure region; A dense three-dimensional voxel model containing all the key anatomical structure regions' labeling information and their spatial coordinate features is stored as the baseline pose model.

[0008] As a further aspect of the present invention, during a CT scan, the image acquisition device continuously and in real-time acquires real-time image data of the target object, including: During the process of the CT scanning equipment starting to execute the scanning program, the scanning bed moving and / or the gantry rotating, the image acquisition device is controlled to work continuously at a set sampling frequency; At each sampling moment, the image acquisition device synchronously acquires a set of three-dimensional coordinate points covering the overall contour of the target object in its current pose. The set of three-dimensional coordinate points acquired continuously at multiple time points in a time series is used as the real-time image data stream.

[0009] As a further aspect of the present invention, the real-time image data is registered and compared with the reference pose model. Through image registration and feature matching algorithms, the pose change parameters of the target object during the real-time scanning process are calculated, including: Extract the set of three-dimensional coordinate points corresponding to the current sampling time from the real-time image data stream, and use it as the current real-time point cloud; The current real-time point cloud is subjected to the same preprocessing as when the reference pose model was established, including denoising and spatial rasterization, to generate the real-time voxel model at the current moment; An iterative nearest neighbor registration algorithm is used to spatially align the real-time voxel model with the dense 3D voxel model of the reference pose model, and the spatial transformation matrix from the real-time coordinate system to the reference coordinate system is calculated. The geometric center coordinates of the key anatomical structure region identified in the real-time voxel model are aligned with the geometric center coordinates of the corresponding key anatomical structure region in the baseline pose model based on the spatial transformation matrix. After alignment, calculate the relative displacement vector of the geometric center point of each corresponding key anatomical region in three-dimensional space; Calculate the magnitude of the relative displacement vector as the displacement amplitude parameter of the key anatomical structure region; Calculate the direction and velocity of the movement of the geometric center point of the key anatomical structure region over multiple consecutive sampling times; The displacement amplitude parameters, movement direction and velocity information of all key anatomical structural regions are summarized to form the attitude change parameters.

[0010] As a further aspect of the present invention, based on the structural parameters and scanning protocol of the CT scanning device, a multi-layered virtual security boundary is established inside and around the CT scanning aperture, including: Obtain the aperture geometry parameters of the CT scanning equipment gantry, including the inner radius of the aperture, the axial length of the aperture, and the position of the center axis of the aperture; Obtain the planned movement trajectory and speed of the scanning bed, as well as the gantry rotation range, as defined in the current scanning protocol; Inside the aperture, with the inner surface of the aperture as a reference, a preset first safety distance is offset inward to generate a coaxial virtual cylindrical surface, which serves as the inner virtual safety boundary. The area inside the inner virtual safety boundary is defined as the danger zone. Inside the aperture, with the inner surface of the aperture as a reference, a preset second safety distance is offset inward. The second safety distance is greater than the first safety distance, generating another coaxial virtual cylindrical surface as a middle virtual safety boundary. The annular space between the middle virtual safety boundary and the inner virtual safety boundary is defined as a warning area. Outside the aperture, along the direction of scan bed movement, in the adjacent area outside the planned scanning range of the target object, a virtual warning boundary is defined. The area outside the warning boundary is defined as the safe zone, and the area inside is defined as the attention zone. The inner virtual security boundary, the middle virtual security boundary, and the warning boundary together constitute the multi-layered virtual security boundary.

[0011] As a further aspect of the present invention, the attitude change parameters are dynamically compared with the multi-layer virtual safety boundary to determine the real-time attitude of the target object relative to the multi-layer virtual safety boundary, and a corresponding safety status signal is generated, including: From the posture change parameters, extract the real-time three-dimensional coordinates of the geometric center point of the key anatomical structure region on the target object that has moved; Calculate the shortest spatial distance from the real-time 3D coordinates to the inner virtual security boundary; Calculate the shortest spatial distance from the real-time 3D coordinates to the mid-level virtual security boundary; Determine whether the real-time three-dimensional coordinates are located within the attention area defined by the warning boundary; Set the security status judgment logic: If the shortest spatial distance from the real-time three-dimensional coordinates to the inner virtual security boundary is less than zero, it is determined that the target object has invaded the dangerous area, and the security status signal is "dangerous status"; If the shortest spatial distance from the real-time 3D coordinates to the inner virtual security boundary is greater than zero, but the shortest spatial distance to the middle virtual security boundary is less than zero, then the target object is determined to have entered the warning area, and the security status signal is "warning status". If the shortest spatial distance from the real-time 3D coordinates to the middle-layer virtual security boundary is greater than zero, but is located within the attention area, then the target object is determined to be in the attention state, and the security state signal is "attention state". If neither of the above applies, then the security state signal is "safe state". Combining the movement direction and speed information in the posture change parameters, if any key anatomical structure region is detected to be moving toward the inner virtual safety boundary or the middle virtual safety boundary at a speed exceeding a preset threshold, the corresponding safety status signal is upgraded by one level or marked as "trend warning", and the final safety status signal is output.

[0012] As a further aspect of the present invention, based on the security status signal, triggering early warning operations corresponding to different security status levels includes: The early warning operation is completed collaboratively by multiple implementing agencies; When the safety status signal is "attention status", no external visible or audible warning signals are triggered, and only a visual prompt is given on the system control interface with a specific color or icon. When the safety status signal is "warning status", a primary warning operation is triggered. The primary warning operation includes at least activating the yellow flashing warning light installed in the scanning room and playing a gentle voice prompt to the CT operator through the voice broadcast system. When the safety status signal is "dangerous" or marked as "trend warning", advanced warning and emergency operation are triggered. The advanced warning and emergency operation includes at least activating the red rotating warning light and high-frequency alarm installed in the scanning room, playing a clear stop movement command to the target object through the voice broadcast system, and sending a digital control signal containing an emergency stop command to the control system of the CT scanning equipment.

[0013] As a further aspect of the present invention, in the advanced early warning and emergency operation, sending a digital control signal containing an emergency stop command to the control system of the CT scanning equipment includes: The digital control signal is sent to the central scanning control system of the CT scanning equipment through a preset data interface protocol; The digital control signal contains specific instruction codes that instruct the CT scanning device to perform a series of synchronized emergency stop actions; The emergency stop actions include: immediately stopping the exposure of the X-ray tube and stopping the output of the high voltage generator; immediately sending a braking command to the scanning bed drive motor to bring the scanning bed to a smooth stop within the shortest safe distance; and sending a stop command to the gantry drive system to stop the rotation of the gantry. The emergency stop action has a higher priority than the regular scan control commands of the CT scanner.

[0014] As a further aspect of the present invention, the present invention also includes an image detection-based CT scan assistance system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the image detection-based CT scan assistance method described above.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Before a CT scan begins, baseline image data of the target object in a preset standard safe posture is acquired. A baseline posture model reflecting the object's surface contour and the location of key anatomical structures is extracted and established. During the scan, real-time image data of the target object is continuously acquired. The real-time image data and the baseline posture model are registered and compared using image registration and feature matching algorithms to calculate the posture change parameters of the target object during the real-time scan. This technology can accurately capture subtle posture changes of the target object during the scan, quantify posture deviation, and avoid the problems of fuzzy posture monitoring and inability to quantify the degree of deviation in conventional technologies. It can reduce scan image artifacts caused by posture deviation, lower the scan repetition rate, and provide accurate posture reference for safety assessment by monitoring the target object's posture status in real time.

[0016] Based on the structural parameters and scanning protocol of the CT scanning equipment, a multi-layered virtual safety boundary is established inside and around the CT scanning aperture. The calculated attitude change parameters are dynamically compared with this multi-layered virtual safety boundary to determine the real-time safety status of the target object relative to the boundary, generating corresponding safety status signals. Based on these signals, warning operations corresponding to different safety status levels are triggered. This technology is adaptable to the structural characteristics and scanning requirements of different CT equipment. The multi-layered virtual safety boundary comprehensively covers safety risk points during the scanning process, and the graded warnings provide corresponding alerts based on the severity of attitude deviations. This avoids the problems of delayed warnings and limited warning methods in conventional technologies, allowing for timely intervention in unsafe postures, reducing the risk of collisions between the equipment and the target object, ensuring the safety of the scanning process, and improving the versatility of the auxiliary technology by adapting to different scanning protocols. Attached Figure Description

[0017] Figure 1 This is a flowchart of an image detection-based CT scan assistance method according to the present invention; Figure 2 A flowchart for acquiring baseline image data; Figure 3 A flowchart for acquiring real-time image data; Figure 4 A three-dimensional distribution and safety boundary map of key anatomical structures in a CT scan; Figure 5 This is a graph showing the execution time of the CT emergency stop action. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0020] See Figure 1 Before the CT scan begins, a baseline image of the target object in a preset standard safe posture is acquired using an image acquisition device. This baseline image data is processed to extract and establish a baseline posture model reflecting the object's surface contour and key anatomical structure positions in the standard safe posture. During the CT scan, real-time image data of the target object is continuously acquired using the image acquisition device. The real-time image data is registered and compared with the baseline posture model. Image registration and feature matching algorithms are used to calculate the posture change parameters of the target object during the real-time scan. Based on the structural parameters and scanning protocol of the CT scanning equipment, multiple virtual safety boundaries are established inside and around the CT scanning aperture. The posture change parameters are dynamically compared with these multiple virtual safety boundaries to determine the real-time safety status of the target object relative to these boundaries and generate corresponding safety status signals. Based on these safety status signals, warning operations corresponding to different safety status levels are triggered.

[0021] In one embodiment of the present invention, the image acquisition device is mounted on the fixed housing of the CT scanning gantry, and its field of view covers the entire effective area of ​​the scanning bed within its axial movement range. (See also...) Figure 2After the CT scanning equipment completes the positioning of the target object, and the target object is indicated and confirmed to be in a preset standard safe supine posture, the image acquisition device is automatically activated by the control software. The image acquisition device is a three-dimensional depth camera based on the time-of-flight principle. The three-dimensional depth camera performs a complete scan of the target object lying still on the scanning bed. The scanning process is completed in one pass by a robotic arm that drives the three-dimensional depth camera to move at a constant speed along the axis of the scanning bed, acquiring a set of three-dimensional coordinate points covering the overall outline of the target object from head to toe. This set of three-dimensional coordinate points is marked as reference image data and stored in a cache. In some embodiments, the image acquisition device can also be multiple laser scanners fixed on the ceiling track of the scanning room. Multiple laser scanners scan the stationary target object simultaneously from different angles. The multiple sets of point cloud data generated are fused by a coordinate system to form a complete set of three-dimensional coordinate points as reference image data.

[0022] The reference image data is processed to extract and establish a reference pose model. In specific implementation, the set of three-dimensional coordinate points in the reference image data is subjected to noise filtering. The processing includes statistical filtering and radius filtering. Statistical filtering removes isolated noise point coordinates with a point density below a threshold in the neighborhood. Radius filtering further removes outlier point coordinates with fewer than a threshold number of points in a space centered at each point and with a specified radius. The filtered set of three-dimensional coordinate points is input into the spatial rasterization processing module. The spatial rasterization processing module defines a unified three-dimensional voxel grid. The resolution parameters of the voxel grid are preset according to the scanning accuracy requirements. The voxel grid cell to which each three-dimensional coordinate point belongs is calculated. Empty voxel grid cells are assigned a value of zero, and voxel grid cells containing at least one coordinate point are marked as one, thereby constructing a dense three-dimensional voxel model representing the surface morphology of the target object. The dense three-dimensional voxel model is a binary three-dimensional matrix. In a dense 3D voxel model, a pre-trained 3D convolutional neural network feature recognition algorithm identifies and marks predefined key anatomical structures, including the head, left shoulder, right shoulder, left arm, right arm, chest, left lower limb, and right lower limb. All voxel indices in voxel space are extracted for each identified key anatomical structure region, and its 3D bounding box is calculated to obtain its spatial coordinate range. The geometric center coordinates of the region are then calculated based on all voxel indices. The calculation of the geometric center coordinates can be determined based on the mean of the voxel indices. The coordinate calculation follows the following relationship:

[0023] in: This represents the coordinate vector of the geometric center point of the k-th key anatomical structure region. This indicates the number of voxels belonging to this region. This represents the coordinate vector of the center position of the i-th voxel in three-dimensional space. Optionally, the feature recognition algorithm can also use a traditional geometric segmentation method based on curvature and normal vectors instead of a three-dimensional convolutional neural network. A dense three-dimensional voxel model containing the labeling information of all key anatomical structure regions, their spatial coordinate range, and the coordinates of their geometric center points is associated with the scan identifier of the target object and stored in a database as the baseline pose model for this scanning task. In some embodiments, the data structure of the baseline pose model also includes a graph structure describing the topological connections between the key anatomical structure regions. The establishment of the baseline pose model provides a static and accurate spatial reference framework for subsequent real-time comparisons.

[0024] In one embodiment of the present invention, during a CT scan, real-time image data of the target object is continuously and in real-time acquired by an image acquisition device, see reference. Figure 3 The CT scanning equipment begins executing the scanning program. The scanning bed moves axially at a speed of 15 millimeters per second, while the gantry rotates at a rate of two revolutions per second. The image acquisition device operates continuously at a sampling frequency of 10 hertz. At each sampling moment, the image acquisition device synchronously triggers and acquires a depth image covering the overall contour of the target object in its current pose. The depth image is then transformed to generate a corresponding set of three-dimensional coordinate points. The sets of three-dimensional coordinate points acquired sequentially at multiple sampling moments are arranged in timestamp order in a memory buffer and transmitted and processed as a real-time image data stream. Each data packet in the real-time image data stream contains a timestamp, the number of point clouds, and a three-dimensional coordinate array. In some embodiments, the sampling frequency can be dynamically adjusted according to the scanning bed's moving speed. When the scanning bed's moving speed is less than 10 millimeters per second, the sampling frequency is set to 5 hertz; when the scanning bed's moving speed is greater than 20 millimeters per second, the sampling frequency is set to 20 hertz.

[0025] Real-time image data is registered and compared with a baseline pose model. Image registration and feature matching algorithms are used to calculate the pose change parameters of the target object during real-time scanning. In practice, the set of 3D coordinate points corresponding to the most recent sampling moment is extracted from the real-time image data stream based on the system's current clock, and this set is marked as the current real-time point cloud. The current real-time point cloud undergoes the same preprocessing as when the baseline pose model was established. Preprocessing includes removing isolated noise coordinates using the same statistical filtering parameters and performing spatial rasterization using the same voxel size parameters as the baseline pose model to generate the current real-time voxel model, which is a binary 3D matrix with the same resolution. An iterative nearest neighbor registration algorithm is employed to spatially align the real-time voxel model with the dense 3D voxel model of the baseline pose model. The algorithm uses voxels with a value of one in the real-time voxel model as the source set and voxels with a value of one in the dense 3D voxel model of the baseline pose model as the target set. By iteratively solving for the optimal rotation matrix and translation vector, the spatial transformation matrix from the real-time coordinate system to the baseline coordinate system is calculated. The geometric center coordinates of key anatomical structures identified in the real-time voxel model using the same feature recognition algorithm are aligned with the corresponding geometric center coordinates of key anatomical structures in the baseline pose model based on the spatial transformation matrix. This alignment operation involves multiplying the geometric center coordinates of the real-time key anatomical structures by the spatial transformation matrix. After alignment, the relative displacement vector of the geometric center point of each corresponding key anatomical structure in 3D space is calculated. This relative displacement vector is the difference vector between the real-time coordinates and the baseline coordinates. The magnitude of the relative displacement vector is calculated as the displacement amplitude parameter for the corresponding key anatomical structure. This displacement amplitude parameter can be understood as a scalar value. The direction and velocity of the movement of the geometric center point of the key anatomical structure region are calculated over multiple consecutive sampling times. In practice, the velocity calculation can be determined based on the positional changes of the geometric center point in five consecutive frames of data.

[0026] in: This represents the velocity vector of the k-th key anatomical structure region at time t. This represents the coordinate vector of the geometric center point of the region at time t. This represents the coordinate vector of the geometric center point at the fourth sampling time before time t. This represents the sampling time interval. Optionally, the direction of movement is represented by a unit vector of the displacement vector. The displacement amplitude parameters, direction of movement, and velocity information of all key anatomical structural regions are aggregated into a structured data package as attitude change parameters. The data package contains an identifier for each region, a three-dimensional displacement vector, displacement amplitude, velocity vector, and a timestamp. In some embodiments, the attitude change parameters also include an overall attitude offset evaluation value, which is calculated by the weighted sum of squares of the displacement amplitude parameters of all key anatomical structural regions. It can be understood that the attitude change parameters are the core input for subsequent safety assessments.

[0027] In one embodiment of the present invention, based on the structural parameters and scanning protocol of the CT scanning equipment, multiple virtual safety boundaries are established inside and around the CT scanning aperture. The aperture geometric parameters of the CT scanning equipment gantry are obtained, including the inner radius, axial length, and central axis position. In a specific scenario, the inner radius of a standard CT scanning equipment aperture is 350 mm, the axial length is 800 mm, and the central axis position is defined by a linear equation in the equipment coordinate system. The planned movement trajectory and speed of the scanning bed, as well as the gantry rotation range, defined in the current scanning protocol are obtained. For a single chest scan, the protocol defines the scanning bed to translate 400 mm along the equipment axis from its starting position at a speed of 20 mm per second, and the gantry rotation range is set from 0 degrees to 240 degrees.

[0028] Inside the aperture, using the inner surface of the aperture as a reference, a pre-set first safety distance is offset inward to generate a coaxial virtual cylindrical surface as the inner virtual safety boundary. The first safety distance is set to 50 mm based on equipment manufacturing tolerances and safety redundancy. The area within the inner virtual safety boundary is defined as the danger zone, which is the space within 50 mm of the physical inner wall of the rack. Inside the aperture, using the inner surface of the aperture as a reference, another pre-set second safety distance is offset inward to generate another coaxial virtual cylindrical surface as the middle virtual safety boundary. The second safety distance is greater than the first safety distance and is set to 100 mm. The annular space between the middle virtual safety boundary and the inner virtual safety boundary is defined as the warning zone, which is the annular space between 50 mm and 100 mm from the physical inner wall of the rack. Outside the aperture, along the scanning bed movement direction, a virtual warning boundary is defined in the adjacent area outside the planned scanning range of the target object. The planned scanning range extends 400 mm from the head starting point towards the feet. The warning boundary is set at a virtual plane perpendicular to the scanning bed movement direction, extending 50 mm from the end of the planned scanning range towards the feet. The area outside the warning boundary is defined as the safe zone, and the area inside is defined as the attention zone. The attention zone covers the planned scanning range and its 50 mm outer extension. The inner virtual safe boundary, the middle virtual safe boundary, and the warning boundary together constitute a multi-layered virtual safe boundary. The multi-layered virtual safe boundary is stored in the computer in the form of a data structure, including the geometric parameters and spatial position equations of each boundary. In some embodiments, the inner virtual safe boundary and the middle virtual safe boundary are strictly cylindrical surface models, whose mathematical expressions can be described in the scanning bed coordinate system. The equation of the inner virtual safe boundary cylindrical surface is expressed as:

[0029] in: This represents the radial coordinates of any point on the inner virtual safety boundary cylindrical surface in cylindrical coordinates. This indicates the inner radius of the aperture of the CT scanner gantry. This represents the preset first safety distance. The radial coordinates of the middle-layer virtual safety boundary are... ,in This represents the preset second safety distance. It can be understood that the above equation defines the radial constraint of the boundary. The warning boundary is defined by a planar equation, with the plane normal along the scanning bed axis, and its position determined by the endpoint of the scanning range and the extension distance. In some embodiments, for scanning devices with non-cylindrical apertures, the inner and middle virtual safety boundaries are complex surfaces generated by equidistant offsets from the three-dimensional model of the aperture's inner surface. Optionally, the warning boundary is not a single plane, but rather consists of a three-dimensional surface surrounding the planned scanning area, used to more precisely define the extent of the attention area. The spatial relationships, geometric parameters, and corresponding region definitions of the inner, middle, and warning boundaries are encapsulated in a configuration file, loaded during each scan initialization.

[0030] In one embodiment of the present invention, the real-time judgment and signal generation of the safety status involves dynamically comparing the attitude change parameters with the multi-layer virtual safety boundary to determine the real-time safety status of the target object relative to the multi-layer virtual safety boundary and generating a corresponding safety status signal. From the attitude change parameters, the real-time three-dimensional coordinates of the geometric center points of the key anatomical structural regions on the target object that have moved are extracted. For example, at a certain sampling moment, the extracted real-time three-dimensional coordinates include the coordinates of the geometric center point of the left arm region (120, 300, 450) and the coordinates of the geometric center point of the head region (0, 250, 1200), with the coordinate unit being millimeters. The origin of the coordinate system is located at the center point of the frame. The shortest spatial distance from the real-time three-dimensional coordinates to the inner virtual safety boundary is calculated. The distance calculation transforms the spatial coordinates of the point to a cylindrical coordinate system with the frame center axis as the Z-axis. The radial coordinates of the point and the radial coordinates of the inner virtual safety boundary cylindrical surface are calculated. The difference. Calculate the shortest spatial distance from the real-time 3D coordinates to the middle-layer virtual safety boundary. The calculation method is to compare the radial coordinates of the calculation point with the radial coordinates of the cylindrical surface of the middle-layer virtual safety boundary. The difference. To determine whether the real-time 3D coordinates are within the attention area defined by the warning boundary, the method is to check whether the axial coordinate of the check point is less than the axial coordinate value of the warning boundary plane.

[0031] The safety status judgment logic is set up and executed based on preset distance thresholds and area definitions. If the shortest spatial distance from the real-time 3D coordinates to the inner virtual safety boundary is less than zero, the target object is determined to have entered the danger zone, and the safety status signal is "dangerous state". If the shortest spatial distance from the real-time 3D coordinates to the inner virtual safety boundary is greater than zero, but the shortest spatial distance to the middle virtual safety boundary is less than zero, the target object is determined to have entered the warning zone, and the safety status signal is "warning state". If the shortest spatial distance from the real-time 3D coordinates to the middle virtual safety boundary is greater than zero, but it is located within the attention zone, the target object is determined to be in the attention state, and the safety status signal is "attention state". If none of the above conditions are met, i.e., the point is located outside the attention zone and the distance to the middle virtual safety boundary is greater than zero, the safety status signal is "safe state". Combining the movement direction and speed information in the attitude change parameters, if any key anatomical structure area is detected moving towards the inner virtual safety boundary or the middle virtual safety boundary at a speed exceeding a preset threshold, the corresponding safety status signal is upgraded by one level or marked as "trend warning", and the final safety status signal is output. In some embodiments, the speed threshold is set to 10 millimeters per second, and the direction of movement is determined based on whether the projection of the speed vector onto the direction vector from the point to the nearest point on the boundary is positive and greater than the threshold. This can be understood as the shortest spatial distance... The calculation formula can be expressed as:

[0032] in: Represents the real-time three-dimensional coordinates of the geometric center point of key anatomical structures. This indicates the calculated virtual security boundary (inner or middle layer). Function calculation points to the border The Euclidean distance, for the boundary of a cylindrical surface, is the distance from the point... The distance to the axis of the cylinder. This represents the radius of the cylinder surface of the virtual security boundary (i.e. or Optionally, the security status determination logic is implemented through a status lookup table, taking the distance comparison result and the region determination result as input. See Table 1 for a simplified example of the determination logic.

[0033] Table 1: Logic Table for Determining Security Status

[0034] In some embodiments, the generation of the safety status signal is periodic, consistent with the sampling frequency of the image acquisition device. Each processing cycle calculates and outputs a safety status signal based on the latest attitude change parameters and fixed multi-layered virtual safety boundaries. Movement direction and velocity information are extracted from the attitude change parameters. If the velocity value exceeds 10 millimeters per second and the movement direction points towards the boundary, the current basic status signal is upgraded by one level, for example, from "attention status" to "warning status," or a "trend warning" flag is superimposed on the original status signal. It can be understood that the final output safety status signal is a data structure containing the status level, the identifier of the key anatomical structure region that triggered the status, and the presence or absence of the "trend warning" flag.

[0035] See Figure 4 This is a 3D distribution and safety boundary map of key anatomical structures in a CT scan. It visually displays the 3D spatial location of the target object's key anatomical structures and the two layers of virtual safety boundaries within the CT aperture. The head, chest cavity, right arm, left arm, and lower limb are each represented by a point that is the geometric center of their respective region. The radial distance of all anatomical structures is significantly greater than the middle layer safety boundary, and their Z-coordinates are all within the planned scan range, indicating a safe state. The inner / middle layer boundaries are coaxial cylindrical surfaces used to assess radial intrusion risk; the axial boundaries are used to determine whether the target object exceeds the scan range. The clear spatial distribution of the target object within the CT aperture facilitates a direct assessment of posture deviation risk. By calculating the distance from each anatomical structure to the boundary in real time, the safety status can be quickly determined.

[0036] In one embodiment of the present invention, based on a safety status signal, a warning operation corresponding to different safety status levels is triggered. The warning operation is completed collaboratively by multiple actuators, including a warning light system, a voice broadcast system, a system control interface, and a data interface module of the CT scanning equipment installed in the scanning room. In a specific implementation, when the safety status signal is "attention status," no externally visible or audible warning signal is triggered. Instead, a visual prompt is provided on the system control interface using a specific color or icon. The system control interface continuously displays a blue static icon in the patient status bar of the graphical user interface, with the text "Normal Posture" displayed next to the icon.

[0037] When the safety status signal is "Warning Status," a primary warning operation is triggered. This operation includes at least activating a flashing yellow warning light installed in the scanning room and playing a gentle voice prompt to the CT operator via a voice broadcast system. In specific implementations, the flashing yellow warning light operates at a frequency of once per second, and the voice broadcast system plays a pre-recorded voice message, "Warning: Patient's posture is approaching the warning area, please pay attention," with the voice volume set to 60 decibels. In some embodiments, the primary warning operation also includes displaying a non-modal prompt window on the CT operator's main control screen. The window has a yellow background and displays the specific information, "Warning: Patient's left arm is approaching the warning boundary." Optionally, the content of the voice prompt can be dynamically synthesized based on the specific anatomical structure of the intrusion warning area; for example, when the left arm enters the warning area, the voice message might be, "Warning: Patient's left arm is approaching the warning area."

[0038] When the safety status signal is "dangerous" or marked as "trend warning," advanced warning and emergency operations are triggered. These operations include at least activating a red rotating warning light and a high-frequency alarm installed in the scanning room, playing a clear stop-motion command to the target object via a voice broadcast system, and sending a digital control signal containing an emergency stop command to the CT scanner's control system. In specific implementations, the red rotating warning light rotates at 120 revolutions per minute, the high-frequency alarm emits intermittent ringing at a frequency of 2500 Hz and a sound pressure level of 85 dB, and the voice broadcast system plays the command "Please stop moving immediately and remain still" to the target object on the scanning bed, repeated three times at a clear and gentle pace. In some embodiments, the advanced warning and emergency operations also simultaneously send the highest-level pop-up alarm to the operator console and record the event log.

[0039] In advanced early warning and emergency operations, a digital control signal containing an emergency stop command is sent to the control system of the CT scanner. This digital control signal is transmitted to the central scanning control system of the CT scanner via a preset data interface protocol. The data interface protocol uses the TCP / IP-based DICOM communication protocol, with a signal transmission delay requirement of less than 100 milliseconds. The digital control signal contains a specific command code that instructs the CT scanner to perform a series of synchronized emergency stop actions. The command code is a 32-bit integer, its value generated based on the type and urgency of the safety status signal. It can be understood that the generation of the command code follows a mapping relationship; for example, "dangerous status" maps to code 0xE001, and "trend warning" maps to code 0xE002. The data structure of the digital control signal can be represented as follows:

[0040] in: This represents a complete digital control signal data packet. Indicates the instruction code field. This indicates the signal priority field. The timestamp field indicates the time of signal generation. The instruction code field... Storing specific values ​​such as 0xE001, priority field Fixed at the highest priority value of 255, timestamp field The system time for signal generation is recorded. Emergency stop actions include: immediately stopping X-ray tube exposure and halting the high-voltage generator output; immediately sending a braking command to the scanning bed drive motor to bring the scanning bed to a smooth stop within the shortest safe distance; and sending a stop command to the gantry drive system to stop the gantry's rotation. In practice, after the scanning bed braking command is triggered, the scanning bed drive motor initiates electrical braking within 200 milliseconds to ensure the scanning bed comes to a complete stop within a sliding distance of no more than 10 millimeters; upon receiving the stop command, the gantry drive system stops rotation within 500 milliseconds using dynamic braking. Emergency stop actions have higher priority than regular scanning control commands of the CT scanner. Upon receiving the digital control signal, the central scanning control system will interrupt any ongoing regular scanning command sequence and prioritize the emergency stop action.

[0041] See Figure 5 This is a chart analyzing the execution time of emergency stops in CT scans, visually demonstrating the response delays of different emergency braking operations. The X-ray tube and high-voltage generator stops have extremely short execution times; these are electrical cut-off operations that quickly terminate radiation output, ensuring patient radiation safety. Scanning bed braking is a mechanical braking operation, requiring overcoming inertia, resulting in a slightly longer response time. Gantry rotation stopping is the most time-consuming action, as the gantry's large mass and strong inertia necessitate a longer period for smooth deceleration. The execution sequence is highly consistent with clinical safety priorities: first, the radiation source is cut off; then, the scanning bed is braked; finally, the gantry rotation is stopped, minimizing the risk of radiation and mechanical collisions. The rapid stopping of the X-ray tube and high-voltage generator keeps additional radiation exposure to extremely low levels. The difference in braking time between the scanning bed and gantry reflects the inertial characteristics of different moving parts and the corresponding safety braking strategies.

[0042] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A CT scan-assisted method based on image detection, characterized in that, The method includes: Before the CT scan begins, baseline image data of the target object on the CT scan bed in a preset standard safe posture is acquired by an image acquisition device. The reference image data is processed to extract and establish a reference posture model that reflects the surface contour and key anatomical structure position of the target object in the standard safe posture. During a CT scan, the image acquisition device continuously and in real-time acquires real-time image data of the target object. The real-time image data is registered and compared with the reference pose model. The pose change parameters of the target object during the real-time scanning process are calculated by image registration and feature matching algorithms. Based on the structural parameters and scanning protocol of the CT scanning equipment, a multi-layered virtual security boundary is established inside and around the CT scanning aperture; The attitude change parameters are dynamically compared with the multi-layer virtual safety boundary to determine the real-time attitude of the target object relative to the multi-layer virtual safety boundary and generate a corresponding safety status signal. Based on the security status signal, a warning operation corresponding to different security status levels is triggered.

2. The CT scan-assisted method based on image detection according to claim 1, characterized in that, Before the CT scan begins, baseline image data of the target object on the CT scan bed in a preset standard safe posture is acquired using an image acquisition device, including: After the CT scanning equipment has positioned the target object and the target object is in a preset standard safe posture, the image acquisition device installed on the CT scanning gantry or a fixed position in the scanning room is activated. The image acquisition device is a device for acquiring three-dimensional spatial coordinate information. The image acquisition device performs a complete scan on the target object in a stationary state to acquire a set of three-dimensional coordinate points covering the overall outline of the target object on the CT scanning bed. The set of three-dimensional coordinate points constitutes the reference image data.

3. The CT scan-assisted method based on image detection according to claim 2, characterized in that, The reference image data is processed to extract and establish a reference pose model reflecting the surface contour and key anatomical structure positions of the target object in the standard safe pose, including: The set of three-dimensional coordinate points in the reference image data is subjected to denoising filtering to remove isolated noise point coordinates; The denoised set of three-dimensional coordinate points is spatially rasterized to construct a dense three-dimensional voxel model representing the surface morphology of the target object. In the dense three-dimensional voxel model, a feature recognition algorithm is used to identify and mark the preset key anatomical structure regions, which include at least the head region, shoulder region, arm region, chest region, and lower limb region. Extract the spatial coordinate range and geometric center point coordinates of the key anatomical structure region; A dense three-dimensional voxel model containing all the key anatomical structure regions' labeling information and their spatial coordinate features is stored as the baseline pose model.

4. The CT scan-assisted method based on image detection according to claim 3, characterized in that, During a CT scan, the image acquisition device continuously and in real-time acquires real-time image data of the target object, including: During the process of the CT scanning equipment starting to execute the scanning program, the scanning bed moving and / or the gantry rotating, the image acquisition device is controlled to work continuously at a set sampling frequency; At each sampling moment, the image acquisition device synchronously acquires a set of three-dimensional coordinate points covering the overall contour of the target object in its current pose. The set of three-dimensional coordinate points acquired continuously at multiple time points in a time series is used as the real-time image data stream.

5. The CT scan-assisted method based on image detection according to claim 4, characterized in that, The real-time image data is registered and compared with the baseline pose model. Using image registration and feature matching algorithms, the pose change parameters of the target object during the real-time scanning process are calculated, including: Extract the set of three-dimensional coordinate points corresponding to the current sampling time from the real-time image data stream, and use it as the current real-time point cloud; The current real-time point cloud is subjected to the same preprocessing as when the reference pose model was established, including denoising and spatial rasterization, to generate the real-time voxel model at the current moment; An iterative nearest neighbor registration algorithm is used to spatially align the real-time voxel model with the dense 3D voxel model of the reference pose model, and the spatial transformation matrix from the real-time coordinate system to the reference coordinate system is calculated. The geometric center coordinates of the key anatomical structure region identified in the real-time voxel model are aligned with the geometric center coordinates of the corresponding key anatomical structure region in the baseline pose model based on the spatial transformation matrix. After alignment, calculate the relative displacement vector of the geometric center point of each corresponding key anatomical region in three-dimensional space; Calculate the magnitude of the relative displacement vector as the displacement amplitude parameter of the key anatomical structure region; Calculate the direction and velocity of the movement of the geometric center point of the key anatomical structure region over multiple consecutive sampling times; The displacement amplitude parameters, movement direction and velocity information of all key anatomical structural regions are summarized to form the attitude change parameters.

6. The CT scan-assisted method based on image detection according to claim 5, characterized in that, Based on the structural parameters and scanning protocol of the CT scanning equipment, multiple layers of virtual security boundaries are established inside and around the CT scanning aperture, including: Obtain the aperture geometry parameters of the CT scanning equipment gantry, including the inner radius of the aperture, the axial length of the aperture, and the position of the center axis of the aperture; Obtain the planned movement trajectory and speed of the scanning bed, as well as the gantry rotation range, as defined in the current scanning protocol; Inside the aperture, with the inner surface of the aperture as a reference, a preset first safety distance is offset inward to generate a coaxial virtual cylindrical surface, which serves as the inner virtual safety boundary. The area inside the inner virtual safety boundary is defined as the danger zone. Inside the aperture, with the inner surface of the aperture as a reference, a preset second safety distance is offset inward. The second safety distance is greater than the first safety distance, generating another coaxial virtual cylindrical surface as a middle virtual safety boundary. The annular space between the middle virtual safety boundary and the inner virtual safety boundary is defined as a warning area. Outside the aperture, along the direction of scan bed movement, in the adjacent area outside the planned scanning range of the target object, a virtual warning boundary is defined. The area outside the warning boundary is defined as the safe zone, and the area inside is defined as the attention zone. The inner virtual security boundary, the middle virtual security boundary, and the warning boundary together constitute the multi-layered virtual security boundary.

7. The CT scan-assisted method based on image detection according to claim 6, characterized in that, The attitude change parameters are dynamically compared with the multi-layer virtual safety boundary to determine the real-time attitude safety status of the target object relative to the multi-layer virtual safety boundary, and a corresponding safety status signal is generated, including: From the posture change parameters, extract the real-time three-dimensional coordinates of the geometric center point of the key anatomical structure region on the target object that has moved; Calculate the shortest spatial distance from the real-time 3D coordinates to the inner virtual security boundary; Calculate the shortest spatial distance from the real-time 3D coordinates to the mid-level virtual security boundary; Determine whether the real-time three-dimensional coordinates are located within the attention area defined by the warning boundary; Set the security status judgment logic: If the shortest spatial distance from the real-time three-dimensional coordinates to the inner virtual security boundary is less than zero, it is determined that the target object has invaded the dangerous area, and the security status signal is "dangerous status"; If the shortest spatial distance from the real-time 3D coordinates to the inner virtual security boundary is greater than zero, but the shortest spatial distance to the middle virtual security boundary is less than zero, then the target object is determined to have entered the warning area, and the security status signal is "warning status". If the shortest spatial distance from the real-time 3D coordinates to the middle-layer virtual security boundary is greater than zero, but is located within the attention area, then the target object is determined to be in the attention state, and the security state signal is "attention state"; if neither of the above applies, then the security state signal is "safe state". Combining the movement direction and speed information in the posture change parameters, if any key anatomical structure region is detected to be moving toward the inner virtual safety boundary or the middle virtual safety boundary at a speed exceeding a preset threshold, the corresponding safety status signal is upgraded by one level or marked as "trend warning", and the final safety status signal is output.

8. The CT scan-assisted method based on image detection according to claim 7, characterized in that, Based on the security status signal, trigger warning operations corresponding to different security status levels, including: The early warning operation is completed collaboratively by multiple implementing agencies; When the safety status signal is "attention status", no external visible or audible warning signals are triggered, and only a visual prompt is given on the system control interface with a specific color or icon. When the safety status signal is "warning status", a primary warning operation is triggered. The primary warning operation includes at least activating the yellow flashing warning light installed in the scanning room and playing a gentle voice prompt to the CT operator through the voice broadcast system. When the safety status signal is "dangerous" or marked as "trend warning", advanced warning and emergency operation are triggered. The advanced warning and emergency operation includes at least activating the red rotating warning light and high-frequency alarm installed in the scanning room, playing a clear stop movement command to the target object through the voice broadcast system, and sending a digital control signal containing an emergency stop command to the control system of the CT scanning equipment.

9. A CT scan-assisted method based on image detection according to claim 8, characterized in that, In the advanced early warning and emergency operation, a digital control signal containing an emergency stop command is sent to the control system of the CT scanning equipment, including: The digital control signal is sent to the central scanning control system of the CT scanning equipment through a preset data interface protocol; The digital control signal contains specific instruction codes that instruct the CT scanning device to perform a series of synchronized emergency stop actions; The emergency stop actions include: immediately stopping the exposure of the X-ray tube and stopping the output of the high voltage generator; immediately sending a braking command to the scanning bed drive motor to bring the scanning bed to a smooth stop within the shortest safe distance; and sending a stop command to the gantry drive system to stop the rotation of the gantry. The emergency stop action has a higher priority than the regular scan control commands of the CT scanner.

10. A CT scan assist system based on image detection, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the image detection-based CT scan assistance method according to any one of claims 1 to 9.

Images