Method and system for evaluating postoperative state of critical spinal patient

Abstract

The invention discloses a severe spinal patient postoperative state evaluation method and system, and relates to the technical field of medical data analysis, and the method comprises the steps: judging the motion balance state of a patient based on the gravity center change track of the patient; constructing a vital sign tendency chart based on the vital sign monitoring data, and analyzing the vital sign intensity of the patient from the vital sign tendency chart; extracting urine sediment characteristics from the urine sample image of the patient, and predicting the urine state of the patient in the future time period based on the urine sediment characteristics; weights are distributed to the action balance state, the vital sign intensity and the urine state of the patient in a self-adaptive mode, and the postoperative state of the severe spine patient is evaluated in a weighted summation mode. According to the method, the weights are distributed to the action balance state, the vital sign intensity and the urine state of the patient in a self-adaptive mode, the postoperative state of the severe spine patient is comprehensively evaluated in a weighted summation mode, and the rehabilitation condition of the patient can be reflected more comprehensively and accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data analysis, and in particular to a method and system for evaluating the postoperative status of critically ill spinal patients. Background Art

[0002] Severe spinal disease refers to patients who experience severe functional impairment, neurological damage, spinal deformity, and other conditions due to spinal disease or spinal-related trauma. These patients often present with complex symptoms and may also have other systemic issues, such as respiratory, circulatory, and neurological problems. The primary purpose of postoperative status assessment for patients with severe spinal disease is to comprehensively and accurately monitor their recovery progress, promptly identify and intervene in potential complications, and promote postoperative recovery.

[0003] Patients undergoing spinal surgery often face a variety of physiological, functional, and psychological issues, such as neurological recovery, motor function, balance, mental state, and visceral function. A single-dimensional assessment (e.g., focusing solely on vital signs) can only reflect a portion of the physiological state and cannot reveal the patient's specific functional recovery, such as balance, neurological recovery, or recovery of limb mobility, all of which are closely related to the patient's overall recovery. A single vital sign assessment cannot assess whether the patient can resume daily living abilities or whether there are potential issues such as nerve damage or muscle weakness, and therefore cannot fully reflect the patient's overall recovery.

[0004] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention

[0005] In response to the problems in the related art, the present invention proposes a method and system for evaluating the postoperative status of critically ill spinal patients to overcome the above-mentioned technical problems existing in the existing related art.

[0006] To this end, the specific technical solutions adopted in the present invention are as follows:

[0007] According to one aspect of the present invention, a method for evaluating the postoperative status of a critically ill spinal patient is provided, the method comprising:

[0008] Collecting the patient's motion posture images, extracting the patient's center of gravity change trajectory from the motion posture images, and judging the patient's motion balance state based on the patient's center of gravity change trajectory;

[0009] Collecting the patient's vital sign monitoring data, constructing a vital sign trend chart based on the vital sign monitoring data, and analyzing the patient's vital sign intensity from the vital sign trend chart;

[0010] collecting a urine sample image of the patient, extracting urine sediment features from the urine sample image of the patient, and predicting the patient's urine status in a future time period based on the urine sediment features;

[0011] Adaptively assign weights to the patient's movement balance, vital sign strength, and urine status, and use the weighted summation method to evaluate the postoperative status of critically ill spinal patients;

[0012] The step of collecting the patient's motion and posture images and extracting the patient's center of gravity change trajectory from the motion and posture images includes:

[0013] Collect the patient's posture images when performing limb activities, and use deep learning algorithms to extract the patient's body center of gravity from the posture images;

[0014] The hidden Markov model is used to analyze the horizontal axis deviation and vertical axis deviation of the patient's body center of gravity position within a preset time period, and the change trajectory of the body center of gravity position is generated based on the deviation analysis results;

[0015] The adaptive weights assigned to the patient's motion balance, vital sign intensity, and urine status include:

[0016] Evaluation indicators are formulated for the patient's movement balance state, vital sign intensity and urine status, and the bat algorithm is used to adaptively assign weights to each evaluation indicator.

[0017] Preferably, judging the patient's movement balance state based on the trajectory of the patient's center of gravity change includes:

[0018] Mapping the changing trajectory of the body center of gravity position to the three-dimensional virtual space;

[0019] The patient's movements and environment are interacted with in a three-dimensional virtual space, and the patient's movement balance state when avoiding obstacles is detected in real time.

[0020] Preferably, using a hidden Markov model to analyze the horizontal axis direction offset and the vertical axis direction offset of the patient's body center of gravity position within a preset time period, and generating a change trajectory of the body center of gravity position based on the offset analysis results includes:

[0021] Discretize the center of gravity of the patient's body in space and construct a graph structure. Connect the nodes in the graph structure to obtain the patient's initial trajectory within a preset time period.

[0022] Mapping the initial trajectory of the patient in a preset time period into a horizontal axis offset and a vertical axis offset, and calculating the horizontal offset and the vertical offset of the patient in each time period;

[0023] A hidden Markov model is established based on the predefined initial state distribution, state transition probability matrix and probability density vector, and the horizontal offset and vertical offset are input into the hidden Markov model to output the action observation;

[0024] Based on the motion observations, the Viterbi decoding algorithm is used to backtrack and generate the trajectory of the patient's center of gravity position changes.

[0025] Preferably, collecting a urine sample image of the patient, extracting urine sediment features from the urine sample image of the patient, and predicting the patient's urine status in a future time period based on the urine sediment features includes:

[0026] collecting a urine sample image of the patient, performing image enhancement processing on the collected urine sample image, and obtaining an optimized urine sample image;

[0027] Extracting and optimizing urine sediment features in the urine sample image, and segmenting the area to which the urine sediment belongs to obtain a sediment segmentation image;

[0028] The principal component analysis technique is used to extract sediment analysis data from the sediment segmentation image, and a training dataset is constructed based on the sediment analysis data;

[0029] The patient's historical urine characteristics are obtained to build a prediction model, and the prediction model is trained using a training data set. The trained prediction model is used to predict the patient's urine status in future time periods.

[0030] Preferably, extracting and optimizing urine sediment features in the urine sample image and segmenting the region to which the urine sediment belongs to obtain a sediment segmentation image includes:

[0031] Calculating and optimizing the color space components of pixels in the urine sample image, comparing the color space components with a preset threshold, and selecting pixels corresponding to color space components greater than the preset threshold as urine pixels;

[0032] Construct a normal distribution function of urine pixels in color space, identify the color space features of urine based on the normal distribution function, and extract urine sediment features from the color space features;

[0033] The area containing urine sediment features in the optimized urine sample image is segmented to obtain a sediment segmentation image, and the noise in the sediment segmentation image is removed using the similarity distance method.

[0034] Preferably, removing noise in the sediment segmentation image using the similarity distance method includes:

[0035] Scan the pixels in the sediment segmentation image one by one, select the pixels that meet the preset conditions as reference points, and calculate the component value of the current reference point in the HSV space;

[0036] Scan the next pixel and calculate the component value of the current pixel in the HSV space according to the preset conditions;

[0037] Calculate the spatial distance between the reference point and the pixel point, and compare the spatial distance with a preset threshold. If the spatial distance is less than or equal to the preset threshold, it indicates that the colors of the reference point and the pixel point are similar, and the reference point and the pixel point are retained as the region point of the sediment segmentation image.

[0038] The pixel scanning process is repeated with the pixel point as the reference point until all the pixel points in the sediment segmentation image are scanned, so as to remove the noise points in the sediment segmentation image.

[0039] Preferably, the weighted sum method is used to evaluate the postoperative status of critically ill spinal patients, including:

[0040] The patient's total score was calculated using a weighted summation method and input into a predefined evaluation model;

[0041] The evaluation model outputs the postoperative status assessment level of critically ill spinal patients, and intervention measures are formulated according to the postoperative status assessment level of critically ill spinal patients.

[0042] Preferably, using the bat algorithm to adaptively assign weights to each evaluation indicator includes:

[0043] The evaluation index is used as the bat individual, and the bat algorithm parameters are initialized. The bat algorithm parameters include the number of bat individuals, bat position, bat speed, bat loudness and bat emission rate;

[0044] Calculate the fitness function value corresponding to the bat according to the fitness function, and select the optimal value from the fitness function values;

[0045] Generate a random number. If the random number is greater than the pulse emission rate, a local solution is generated near the optimal value. If the random number is less than the bat loudness, the local solution is used as the new optimal solution, and the bat loudness and bat emission rate are updated.

[0046] The iterative update process of the optimal solution is repeated until the maximum number of iterations is met, and the global optimal solution is output and used as the optimal weight of each evaluation indicator.

[0047] Preferably, the expression of the normal distribution function is:

[0048]

[0049] Where, f represents the normal distribution function; Cb represents a blue difference component; C r represents a red difference component; D represents a reference point; F represents a pixel point; E represents a spatial distance between the reference point and the pixel point.

[0050] According to another aspect of the present application, there is also provided a postoperative state evaluation system for a severe spinal patient, comprising a motion analysis module, a vital sign analysis module, a urine analysis module and a state comprehensive evaluation module.

[0051] The motion analysis module is configured to acquire a motion posture image of the patient, extract a center of gravity change trajectory of the patient from the motion posture image, and determine a motion balance state of the patient based on the center of gravity change trajectory of the patient.

[0052] The vital sign analysis module is configured to acquire vital sign monitoring data of the patient, construct a vital sign trend graph based on the vital sign monitoring data, and analyze a vital sign intensity of the patient from the vital sign trend graph.

[0053] The urine analysis module is configured to acquire a urine sample image of the patient, extract a urine precipitate feature from the urine sample image of the patient, and predict a urine state of the patient in a future time period based on the urine precipitate feature.

[0054] The state comprehensive evaluation module is configured to adaptively assign weights to the motion balance state, the vital sign intensity and the urine state of the patient respectively, and evaluate the postoperative state of the severe spinal patient by using a weighted summation method.

[0055] The present application has the following advantages:

[0056] 1. The present application uses deep learning technology to extract the position of the center of gravity from the motion image, which can accurately capture the change of the center of gravity in each motion process of the patient, and analyze the horizontal and vertical axis direction deviation of the patient in a preset time period through a hidden Markov model, reflecting the balance performance of the patient in the activity.

[0057] 2、The urine sediment characteristics are extracted through image processing technology, the optimized urine sample image can be obtained, the precision of the sediment characteristic extraction is improved, the key sediment characteristics are accurately positioned and extracted by segmenting the sediment area, the characteristics can reflect the health conditions of the patient's kidney and urinary system, important basis is provided for postoperative management, the urine state change in the future period of the patient can be predicted, the urine state change can be monitored in time during the postoperative stage of the patient for early intervention, and a more accurate and personalized monitoring tool is provided for the postoperative state evaluation of the severe spinal column patient.

[0058] 3、The postoperative state of the severe spinal column patient is comprehensively evaluated by adaptively assigning weights to the action balance state, vital sign intensity and urine state of the patient and using weighted summation, the rehabilitation condition of the patient can be more comprehensively and accurately reflected, the one-sidedness caused by a single index is avoided through the multi-dimensional combined method, potential problems can be found and intervened in time in different stages of postoperative recovery, and better rehabilitation effect is achieved for the patient. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0060] Figure 1 is a flowchart of a postoperative state evaluation method for a severe spinal column patient according to an embodiment of the present application;

[0061] Figure 2 is a principle block diagram of a postoperative state evaluation system for a severe spinal column patient according to an embodiment of the present application.

[0062] In the drawings:

[0063] 1, action analysis module; 2, vital sign analysis module; 3, urine analysis module; 4, state comprehensive evaluation module. DETAILED DESCRIPTION

[0064] To further illustrate the embodiments, the present application provides drawings which are part of the disclosure of the present application, mainly used to illustrate the embodiments, and can explain the operating principle of the embodiments in cooperation with the related description of the specification. Those skilled in the art should understand other possible embodiments and advantages of the present application by referring to these contents. The components in the drawings are not drawn to scale, and similar component symbols are usually used to represent similar components.

[0065] According to an embodiment of the present invention, a method and system for evaluating the postoperative status of critically ill spinal patients are provided.

[0066] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to an embodiment of the present invention, a method for evaluating the postoperative status of a critically ill spinal patient includes:

[0067] S1. Collecting a patient's motion posture image, extracting the patient's center of gravity change trajectory from the motion posture image, and judging the patient's motion balance state based on the patient's center of gravity change trajectory.

[0068] The steps of collecting the patient's motion posture image, extracting the patient's center of gravity change trajectory from the motion posture image, and judging the patient's motion balance state based on the patient's center of gravity change trajectory include:

[0069] The patient's posture images when performing limb activities are collected, and the patient's body center of gravity position is extracted from the posture images using a deep learning algorithm.

[0070] It should be noted that collecting the patient's posture images when performing limb activities and using deep learning algorithms to extract the patient's body center of gravity from the posture images include:

[0071] Collect images of the patient's movements and postures when performing limb activities, and shoot them using a camera or depth camera to ensure that the collected images have sufficient resolution and clarity for subsequent processing.

[0072] The image data is input into a deep learning model. Deep learning architectures such as convolutional neural networks (CNN) or human posture recognition models, such as OpenPose and HRNet, can effectively extract the human body's joint information from the image and use the position information of these joints to reconstruct the human skeleton model.

[0073] Based on the skeleton model, the position of each joint is calculated to obtain the center of gravity of the body position. The calculation of the center of gravity of the body position usually depends on the mass distribution of the body. It can be estimated by taking a weighted average based on the relative position and weight of each joint to estimate the center of gravity of the entire body.

[0074] The calculated center of gravity position is combined with actual body balance, stability and other indicators for further analysis and evaluation, and an optimization algorithm is used for iterative adjustment to ensure that the extracted center of gravity position reflects the actual movement process.

[0075] The hidden Markov model is used to analyze the horizontal axis deviation and vertical axis deviation of the patient's body center of gravity position within a preset time period, and the change trajectory of the body center of gravity position is generated based on the deviation analysis results.

[0076] The hidden Markov model is used to analyze the horizontal axis deviation and vertical axis deviation of the patient's body center of gravity position within a preset time period, and the change trajectory of the body center of gravity position is generated based on the deviation analysis results, including:

[0077] Discretize the center of gravity of the patient's body in space and construct a graph structure. Connect the nodes in the graph structure to obtain the patient's initial trajectory within a preset time period.

[0078] Mapping the initial trajectory of the patient in a preset time period into a horizontal axis offset and a vertical axis offset, and calculating the horizontal offset and the vertical offset of the patient in each time period;

[0079] A hidden Markov model is established based on the predefined initial state distribution, state transition probability matrix and probability density vector, and the horizontal offset and vertical offset are input into the hidden Markov model to output the action observation;

[0080] Based on the motion observations, the Viterbi decoding algorithm is used to backtrack and generate the trajectory of the patient's center of gravity position changes.

[0081] To facilitate understanding of the above technical solution of the present invention, the following is a detailed description of how the present invention uses a hidden Markov model to analyze the horizontal axis deviation and vertical axis deviation of the patient's body center of gravity position within a preset time period, and generates a change trajectory of the body center of gravity position based on the deviation analysis results:

[0082] Step 1: Discretize the patient's position in space and construct a graph structure:

[0083] The patient's body center of gravity position data is collected. This data is usually obtained through a depth camera, sensor, or motion capture system, and the coordinates (X, Y) of the patient's body center of gravity in space are recorded in each time period. These continuous body center of gravity positions are discretized, and the space is divided into several small areas (such as grid division), so that each grid represents a state. This process can be implemented using a simple discretization function. Assume that the space is divided into grids, and the coordinates (i, j) of each grid represent a state. In this way, a graph structure is obtained, in which the nodes of the graph represent the discretized body center of gravity positions, and the edges represent the transitions between adjacent states.

[0084] Step 2: Map the patient's initial trajectory within the preset time period into horizontal and vertical axis offsets:

[0085] After discretization, the patient's body center of gravity position in each time period can be obtained by calculating the difference between the current position and the previous position to obtain the offset. Specifically, assuming that at time t, the patient's body center of gravity position is (X t , Yt ), and at time t-1, the patient's body center of gravity is (X t-1 , Y t-1 ), then the offset in the horizontal direction is ΔX=X t -X t-1 , the offset in the vertical direction is ΔY=Y t -Y t-1 .

[0086] If the patient's center of gravity is at (3, 4) at time t and at (3, 3) at time t-1, then the lateral displacement ΔX = 3-3 = 0 and the longitudinal displacement ΔY = 4-3 = 1. This indicates that the patient has only shifted in the longitudinal direction during the period t.

[0087] Step 3: Establish a hidden Markov model based on the predefined initial state distribution, state transition probability matrix and probability density vector:

[0088] In a hidden Markov model, we first need to define an initial state distribution. This distribution represents the possible positions of the patient's center of gravity at the initial time (usually t = 0). Next, we construct a state transition probability matrix, which represents the probability of transitioning from one body position to another. Finally, we define a probability density vector, which represents the distribution of the patient's horizontal and vertical displacements given a given state.

[0089] In this process, it is assumed that statistical information is extracted from the acquired body position data to estimate the initial state distribution, the state transition probability matrix, and the probability density of the offset. For example, if the probability of the patient moving from position (3, 4) to (3, 5) is high, the state transition matrix will reflect this.

[0090] Step 4: Input the horizontal offset and vertical offset into the hidden Markov model and output the action observation:

[0091] The horizontal and vertical offsets for each time period are input as observations into the Hidden Markov Model. Based on the input offsets and current state information, the Hidden Markov Model outputs the model's state prediction for each time period, i.e., the observation (posture state), using the state transition probability matrix and the observation probability density vector.

[0092] If the patient's lateral offset at a certain moment is 1 and the longitudinal offset is -1, based on the model's state transition probability matrix and observation probability density vector, the hidden Markov model will predict that in the next time period, the patient's body center of gravity may shift to a new position (for example, (3, 5) or (4, 4)) and provide the probability of this state.

[0093] Step 5: Based on the motion observations, use the Viterbi decoding algorithm to back-generate the trajectory of the patient's center of gravity position changes:

[0094] The Viterbi decoding algorithm is used to backtrack the most likely sequence of hidden states (i.e., the trajectory of the center of gravity changes) given a given observation sequence (offset sequence). The algorithm selects the most likely path of center of gravity changes by maximizing the probability of state transitions and the probability of observations. Specifically, the Viterbi algorithm calculates the optimal path for each possible state and, through backtracking steps, outputs the trajectory of the patient's center of gravity changes over a preset time period.

[0095] The Viterbi decoding algorithm is used to back-generate the trajectory of the patient's center of gravity position change, including:

[0096] Based on the patient's posture images while performing limb movements, joint position information or motion features are obtained at each moment of movement. An observation sequence is constructed based on these observations. Each observation corresponds to the patient's body position or joint position data at a different time point. Joint coordinate information is typically extracted using a deep learning model (such as OpenPose or other human pose estimation algorithms).

[0097] A hidden state model is designed to represent the trajectory of the patient's center of gravity. Assume that at each moment, the patient's center of gravity may be in different states, such as sitting, standing, and bending over. Each state corresponds to a specific center of gravity position. A state transition probability matrix is ​​constructed to describe the transition patterns between different body states, such as the transition probability from standing to bending over.

[0098] A dynamic time window is used to capture the changes in movements between different time points, and the Viterbi algorithm is used for optimization. Given an observation sequence and transition probability, the optimal body center of gravity position sequence is determined by maximizing the posterior probability of the state at each moment.

[0099] The Viterbi algorithm calculates the optimal path at each time step based on the current observation data and, through backtracking, infers the most likely center of gravity state at each moment, ultimately generating a complete trajectory of body center of gravity position changes. To further improve algorithm accuracy, the state transition probabilities and observation probabilities can be adjusted based on individual patient characteristics (such as weight and height) to better fit the patient's actual movement pattern. Once generated, the center of gravity trajectory can be used to assess the patient's center of gravity stability and balance during movement, as well as provide data support for rehabilitation treatment.

[0100] The changing trajectory of the center of gravity position of the body is mapped to the three-dimensional virtual space; the interaction between the patient's movements and the environment is performed in the three-dimensional virtual space, and the patient's movement balance state when avoiding obstacles is detected in real time.

[0101] It should be noted that, to facilitate understanding of the above-mentioned technical solutions of the present invention, the following describes in detail the actual process of mapping the trajectory of the center of gravity position of the body into a three-dimensional virtual space; performing the interaction between the patient's movements and the environment in the three-dimensional virtual space; and detecting the patient's movement balance state in real time when avoiding obstacles:

[0102] Step 1: Mapping to 3D virtual space:

[0103] Use a 3D modeling tool (such as Unity 3D or Unreal Engine) to create a virtual environment and design a virtual scene. The virtual environment can include a room for patient training, obstacles (such as tables, chairs, walls, etc.), and other interactive elements (such as buttons, target objects, etc.).

[0104] The calculated center of gravity position (X, Y, Z coordinates) is connected to the coordinate system in the virtual space. To do this, the collected 3D coordinate data can be mapped to the corresponding coordinate system in the virtual space (such as the world coordinate system in Unity). By transferring the center of gravity position at each time step to the virtual space, the trajectory of the patient's center of gravity changes can be visualized in the virtual environment.

[0105] Step 2: Patient’s movements interact with the environment:

[0106] The deep learning model identifies the patient's movements and maps them to corresponding actions in the virtual space. For example, when a patient is doing gait training, the model identifies whether the patient is taking a step, turning, or other movement, and updates their movements in real time in the virtual environment.

[0107] Objects in the virtual environment can be set as obstacles (such as chairs and walls) or target objects (such as balls and signs), and the patient's center of gravity can collide or interact with these objects. If the patient's movement causes the center of gravity to exceed the set range, the virtual environment can simulate actual feedback, such as obstacle collision, to prompt the patient to avoid the obstacle.

[0108] Step 3: Real-time detection of the patient's balance state during obstacle avoidance:

[0109] By tracking the changes in the patient's center of gravity in real time, the patient's balance status is assessed using their trajectory. If the center of gravity deviates from its previously stable range, the patient may be losing balance. To monitor balance in real time, a threshold balance zone is set. When the center of gravity crosses this zone, the patient is considered to be losing balance.

[0110] The system monitors the patient's balance in real time in the virtual space and provides feedback. When the patient's body center of gravity is detected to be unstable or unbalanced, a real-time prompt in the virtual environment is triggered.

[0111] S2. Collect the patient's vital sign monitoring data, construct a vital sign trend chart based on the vital sign monitoring data, and analyze the patient's vital sign intensity from the vital sign trend chart.

[0112] It should be noted that, in order to facilitate understanding of the above technical solutions of the present invention, the following is a detailed description of the present invention's actual process of collecting a patient's vital sign monitoring data, constructing a vital sign trend chart based on the vital sign monitoring data, and analyzing the patient's vital sign intensity from the vital sign trend chart:

[0113] Step 1: Collect the patient's vital signs monitoring data:

[0114] First, determine the types of vital signs that need to be monitored. Common vital signs include: heart rate (HR), blood pressure (BP, including systolic and diastolic pressure), blood oxygen saturation (SpO2), respiratory rate (RR), body temperature (Temp), etc.

[0115] Choose the appropriate monitoring device based on the type of vital signs you choose. Common devices include:

[0116] Heart rate and blood oxygen saturation can be monitored by a pulse oximeter (such as a non-invasive handheld pulse oximeter) or a smart bracelet (such as a Fitbit).

[0117] Blood pressure can be obtained using an automatic blood pressure monitor (such as an Omron electronic blood pressure monitor).

[0118] Respiratory rate and body temperature can be measured using a smart thermometer or respiratory monitoring device.

[0119] Set the time interval for data collection, usually recording vital signs data at a certain interval (such as every minute or every five minutes) to ensure that the monitoring data has sufficient time sequence and continuity.

[0120] Step 2: Build a vital signs trend chart:

[0121] The collected raw vital signs data are sorted and cleaned to ensure that there are no missing values ​​or abnormal values ​​(such as unreasonably high heart rate and blood pressure values ​​outside the normal range). If missing values ​​are found, interpolation methods (such as linear interpolation) can be used to fill them.

[0122] Visualize the organized data as a trend chart. Use common data visualization tools, such as Matplotlib (Python) or Excel, to plot a trend chart based on time (horizontal axis) and each vital sign data (vertical axis). Use a different color and style for each vital sign to make it easier to distinguish.

[0123] Plot vital signs data such as heart rate, blood pressure, blood oxygen saturation, and body temperature on different subgraphs or the same chart within the same time period to intuitively observe the changing trends of each vital sign.

[0124] Step 3: Analyze the vital signs trend chart:

[0125] Analyze the changing patterns of each vital sign based on the vital sign trend chart. Determine if there are any unusual fluctuations or trends. For example, a sudden increase or decrease in heart rate may indicate that the patient is experiencing some kind of stress or tension, while fluctuations in blood oxygen saturation may indicate respiratory issues.

[0126] By calculating the fluctuation amplitude of each vital sign within a certain time window, the intensity of the vital sign can be determined. For example, calculating the standard deviation of heart rate or the range of variation in blood oxygen saturation can reflect the intensity of the patient's vital sign fluctuations.

[0127] Set thresholds to monitor for abnormal vital signs. For example, if the blood oxygen saturation falls below 90% or the heart rate rises above 120 bpm, an alarm system can be triggered to alert caregivers or doctors.

[0128] Step 4: Vital Signs Strength Assessment and Decision Support:

[0129] Establish a scoring system: A scoring system is established based on the fluctuation range or standard deviation of the intensity of vital signs. For example, if the standard deviation of heart rate is less than 3 bpm, the score is 1; if the blood oxygen fluctuation range is less than 3%, the score is 1; if the body temperature fluctuation is less than 0.5°C, the score is 1. A comprehensive score is used to assess the intensity of the patient's vital signs.

[0130] Develop an intervention strategy: Based on the assessment results, determine whether intervention measures are necessary. If the score is high, it indicates that the patient's vital signs are fluctuating significantly and immediate medical intervention may be required, such as intensive monitoring or adjustment of treatment plans.

[0131] S3. Collecting a urine sample image of the patient, extracting urine sediment features from the patient's urine sample image, and predicting the patient's urine status in a future time period based on the urine sediment features.

[0132] The steps of collecting a patient's urine sample image, extracting urine sediment features from the patient's urine sample image, and predicting the patient's urine status in a future time period based on the urine sediment features include:

[0133] A urine sample image of a patient is collected, and image enhancement processing is performed on the collected urine sample image to obtain an optimized urine sample image.

[0134] It should be noted that the image of the urine sample is captured using a camera, a microscope camera or other medical imaging equipment (such as a digital microscope).

[0135] Obtain urine samples from patients and process them using clear microscope slides and cover slips, ensuring the samples are clean and free from contamination.

[0136] Depending on the nature of the urine sample (e.g., crystals, sediment, or cells in the urine), set the appropriate focus and exposure time to ensure that details in the image are clearly visible. If a microscope is used, an appropriate magnification is usually selected to capture the tiny particles and components in the urine.

[0137] Most urine sample image analyses rely on shape and structure rather than color information. Therefore, it is first necessary to convert the color image into a grayscale image. Grayscale conversion helps reduce computational complexity and highlight the texture and morphology of the image.

[0138] By replacing the value of each pixel with the median value of its surrounding neighborhood pixels, salt and pepper noise can be effectively removed.

[0139] Image brightness or contrast may be uneven. Normalization helps standardize the image brightness range, making the images more comparable. This step improves the effectiveness of subsequent image enhancement. Urine sediment features are extracted and optimized in the urine sample image, and the region containing the urine sediment is segmented to produce a sediment segmentation image.

[0140] The extraction and optimization of urine sediment features in the urine sample image and segmentation of the area to which the urine sediment belongs to obtain a sediment segmentation image include:

[0141] Calculating and optimizing the color space components of pixels in the urine sample image, comparing the color space components with a preset threshold, and selecting pixels corresponding to color space components greater than the preset threshold as urine pixels;

[0142] A normal distribution function of urine pixels in color space is constructed, the color space features of urine are identified based on the normal distribution function, and the urine sediment features are extracted from the color space features.

[0143] It should be noted that the color space components of the pixels in the urine sample image are calculated and optimized, and urine pixels are selected based on the comparison between the color space components and the preset threshold. Finally, the normal distribution function of the urine pixels in the color space is constructed to identify the color space features of urine and extract the features of urine sediment. The following detailed steps can be followed:

[0144] Step 1: Calculate and optimize the color space components of the pixels in the urine sample image:

[0145] Selecting color space: Commonly used color spaces include RGB (Red, Green, Blue), HSV (Hue, Saturation, Value), Lab (Lightness, Chroma a, Chroma b), etc. To extract the color features of urine, HSV or Lab color space is usually chosen because they better reflect the hue, saturation, and brightness features.

[0146] Color space conversion: Convert the optimized urine sample image from RGB color space to HSV color space. In the HSV color space, hue (H) reflects the color type, saturation (S) reflects the vividness of the color, and value (V) reflects the brightness of the color.

[0147] Assumed conversion results:

[0148] The HSV components of a pixel in the image are (H: 30°, S: 80%, V: 70%).

[0149] Calculate the color space components of each pixel: For each pixel in the image, extract its H, S, V values or L, a, b values (according to the selection of color space), and construct the color space components of the pixel.

[0150] The color space conversion is performed on the optimized urine sample image. For a pixel of a urine sample in the image, the obtained HSV values are:

[0151] H: 35°, S: 85%, V: 75% (this is the hue, saturation, and brightness of the urine feature).

[0152] Step two, compare the color space components with the preset threshold, and select the color space components greater than the preset threshold as the urine pixel points:

[0153] According to the experiment and the actual characteristics of the urine sample, set appropriate threshold values. For example, you can set the threshold range through prior knowledge, and select the pixel points with hue H, saturation S, and brightness V within a certain range as urine pixel points.

[0154] Compare the color space components of each pixel with the set threshold value. If the color space components (H, S, V) of the pixel meet the preset range, the pixel point is selected as a urine pixel point.

[0155] For a pixel point in the image, the obtained HSV values are (H: 35°, S: 85%, V: 75%). Since the color space components of the pixel fall within the set threshold range, the pixel point is selected as a urine pixel point.

[0156] Step three, construct the normal distribution function of the urine pixel points in the color space:

[0157] Urine pixels: Extract qualified urine pixels (i.e., pixels that meet the color space threshold) from the entire image.

[0158] Calculate statistical features: For each selected urine pixel, calculate the mean and standard deviation of its color space components. The mean and standard deviation of each color space component (such as H, S, and V) can be calculated to determine the distribution of urine color in color space.

[0159] Construct a normal distribution model: Based on the calculated mean and standard deviation, assume that the color space components of these urine pixels obey a normal distribution. For each component (H, S, V), a normal distribution function can be constructed separately:

[0160] H~N(35°, 5°);

[0161] S~N (80%, 7%);

[0162] V~N (70%, 8%);

[0163] For example, 1,000 urine pixels that meet the criteria were extracted from the optimized urine sample image. The calculation results show:

[0164] The distribution of hues is: mean 35°, standard deviation 5°;

[0165] The distribution of saturation is: mean 80%, standard deviation 7%;

[0166] The distribution of brightness is: mean 70%, standard deviation 8%;

[0167] Therefore, hue H, saturation S, and brightness V can be modeled as normal distributions, respectively.

[0168] Step 4: Identify the color space characteristics of urine:

[0169] Color space feature extraction: Based on the constructed normal distribution function, the color space features of urine can be further analyzed. For example, a shift in hue may indicate a change in urine color (such as a darker or lighter yellow), while changes in saturation and brightness may reflect urine concentration or health status.

[0170] Feature recognition: The normal distribution model is used to classify new urine samples, and the color space features of urine are identified by calculating the distance between the color space components (H, S, V) of each pixel and the normal distribution.

[0171] Step 5: Extract urine sediment features from color space features:

[0172] For example, sediment feature extraction: Urine sediment often presents special color or texture characteristics. After identifying the color space features of urine, further image processing techniques such as edge detection, texture analysis, etc. can be used to extract the features of the sediment.

[0173] Sediment region positioning: By analyzing the urine pixels in the image, combined with color space features, the sediment region can be located. Sediment usually shows a unique shape and texture, which may appear as a more dense pixel area in the image.

[0174] By extracting the color space features of the urine sample, it is found that the hue of the sediment region in the image is 38°, the saturation is 85%, and the brightness is 72%. This region matches the pre-set sediment features, confirming that this region is the sediment region.

[0175] where the expression of the normal distribution function is:

[0176]

[0177] In the formula, f represents the normal distribution function; C b represents the blue difference component; C r represents the red difference component; D represents the reference point; F represents the pixel point; E represents the spatial distance between the reference point and the pixel point.

[0178] Segment the region to which the optimized urine sediment feature in the urine sample image belongs to obtain a sediment segmentation image, and remove noise points in the sediment segmentation image using a similarity distance method.

[0179] wherein removing noise points in the sediment segmentation image using the similarity distance method comprises:

[0180] Scan the pixel points in the sediment segmentation image one by one, and select the pixel points that meet the pre-set conditions as the reference point D, and calculate the component value of the current reference point in the HSV space;

[0181] Scan the next pixel point F, and calculate the component value of the current pixel point in the HSV space according to the pre-set condition.

[0182] It should be noted that the pre-set condition is usually a standard for defining color similarity, image features, etc., which may be determined based on the following factors:

[0183] Selection of color space: In the HSV space, H represents hue, S represents saturation, and V represents brightness. These components can be used to describe color and image brightness, so the pre-set condition usually defines which HSV components or their combinations have an important impact on a specific target object (such as sediment).

[0184] Image characteristics: Select appropriate HSV components based on the sediment's color, shape, texture, and other characteristics. For example, if the target object has a relatively single color, then the H (hue) preset condition may be more important.

[0185] The spatial distance E between the reference point D and the pixel point F is calculated and compared with the preset threshold. If the spatial distance is less than or equal to the preset threshold, it indicates that the colors between the reference point and the pixel point are similar and are retained as regional points of the sediment segmentation image.

[0186] It should be noted that the preset threshold is typically used to determine whether pixels belong to the same category (e.g., sediment area). For example, the color range of different sediments may vary, so the threshold should be determined based on the actual distribution of sediment colors. For example, if the hue of the target object varies slightly, the threshold in the hue dimension can be set smaller; if the color variation is large, the threshold should be larger.

[0187] The pixel scanning process is repeated with pixel F as the reference point until all the pixels in the sediment segmentation image are scanned, thereby removing noise in the sediment segmentation image.

[0188] The principal component analysis technique is used to extract sediment analysis data from the sediment segmentation image, and a training dataset is constructed based on the sediment analysis data;

[0189] The patient's historical urine characteristics are obtained to build a prediction model, and the prediction model is trained using a training data set. The trained prediction model is used to predict the patient's urine status in the future time period.

[0190] It should be noted that the patient's urine sample image is collected and image preprocessing, such as image enhancement, is performed to improve image quality. Image processing technology is used to segment the sediment from the optimized urine sample image, and the sediment features in the urine are identified and extracted. Principal component analysis (PCA) technology is applied to extract sediment analysis data from the sediment segmentation image, and the extracted data is used to construct a training data set. On this basis, the patient's historical urine feature data, including urine composition, concentration and other related information, is collected, and the historical data is integrated with the extracted sediment analysis data to construct a complete training data set. This training data set is used to build a prediction model, and the model is trained using common machine learning algorithms (such as regression analysis, support vector machines or neural networks). After training, the prediction model is used to predict the patient's urine status in the future time period, so as to detect potential urine changes or related health problems in advance, providing strong support for the postoperative status assessment of critically ill spinal patients.

[0191] S4. Adaptively assign weights to the patient's movement balance, vital sign strength, and urine status, and use the weighted summation method to evaluate the postoperative status of critically ill spinal patients;

[0192] The patient's motion balance, vital sign strength, and urine status are adaptively assigned weights, and the postoperative status of critically ill spinal patients is evaluated using a weighted summation method, including:

[0193] Evaluation indicators are formulated for the patient's movement balance state, vital sign intensity and urine status, and the bat algorithm is used to adaptively assign weights to each evaluation indicator.

[0194] Among them, the bat algorithm is used to adaptively assign weights to each evaluation indicator, including:

[0195] The evaluation index is used as the bat individual, and the bat algorithm parameters are initialized. The bat algorithm parameters include the number of bat individuals, bat position, bat speed, bat loudness and bat emission rate;

[0196] Calculate the fitness function value corresponding to the bat according to the fitness function, and select the optimal value from the fitness function values;

[0197] Generate a random number. If the random number is greater than the pulse emission rate, a local solution is generated near the optimal value. If the random number is less than the bat loudness, the local solution is used as the new optimal solution, and the bat loudness and bat emission rate are updated.

[0198] The iterative update process of the optimal solution is repeated until the maximum number of iterations is met, and the global optimal solution is output and used as the optimal weight of each evaluation indicator.

[0199] The patient's total score was calculated using a weighted summation method and the total score was input into a predefined evaluation model;

[0200] The evaluation model outputs the postoperative status assessment level of critically ill spinal patients, and intervention measures are formulated according to the postoperative status assessment level of critically ill spinal patients.

[0201] It should be noted that, in order to facilitate understanding of the above technical solutions of the present invention, the following is a detailed description of the present invention in actual practice, which adaptively assigns weights to the patient's movement balance state, vital sign intensity, and urine status, and uses a weighted summation method to evaluate the postoperative status of critically ill spinal patients:

[0202] Step 1: Develop assessment indicators for the patient's motor balance, vital sign strength, and urine status.

[0203] Movement balance status evaluation indicators:

[0204] Balance ability: Assess the patient's stability during basic activities such as standing and walking after surgery. This can be measured using stance time (in seconds) and gait analysis (e.g., gait cycle time).

[0205] Balance score: A standardized scale such as the Berg balance scale is used, with higher scores indicating better balance. Common scores range from 0 to 56.

[0206] Vital signs intensity assessment indicators:

[0207] Heart rate: Measure the patient's heart rate after surgery. A heart rate below 60 beats / min indicates a risk of hypotension, while a high heart rate (>100 beats / min) may indicate postoperative complications.

[0208] Blood oxygen saturation: The normal range is 95%-100%, and levels below 90% may indicate insufficient oxygenation.

[0209] Respiratory rate: The normal respiratory rate for adults is 12-16 times / minute. A respiratory rate that is too fast (>20 times / minute) or too slow (<8 times / minute) requires attention.

[0210] Urine status assessment indicators:

[0211] Urine output: Whether urine output is sufficient. Postoperative patients should have an appropriate amount of urine output. Generally, adults have a daily urine output of 800-2000 ml. Less than 400 ml after surgery may indicate renal insufficiency.

[0212] Urine color and transparency: Healthy urine is usually transparent yellow. If it is turbid or has an abnormal color (such as red, brown, etc.), it may indicate a urinary tract infection or kidney problems.

[0213] Step 2: Use the bat algorithm to adaptively assign weights to evaluation indicators:

[0214] Initialize bat algorithm parameters:

[0215] Number of bat individuals: set to 20 bat individuals.

[0216] Bat Position: The initial position of each individual bat is random, representing the initial weight of the evaluation indicator. The initial weight can be randomly generated between 0 and 1, and the sum of all weights is 1.

[0217] Bat speed: The speed is initially 0 and then adjusted according to the update rule of the optimization process.

[0218] Bat Loudness: Usually set to an initial value of 1, representing the ability of bats to search the solution space.

[0219] Pulse Rate: Set to an initial value of 0.5, it determines the randomness of the bat's search.

[0220] Construct fitness function:

[0221] The fitness function evaluates the quality of each bat position (i.e., weight combination). For each set of weights (e.g., action balance, vital signs, urine status), a total score is calculated using a weighted sum approach, and the fitness is evaluated by the deviation from the actual clinical outcome. The fitness function can be:

[0222] Calculate the fitness function value for each bat, and select the optimal solution based on the fitness value.

[0223] If the random number is greater than the pulse emission rate, the bat will perform a local search around the current optimal solution to generate a new solution.

[0224] If the random number is less than the bat loudness, the bat will generate a new solution, and select a new optimal solution based on the fitness evaluation.

[0225] Repeat the process until the maximum number of iterations (e.g., 500) is reached.

[0226] For example, by using the bat algorithm, the optimal weights for each evaluation indicator are obtained:

[0227] Weight of action balance status: 0.45;

[0228] Weight of vital sign intensity: 0.35;

[0229] Weight of urine status: 0.20;

[0230] Step three, calculate the total score of the patient by weighted sum:

[0231] Weighted sum: Use the previously determined evaluation indicator weights (such as the weights output by the bat algorithm) to perform a weighted sum of the patient's evaluation indicators to obtain the total score.

[0232] Step four, input the total score into the pre-defined evaluation model:

[0233] Evaluation model design: Use machine learning methods such as regression analysis, decision tree, support vector machine (SVM) to establish a postoperative state evaluation model. This model is trained based on the patient's total score and historical data (e.g., postoperative recovery), and outputs the patient's postoperative state level.

[0234] Step five, develop intervention measures:

[0235] Intervention measures for severe spinal patients after surgery can be adjusted according to the different evaluation levels, for example:

[0236] Severe state (score 0-40): Strengthen vital sign monitoring, and if necessary, perform drug intervention to prevent complications.

[0237] Moderate state (score 40-70): strengthen physical therapy and rehabilitation training, maintain stable vital signs, adjust diet and urination management.

[0238] Good condition (score 70-100): resume routine monitoring, continue rehabilitation training, gradually resume normal activities.

[0239] As Figure 2 shown, according to another embodiment of the present application, a postoperative state evaluation system for severe spinal patients is also provided, which comprises a motion analysis module 1, a vital sign analysis module 2, a urine analysis module 3 and a state comprehensive evaluation module 4;

[0240] The motion analysis module 1 is used to collect the motion posture image of the patient, extract the center of gravity change trajectory of the patient from the motion posture image, and judge the motion balance state of the patient based on the center of gravity change trajectory of the patient;

[0241] The vital sign analysis module 2 is used to collect the vital sign monitoring data of the patient, construct a vital sign trend graph based on the vital sign monitoring data, and analyze the vital sign intensity of the patient from the vital sign trend graph;

[0242] The urine analysis module 3 is used to collect the urine sample image of the patient, extract the urine sediment features from the urine sample image of the patient, and predict the urine state of the patient in the future time period based on the urine sediment features;

[0243] The state comprehensive evaluation module 4 is used to adaptively assign weights to the motion balance state, vital sign intensity and urine state of the patient respectively, and evaluate the postoperative state of the severe spinal patient by using the weighted summation method.

[0244] In summary, with the help of the above technical solutions of the present invention, the present invention uses deep learning technology to extract the position of the center of gravity of the body from the action image, which can accurately capture the changes in the center of gravity of the patient during each movement. It also uses the hidden Markov model to analyze the patient's horizontal and vertical axis deviations within a preset time period, reflecting the patient's balance performance during the activity. These trajectories are mapped into a three-dimensional virtual space, and the patient's movement balance state during obstacle avoidance or movement is detected in real time, thereby not only improving the accuracy of the assessment, but also being able to better assess the patient's actual rehabilitation progress and movement obstacle avoidance ability through simulation in a three-dimensional virtual environment. The present invention uses image processing technology to extract urine sediment features, which can obtain optimized urine sample images, thereby improving the accuracy of sediment feature extraction, and by segmenting the sediment area, further accurately locate and extract key sediment features. These features can reflect the health status of the patient's kidneys and urinary system, providing an important basis for postoperative management. At the same time, it predicts the changes in the patient's urine status in the future time period, and can timely monitor the changes in the urine status of the patient in the postoperative stage for early intervention, thereby providing a more accurate and personalized monitoring tool for the postoperative status assessment of critically ill spinal patients. The present invention adaptively assigns weights to the patient's movement balance state, vital sign intensity and urine status, and uses weighted summation to comprehensively evaluate the postoperative status of critically ill spinal patients. This can more comprehensively and accurately reflect the patient's recovery status. Through this multi-dimensional combination method, the one-sidedness brought by a single indicator is avoided, ensuring that potential problems are discovered and intervened in time at different stages of postoperative recovery, thereby helping patients achieve better recovery effects.

[0245] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating the postoperative status of patients with severe spinal disease, characterized in that: The method includes: Collect the patient's posture images when performing limb activities, and use deep learning algorithms to extract the patient's body center of gravity from the posture images; A hidden Markov model is used to analyze the horizontal and vertical axis deviations of the patient's center of gravity position within a preset time period, and a trajectory of the center of gravity position change is generated based on the deviation analysis results. The patient's movement balance state is determined based on the trajectory of the center of gravity change of the patient; Collecting the patient's vital sign monitoring data, constructing a vital sign trend chart based on the vital sign monitoring data, and analyzing the patient's vital sign intensity from the vital sign trend chart; collecting a urine sample image of the patient, extracting urine sediment features from the urine sample image of the patient, and predicting the patient's urine status in a future time period based on the urine sediment features; The bat algorithm is used to adaptively assign weights to the patient's motion balance status, vital sign intensity, and urine status, and the weighted summation method is used to evaluate the postoperative status of critically ill spinal patients.

2. A method for evaluating the postoperative status of critically ill spinal patients according to claim 1, characterized in that: The determining of the patient's movement balance state based on the patient's center of gravity change trajectory includes: Mapping the changing trajectory of the body center of gravity position to the three-dimensional virtual space; The patient's movements and environment are interacted with in a three-dimensional virtual space, and the patient's movement balance state when avoiding obstacles is detected in real time.

3. A method for evaluating the postoperative status of critically ill spinal patients according to claim 2, characterized in that: The method of analyzing the horizontal axis deviation and the vertical axis deviation of the patient's body center of gravity position within a preset time period using a hidden Markov model and generating a change trajectory of the body center of gravity position based on the deviation analysis results includes: Discretize the center of gravity of the patient's body in space and construct a graph structure. Connect the nodes in the graph structure to obtain the patient's initial trajectory within a preset time period. Mapping the initial trajectory of the patient in a preset time period into a horizontal axis offset and a vertical axis offset, and calculating the horizontal offset and the vertical offset of the patient in each time period; A hidden Markov model is established based on the predefined initial state distribution, state transition probability matrix and probability density vector, and the horizontal offset and vertical offset are input into the hidden Markov model to output the action observation; Based on the motion observations, the Viterbi decoding algorithm is used to backtrack and generate the trajectory of the patient's center of gravity position changes.

4. A method for evaluating the postoperative status of critically ill spinal patients according to claim 3, characterized in that: The collecting of the patient's urine sample image, extracting urine sediment features from the patient's urine sample image, and predicting the patient's urine status in a future time period based on the urine sediment features include: collecting a urine sample image of the patient, performing image enhancement processing on the collected urine sample image, and obtaining an optimized urine sample image; Extracting and optimizing urine sediment features in the urine sample image, and segmenting the area to which the urine sediment belongs to obtain a sediment segmentation image; The principal component analysis technique is used to extract sediment analysis data from the sediment segmentation image, and a training dataset is constructed based on the sediment analysis data; The patient's historical urine characteristics are obtained to build a prediction model, and the prediction model is trained using a training data set. The trained prediction model is used to predict the patient's urine status in the future time period.

5. A method for evaluating the postoperative status of critically ill spinal patients according to claim 4, characterized in that: The step of extracting and optimizing urine sediment features in the urine sample image and segmenting the region to which the urine sediment belongs to obtain a sediment segmentation image includes: Calculating and optimizing the color space components of pixels in the urine sample image, comparing the color space components with a preset threshold, and selecting pixels corresponding to color space components greater than the preset threshold as urine pixels; Construct a normal distribution function of urine pixels in color space, identify the color space features of urine based on the normal distribution function, and extract urine sediment features from the color space features; The area containing urine sediment features in the optimized urine sample image is segmented to obtain a sediment segmentation image, and the noise in the sediment segmentation image is removed using the similarity distance method.

6. A method for evaluating the postoperative status of critically ill spinal patients according to claim 5, characterized in that: The method of removing noise points in the sediment segmentation image by using the similarity distance method includes: Scan the pixels in the sediment segmentation image one by one, select the pixels that meet the preset conditions as reference points, and calculate the component value of the current reference point in the HSV space; Scan the next pixel and calculate the component value of the current pixel in the HSV space according to the preset conditions; Calculate the spatial distance between the reference point and the pixel point, and compare the spatial distance with a preset threshold. If the spatial distance is less than or equal to the preset threshold, it indicates that the colors of the reference point and the pixel point are similar, and the reference point and the pixel point are retained as the region point of the sediment segmentation image. The pixel scanning process is repeated with the pixel point as the reference point until all the pixel points in the sediment segmentation image are scanned, so as to remove the noise points in the sediment segmentation image.

7. A method for evaluating the postoperative status of critically ill spinal patients according to claim 6, characterized in that: The bat algorithm is used to adaptively assign weights to the patient's motion balance state, vital sign intensity, and urine state, respectively, including: The evaluation index is used as the bat individual, and the bat algorithm parameters are initialized. The bat algorithm parameters include the number of bat individuals, bat position, bat speed, bat loudness and bat emission rate; Calculate the fitness function value corresponding to the bat according to the fitness function, and select the optimal value from the fitness function values; Generate a random number. If the random number is greater than the pulse emission rate, a local solution is generated near the optimal value. If the random number is less than the bat loudness, the local solution is used as the new optimal solution, and the bat loudness and bat emission rate are updated. The iterative update process of the optimal solution is repeated until the maximum number of iterations is met, and the global optimal solution is output and used as the optimal weight of each evaluation indicator.

8. A method for evaluating the postoperative status of critically ill spinal patients according to claim 7, characterized in that: The method of using weighted summation to evaluate the postoperative status of critically ill spinal patients includes: The patient's total score was calculated using a weighted summation method and the total score was input into a predefined evaluation model; The evaluation model outputs the postoperative status assessment level of critically ill spinal patients, and intervention measures are formulated according to the postoperative status assessment level of critically ill spinal patients.

9. A method for evaluating the postoperative status of critically ill spinal patients according to claim 8, characterized in that: The expression of the normal distribution function is: Where, f represents the normal distribution function; C b represents the blue differential component; C r represents the red differential component; D represents the reference point; F represents the pixel point; and E represents the spatial distance between the reference point and the pixel point.

10. A system for evaluating the postoperative status of patients with severe spinal disease, used to implement the method for evaluating the postoperative status of patients with severe spinal disease according to any one of claims 1 to 9, characterized in that: The system includes a motion analysis module, a vital sign analysis module, a urine analysis module, and a comprehensive status assessment module; A motion analysis module is used to collect the patient's motion posture images, extract the patient's center of gravity change trajectory from the motion posture images, and determine the patient's motion balance state based on the patient's center of gravity change trajectory; A vital sign analysis module is used to collect the patient's vital sign monitoring data, construct a vital sign trend chart based on the vital sign monitoring data, and analyze the patient's vital sign intensity from the vital sign trend chart; A urine analysis module is used to collect urine sample images of patients, extract urine sediment features from the urine sample images of patients, and predict the urine status of patients in the future time period based on the urine sediment features; The comprehensive status assessment module is used to adaptively assign weights to the patient's movement balance status, vital sign intensity and urine status, and use weighted summation to evaluate the postoperative status of critically ill spinal patients.

Images