A method, device and medium for intelligent monitoring of postoperative posture of a neurosurgical patient

Abstract

This invention discloses a method, device, and medium for intelligent monitoring of postoperative posture in neurosurgical patients, relating to the field of human behavioral feature recognition technology. The method includes: acquiring video images of the bedridden area to form a continuous image frame sequence; analyzing the human posture of the continuous image frame sequence and determining the spatial coordinates of effective anatomical key points to form a posture evolution sequence; dividing the posture evolution sequence into continuous time segments; calculating the spatial displacement amplitude of effective anatomical key points between adjacent continuous image frames and selecting stable posture time periods; forming anatomical key point pairs within the stable posture time periods and calculating the spatial distance between these pairs; based on the change in the spatial distance between anatomical key point pairs with acquisition time within the stable posture time periods, forming a sliding window cumulative change operator; and identifying the posture behavior change state time periods and the posture behavior maintenance state time periods by comparing the sliding window cumulative change operator frame by frame. This achieves posture monitoring.

Description

Technical Field

[0001] This invention relates to the field of human behavioral feature recognition technology, and in particular to a method, device and medium for intelligent monitoring of postoperative posture of neurosurgical patients. Background Technology

[0002] Neurosurgical patients typically require prolonged bed rest during postoperative recovery. During this period, changes in body posture are closely related to the progress of neurological function recovery, complication risk control, and nursing intervention strategies. In the field of medical information processing, the monitoring and analysis of body posture during bed rest has increasingly adopted technologies such as video perception, posture modeling, and behavioral feature recognition. By modeling changes in key anatomical regions of the human body in spatial and temporal dimensions, an objective description of the patient's bedridden behavioral state can be achieved. With the development of computer vision and intelligent analysis methods, extracting posture-related behavioral features based on continuous image information has become an important direction in postoperative nursing monitoring research, providing data support for clinical assessment and decision support.

[0003] However, existing methods still have two limitations: First, some behavioral feature recognition methods focus on judging a single posture state or short-term action, paying insufficient attention to the evolution of posture over time during bed rest, making it difficult to fully characterize the collaborative changes between different human body structural regions from a behavioral perspective; Second, in actual ward environments, video acquisition conditions are complex and varied, and the spatial representation of human anatomical positions is easily affected by factors such as occlusion and changes in lighting, which means that posture analysis based on behavioral feature recognition has room for improvement in terms of stability and continuity. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for intelligent monitoring of postoperative posture in neurosurgical patients, which solves the problems of difficulty in continuously characterizing postoperative posture and behavior features in the time dimension and difficulty in stably identifying the coordinated changes in different human body structural regions in the postoperative bed rest scenario.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for intelligent monitoring of the postoperative posture of neurosurgical patients, which includes: acquiring video images of the bed resting area to form a continuous image frame sequence; performing human posture analysis on the continuous image frame sequence and determining the spatial coordinates of effective anatomical key points; and arranging the spatial coordinates of effective anatomical key points in the order of acquisition time to form a posture evolution sequence.

[0008] The pose evolution sequence is divided into continuous time segments. The spatial displacement amplitude of effective anatomical key points between adjacent consecutive image frames is calculated and the pose stable state time segment is selected. Anatomical key point pairs are formed within the pose stable state time segment and the spatial distance between the anatomical key point pairs is calculated.

[0009] Based on the change in spatial distance between anatomical key points and acquisition time during the stable posture period, a sliding window cumulative change operator is formed. By comparing the sliding window cumulative change operator frame by frame, the time periods of posture behavior change and posture behavior maintenance are identified.

[0010] The analysis focused on the time periods of postural behavior change and the time periods of postural behavior maintenance. Based on the changes in spatial distance values ​​at anatomical key points, the analysis distinguished the relationship between changes in different human structural levels and arranged them in chronological order of collection to form a postoperative postural structure behavior evolution record.

[0011] As a preferred embodiment of the method for intelligent monitoring of postoperative posture of neurosurgical patients according to the present invention, the method of acquiring video images of the bed rest area to form a continuous image frame sequence includes, during the period when the neurosurgical patient has completed surgery and is in postoperative bed rest recovery, acquiring video of the bed rest area through an imaging device set above the bed rest area and an imaging device set to the side of the bed rest area.

[0012] The video capture field of view covers the head, neck, trunk, and limb areas involved in postoperative posture changes of neurosurgical patients;

[0013] The acquired video images are arranged in the order of their acquisition time to form a continuous image frame sequence.

[0014] As a preferred embodiment of the method for intelligent monitoring of postoperative posture of neurosurgical patients according to the present invention, the step of performing human posture analysis on a continuous image frame sequence and determining the spatial coordinates of effective anatomical key points, and arranging the spatial coordinates of effective anatomical key points in the order of acquisition time to form a posture evolution sequence includes performing human posture analysis processing on each frame of the continuous image frame sequence to extract the spatial coordinates of anatomical key points of the head center position, neck anatomical position, trunk anatomical position and pelvic anatomical position.

[0015] For each frame of the image, the spatial coordinates of the anatomical key points are extracted and a validity determination is performed to determine whether the spatial coordinates of the anatomical key points are completely present and whether the change range of the spatial coordinates of the anatomical key points in adjacent consecutive image frames is within a reasonable range of human posture changes.

[0016] When the spatial coordinates of anatomical key points fail the validity judgment process, record the invalid anatomical key point spatial coordinates.

[0017] The spatial coordinates of the anatomical key points processed through validity determination are used as the valid spatial coordinates of the anatomical key points. The valid spatial coordinates of the anatomical key points of each frame are arranged in the order of acquisition time to form a pose evolution sequence.

[0018] As a preferred embodiment of the method for intelligent monitoring of postoperative posture of neurosurgical patients according to the present invention, the calculation of the spatial displacement amplitude of effective anatomical key points between adjacent consecutive image frames and the selection of the time period of stable posture includes dividing the posture evolution sequence into multiple consecutive time segments according to the acquisition time order, and each consecutive time segment is composed of a fixed number of adjacent consecutive image frames.

[0019] Within each consecutive time segment, for each effective anatomical key point, the spatial coordinates of the effective anatomical key point are obtained in adjacent consecutive image frames. The spatial displacement amplitude of the effective anatomical key point between adjacent consecutive image frames is obtained by the straight-line distance between the two spatial coordinate positions of the effective anatomical key point in adjacent consecutive image frames.

[0020] If there are no invalid spatial coordinate markers for anatomical key points within a continuous time segment, and if there is no situation within the acquisition time range covered by the continuous time segment where the spatial displacement amplitude in the next frame is not less than that in the previous frame for all adjacent consecutive image frames, then the continuous time segment is selected as a time period of stable attitude.

[0021] As a preferred embodiment of the method for intelligent monitoring of postoperative posture of neurosurgical patients according to the present invention, wherein: the step of forming a combination of anatomical key point pairs and calculating the spatial distance of anatomical key point pairs during the time period of stable posture includes, during the time period of stable posture, combining the spatial coordinates of effective anatomical key points based on the spatial positional relationship of effective anatomical key points in the human body structure to form a combination of anatomical key point pairs.

[0022] For each pair of anatomical key points during the stable posture period, the spatial coordinates of the effective anatomical key points at both ends of the pair are obtained in the same image frame, and the spatial distance between the two ends is obtained by the straight-line distance between the spatial coordinates.

[0023] As a preferred embodiment of the method for intelligent monitoring of postoperative posture of neurosurgical patients according to the present invention, the step of forming a sliding window cumulative change operator based on the change of spatial distance between anatomical key point pairs during the posture stability period includes: traversing each frame of images in the posture stability period according to the acquisition time order during the posture stability period; using the spatial distance between each anatomical key point pair in the starting frame of the posture stability period as the reference spatial distance; and constructing the logarithmic deviation of the anatomical key point pair based on the difference between the spatial distance between each anatomical key point pair in any frame of images and the reference spatial distance for any frame of images during the posture stability period.

[0024] For each frame of image, a sliding window is selected with the current frame as the terminating frame. The logarithmic deviation of each frame within the window coverage area is accumulated in a time-weighted manner to obtain the sliding window cumulative change operator.

[0025] As a preferred embodiment of the method for intelligent monitoring of postoperative posture of neurosurgical patients according to the present invention, the step of identifying the time period of posture behavior change state and the time period of posture behavior maintenance state by comparing the cumulative change operator of the sliding window frame by frame includes performing frame-by-frame comparison of the values ​​of the cumulative change operator of the sliding window corresponding to adjacent consecutive image frames in the order of acquisition time within the acquisition time range of the posture stable state time period.

[0026] During the frame-by-frame comparison process, when the cumulative change operator of the sliding window of the next frame image is always not less than the cumulative change operator of the sliding window of the previous frame image, the attitude stable state time period is identified as the attitude behavior change state time period. When there is a situation where the cumulative change operator of the sliding window of the next frame image is less than the cumulative change operator of the sliding window of the previous frame image, the attitude stable state time period is identified as the attitude behavior maintenance state time period.

[0027] As a preferred embodiment of the method for intelligent monitoring of postoperative posture of neurosurgical patients according to the present invention, the step of distinguishing the change relationship of different human body structural levels based on the change process of spatial distance values ​​of anatomical key points and arranging them in the order of collection time to form a postoperative posture structure behavior evolution record includes selecting the time period of posture behavior change state and the time period of posture behavior maintenance state as the analysis object, and distinguishing the posture change relationship of different human body structural levels based on the change process of spatial distance values ​​of anatomical key points on the collection time axis.

[0028] The human body structural hierarchy includes at least local structural hierarchy characterized by different anatomical key point pairs, axial structural hierarchy, and cross-regional structural hierarchy;

[0029] The posture change relationships corresponding to each structural level are temporally correlated with the start and end times of the posture behavior maintenance period, arranged in the order of acquisition time, and the overlapping change intervals in time are integrated to form a postoperative posture structure behavior evolution record.

[0030] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the method for intelligent monitoring of postoperative posture of neurosurgical patients as described in the first aspect of the present invention.

[0031] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for intelligent monitoring of postoperative posture of neurosurgical patients as described in the first aspect of the present invention.

[0032] The beneficial effects of this invention are as follows: by determining the validity of the spatial coordinates of anatomical key points and constructing a posture evolution sequence, the continuity and consistency of the temporal data required for the recognition of posture behavior features in bedridden scenarios are realized; by forming a sliding window cumulative change operator based on the spatial distance changes of anatomical key points and combining the posture change relationship with the human body structure hierarchy, the joint characterization of postoperative posture adjustment in the time dimension and structural hierarchy dimension is realized, forming a postoperative posture structure behavior evolution record that can reflect the posture behavior evolution process. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 A flowchart for a method of intelligent postoperative posture monitoring for neurosurgical patients.

[0035] Figure 2 This is a flowchart for selecting time periods of stable posture and calculating spatial distances for key points in the anatomy.

[0036] Figure 3 A flowchart for calculating the cumulative change operator for the sliding window and recognizing the attitude behavior state.

[0037] Figure 4 A flowchart for distinguishing the hierarchical changes in human body structure and recording the evolution of postoperative posture, structure, and behavior.

[0038] Figure 5The time-series response curves of the sliding window cumulative change operator in the attitude behavior change segment and the hold segment. Detailed Implementation

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0042] Reference Figures 1-5 As one embodiment of the present invention, this embodiment provides a method for intelligent monitoring of postoperative posture of neurosurgical patients, comprising the following steps:

[0043] S1. Collect video images of the bedridden area to form a continuous image frame sequence. Analyze the human posture of the continuous image frame sequence and determine the spatial coordinates of effective anatomical key points. Arrange the spatial coordinates of effective anatomical key points in the order of collection time to form a posture evolution sequence.

[0044] Furthermore, during the postoperative recovery period after neurosurgical surgery, the patient's postoperative bed rest posture was continuously captured via video.

[0045] The video acquisition method includes setting up an imaging device above the bed area and setting up an imaging device to the side of the bed area to acquire continuous video images.

[0046] The video acquisition process is set up for the bedside area where the neurosurgical patient is located. The video acquisition field of view covers the head area, neck area, trunk area and limb areas involved in postoperative posture changes of the neurosurgical patient. The video acquisition process forms continuous video images to reflect the entire process of postoperative bedside posture changes of the neurosurgical patient.

[0047] Furthermore, the continuously acquired video images are arranged in chronological order of acquisition time to form a continuous image frame sequence.

[0048] For each frame in a continuous image frame sequence, human pose analysis processing is performed to extract the spatial coordinates of anatomical key points from each frame. The spatial coordinates of anatomical key points include the spatial coordinates of the head center position, the spatial coordinates of the neck anatomical position, the spatial coordinates of the trunk anatomical position, and the spatial coordinates of the pelvic anatomical position. The position of each anatomical key point in the human anatomical structure has a definite meaning.

[0049] During the human pose analysis process, the spatial coordinates of the anatomical key points extracted in each frame of the image are subjected to validity determination. The validity determination process includes determining whether the spatial coordinates of the anatomical key points are completely present, and determining whether the change range of the spatial coordinates of the anatomical key points in adjacent consecutive image frames is within a reasonable range of human pose changes.

[0050] When a certain anatomical key point in a certain frame of an image does not obtain complete spatial coordinate information, and the change in spatial coordinates of the anatomical key point in adjacent consecutive image frames exceeds the reasonable range of human posture changes, an invalid anatomical key point spatial coordinate marker is recorded. The invalid anatomical key point spatial coordinate marker includes the image frame acquisition time information, anatomical key point category information, and invalid reason information. The invalid reason information includes spatial coordinate missing and spatial coordinate jump.

[0051] When complete spatial coordinate information of a certain anatomical key point in a certain frame of an image is obtained, and the change in spatial coordinates of the anatomical key point in adjacent consecutive image frames does not exceed the reasonable range of human posture changes, the spatial coordinates of the anatomical key point in the image frame are marked as valid anatomical key point spatial coordinates.

[0052] It should be noted that the reasonable range of human posture changes refers to the range of continuous spatial displacement that anatomical key points can form between adjacent consecutive image frames. This range is limited by both the time interval between adjacent image frames during video acquisition and the spatial resolution of the imaging device. In postoperative bed rest recovery scenarios, the time interval between adjacent image frames is typically on the order of tens of milliseconds. Within this timescale, human posture changes manifest as continuous, slow spatial movement, and the spatial displacement amplitude of anatomical key points between adjacent consecutive image frames is limited. In postoperative bed rest recovery scenarios, when the imaging device operates with a common video acquisition configuration, the following are examples of reasonable range values ​​for human posture changes. The spatial displacement between adjacent consecutive image frames is no more than approximately four centimeters. Common video acquisition configurations include a video frame rate in the range of 25 to 30 frames per second and an image resolution in the range of 1,920 pixels horizontally and 1,080 pixels vertically. When the video frame rate or image resolution changes, the reasonable range of human posture changes is re-determined based on the video frame rate and the spatial resolution of the imaging device. When the spatial displacement range is smaller than the example value, it is easy to misjudge real rapid posture changes as abnormal displacements. When the spatial displacement range is larger than the example value, it is easy to include discontinuous jumps caused by occlusion or key point misalignment within the reasonable range.

[0053] It should also be noted that in postoperative bed rest recovery scenarios, the acquisition time interval between adjacent consecutive image frames is within a small time scale, and changes in human posture usually manifest as a continuous and slow spatial displacement process. If abnormal abrupt changes in the spatial coordinates of anatomical key points occur in adjacent consecutive image frames, these abnormal changes are unlikely to be formed by real human posture changes, but usually originate from non-posture factors such as occlusion, changes in illumination, recognition misalignment, or loss of key point tracking. Based on this, performing validity judgment on the spatial coordinates of anatomical key points by analyzing the spatial displacement changes in adjacent consecutive image frames helps to eliminate unstable data sources during the posture evolution sequence construction stage.

[0054] Furthermore, the spatial coordinates of the effective anatomical key points in each frame of the continuous image frame sequence are arranged in the order of acquisition time to form a pose evolution sequence. The pose evolution sequence includes the spatial coordinates of the effective anatomical key points, the spatial coordinates of the invalid anatomical key points, and the corresponding acquisition time information of each frame image.

[0055] S2. Divide the posture evolution sequence into continuous time segments, calculate the spatial displacement amplitude of effective anatomical key points between adjacent consecutive image frames, and select the posture stable state time segment. Form anatomical key point pair combinations within the posture stable state time segment and calculate the spatial distance between anatomical key point pairs.

[0056] Furthermore, based on the effective anatomical key point spatial coordinates of each frame image in the pose evolution sequence, pose stability state recognition is performed.

[0057] The pose evolution sequence is divided into multiple consecutive time segments according to the acquisition time order. Each consecutive time segment consists of a fixed number of adjacent consecutive image frames.

[0058] It should be noted that the fixed quantity is determined by the target time length of the continuous time segment and the acquisition time interval. For example, the target time length of the continuous time segment is two seconds. The acquisition time interval is obtained by subtracting the acquisition time information of two adjacent frames in the continuous image frame sequence. In the continuous image frame sequence, starting from the starting frame image of any continuous time segment, the acquisition time interval is accumulated frame by frame along the acquisition time sequence. When the accumulated acquisition time interval first reaches two seconds, the number of image frames participating in the accumulation is taken as the fixed quantity. The fixed quantity ensures that the acquisition time length covered by the continuous time segment is not less than two seconds.

[0059] It should also be noted that the target time length of the continuous time segment is used to cover a complete bed rest posture adjustment process on a time scale. During postoperative bed rest recovery, patient posture changes usually go through continuous stages such as initiation, adjustment and termination. A two-second time span can simultaneously include the initial response and the final convergence process of posture changes, and reduce the interference of short-term fluctuations caused by single-frame occlusion and instantaneous misalignment of key points on the posture stability recognition results. When the target time length is less than the example value, the bed rest posture adjustment process is easily divided into multiple short segments, and the time boundary of posture stability recognition is easily dispersed. When the target time length is greater than the example value, the bed rest posture adjustment process and the bed rest posture maintenance process are easily included in the same continuous time segment, and the time boundary of posture stability recognition is easily blurred.

[0060] Furthermore, for each continuous time segment, the spatial coordinates of the effective anatomical key points in each frame of the image within the continuous time segment are obtained frame by frame.

[0061] Within a continuous time segment, for each effective anatomical key point, the spatial displacement amplitude between adjacent consecutive image frames is calculated frame by frame. The process of obtaining the spatial displacement amplitude includes obtaining the spatial coordinates of the effective anatomical key point in adjacent consecutive image frames for the same effective anatomical key point, and obtaining the spatial displacement amplitude of the effective anatomical key point in adjacent consecutive image frames by the straight-line distance between the two spatial coordinate positions.

[0062] Based on the spatial displacement amplitude of all effective anatomical key points within a continuous time segment between adjacent consecutive image frames, attitude stability state recognition is performed. Attitude stability state recognition includes performing frame-by-frame comparison on the acquisition time range covered by the continuous time segment. For each pair of adjacent consecutive image frames, the frame-by-frame comparison determines whether the spatial displacement amplitude in the later image frame is not less than the spatial displacement amplitude in the previous image frame.

[0063] When invalid anatomical key point spatial coordinate markers exist within a continuous time segment, unstable markers are recorded within the continuous time segment. The time interval of the unstable markers is determined by the earliest and latest acquisition time information of the continuous time segment containing invalid anatomical key point spatial coordinate markers. When, within the acquisition time range covered by the continuous time segment, for all adjacent consecutive image frames, the spatial displacement amplitude in the subsequent image is not less than the spatial displacement amplitude in the previous image, continuous attitude change markers are recorded within the continuous time segment. The time interval of the continuous attitude change markers is determined by the start and end acquisition times of the continuous time segment.

[0064] When there are no invalid spatial coordinate markers for anatomical key points within a continuous time segment, and there is no situation within the acquisition time range covered by the continuous time segment where the spatial displacement amplitude in the next frame is not less than that in the previous frame for all adjacent consecutive image frames, the continuous time segment is identified as a time period of stable attitude.

[0065] It should be noted that in the postoperative bed rest recovery scenario, the stable posture state is not the same as the complete stillness of the spatial position of the anatomical key points, but rather manifests as small back-and-forth adjustments around a certain posture position; in this state, the spatial displacement amplitude of the anatomical key points between adjacent consecutive image frames usually exhibits non-monotonic variation characteristics, that is, the spatial displacement amplitude decreases in some adjacent frames.

[0066] Therefore, when the spatial displacement amplitude sequence does not show a continuous unidirectional increasing characteristic within the acquisition time range covered by the continuous time segment, and no invalid anatomical key point spatial coordinate markers are detected, it can be determined that the attitude change in the continuous time segment is a natural adjustment process around the stable attitude position, rather than a continuous attitude change process, thus identifying the continuous time segment as a period of stable attitude state.

[0067] The attitude stability period does not record the unstable marker time interval and the continuous attitude change marker time interval. The attitude stability period includes the acquisition time information of each frame image within the attitude stability period, as well as the spatial coordinates of the effective anatomical key points of each frame image within the attitude stability period.

[0068] Furthermore, during the period of stable posture, based on the spatial positional relationship of anatomical key points in the human body structure, the spatial coordinates of anatomical key points are combined and processed to select pairs of anatomical key points that reflect the relative spatial relationship between the head, neck, trunk and pelvis, forming a combination of anatomical key point pairs.

[0069] The key anatomical points are paired together, including the central position of the head and the anatomical position of the neck, the anatomical position of the neck and the anatomical position of the trunk, the anatomical position of the trunk and the anatomical position of the pelvis, the central position of the head and the anatomical position of the trunk, and the anatomical position of the neck and the anatomical position of the pelvis.

[0070] For each pair of anatomical key points during the attitude stabilization period, the spatial distance between the anatomical key points is obtained by using the effective spatial coordinates of the anatomical key points in each frame of the image during the attitude stabilization period. The process of obtaining the spatial distance between the anatomical key points includes extracting the effective spatial coordinates of the two ends of the anatomical key point pair in the same frame image, and obtaining the spatial distance between the two ends of the spatial coordinate position by the straight line distance.

[0071] S3. Based on the changes in spatial distance between anatomical key points and acquisition time during the stable posture period, a sliding window cumulative change operator is formed. By comparing the sliding window cumulative change operator frame by frame, the time periods of posture behavior change and posture behavior maintenance are identified.

[0072] Furthermore, during the attitude stabilization period, each frame of the image is traversed in chronological order of acquisition. The spatial distance between each pair of anatomical keypoints in the first frame of the attitude stabilization period is used as the reference spatial distance. For any frame of the attitude stabilization period, the logarithmic deviation of each pair of anatomical keypoints is constructed based on the difference between the spatial distance between each pair of anatomical keypoints in any frame and the reference spatial distance, expressed as:

[0073] ;

[0074] in, Indicates the first The key anatomical points of the group were located during the time period of postural stability. Logarithmic deviation at the frame image This indicates the number of the anatomical keypoint pair within the anatomical keypoint pair combination. Indicates the image frame number within the time period of stable attitude. Represents the natural logarithm operation. This represents the absolute value operation.

[0075] Indicates the first Group anatomical key points in the first Spatial distance in frame image Indicates the first Spatial distances of key anatomical points in the initial frame image during the pose-steady period. It represents a very small positive number and is used to avoid the denominator being zero.

[0076] Furthermore, during the attitude-stable state time period, for each frame, a sliding window is selected with the current frame as the terminating frame. The logarithmic deviation of each frame within the window coverage area is accumulated in a time-weighted manner to obtain the sliding window cumulative change operator, expressed as:

[0077] ;

[0078] in, Indicates the first time within the stable attitude period The sliding window cumulative change operator at the frame image. This indicates the image frame number within the coverage area of ​​the sliding window. This indicates the number of image frames covered by the sliding window, with the sliding window coverage ranging from the [number]th frame to the [number]th frame. Frame image extended to the Frame image, Indicates the first The acquisition time interval between a frame image and the immediately preceding frame image, and the starting frame image of the sliding window. The value is zero. This indicates the number of anatomical keypoint pairs in the combination. Indicates the first Group anatomical key points in the first Logarithmic deviation at the frame image.

[0079] It should be noted that the sliding window coverage time is limited by both the video capture frame rate and the number of image frames. The sliding window coverage time is equal to the ratio of the number of image frames covered by the sliding window to the video capture frame rate. In the postoperative bed rest recovery scenario, the example value for the sliding window coverage time is two seconds. This example value is chosen because a two-second time span can cover the continuous image frame changes during a short-term posture fine-tuning process of the patient, while avoiding mistaking short-cycle jitters caused by breathing fluctuations and mattress rebound as posture adjustments. When the sliding window coverage time is less than the example value, the number of image frames contained in the sliding window decreases, the sliding window cumulative change operator is more sensitive to short-term noise, and the sliding window cumulative change operator is more likely to produce frequent fluctuations in the comparison of adjacent consecutive image frames. When the sliding window coverage time is greater than the example value, the number of image frames contained in the sliding window increases, the response of the sliding window cumulative change operator to posture changes is over-smoothed, and the start and end boundaries of the posture behavior change state time period are more likely to have time lag. When the video capture frame rate is in the range of 25 to 30 frames per second, the example value for the number of image frames covered by the sliding window is in the range of 50 to 60 frames.

[0080] It should also be noted that, within the attitude stability period, along the acquisition time range covered by the attitude stability period, the values ​​of the sliding window cumulative change operator are compared frame by frame according to the acquisition time sequence.

[0081] When, within the acquisition time range covered by the attitude stability period, for each pair of adjacent consecutive image frames, the value of the sliding window cumulative change operator in the later image frame is not less than the value in the earlier image frame, the attitude stability period is identified as the attitude behavior change period. When, within the acquisition time range covered by the attitude stability period, there exists any pair of adjacent consecutive image frames where the value of the sliding window cumulative change operator in the later image frame is less than the value in the earlier image frame, the attitude stability period is identified as the attitude behavior maintenance period.

[0082] S4. Select the time periods of postural behavior change and the time periods of postural behavior maintenance as the analysis objects. Based on the changes in spatial distance values ​​of anatomical key points, distinguish the relationship of changes in different human body structural levels, and arrange them in the order of collection time to form a postoperative postural structure behavior evolution record.

[0083] Furthermore, during the posture maintenance state period, each frame of the image is traversed in the order of acquisition time. For each pair of anatomical key points in the combination of anatomical key points, the spatial distance value of the anatomical key point pair is obtained frame by frame. For the spatial distance value of the anatomical key point pair in each frame of the posture maintenance state period, spatial resolution discretization processing is performed. The spatial resolution discretization processing includes mapping the spatial distance value of the anatomical key point pair. The mapping processing represents the spatial distance value of the anatomical key point pair as an integer multiple of the minimum resolvable spatial distance obtained by the spatial resolution conversion of the imaging device.

[0084] During the pose maintenance state time period, for each pair of anatomical keypoints, the frequency of occurrence of the spatial distance value of the discretized anatomical keypoint pair in each frame image during the pose maintenance state time period is counted. The spatial distance value of the discretized anatomical keypoint pair with the highest frequency is selected as the maintenance reference distance value. The maintenance reference distance values ​​of each pair of anatomical keypoints in the combination of anatomical keypoints together constitute the maintenance reference distance group.

[0085] During the time period of posture behavior change, the images of each frame in the time period of posture behavior change are traversed in the order of acquisition time, and the spatial distance value of the anatomical key point pair is obtained frame by frame for each anatomical key point pair combination; the spatial resolution discretization processing is performed on the spatial distance value of the anatomical key point pair of each frame image in the time period of posture behavior change.

[0086] Furthermore, during the time period of posture and behavior changes, the spatial distance changes of discrete anatomical key point pairs formed by the head center position and neck anatomical position, and the neck anatomical position and trunk anatomical position, are extracted in the time sequence of acquisition to form a record of changes at the head and neck structural level; the spatial distance changes of discrete anatomical key point pairs formed by the trunk anatomical position and pelvic anatomical position are extracted in the time sequence of acquisition to form a record of changes at the trunk axial structural level; and the spatial distance changes of discrete anatomical key point pairs formed by the head center position and trunk anatomical position, and the neck anatomical position and pelvic anatomical position, are extracted in the time sequence of acquisition to form a record of changes at the cross-regional structural level.

[0087] During the time period of posture and behavior change, for the records of changes in the head and neck structure level, the records of changes in the trunk axial structure level, and the records of changes in the cross-regional structure level, the relationship between the spatial distance values ​​of discrete anatomical keypoint pairs and the values ​​of the maintained reference distances in the maintained reference distance group is compared frame by frame. When the spatial distance value of any set of anatomical keypoint pairs within the same structural level changes in any frame, the spatial distance value of the discrete anatomical keypoint pairs is compared with the maintained reference distance value of the same set of anatomical keypoint pairs in the maintained reference distance group. When the spatial distance value of the discrete anatomical keypoint pairs is not equal to the maintained reference distance value, the image frames with the unequal relationship are marked as structural level deviation frames. When structural level deviation frames appear consecutively in the acquisition time sequence, the acquisition time range covered by the structural level deviation frames is recorded as the structural level deviation time interval.

[0088] Furthermore, the distribution of the time intervals of deviation from the head and neck structure hierarchy, the time intervals of deviation from the trunk axial structure hierarchy, and the time intervals of deviation from the cross-regional structure hierarchy in the acquisition time sequence was compared to determine the occurrence of different structural hierarchy deviation time intervals within the same stable posture time period, as well as the order of occurrence of different structural hierarchy deviation time intervals in the acquisition time sequence.

[0089] When there is only a time interval of head and neck structural level deviation within the time period of posture behavior change, and there is no time interval of trunk axial structural level deviation or cross-regional structural level deviation within the time period of posture behavior change, the posture behavior change is classified as local head and neck adjustment behavior, and the time interval of occurrence of local head and neck adjustment behavior in the acquisition time sequence is recorded.

[0090] When there is only a time interval of deviation of the trunk axial structure level within the time period of posture behavior change, and there is no time interval of deviation of the head and neck structure level or the time interval of deviation of the cross-regional structure level within the time period of posture behavior change, the posture behavior change is classified as trunk axial adjustment behavior, and the time interval of occurrence of trunk axial adjustment behavior in the acquisition time sequence is recorded.

[0091] When there are both head and neck structural level deviation time intervals and trunk axial structural level deviation time intervals within the time period of posture behavior change, and there are no cross-regional structural level deviation time intervals within the time period of posture behavior change, the posture behavior change is classified as segmented posture adjustment behavior, and the occurrence time intervals of the head and neck structural level deviation time intervals and the trunk axial structural level deviation time intervals are recorded in the order of acquisition time.

[0092] When there is a cross-regional structural hierarchy deviation time interval during the posture behavior change state period, the posture behavior change is classified as overall posture adjustment behavior, and the occurrence time interval of the cross-regional structural hierarchy deviation time interval in the acquisition time sequence is recorded; the existence of the head and neck structure hierarchy deviation time interval and the trunk axial structure hierarchy deviation time interval during the posture behavior change state period does not change the classification result of the overall posture adjustment behavior.

[0093] It should be noted that cross-regional structural levels are characterized by anatomical key points located in different anatomical regions of the human body, and changes in their spatial distance reflect the coordinated displacement of multiple structural regions within the same time frame. Compared with local structural levels or single-axial structural levels, changes in the spatial distance of cross-regional structural levels are usually accompanied by overall human posture rearrangement or center of gravity shift. Based on the hierarchical relationship of human anatomical structures, when cross-regional structural levels show continuous deviation, the consistency of human posture changes at the overall structural level is stronger. Therefore, it can be used to distinguish overall posture adjustment behavior, without being affected by the presence or absence of changes in other structural levels.

[0094] Furthermore, within the stable posture period, the start and end times of the acquisition and the reference distance for maintaining the posture behavior are summarized, and the time intervals of occurrence of local head and neck adjustment behavior, trunk axial adjustment behavior, segmented posture adjustment behavior, and overall posture adjustment behavior are summarized. At the same time, the unstable marked time intervals and the continuous posture change marked time intervals are summarized. All occurrence time intervals are sorted according to the start acquisition time and the overlap of time intervals is processed to form a postoperative posture structure behavior evolution record.

[0095] Postoperative postural structural behavioral evolution records are used to describe the behavioral evolution of postural adjustments in neurosurgical patients at different human structural levels during postoperative bed rest.

[0096] It should also be noted that, Figure 5 This method is used to illustrate the effectiveness of intelligent postoperative posture monitoring in neurosurgical patients in depicting the time dimension of the postoperative posture adjustment process under continuous bed rest monitoring, and to provide a direct comparison of the results of distinguishing between the time periods of posture behavior change and the time periods of posture behavior maintenance. Figure 5 The horizontal axis represents the acquisition time, and the vertical axis represents the normalized deviation change value obtained by constructing the spatial distance changes of the anatomical key points; the light blue shaded area represents the true value range of the time period of posture behavior change state, and the unshaded area represents the true value range of the time period of posture behavior maintenance state; the consistency between the shaded area and the curve trend is used to reflect the ability of this invention to recordable express the posture behavior state switching.

[0097] Figure 5The curves of different colors are all time-series response curves, used to compare the differences between sliding window cumulative processing and non-cumulative processing. The three sliding window cumulative change operator curves of different colors correspond to sliding window lengths of 1s, 2s, and 3s, respectively. The sliding window cumulative change operator is formed by the change of spatial distance between anatomical keypoint pairs in the pose evolution sequence over time. Sliding window cumulative processing accumulates short-term deviation changes within the sliding window coverage time range, thus highlighting the continuous change trend. The average deviation without sliding window curve is derived from the spatial distance sequence of anatomical keypoint pairs: within each frame, the deviation of the spatial distance of each anatomical keypoint pair combination relative to the reference distance is normalized to obtain the deviation change of each keypoint pair combination. Then, the deviation change of all keypoint pair combinations is averaged to obtain the average deviation of the frame. The average deviation reflects the instantaneous deviation level at the current moment and is used to compare with the cumulative response of the sliding window cumulative change operator.

[0098] The comparison between the curves and the true value intervals shows that during the period of posture behavior change, the average deviation often initially increases or fluctuates more, reflecting the immediate response of anatomical key points to the spatial distance starting to deviate from the reference state. The sliding window cumulative change operator then exhibits a more continuous upward trend, reflecting the enhancement effect formed by the continuous accumulation of deviation changes within the sliding window coverage time. During the period of posture behavior maintenance, the average deviation may show short-term spikes or rapid declines, reflecting short-term disturbances caused by respiratory micro-movements, local occlusion, and posture resolution errors. The sliding window cumulative change operator is closer to a stable level, demonstrating the inhibitory effect of sliding window accumulation on short-term disturbances, making the difference between the change segment and the maintenance segment clearer in the time dimension. These results demonstrate that this invention, by validating the spatial coordinates of anatomical key points and constructing a posture evolution sequence, achieves continuous and consistent expression of the temporal data required for posture behavior recognition in bedridden scenarios. Furthermore, by forming a sliding window cumulative change operator based on spatial distance changes using anatomical key points, it achieves a joint characterization of postoperative posture adjustments in the time dimension, forming a postoperative posture structure behavior evolution record that reflects the posture behavior evolution process.

[0099] The sliding window length has a direct impact on the curve characteristics. When the sliding window length is 1s, the cumulative change operator is more sensitive to the start of the change, and the upward trend is more likely to appear earlier at the beginning of the attitude behavior change state time period. However, the curve is more likely to fluctuate with short-term fluctuations, and local upward segments are more likely to appear during the attitude behavior hold state time period. When the sliding window length is 3s, the curve is smoother, the stability level of the hold segment is more obvious, and short-term spikes are more strongly suppressed. However, the start response of the change segment is later, and the cumulative lag is more obvious. When the sliding window length is 2s, the response timeliness and smoothness are more balanced, the upward trend in the change segment is more continuous, the stability level in the hold segment is clearer, and it is more conducive to forming a stable boundary between the change segment and the hold segment in the frame-by-frame comparison process. Figure 5 This invention visually demonstrates the ability of the present invention to simultaneously obtain instantaneous deviation response and cumulative change response on the same continuous monitoring sequence, and proves that the sliding window cumulative change operator can record the evolution process of postoperative posture adjustment in a quantifiable and comparable manner, supporting the reliable identification of the time period of posture behavior change state and the time period of posture behavior maintenance state.

[0100] This embodiment also provides a computer device applicable to the method of intelligent postoperative posture monitoring of neurosurgical patients, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the method of intelligent postoperative posture monitoring of neurosurgical patients as proposed in the above embodiment.

[0101] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0102] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for intelligent monitoring of postoperative posture of neurosurgical patients as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0103] In summary, this invention achieves continuous and consistent expression of temporal data required for posture behavior feature recognition in bedridden scenarios by: determining the validity of spatial coordinates of anatomical key points and constructing a posture evolution sequence; and by forming a sliding window cumulative change operator based on changes in spatial distance from anatomical key points and combining it with the hierarchical division of human body structure to define posture change relationships, thereby achieving a joint characterization of postoperative posture adjustment in both the temporal and structural hierarchical dimensions, forming a postoperative posture structure behavior evolution record that reflects the evolution process of posture behavior.

[0104] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent monitoring of postoperative posture in neurosurgical patients, characterized in that: include, Video images of the bedridden area are collected to form a continuous image frame sequence. The human posture of the continuous image frame sequence is analyzed and the spatial coordinates of effective anatomical key points are determined. The spatial coordinates of effective anatomical key points are arranged in the order of collection time to form a posture evolution sequence. The pose evolution sequence is divided into continuous time segments. The spatial displacement amplitude of effective anatomical key points between adjacent consecutive image frames is calculated and the pose stable state time segment is selected. Anatomical key point pairs are formed within the pose stable state time segment and the spatial distance between the anatomical key point pairs is calculated. Based on the change in spatial distance between anatomical key points and acquisition time during the stable posture period, a sliding window cumulative change operator is formed. By comparing the sliding window cumulative change operator frame by frame, the time periods of posture behavior change and posture behavior maintenance are identified. The analysis focused on the time periods of postural behavior change and the time periods of postural behavior maintenance. Based on the changes in spatial distance values ​​at anatomical key points, the analysis distinguished the relationship between changes in different human structural levels and arranged them in chronological order of collection to form a postoperative postural structure behavior evolution record.

2. The method for intelligent monitoring of postoperative posture in neurosurgical patients as described in claim 1, characterized in that: The process of acquiring video images of the bed rest area to form a continuous image frame sequence includes, during the period when a neurosurgical patient has completed surgery and is in postoperative bed rest recovery, acquiring video of the bed rest area through an imaging device positioned above the bed rest area and an imaging device positioned to the side of the bed rest area. The video capture field of view covers the head, neck, trunk, and limb areas involved in postoperative posture changes of neurosurgical patients; The acquired video images are arranged in the order of their acquisition time to form a continuous image frame sequence.

3. The method for intelligent monitoring of postoperative posture in neurosurgical patients as described in claim 2, characterized in that: The step of performing human pose analysis on a continuous image frame sequence and determining the spatial coordinates of effective anatomical key points, and arranging the spatial coordinates of effective anatomical key points in the order of acquisition time to form a pose evolution sequence includes performing human pose analysis processing on each frame of the continuous image frame sequence and extracting the spatial coordinates of anatomical key points of the head center position, neck anatomical position, trunk anatomical position and pelvic anatomical position. For each frame of the image, the spatial coordinates of the anatomical key points are extracted and a validity determination is performed to determine whether the spatial coordinates of the anatomical key points are completely present and whether the change range of the spatial coordinates of the anatomical key points in adjacent consecutive image frames is within a reasonable range of human posture changes. When the spatial coordinates of anatomical key points fail the validity judgment process, record the invalid anatomical key point spatial coordinates. The spatial coordinates of the anatomical key points processed through validity determination are used as the valid spatial coordinates of the anatomical key points. The valid spatial coordinates of the anatomical key points of each frame are arranged in the order of acquisition time to form a pose evolution sequence.

4. The method for intelligent monitoring of postoperative posture in neurosurgical patients as described in claim 3, characterized in that: The calculation of the spatial displacement amplitude of the effective anatomical key points between adjacent consecutive image frames and the selection of the time period of the attitude stability state include dividing the attitude evolution sequence into multiple consecutive time segments according to the acquisition time order, and each consecutive time segment consists of a fixed number of adjacent consecutive image frames. Within each consecutive time segment, for each effective anatomical key point, the spatial coordinates of the effective anatomical key point are obtained in adjacent consecutive image frames. The spatial displacement amplitude of the effective anatomical key point between adjacent consecutive image frames is obtained by the straight-line distance between the two spatial coordinate positions of the effective anatomical key point in adjacent consecutive image frames. If there are no invalid spatial coordinate markers for anatomical key points within a continuous time segment, and if there is no situation within the acquisition time range covered by the continuous time segment where the spatial displacement amplitude in the next frame is not less than that in the previous frame for all adjacent consecutive image frames, then the continuous time segment is selected as a time period of stable attitude.

5. The method for intelligent monitoring of postoperative posture in neurosurgical patients as described in claim 4, characterized in that: The process of forming a combination of anatomical key point pairs and calculating the spatial distance between the anatomical key point pairs during the time period of stable posture includes, during the time period of stable posture, combining the spatial coordinates of the effective anatomical key points based on the spatial positional relationship of the effective anatomical key points in the human body structure to form a combination of anatomical key point pairs. For each pair of anatomical key points during the stable posture period, the spatial coordinates of the effective anatomical key points at both ends of the pair are obtained in the same image frame, and the spatial distance between the two ends is obtained by the straight-line distance between the spatial coordinates.

6. The method for intelligent monitoring of postoperative posture in neurosurgical patients as described in claim 5, characterized in that: The step of forming a sliding window cumulative change operator based on the change of spatial distance between anatomical key point pairs with acquisition time during the attitude stability period includes: traversing each frame of the image in the attitude stability period according to the acquisition time order during the attitude stability period; using the spatial distance between each anatomical key point pair in the starting frame of the attitude stability period as the reference spatial distance; and constructing the logarithmic deviation of the anatomical key point pair based on the difference between the spatial distance between each anatomical key point pair in any frame of the attitude stability period and the reference spatial distance for any frame of the image. For each frame of image, a sliding window is selected with the current frame as the terminating frame. The logarithmic deviation of each frame within the window coverage area is accumulated in a time-weighted manner to obtain the sliding window cumulative change operator.

7. The method for intelligent monitoring of postoperative posture in neurosurgical patients as described in claim 6, characterized in that: The step of identifying the time period of attitude behavior change state and the time period of attitude behavior maintenance state by comparing the sliding window cumulative change operator frame by frame includes performing frame-by-frame comparison of the values ​​of the sliding window cumulative change operator corresponding to adjacent consecutive image frames in the acquisition time range covered by the attitude stable state time period. During the frame-by-frame comparison process, when the cumulative change operator of the sliding window of the next frame image is always not less than the cumulative change operator of the sliding window of the previous frame image, the attitude stable state time period is identified as the attitude behavior change state time period. When there is a situation where the cumulative change operator of the sliding window of the next frame image is less than the cumulative change operator of the sliding window of the previous frame image, the attitude stable state time period is identified as the attitude behavior maintenance state time period.

8. The method for intelligent monitoring of postoperative posture in neurosurgical patients as described in claim 7, characterized in that: The process of distinguishing the changes in spatial distance values ​​based on anatomical key points and arranging them in the order of collection time to form a postoperative posture structure behavior evolution record includes selecting the time period of posture behavior change state and the time period of posture behavior maintenance state as the analysis object, and distinguishing the posture change relationship of different structural levels of the human body based on the changes in spatial distance values ​​based on anatomical key points on the collection time axis. The human body structural hierarchy includes at least local structural hierarchy characterized by different anatomical key point pairs, axial structural hierarchy, and cross-regional structural hierarchy; The posture change relationships corresponding to each structural level are temporally correlated with the start and end times of the posture behavior maintenance period, arranged in the order of acquisition time, and the overlapping change intervals in time are integrated to form a postoperative posture structure behavior evolution record.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for intelligent monitoring of postoperative posture of neurosurgical patients as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for intelligent monitoring of postoperative posture of neurosurgical patients as described in any one of claims 1 to 8.

Images