A method of magnetic resonance patient assisted transfer based on visual localization

Abstract

The application provides a magnetic resonance patient auxiliary transfer method based on visual positioning, and relates to the technical field of medical auxiliary transfer. The application first acquires continuous transfer images in the process that a patient is translated from a transfer bed to a magnetic resonance examination bed, determines a shoulder position of the patient and a pelvic position of the patient, determines a shoulder movement amount and a pelvic movement amount according to position changes in adjacent images, and calculates a movement difference value between the two movement amounts; when the movement difference value exceeds a preset synchronization range, a faster moving part is determined according to the size relationship between the two movement amounts, and a corresponding movement slowing down prompt is triggered. The application can identify an asynchronous state of the shoulder movement and the pelvic movement of the patient, and realizes directional prompting on the faster moving part.

Description

Technical Field

[0001] This invention relates to the field of medical assisted transport technology, and in particular to a method for assisted transport of magnetic resonance imaging (MRI) patients based on visual positioning. Background Technology

[0002] During MRI scans, patients with limited mobility typically require a transport bed for pre- and post-examination movement, transferring them to an examination bed that can access the magnet cavity. Chinese patent application CN117838447A discloses a patient transport cart and a patient transport method. The patient transport cart is equipped with a bed support plate and a movable bed plate, which can move along the length of the bed support plate. After the patient transport cart arrives at the MRI system, the movable bed plate carries the patient into the MRI system, thereby reducing the number of times the patient needs to be moved between the transport cart and the examination equipment.

[0003] Image acquisition technology has begun to be used for position detection during patient transfer. Chinese Patent CN117883245B discloses a medical electric transfer method, controller, and storage medium. It acquires human body image information of the patient through an image acquisition device, determines the spatial coordinates of the patient's body parts, adjusts the transfer cart and extension plate according to the relative position of the patient's body and the extension plate, and stops the drive belt after the patient moves to the center position of the extension plate. Chinese Patent CN108652625B discloses an image recognition method and system for ensuring the safety of magnetic resonance imaging (MRI) scans. It determines the positional relationship between the patient's edge and the edge of the scanning bed through image recognition, and controls the movement or stopping of the scanning bed based on whether the patient is within the safe scanning area. Existing visual detection is mainly used for matching the position of the patient with the supporting components, determining the centering of the patient after transfer, or determining the safe area of ​​the MRI scanning bed.

[0004] However, current visually assisted transfer methods do not continuously determine the distance the patient's shoulder and pelvis move along the translational direction during the patient's transfer from the transport bed to the MRI examination bed. When the distances the shoulder and pelvis move are inconsistent, current methods struggle to promptly identify body parts that need to slow down or continue moving, and cannot provide corresponding assisted transfer cues for each. Summary of the Invention

[0005] The purpose of this invention is to provide a visual positioning-based assisted transport method for MRI patients, which can solve the problem of difficulty in timely identification of asynchronous shoulder and pelvic movements during the translation of the patient from the transport bed to the MRI examination bed, and realize directional slowing prompts for faster-moving parts.

[0006] To achieve the above objectives, the present invention provides the following solution: A visual positioning-based method for assisted transport of MRI patients includes: Acquire continuous transport images of the patient during the translation process from the transport bed to the magnetic resonance examination bed, identify valid transport images from the continuous transport images, and determine the patient's shoulder position and pelvic position in each valid transport image; Based on the changes in the patient's shoulder position and pelvic position in two adjacent effective transport images, the amount of shoulder movement along the translation direction and the amount of pelvic movement along the translation direction are determined respectively. When both shoulder and pelvic movements are greater than zero, the absolute value of the difference between the shoulder and pelvic movements is defined as the movement difference. When the movement difference exceeds the preset synchronization range, the faster-moving part is determined based on the relationship between the shoulder movement and the pelvic movement, and a movement slowdown prompt is triggered for the faster-moving part.

[0007] Preferably, determining the patient's shoulder position in each valid transport image includes: Identify the left and right shoulder positioning points in each valid transport image; The midpoint between the left and right shoulder positioning points is determined as the patient's shoulder position.

[0008] Preferably, determining the patient's pelvic position in each valid transport image includes: Identify the left and right hip positioning points in each valid transport image; The midpoint between the left hip positioning point and the right hip positioning point is determined as the patient's pelvic position.

[0009] Preferably, determining the valid transport image from the continuous transport images includes: Identify the left shoulder, right shoulder, left hip, and right hip positioning points in each continuous transport image, as well as the recognition confidence level for each positioning point; Four consecutive transport images with recognition confidence levels all greater than or equal to a preset confidence level are identified as valid transport images; The valid transported images are arranged according to their timestamps.

[0010] Preferably, based on the changes in the patient's shoulder position and pelvic position in two temporally adjacent effective transport images, the amount of shoulder movement along the translational direction and the amount of pelvic movement along the translational direction are determined, including: The direction perpendicular to the edge of the transport bed near the MRI examination bed, and pointing from the transport bed to the MRI examination bed, is defined as the translation direction; The directional projection value of the displacement vector between the patient's shoulder position in two temporally adjacent effective transport images in the translation direction is determined as the shoulder movement amount; The directional projection of the displacement vector between the patient's pelvic position in two temporally adjacent effective transport images in the translation direction is determined as the pelvic displacement.

[0011] Preferably, the preset synchronization range is a numerical range that is greater than or equal to zero and less than or equal to the synchronization threshold; The synchronization threshold is the product of the actual acquisition time interval between two temporally adjacent valid transport images and the preset allowable difference in movement speed; The actual acquisition time interval is determined based on the timestamps of two adjacent valid transport images.

[0012] Preferably, the faster-moving area is determined based on the relationship between the shoulder movement and the pelvic movement, including: When the shoulder movement is greater than the pelvic movement, the patient's shoulder is identified as the faster-moving part. When the pelvic movement is greater than the shoulder movement, the patient's pelvis is identified as the faster-moving part.

[0013] Preferably, triggering a slowdown prompt for fast-moving parts includes: Based on two consecutive sets of time-adjacent effective transport images, the motion difference and the faster-moving parts corresponding to each set of effective transport images are determined. When the motion difference corresponding to each set of valid transport images that has reached the preset trigger number exceeds the preset synchronization range, and the corresponding fast-moving parts are the same, a motion slowing prompt is triggered for the fast-moving parts.

[0014] Preferably, the slow-motion prompt is a graphic prompt displayed on the display device; The display device shows the current valid transport image, with graphic prompts superimposed on the positions corresponding to the faster-moving parts in the current valid transport image.

[0015] Preferably, after triggering the slow-down prompt for a fast-moving part, the method further includes: Based on two consecutive effective transport images taken at multiple time intervals, the shoulder movement, pelvic movement, and movement difference corresponding to each group of effective transport images were determined. When the motion difference corresponding to each group of valid transfer images that has reached the preset release number is within the preset synchronization range, the motion slowdown prompt is released.

[0016] The present invention discloses the following beneficial effects: (1) This invention acquires continuous transport images of the patient during the translation process from the transport bed to the MRI examination bed, and determines the position of the patient's shoulder and pelvis in each continuous transport image, so that the positional changes of the patient's upper body and pelvis during the translation process can be continuously recorded. Compared with methods that only determine the patient's position before transport or determine whether the patient is centered after transport, this invention can acquire the motion state of two body parts before the patient's translation is completed, providing positional data for determining whether the patient's shoulder and pelvis move synchronously.

[0017] (2) Based on the changes in the patient's shoulder and pelvic positions in two consecutive adjacent transport images, this invention determines the amount of shoulder movement along the translational direction and the amount of pelvic movement along the translational direction, respectively, and defines the absolute value of the difference between the two movement amounts as the movement difference value. Through the above processing, the problem of nurses relying on visual judgment to determine whether the patient's body is moving synchronously is transformed into a quantitative comparison between the two movement amounts. When the shoulder movement is 35 mm and the pelvic movement is 20 mm, the movement difference value is 15 mm, which can directly reflect the inconsistency in the movement of the patient's shoulder and pelvis during the corresponding image acquisition period, avoiding the difficulty in distinguishing local movement states by making judgments based solely on the patient's overall position.

[0018] (3) When the movement difference exceeds the preset synchronization range, this invention determines the faster-moving part based on the relationship between the shoulder movement and the pelvic movement, and triggers a slowdown prompt for the faster-moving part. When the shoulder movement is large, the slowdown prompt points to the patient's shoulder; when the pelvic movement is large, the slowdown prompt points to the patient's pelvis. Therefore, the assisted transport prompt can correspond to specific body parts with asynchronous movement, enabling caregivers to slow down the faster-moving parts in a timely manner during translation, reducing the difference between the subsequent shoulder and pelvic movement, and reducing the need for overall position adjustment after the patient's shoulder and pelvis have continued to move asynchronously. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of the method provided in an embodiment of the present invention; Figure 2 This is a graph showing the changes in the patient's shoulder position and pelvic position along the translation direction, provided in an embodiment of the present invention. Figure 3This is a schematic diagram illustrating the location of body parts in an effective transport image provided in an embodiment of the present invention; Figure 4 A schematic diagram of the structure of a vision-based magnetic resonance imaging patient-assisted transport system provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the patient assisted transport setup provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, this embodiment provides a visual positioning-based assisted transport method for magnetic resonance imaging (MRI) patients, including: Step 100: Acquire continuous transport images of the patient during the translation process from the transport bed to the MRI examination bed, identify the valid transport images from the continuous transport images, and determine the patient's shoulder position and pelvic position in each valid transport image; Step 200: Based on the changes in the patient's shoulder position and pelvic position in two adjacent effective transport images, determine the amount of shoulder movement along the translation direction and the amount of pelvic movement along the translation direction, respectively. Step 300: When both shoulder movement and pelvic movement are greater than zero, the absolute value of the difference between shoulder movement and pelvic movement is determined as the movement difference. Step 400: When the movement difference exceeds the preset synchronization range, determine the faster-moving part based on the relationship between the shoulder movement and the pelvic movement, and trigger a slowdown prompt for the faster-moving part.

[0024] Before the patient transfer begins, move the transport bed to one side of the MRI examination bed, lock the transport bed's walking mechanism, and adjust the relative height of the two bed surfaces. Place the transfer pad under the patient's body. Nursing staff apply pressure near the patient's upper body and pelvis to move the patient along the path between the two bed surfaces.

[0025] The image acquisition device is positioned above or diagonally above the patient's translational area, covering the patient's initial position on the transport bed, the receiving position on the MRI examination bed, and the movement area between the two positions. The image acquisition device maintains its installation position and imaging direction throughout a single patient transport. When the image acquisition device is located within the controlled area of ​​the MRI examination, an image acquisition device that meets the equipment requirements of the MRI examination site is used.

[0026] This embodiment employs an image acquisition device capable of simultaneously acquiring color and depth images. The color image is used to determine the pixel positions of key human body points, while the depth image is used to determine the distance from these key points to the image acquisition device. The simultaneously acquired color and depth images use the same timestamp.

[0027] After the image acquisition device is installed, its internal parameters are calibrated, and the coordinate transformation relationship from the image acquisition device coordinate system to the bed surface coordinate system is determined. Let the pixel coordinates of the human body key points in the color image be... The corresponding depth value is The coordinates of the key points of the human body in the coordinate system of the image acquisition device are: : In the formula, and These represent the focal length parameters of the image acquisition device along the horizontal and vertical directions, respectively. and This represents the coordinates of the principal point in the image.

[0028] The depth value is obtained from a local region surrounding the pixel location of the human body keypoint. Let the... Aspect Ratio Depth Image , No. The local area corresponding to the key points of an individual's body is The depth value of the key points of the human body for: This embodiment will Set as a 5-pixel by 5-pixel region centered on the human body keypoint pixel location. If the effective depth values ​​in a local region are less than nine, the corresponding human body keypoint is not used in the location calculation. Using the median can suppress the effects of sheet wrinkles, local reflections, and isolated depth outliers.

[0029] The coordinates of key human body points in the bed coordinate system are: : In the formula, The rotation matrix represents the rotation from the image acquisition device coordinate system to the bed surface coordinate system. This represents the translation vector from the image acquisition device coordinate system to the bed surface coordinate system. Rotation matrix. Translation vector Determined based on a calibration point set at a known location on the bed surface, and remains unchanged during a single patient assisted transport.

[0030] Subsequent calculations used the coordinates of key human body points in the bed coordinate system. The patient's shoulder position, pelvic position, and the amount of movement of the two body parts were represented using the same coordinate scale, with distances in millimeters.

[0031] Human body keypoints are determined by a pre-trained keypoint detection network. The input to the keypoint detection network is a color image, and the output is seventeen human body keypoint response maps. The seventeen human body keypoints include the nose, left eye, right eye, left ear, right ear, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, left hip, right hip, left knee, right knee, left ankle, and right ankle.

[0032] The keypoint detection network consists of an image feature extraction layer, a feature upsampling layer, and a response map generation layer. The image feature extraction layer uses a residual convolutional structure to extract multi-scale image features; the feature upsampling layer progressively amplifies the image features; and the response map generation layer uses a one-pixel multi-pixel convolutional kernel to generate seventeen human keypoint response maps. The input image size can be set to 256 pixels by 192 pixels, and the response map size can be set to 64 pixels by 48 pixels.

[0033] During the training of the keypoint detection network, seventeen human keypoints are labeled on each training image, and a target response map is generated based on the labeled locations. Let the first... The predicted response map of individual human key points is as follows The target response diagram is as follows The width and height of the response map are respectively and Training loss for: The same loss calculation method was used for all seventeen human body keypoints, and no weighting coefficients were set between different human body keypoints. After the keypoint detection network was trained, the network parameters were fixed.

[0034] Let the first The first continuous transport image The response diagram corresponding to each individual's key points is as follows . No. The location of the individual human key points in the response map is as follows: No. Confidence level of identification of key points of an individual for: The locations of human key points in the response image are converted to a color image according to the ratio of the response image size to the color image size to obtain the pixel coordinates of the human key points. Combined with the corresponding depth values, internal parameters of the image acquisition device, and coordinate transformation relationships, the positions of the human key points in the bed surface coordinate system are obtained.

[0035] This embodiment uses the left shoulder positioning point, right shoulder positioning point, left hip positioning point, and right hip positioning point. Figure 3 This is a schematic diagram illustrating the positioning of body parts in an effective transport image provided by an embodiment of the present invention. The left side of the figure shows the transport bed, and the right side shows the MRI examination bed. Horizontal arrows indicate the translational direction from the transport bed to the MRI examination bed. The patient is in a supine position within the image acquisition range, and the patient's face is represented by an obscured area. The patient's left shoulder, right shoulder, left hip, and right hip positioning points are each marked with a solid circle with an outer ring, and each positioning point is connected to its corresponding name by a dashed line. The patient's shoulder position and pelvic position are each marked with a quarter circle. The patient's shoulder position is located between the left and right shoulder positioning points, and the patient's pelvic position is located between the left and right hip positioning points.

[0036] Let the four positioning points be at the th The recognition confidence scores in the continuous transport images are respectively , , and Minimum recognition confidence for: when Greater than or equal to the preset reliability And when all four positioning points have valid depth values, the first... A series of continuous transport images are determined to be valid transport images. If the minimum recognition confidence is less than the preset confidence, or if any positioning point lacks a valid depth value, the corresponding continuous transport image will not be included in the calculation of shoulder and pelvic movement.

[0037] During patient transfer, the caregiver's arm may briefly obscure the patient's shoulder. In one image acquisition, the right shoulder positioning point in the 60th consecutive transfer image was obscured by the caregiver's forearm, causing the recognition confidence of the right shoulder positioning point to drop to 0.41, below the preset confidence of 0.60. The 60th consecutive transfer image was not determined as a valid transfer image. In the 61st consecutive transfer image, the recognition confidence of all four positioning points recovered to above 0.80, and the 61st consecutive transfer image continued to be used in subsequent calculations.

[0038] Table 1 Examples of effective transport image selection

[0039] Valid transport images are arranged according to timestamps. Two valid transport images with adjacent timestamps are used to determine shoulder and pelvic movement. Images with a timestamp of 0.40 seconds in Table 1 are not included in the calculation; therefore, two valid transport images with timestamps of 0.20 seconds and 0.44 seconds are adjacent, and the actual acquisition time interval is 0.24 seconds.

[0040] Let the first The left shoulder positioning point, right shoulder positioning point, left hip positioning point, and right hip positioning point in the effective transport image are respectively , , and The patient's shoulder position for: Patient's pelvic position for: In a valid transport image, the bed coordinates of the left shoulder positioning point are (412.0, 318.0, 236.0) mm, the bed coordinates of the right shoulder positioning point are (608.0, 320.0, 238.0) mm, and the patient's shoulder position is (510.0, 319.0, 237.0) mm. The bed coordinates of the left hip positioning point are (438.0, 516.0, 184.0) mm, the bed coordinates of the right hip positioning point are (582.0, 518.0, 186.0) mm, and the patient's pelvis position is (510.0, 517.0, 185.0) mm.

[0041] Using the midpoint of the left and right shoulder positioning points to represent the patient's shoulder position can reduce positional changes caused by localized unilateral shoulder movement. Using the midpoint of the left and right hip positioning points to represent the patient's pelvic position can reduce the impact of unilateral hip positioning deviations on the overall pelvic position.

[0042] In step 200, the translation direction is determined based on the edge of the transfer bed near the MRI examination bed. Let the two positions on the edge of the bed be... and Bed surface edge direction vector for: like , Then the candidate direction vector perpendicular to the edge of the bed surface for: Let the center position of the transfer bed surface be... The center position of the MRI examination bed surface is .from and The direction pointing towards the MRI examination table is selected and normalized to obtain the unit vector of the translation direction. : When the dot product result is zero, redetermine the two positions on the edge of the bed or the two center positions of the bed.

[0043] Before a patient translation, the two positions on the edge of the bed are (180, 120) mm and (180, 1920) mm, respectively. The direction vector of the bed edge is (0, 1800), and the two candidate vertical directions are (-1800, 0) and (1800, 0), respectively. The center of the transport bed is located to the right of the center of the MRI examination bed; therefore, the candidate direction pointing towards the MRI examination bed is selected as the translation direction.

[0044] Project the patient's shoulder position and pelvic position onto the translation direction to obtain the shoulder position. and pelvic position : Let the two adjacent effective transport images be the first and second images respectively. The first effective transport image and the first 10 valid transported images, with timestamps of respectively and Actual data collection time interval for: shoulder movement for: pelvic movement for: This indicates that the patient's shoulder is moving towards the MRI examination table; This indicates that the patient's pelvis has moved toward the MRI examination table.

[0045] Before the patient begins translation, the nursing staff will pause briefly to adjust the transfer pad. If the shoulder movement is 1 mm and the pelvic movement is 0 mm in two adjacent valid transfer images, no synchronization judgment is performed. If local correction is needed during patient translation, the shoulder movement may be 4 mm and the pelvic movement may be -2 mm. If the pelvic movement is not greater than zero, the current two valid transfer images are not included in the movement difference calculation.

[0046] Step 300 is executed only when both shoulder and pelvic movement are greater than zero. Pausing movement, reversing the entire body movement, or reversing the movement of a single body part will not trigger the slowdown prompt.

[0047] In step 300, the absolute value of the difference between the shoulder movement and the pelvic movement is determined as the movement difference. : The motion difference represents the difference in the amount of movement of the patient's shoulder and pelvis along the translational direction within the actual data acquisition time interval. The motion difference only indicates the magnitude of the difference, not the location of faster movement.

[0048] Based on the actual data collection time interval Difference between the preset allowed movement speed and the actual speed. Determine the synchronization threshold : The preset synchronization range is a range of values ​​greater than or equal to zero and less than or equal to the synchronization threshold. When the movement difference is less than or equal to the synchronization threshold, the difference between the shoulder movement and the pelvic movement is within the preset synchronization range; when the movement difference is greater than the synchronization threshold, the difference between the shoulder movement and the pelvic movement exceeds the preset synchronization range.

[0049] The preset allowable movement speed difference represents the permissible difference in shoulder and pelvic movement speed per unit time, measured in millimeters per second. In this embodiment, the preset allowable movement speed difference is set to 25 millimeters per second. When the actual acquisition time interval is 0.20 seconds, the synchronization threshold is: After excluding a continuous transport image from the calculation, the actual acquisition time interval between two temporally adjacent valid transport images increases to 0.24 seconds, and the synchronization threshold is: When the actual acquisition time interval changes, the synchronization threshold still corresponds to an allowable movement speed difference of 25 mm / s. If a 0.24-second image interval still uses 5 mm as the synchronization threshold, the calculated allowable movement speed difference is approximately 20.83 mm / s, which differs from the preset value.

[0050] In step 400, when the movement difference is greater than the synchronization threshold, the shoulder movement and pelvic movement are compared. If the shoulder movement is greater than the pelvic movement, the patient's shoulder is identified as the faster-moving part; if the pelvic movement is greater than the shoulder movement, the patient's pelvis is identified as the faster-moving part.

[0051] Let the first The excess state of a group of effective transport images is : when and At that time, the faster-moving parts Record it as the patient's shoulder; when and At that time, the faster-moving parts Record it as the patient's pelvis.

[0052] Let the consecutive over-limit count be... The consecutive over-limit count is: When the consecutive over-limit count reaches the preset trigger number At that time, a slowdown prompt is triggered for the currently fast-moving part. This embodiment will... Set it to 3.

[0053] In the first set of valid transport images during a translation process, the shoulder movement is 17 mm and the pelvic movement is 10 mm. The patient's shoulder is the fastest moving part, and the consecutive over-limit count is 1. In the second set of valid transport images, the shoulder movement is 9 mm and the pelvic movement is 16 mm. The patient's pelvis becomes the fastest moving part, and the consecutive over-limit count is reset to 1. When the third set of valid transport images again confirms that the patient's shoulder is moving fast, the consecutive over-limit count is again reset to 1. When the patient's shoulder and pelvis move alternately at high speeds, no slowing-down prompts for individual body parts are triggered.

[0054] Slowing-down movement indicators are displayed on a monitor visible to caregivers. The monitor shows the currently active transport image. When the patient's shoulder is identified as a fast-moving area, a slowing-down movement graphic is overlaid near the patient's shoulder; when the patient's pelvis is identified as a fast-moving area, a slowing-down movement graphic is overlaid near the patient's pelvis. The slowing-down movement graphics update according to the position of the corresponding body part.

[0055] After the slowing-down movement cue is triggered, continue acquiring valid transport images and calculating shoulder movement, pelvic movement, and movement difference. Let the recovery count be... : The recovery count remains unchanged when the patient is in a paused or reverse adjustment state. After the patient resumes forward translation, the recovery count continues to be updated based on the movement difference.

[0056] When the recovery count reaches the preset release quantity At that time, the slow-down prompt will be deactivated. This embodiment will... Set to 3. Continuous recovery judgment can reduce the repeated appearance and disappearance of prompts when the movement difference is near the synchronization threshold.

[0057] Figure 2 The changes in the patient's shoulder and pelvic positions during the patient's translation are shown. Figure 2 This graph shows the changes in the patient's shoulder and pelvic positions along the translational direction. The horizontal axis represents time in seconds, marked as 0.00, 0.20, 0.44, 0.64, 0.84, 1.08, and 1.28 seconds. The vertical axis represents position along the translational direction in millimeters, ranging from 200 to 330 millimeters. Solid lines and circular data points represent the patient's shoulder position, while dashed lines and circular data points represent the patient's pelvic position. The corresponding shoulder positions at each time point are 210 mm, 229 mm, 250 mm, 268 mm, 284 mm, 303 mm, and 319 mm, respectively, while the corresponding pelvic positions are 208 mm, 220 mm, 234 mm, 246 mm, 260 mm, 277 mm, and 292 mm, respectively. The vertical dashed line at 0.64 seconds indicates the trigger time of the slow movement prompt, and the vertical dashed line at 1.28 seconds indicates the deactivation time of the slow movement prompt.

[0058] Table 2 Examples of shoulder and pelvic movement synchronization judgment

[0059] Taking sequence number 2 as an example, the actual acquisition time interval is 0.44 seconds minus 0.20 seconds, resulting in 0.24 seconds; the shoulder movement is 250 mm minus 229 mm, resulting in 21 mm; the pelvic movement is 234 mm minus 220 mm, resulting in 14 mm; the movement difference is the absolute value of the difference between 21 mm and 14 mm, resulting in 7 mm; the synchronization threshold is 25 mm per second multiplied by 0.24 seconds, resulting in 6 mm. Since the movement difference is greater than the synchronization threshold, and the shoulder movement is greater than the pelvic movement, the patient's shoulder is identified as the faster-moving area.

[0060] The movement differences corresponding to numbers 1 to 3 are 7 mm, 7 mm, and 6 mm, respectively, which are greater than the synchronization thresholds of 5 mm, 6 mm, and 5 mm. In all three judgments, the patient's shoulder was identified as the fastest-moving area. After judgment number 3 is completed, a prompt to slow down the movement of the patient's shoulder is triggered.

[0061] The movement differences corresponding to numbers 4 to 6 are 2 mm, 2 mm, and 1 mm, respectively, which are less than or equal to the synchronization thresholds of 5 mm, 6 mm, and 5 mm. If all three consecutive judgments are within the preset synchronization range, the slowing movement prompt for the patient's shoulder is deactivated after judgment number 6 is completed.

[0062] The data in Table 2 satisfy the following calculation relationships: When the image recognition confidence is insufficient, the corresponding continuous transport images are not included in the movement calculation; when the actual acquisition time interval between effective transport images changes, the synchronization threshold changes with the actual acquisition time interval; the pause state and the reverse adjustment state are not included in the movement difference judgment; after the same body part moves continuously and quickly to reach the preset trigger number, the display device generates a corresponding slowing movement prompt; after the movement difference continuously recovers to the preset synchronization range, the slowing movement prompt is canceled.

[0063] The values ​​in Tables 1 and 2 are used to illustrate the calculation process of this embodiment. The preset reliability, preset allowable movement speed difference, preset trigger number, and preset release number can be set according to the image acquisition conditions and patient assisted transport requirements.

[0064] This embodiment also provides a visual positioning-based magnetic resonance imaging (MRI) patient-assisted transport system. The system includes an image acquisition device, a processing device, and a display device. The image acquisition device is connected to the processing device, and the processing device is connected to the display device. Figure 4 This is a schematic diagram of a vision-based magnetic resonance imaging (MRI) patient-assisted transport system provided in an embodiment of the present invention. The image acquisition device, processing device, and display device are arranged sequentially according to the data transmission direction. The arrow between the image acquisition device and the processing device indicates the transmission direction of continuous transport images, and the arrow between the processing device and the display device indicates the transmission direction of prompting information. The processing device internally shows two processing functions: position calculation and prompting control. Position calculation is used to determine the patient's shoulder position, pelvic position, and corresponding movement amount. Prompt control is used to generate prompts to slow down movement or cancel prompts based on the movement synchronization judgment result. Figure 5 This is a schematic diagram of a patient-assisted transport arrangement provided in an embodiment of the present invention. The transport bed is located to one side of the MRI examination bed, which is positioned facing the magnet cavity. The patient lies supine in the transfer area between the two bed surfaces. The arrows in the diagram indicate the direction of translation from the transport bed to the MRI examination bed. The image acquisition device is located above the patient's translation area. The dashed lines extending downwards on both sides of the image acquisition device define the acquisition range, which covers the patient's body and the transfer area between the transport bed and the MRI examination bed.

[0065] Specifically, the image acquisition device is positioned above or diagonally above the patient's translational area, with its acquisition range covering the transfer bed, the MRI examination bed, and the patient's movement area between the two bed surfaces. The image acquisition device continuously acquires transfer images during the patient's translation from the transfer bed to the MRI examination bed and records a timestamp for each transfer image. When the image acquisition device employs an image acquisition device capable of simultaneously acquiring color and depth images, the color image is used to determine the pixel positions of the patient's body parts, and the depth image is used to determine the distance from the corresponding body part to the image acquisition device.

[0066] The processing device includes a processor and a memory. The memory stores a computer program, internal parameters of the image acquisition device, coordinate transformation parameters from the image acquisition device coordinate system to the bed surface coordinate system, preset confidence level, preset allowable movement speed difference, preset trigger quantity, and preset release quantity. When the processor executes the computer program, it receives continuous transport images sent by the image acquisition device and performs human key point recognition on the continuous transport images.

[0067] The processor identifies left shoulder, right shoulder, left hip, and right hip positioning points from each of the continuous transport images and determines the recognition confidence level for each of the four positioning points. When all four recognition confidence levels are greater than or equal to the preset confidence level, and all four positioning points have valid depth values, the processor determines the corresponding continuous transport image as a valid transport image.

[0068] The processor arranges the valid transport images according to timestamps, determines the midpoint between the left and right shoulder positioning points as the patient's shoulder position, and determines the midpoint between the left and right hip positioning points as the patient's pelvic position. The patient's shoulder and pelvic positions are represented in a bed coordinate system after coordinate transformation.

[0069] The processor determines the translation direction based on the edge of the transport bed near the MRI examination bed, and determines the shoulder movement along the translation direction based on the change in the patient's shoulder position in two temporally adjacent valid transport images; it also determines the pelvic movement along the translation direction based on the change in the patient's pelvic position in two temporally adjacent valid transport images. Both the shoulder movement and the pelvic movement are expressed as directional projections of the corresponding positional changes along the translation direction.

[0070] When both the shoulder movement and the pelvic movement are greater than zero, the processor determines the absolute value of the difference between the two movements as the movement difference. The processor determines a synchronization threshold based on the actual acquisition time interval between two temporally adjacent valid transport images and the preset allowable movement speed difference, and determines the range of values ​​greater than or equal to zero and less than or equal to the synchronization threshold as the preset synchronization range.

[0071] When the movement difference is greater than the synchronization threshold, the processor compares the shoulder movement and the pelvic movement. When the shoulder movement is greater than the pelvic movement, the processor identifies the patient's shoulder as the faster-moving part; when the pelvic movement is greater than the shoulder movement, the processor identifies the patient's pelvis as the faster-moving part.

[0072] The processor continuously judges multiple sets of two valid transport images that are adjacent in time. When the movement difference corresponding to each set of valid transport images that has reached the preset trigger number is greater than the corresponding synchronization threshold, and the fast-moving part is the same in each determination, the processor generates a slow-down prompt message for the fast-moving part and sends the slow-down prompt message to the display device.

[0073] The display device shows the currently active transport image and overlays a slowing-down motion graphic near the patient's shoulder or pelvis in the image. Nursing staff use the slowing-down motion graphic to determine the body parts requiring slowing down. The system does not directly control the movement of the transport bed or the MRI examination bed; patient translation is still performed by the nursing staff.

[0074] After the slowing-down movement prompt is triggered, the processor continues to receive subsequent valid transport images and continues to determine the shoulder movement amount, the pelvic movement amount, and the movement difference. When the movement difference corresponding to each group of valid transport images that has reached the preset release number is within the preset synchronization range, the processor sends a prompt release message to the display device, and the display device stops displaying the slowing-down movement graphic.

[0075] In this embodiment, the image acquisition device acquires positional data during the patient's translation process, the processing device performs effective image filtering, body part position calculation, movement calculation, synchronization status judgment, and prompt status control, and the display device presents the faster-moving parts to the nursing staff. The system can identify differences in movement between the patient's shoulder and pelvis before the patient's translation is complete, and indicate to the nursing staff the specific body parts that need to be slowed down.

[0076] In summary, this embodiment utilizes continuous transport images to determine the patient's shoulder and pelvic positions, and filters out unreliable images based on recognition confidence and depth data. This reduces the impact of nursing staff obstruction, covering obstruction, and missing depth data on movement calculation. Using a unified bed coordinate system and directional projection to calculate shoulder and pelvic movement separately distinguishes between patient translation towards the MRI examination table, paused movement, and reverse adjustment. A synchronization threshold is determined based on the actual acquisition time interval of effective transport images, maintaining a consistent movement speed difference judgment standard even when images are excluded or the acquisition interval changes. A directional slowdown prompt is triggered when the movement difference continuously exceeds the preset synchronization range and the same body part continues to move rapidly. The prompt is deactivated after the movement difference continuously returns to the preset synchronization range. This reduces erroneous prompts caused by single positioning fluctuations or alternating shoulder and pelvic movements, allowing nursing staff to promptly identify body parts requiring slowdown during patient translation, reducing the movement difference between the patient's shoulder and pelvis, and minimizing the need for readjustment of body position after translation.

[0077] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0078] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A visual-guided patient transport method based on magnetic resonance imaging, characterized in that, include: Acquire continuous transport images of the patient during the translation process from the transport bed to the magnetic resonance examination bed, determine the valid transport images from the continuous transport images, and determine the patient's shoulder position and patient's pelvic position in each of the valid transport images; Based on the changes in the patient's shoulder position and pelvic position in two adjacent effective transport images, the amount of shoulder movement along the translation direction and the amount of pelvic movement along the translation direction are determined respectively. When both the shoulder movement and the pelvic movement are greater than zero, the absolute value of the difference between the shoulder movement and the pelvic movement is determined as the movement difference. When the movement difference exceeds the preset synchronization range, the faster-moving part is determined based on the relationship between the shoulder movement and the pelvic movement, and a movement slowdown prompt is triggered for the faster-moving part.

2. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, Determining the patient's shoulder position in each of the effective transport images includes: Identify the left and right shoulder positioning points in each of the valid transport images; The midpoint between the left shoulder positioning point and the right shoulder positioning point is determined as the patient's shoulder position.

3. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, Determining the patient's pelvic position in each of the effective transport images includes: Identify the left hip positioning point and the right hip positioning point in each of the effective transport images; The midpoint between the left hip positioning point and the right hip positioning point is determined as the patient's pelvic position.

4. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, Determining valid transport images from the continuous transport images includes: Identify the left shoulder positioning point, right shoulder positioning point, left hip positioning point, right hip positioning point, and the identification confidence level corresponding to each positioning point in each of the continuous transport images; The four consecutive transport images in which the recognition confidence level is greater than or equal to the preset confidence level are determined as the valid transport images; The valid transport images are arranged according to their timestamps.

5. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, Based on the changes in the patient's shoulder position and pelvic position in two adjacent effective transport images, the amount of shoulder movement along the translational direction and the amount of pelvic movement along the translational direction are determined, including: The direction perpendicular to the edge of the transport bed near the magnetic resonance imaging (MRI) examination bed, and pointing from the transport bed to the MRI examination bed, is defined as the translation direction. The directional projection value of the displacement vector between the patient's shoulder position in two adjacent effective transport images in the translation direction is determined as the shoulder movement amount; The displacement vector between the patient's pelvic position in two temporally adjacent effective transport images is determined as the pelvic movement amount by the directional projection of the displacement vector in the translation direction.

6. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, The preset synchronization range is a numerical range that is greater than or equal to zero and less than or equal to the synchronization threshold. The synchronization threshold is the product of the actual acquisition time interval between two temporally adjacent valid transport images and the preset allowable movement speed difference; The actual acquisition time interval is determined based on the timestamps of two adjacent valid transport images.

7. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, The faster-moving areas are determined based on the relationship between the shoulder movement and the pelvic movement, including: When the shoulder movement is greater than the pelvic movement, the patient's shoulder is identified as the faster-moving part. When the pelvic movement is greater than the shoulder movement, the patient's pelvis is identified as the faster-moving part.

8. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, Triggering a slowdown prompt for the fast-moving part, including: Based on two consecutive sets of time-adjacent effective transport images, the movement difference and the faster-moving part are determined for each set of effective transport images. When the movement difference corresponding to each group of valid transport images that has reached a preset trigger number exceeds the preset synchronization range, and the corresponding fast-moving parts are the same, the movement slowing prompt for the fast-moving parts is triggered.

9. The visual positioning-based magnetic resonance imaging patient-assisted transport method according to claim 1, characterized in that, The slow-down prompt is a graphical prompt displayed on the display device; The display device displays the current effective transfer image, and the graphic prompt is superimposed on the position corresponding to the faster-moving part in the current effective transfer image.

10. The visual-guided patient transport method based on visual positioning according to claim 1, characterized in that, After triggering the slowdown prompt for the faster-moving part, the following is also included: Based on multiple subsequent sets of two adjacent effective transport images, the shoulder movement, pelvic movement, and movement difference corresponding to each set of effective transport images are determined respectively. When the movement difference corresponding to each group of valid transfer images that has reached the preset release number is within the preset synchronization range, the slowdown movement prompt is released.

Images