Intensive care unit image processing method and system based on respiratory movement

Abstract

The invention provides an intensive care unit image processing method and system based on respiratory movement, and belongs to the field of image processing. The problem of low respiratory movement monitoring efficiency of the patient is solved; the method specifically comprises the following steps: S1, acquiring a disease name, a medical examination report and a monitoring video; s2, enhancing the monitoring video, determining periodic breathing information of the patient, and judging whether the breathing frequency of the patient is normal or not; if yes, not processing; if not, an alarm is given immediately, and pathological information is obtained; s3, constructing a lesion index change equation according to the pathological information; predicting the disease change of the patient according to the lesion index change equation to obtain diagnosis information, and feeding back the diagnosis information; s4, continuously updating and analyzing the monitoring video until the patient is healed; by acquiring, analyzing and processing the monitoring video of the patient, whether breathing of the patient is normal or not is judged, real-time alarming is performed, and the burden of medical staff is relieved.

Description

Technical Field

[0001] The invention discloses a respiratory motion-based intensive care unit image processing method and system, and relates to the field of image processing. Background Art

[0002] The existing image processing methods or systems for respiratory motion have the following deficiencies: High requirements for image quality: Existing image processing systems rely on high-quality image data for analysis; if the image quality is poor (such as blur, noise interference, etc.), less image features are provided, which will directly affect the monitoring accuracy and reliability of the system.

[0003] High requirements for operators: Existing image processing systems require professional operators to operate and maintain them; operators need to have certain medical knowledge and image processing technology background to accurately interpret the monitoring results output by the system and formulate corresponding treatment plans.

[0004] High equipment cost: The deployment of existing image processing systems usually requires high-performance image acquisition equipment and computing equipment. The cost of these devices is usually high, which limits the popularity and application of the system in medical institutions. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention aims to provide a respiratory motion-based intensive care unit image processing method and system, aiming to solve the problem of low efficiency in patient respiratory motion monitoring.

[0006] In order to achieve the above object, the present invention is implemented by the following technical scheme: a method for processing images in a ward based on respiratory motion comprises: Step S1: Obtain the patient's disease name and medical examination report, and obtain the patient's monitoring video in the ICU; Step S2: Enhance the monitoring video and mark the chest contour of the patient through background modeling technology; detect the subtle movement of the chest contour of the patient in the monitoring video to determine the patient's respiratory movement; perform frequency domain analysis on the patient's respiratory movement to obtain the periodic information of the patient's breathing and determine whether the patient's respiratory frequency is normal; if so, do not process it; if not, immediately alarm and obtain the medical records of all target hospitals to analyze the patient's pathological conditions; Step S3: extracting the medical record information with the same disease name as the patient as pathological information; constructing a lesion index change equation based on the pathological information; predicting the patient's condition change based on the lesion index change equation, obtaining diagnostic information, and providing feedback; Step S4: Continuously update and analyze the monitoring video according to the lesion condition and diagnosis information until the patient recovers.

[0007] Furthermore, the specific steps of step S2 are as follows: Step S21: defining a processing step for marking the patient's chest contour; Get the resolution of the monitoring video Lb×Hb; Assume that time is t, and take the image at time t in the monitoring video as the monitoring image; Define triple 2-1: (r （x，y） , g （x，y） , b （x，y） ); (r （x，y） , g （x，y） , b （x，y） ) represents the RGB value of the pixel at (x, y) on the monitoring image; r （x，y） , g （x，y） and b （x，y） , respectively represent the color intensity of red, green and blue at the pixel point (x, y) on the monitoring image; the value range of x is: 1~Lb, the value range of y is: 1~Hb; Step S22: taking the monitoring image as image A; The RGB values ​​of all pixels on image A are recorded as (r （1，1） , g （1，1） , b （1，1） )~(r （Lb，Hb） , g （Lb，Hb） , b （Lb，Hb） ); calculate r （1，1） ~r （Lb，Hb） The average value of ar,g （1，1） ~g （Lb，Hb） The average value of ag,b （1，1） ~b （Lb，Hb） The average value ab; The number of Gaussian filters is recorded as N, and the scale of the 1st to Nth Gaussian filters is recorded as d （1） ~d （N） , the standard deviation is denoted as σ （1） ~σ （N） ; Calculate the weights of the 1st to Nth Gaussian filters and get ω （1） ~ω （N） ; Step S23: Transform the monitoring image once to obtain the gray value gr （1，1） ~gr （Lb，Hb） .

[0008] Furthermore, the step S2 further includes: Step S24: Count the number of pixels with gray value 0 and record it as ga （0） ; Similarly, the number of pixels with a grayscale value of 255 is recorded as ga （255） ; Let pj （z） Indicates the frequency of pixels whose gray value is less than or equal to z; ga （i） Represents the number of pixels whose grayscale value is i and whose grayscale value is less than or equal to z; Define calculation formula 2-6: ; Calculate the pixel frequency pj whose gray value is less than or equal to 0 to 255 （0） ~pj （255） ; Step S25: Define calculation formula 2-7: ; Among them, gr （x，y） Indicates (Rr （x，y） , Rg （x，y） , Rb （x，y） ), Indicates that the gray value is less than or equal to gr （x，y） The pixel frequency, xf （x，y） Indicates gr （x，y） The transformation value of Adaptively enhance the contrast of the monitoring image to obtain xf （1，1） ~xf （Lb，Hb） ; xf （1，1） ~xf （Lb，Hb） The corresponding image is taken as the quasi-image; Use the pyrMeanShiftFiltering function in the OpenCV library to align the image for background removal, and then use the adaptivethreshold function to align the image for adaptive threshold segmentation to mark the patient's chest contour; Step S26: Get the current time, recorded as dh hours dm minutes; Extract the patient's monitoring video within dh hour (dm-1) minutes as the target video; take the picture per second in the target video as the target picture; calculate the patient's chest movement information in the past 1 minute; Step S27: Based on the patient's chest transverse expansion and contraction velocity uv in the past 1 minute （1） ~uv （60） and the thoracic longitudinal velocity vv （1） ~vv （60） , analyzing the patient's respiratory movements.

[0009] Furthermore, the specific steps of step S22 are as follows: Step S221: Calculate r （1，1） ~r （Lb，Hb） The mean square error Sr is: ; g（1，1） ~g （Lb，Hb） The mean square error Sg is: ; b （1，1） ~b （Lb，Hb） The mean square error Sb is: ; Step S222: back up image A to obtain a duplicate image; Extract （1） ~d （N） The minimum value dl in the equation 2-2: ; Among them, Fr （k）（x，y） 、Fr （k）（x，y） 、Fr （k）（x，y） , respectively represent the color intensity of red, green and blue at the monitoring image (x, y) after being processed by the k-th Gaussian filter; d （k） and σ （k） , respectively represent the scale and standard deviation of the kth Gaussian filter; Use the 1st to Nth Gaussian filters to process the heavy image and get image A （1） ~Image A （N） ; Step S223: Calculate image A （1） The mean square error of the color intensities of red, green, and blue is sr （1） ,sg （1） and sb （1） ; Image A （1） The RGB values ​​of all pixels are recorded as (r （1）（1，1） , g （1）（1，1） , b （1）（1，1） )~(r （1）（Lb，Hb） , g （1）（Lb，Hb） , b （1）（Lb，Hb） ); calculate r （1）（1，1） ~r （1）（Lb，Hb） The average value of （1） , g （1）（1，1） ~g （1）（Lb，Hb） The average value of ag （1） , b （1）（1，1） ~b （1）（Lb，Hb） The average value of ab （1） .

[0010] Furthermore, the subsequent steps of step S223 are as follows: Step S224: Calculate r （1）（1，1） ~r （1）（Lb，Hb） The mean square error sr （1） : ; g （1）（1，1） ~g （1）（Lb，Hb） The mean square error sg （1） : ; b （1）（1，1） ~b （1）（Lb，Hb） The mean square error sb （1） : ; Step S225: Calculate image A （2） To Image A （N） The mean square error of the color intensities of red, green, and blue is sr （2） ~sr （N） , sg （2） ~sg （N） and sb （2） ~sb （N） ; Let image A （k） The mean square deviation of the color intensity of red, green and blue is sr （k） ,sg （k） and sb （k） ; The weight coefficient of the kth Gaussian filter is zq （k） ; Define calculation formula 2-3: ; Calculate the weight coefficients zq of the 1st to Nth Gaussian filters （1） ~zq （N） ; Step S226: Calculate zq （1） ~zq （N） The sum of is denoted as aq; The weight coefficient of the kth Gaussian filter is denoted as zq （k） , the weight is denoted as ω （1） , define calculation formula 2-4: ω （k） =(zq （k） / aq); Calculate the weights ω of the 1st to Nth Gaussian filters （1） ~ω （N） .

[0011] Furthermore, the specific steps of step S23 are as follows: Step S231: define calculation formula 2-5: ; Among them, Rr （x，y） , Rg （x，y） and Rb （x，y） , respectively represent the color intensity of the pixel point (x, y) on the monitoring image after red, green and blue transformation; Fr（k）（x，y） , Fg （k）（x，y） , Fb （k）（x，y） , respectively represent the color intensity of red, green and blue at the monitoring image (x, y) after being processed by the k-th Gaussian filter; Step S232: Transform the monitoring image once to obtain (Rr （1，1） , Rg （1，1） , Rb （1，1） )~(Rr （Lb，Hb） , Rg （Lb，Hb） , Rb （Lb，Hb） ); Will (Rr （1，1） , Rg （1，1） , Rb （1，1） )~(Rr （Lb，Hb） , Rg （Lb，Hb） , Rb （Lb，Hb） ) is converted into grayscale value, and gr （1，1） ~gr （Lb，Hb） .

[0012] Furthermore, the specific steps of step S26 are as follows: Step S261: Set the picture at the dtth second in the target video as picture Pr （dt） , the (dt+1) second picture is picture Pr （dt+1） ; Step S262: Count the pictures Pr （dt） The number of pixels of the patient's chest contour in st; Put the picture Pr （dt） The pixel points of the patient's chest contour are used as punctuation points; The positions of the 1st to the stth punctuation marks are recorded as (sx （1） ,sy （1） )~(sx （st） ,sy （st） ); In the picture Pr （dt） In the above example, the grayscale values ​​of the 1st to the stth punctuation points are extracted and recorded as Ii (sx （1） ,sy （1） )~Ii(sx （st） ,sy （st） ); Step S263: In the picture Pr （dt+1） Extract the positions corresponding to the 1st to the st punctuation marks (tx （1） ,ty （1） )~(tx （st） ,ty （st） ); Extract (tx （1） ,ty （1） )~(tx（st） ,ty （st） ) is the gray value of the pixel point Io (tx （1） ,ty （1） )~Io(tx （st） ,ty （st） ).

[0013] Furthermore, the subsequent steps of step S263 are as follows: Step S264: Picture Pr （dt） In the example, the position of the rth punctuation mark (sx （r） ,sy （r） ), grayscale Ii (sx （r） ,sy （r） ); Picture Pr （dt+1） In the example, the position of the rth punctuation mark (tx （r） ,ty （r） ), Grayscale Io (tx （r） ,ty （r） ); Define calculation formula 2-8: ; Among them, Ix （r） ,Iy （r） and Iio （r） , respectively represent the horizontal gradient, vertical gradient and brightness gradient of the rth mark; Calculate the lateral gradient Ix from the 1st to the stth mark （1） ~Ix （st） , longitudinal gradient Iy （1） ~Iy （st） , brightness gradient Iio （1） ~Iio （st） ; Step S265: Record the patient's chest transverse expansion and contraction velocity at the dt second as uv （dt） , the longitudinal velocity of the thorax is recorded as vv （dt） ; Define calculation formula 2-9: ; Step S266: Extract the monitoring image of the patient at dh hours (dm-2) minutes and 59 seconds as the reference image; set the value of dt to 1, and calculate the transverse expansion and contraction velocity uv of the patient's chest cavity within dh hours (dm-1) minutes （1） ~uv （60） , chest longitudinal velocity vv （1） ~vv （60） ; The uv （1） ~uv （60） and vv （1） ~vv（60） As chest movement information.

[0014] Furthermore, the specific steps of step S27 are as follows: Step S271: Assume that the patient's chest transverse expansion and contraction speed at the qth second is uv （q） , the longitudinal velocity of the chest cavity is vv （q） ; Get the chest transverse expansion and contraction velocity interval of the first to the hoth healthy person within 1 minute, recorded as [uvl （1，ho） ~uvl （60，ho） , uvm （1，ho） ~uvm （60，ho） ]; the interval of the longitudinal expansion and contraction velocity of the chest cavity, denoted as [vvl （1，ho） ~vvl （60，ho） , vvm （1，ho） ~vvm （60，ho） ]; where ho represents the number of healthy people; Step S272: Assume that the patient's chest transverse expansion and contraction speed at the uth second within 1 minute is uvr （u） , the longitudinal velocity of the chest cavity is vvr （u） , the respiratory frequency is ff; the value range of u is: 1~60; Define the relation S: ; Among them, uvr （s） and vvr （s） , respectively represent the transverse expansion and contraction velocity and longitudinal velocity of the chest cavity at the sth second within 1 minute; ε represents the judgment value; j represents the imaginary unit; Step S273: [uvl （1，ho） ~uvl （60，ho） , uvm （1，ho） ~uvm （60，ho） ] and [vvl （1，ho） ~vvl （60，ho） , vvm （1，ho） ~vvm （60，ho） ] into the relation S, and calculate the breathing frequency range of a healthy person, and we get [fh （1） , fh （ho） ]; The uv （1） ~uv （60） and vv （1） ~vv （60） Substituting into the relation S, calculate the patient's respiratory rate, recorded as hf; Step S274: Determine whether hf is in [fh （1） , fh （ho） ] interval; If it exists, it will not be processed; If not, call the police immediately and analyze the patient's pathological condition.

[0015] A respiratory motion-based ICU image processing system comprises: Image acquisition module: used to obtain the patient's disease name and medical examination report, and obtain the patient's monitoring video in the ICU; Image analysis module: used to enhance the monitoring video and mark the patient's chest contour through background modeling technology; detect the subtle movement of the patient's chest contour in the monitoring video to determine the patient's respiratory movement; perform frequency domain analysis on the patient's respiratory movement to obtain the patient's breathing periodicity information and determine whether the patient's respiratory frequency is normal; if so, no processing will be done; if not, an alarm will be immediately sounded and the medical records of all target hospitals will be obtained to analyze the patient's pathological conditions; Patient monitoring module: used to extract medical record information with the same name as the patient's disease as pathological information; construct a lesion index change equation based on the pathological information; predict the patient's condition change based on the lesion index change equation, obtain diagnostic information, and provide feedback; Continuous monitoring module: used to continuously update and analyze monitoring videos until the patient recovers.

[0016] Compared with the prior art, the present invention has the following beneficial effects: Non-contact monitoring: The present invention adopts a non-contact monitoring method, which does not require direct contact with the patient's body, avoiding the risk of cross-infection that may be caused by traditional monitoring methods; and non-contact monitoring also makes patients more comfortable during the monitoring process, reduces the sense of restraint caused by the monitoring equipment, and is beneficial to the patient's recovery.

[0017] Real-time monitoring and early warning: The present invention can monitor the patient's respiratory status in real time, including key indicators such as respiratory rate and respiratory amplitude; once these indicators exceed the normal range, the system can immediately issue an early warning signal to remind medical staff to take timely measures.

[0018] Data recording and analysis: The present invention can automatically record the patient's respiratory movement data. By analyzing a large amount of data, medical staff can gain an in-depth understanding of the changing patterns of the patient's respiratory status, thus providing a basis for formulating a more scientific and reasonable treatment plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 It is a schematic diagram of the method of the present invention; Figure 2 It is a schematic diagram of the system of the present invention; Figure 3 It is a schematic diagram of a ward of the present invention. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Embodiment 1 See also Figure 1 and Figure 3 , a respiratory motion-based ICU image processing method includes: Step S1: Obtain the patient's disease name and medical examination report, and obtain the patient's monitoring video in the ICU; Step S2: Enhance the monitoring video and mark the patient's chest contour through background modeling technology (such as Gaussian mixture model); detect the subtle movement of the patient's chest contour in the monitoring video to determine the patient's respiratory movement; perform frequency domain analysis on the patient's respiratory movement (using Fourier transform) to obtain the patient's breathing periodicity information and determine whether the patient's respiratory frequency is normal; if so, do not process it (and skip step S3); if not, immediately alarm, obtain the medical records of all target hospitals, and analyze the patient's pathological conditions; It should be noted that the "target hospital" in the present invention means: a municipal hospital that uses the present invention (a respiratory motion-based intensive care unit image processing method and system) to monitor the patient's respiratory motion; The specific steps of step S2 are as follows: Step S21: defining a processing step for marking the patient's chest contour; Get the resolution of the monitoring video, recorded as Lb×Hb; (that is, on the monitoring image, there are Lb pixels horizontally and Hb pixels vertically) Assume that time is t, and take the image at time t in the monitoring video as the monitoring image; Define triple 2-1: (r （x，y） , g （x，y） , b （x，y） ); (numerical representation) (r （x，y） , g （x，y） , b （x，y） ) represents the RGB value of the pixel at (x, y) on the monitoring image; (RGB value has 3 specific values) r （x，y） , g （x，y） and b （x，y）, respectively represent the color intensity of red (channel), green (channel) and blue (channel) of the pixel at (x, y) on the monitoring image; the value range of x is: 1~Lb, the value range of y is: 1~Hb; Step S22: taking the monitoring image as image A; The RGB values ​​of all pixels on image A are recorded as (r （1，1） , g （1，1） , b （1，1） )~(r （Lb，Hb） , g （Lb，Hb） , b （Lb，Hb） ); calculate r （1，1） ~r （Lb，Hb） The average value of ar,g （1，1） ~g （Lb，Hb） The average value of ag,b （1，1） ~b （Lb，Hb） The average value ab; The number of Gaussian filters is recorded as N, and the scale of the 1st to Nth Gaussian filters is recorded as d （1） ~d （N） , the standard deviation is denoted as σ （1） ~σ （N） ; (It should be noted that the value of N is generally 3, and the scale of the Gaussian filter represents the window size or kernel size of the Gaussian filter (common scales of Gaussian filters are: 3×3, 5×5, 7×7, etc.); users or relevant technical personnel can adjust the number and scale of Gaussian filters according to actual needs) Calculate the weights of the 1st to Nth Gaussian filters and get ω （1） ~ω （N） ; Step S221: Calculate r （1，1） ~r （Lb，Hb） The mean square error Sr is: ; g （1，1） ~g （Lb，Hb） The mean square error Sg is: ; b （1，1） ~b （Lb，Hb） The mean square error Sb is: ; Step S222: back up N copies of image A to obtain (N copies) of duplicate images; Extract （1） ~d （N） The minimum value in is denoted as dl; define calculation formula 2-2: ; Among them, Fr （k）（x，y） , Fg （k）（x，y） , Fb（k）（x，y） , respectively represent the color intensity of red (channel), green (channel) and blue (channel) at the monitoring image (x, y) after being processed by the k-th Gaussian filter; d （k） and σ （k） , respectively represent the scale and standard deviation of the kth Gaussian filter, and the value range of k is: 1~N; (According to formula 2-2) Use the 1st to Nth Gaussian filters to process the heavy image and obtain image A （1） ~Image A （N） ; Step S223: Calculate image A （1） The mean square error of the color intensities of red (channel), green (channel), and blue (channel) is obtained by （1） ,sg （1） and sb （1） ; Image A （1） The RGB values ​​of all pixels are recorded as (r （1）（1，1） , g （1）（1，1） , b （1）（1，1） )~(r （1）（Lb，Hb） , g （1）（Lb，Hb） , b （1）（Lb，Hb） ); calculate r （1）（1，1） ~r （1）（Lb，Hb） The average value of （1） , g （1）（1，1） ~g （1）（Lb，Hb） The average value of ag （1） , b （1）（1，1） ~b （1）（Lb，Hb） The average value of ab （1） ; Calculate r （1）（1，1） ~r （1）（Lb，Hb） The mean square error sr （1） : ; g （1）（1，1） ~g （1）（Lb，Hb） The mean square error sg （1） : ; b （1）（1，1） ~b （1）（Lb，Hb） The mean square error sb （1） : ; Step S224: Repeat the calculation of sr （1） ,sg （1） and sb （1） Steps to calculate image A （2） To Image A （N） The mean square error of the color intensities of red (channel), green (channel), and blue (channel) is obtained by（2） ~sr （N） , sg （2） ~sg （N） and sb （2） ~sb （N） ; Let image A （k） The mean square error of the color intensity of red (channel), green (channel) and blue (channel) is sr （k） ,sg （k） and sb （k） ; The weight coefficient of the kth Gaussian filter is zq （k） ; Define calculation formula 2-3: ; (According to formula 2-3) Calculate the weight coefficients zq of the 1st to Nth Gaussian filters （1） ~zq （N） ; Step S225: Calculate zq （1） ~zq （N） The sum of is denoted as aq; The weight coefficient of the kth Gaussian filter is denoted as zq （k） , the weight is denoted as ω （1） , define calculation formula 2-4: ω （k） =(zq （k） / aq); (According to formula 2-4) Calculate the weights ω of the 1st to Nth Gaussian filters （1） ~ω （N） ; Step S23: Define calculation formula 2-5: ; Among them, Rr （x，y） , Rg （x，y） and Rb （x，y） , respectively represent the color intensity of the pixel point red (channel), green (channel) and blue (channel) after transformation at (x, y) on the monitoring image; Fr （k）（x，y） 、Fr （k）（x，y） 、Fr （k）（x，y） , respectively represent the color intensity of red (channel), green (channel) and blue (channel) at the monitoring image (x, y) after being processed by the k-th Gaussian filter; According to formula 2-5, the monitoring image is transformed once to obtain (Rr （1，1） , Rg （1，1） , Rb （1，1） )~(Rr （Lb，Hb） , Rg （Lb，Hb） , Rb （Lb，Hb） ); (Using weighted average algorithm) (Rr （1，1） , Rg （1，1） , Rb （1，1） )~(Rr （Lb，Hb） , Rg （Lb，Hb） , Rb （Lb，Hb） ) is converted into grayscale value (grayscale value range: 0 ~ 255), and gr （1，1） ~gr （Lb，Hb） ; Step S24: Count the number of pixels with gray value 0 and record it as ga （0） ; The number of pixels with a gray value of 1 is denoted as ga （1） ; By analogy, the number of pixels with a grayscale value of 255 is recorded as ga （255） ; Let pj （z） Indicates the frequency of pixels whose grayscale value is less than or equal to z. The value range of z is: 0 to 255; ga （i） Indicates the number of pixels whose grayscale value is i and whose grayscale value is less than or equal to z. The value range of i is: 0~z; Define calculation formula 2-6: ; (According to formula 2-6) calculate the frequency pj of pixels with grayscale values ​​less than or equal to 0 to 255 （0） ~pj （255） ; Step S25: Define calculation formula 2-7: ; Among them, gr （x，y） Indicates (Rr （x，y） , Rg （x，y） , Rb （x，y） ), Indicates that the gray value is less than or equal to gr （x，y） The pixel frequency, xf （x，y） Indicates gr （x，y） The transformation value of Based on gr （1，1） ~gr （Lb，Hb） , (according to formula 2-7) adaptively enhance the contrast of the monitoring image and obtain xf （1，1） ~xf （Lb，Hb） ; xf （1，1） ~xf （Lb，Hb） The corresponding image is taken as the quasi-image; Use the pyrMeanShiftFiltering function in the OpenCV library to align the image for background removal, and then use the adaptivethreshold function to align the image for adaptive threshold segmentation (obtain the patient's body contour; identify the patient's body contour,) and mark the patient's chest contour; Step S26: Get the current time, recorded as dh hours dm minutes; (From the complete monitoring video) extract the patient's monitoring video within dh hours (dm-1) minutes as the target video; take the picture per second in the target video as the target picture; (the number of target pictures is 60) calculate the patient's chest movement information in the past 1 minute; Step S261: Set the target picture at the dtth second in the target video (after processing in steps S21 to S25) as picture Pr （dt） , the target image at (dt+1) seconds is image Pr （dt+1） ; Step S262: Mark the image Pr （dt） The chest contour of the patient was counted (Figure Pr （dt） The number of chest contour pixels in the patient) is denoted as st; Put the picture Pr （dt） The pixel points of the patient's chest contour are used as punctuation points; The positions of the 1st to the stth punctuation marks are recorded as (sx （1） ,sy （1） )~(sx （st） ,sy （st） ); In the picture Pr （dt） Extract the 1st to the stth punctuation marks (i.e. (sx （1） ,sy （1） )~(sx （st） ,sy （st） The gray value of the pixel at ) is recorded as Ii (sx （1） ,sy （1） )~Ii(sx （st） ,sy （st） ); Step S263: (In the picture Pr （dt） (middle) will (sx （1） ,sy （1） )~(sx （st） ,sy （st） ) as punctuation points; (the number of punctuation points is st) In the picture Pr （dt+1） In the above example, extract the positions corresponding to the 1st to the st punctuation marks and record them as (tx （1） ,ty （1） )~(tx（st） ,ty （st） ); (In the picture Pr （dt+1） Extraction (tx （1） ,ty （1） )~(tx （st） ,ty （st） ) is the gray value of the pixel at tx （1） ,ty （1） )~Io(tx （st） ,ty （st） ); Step S264: The picture Pr （dt） In the example, the position of the rth punctuation mark is denoted by (sx （r） ,sy （r） ), grayscale is recorded as Ii (sx （r） ,sy （r） ); Put the picture Pr （dt+1） In the example, the position of the rth punctuation mark is denoted by (tx （r） ,ty （r） ), grayscale is recorded as Io (tx （r） ,ty （r） ); the value range of r is: 1~st; Define calculation formula 2-8: ; Among them, Ix （r） ,Iy （r） and Iio （r） , respectively represent the horizontal (position) gradient, vertical (position) gradient and brightness gradient of the rth mark; (According to formula 2-8) calculate the horizontal (position) gradient Ix from the 1st to the stth mark （1） ~Ix （st） , longitudinal (position) gradient Iy （1） ~Iy （st） , brightness gradient Iio （1） ~Iio （st） ; Step S265: Record the patient's chest transverse expansion and contraction speed (i.e., chest expansion and chest contraction) at the dt second as uv （dt） , the longitudinal velocity of the thorax is recorded as vv （dt） ; Define calculation formula 2-9: ; Step S266: (in the complete monitoring video) extract the monitoring image of the patient at dh hours (dm-2) minutes and 59 seconds as the reference image; set the value of dt (in step S261) to 1, and calculate the transverse expansion and contraction velocity uv of the patient's chest cavity within dh hours (dm-1) minutes （1） ~uv （60） , chest longitudinal velocity vv （1） ~vv （60） ; The uv （1） ~uv （60） and vv （1） ~vv （60） As chest movement information; Step S27: Based on the patient's chest transverse expansion and contraction velocity uv in the past 1 minute （1） ~uv （60） and the thoracic longitudinal velocity vv （1） ~vv （60） , analyze the patient’s respiratory movements; Step S271: Assume that the patient's chest transverse expansion and contraction speed at the qth second is uv （q） , the longitudinal velocity of the chest cavity is vv （q） ; Get the chest transverse expansion and contraction velocity interval of the first to the hoth healthy person within 1 minute, recorded as [uvl （1，ho） ~uvl （60，ho） , uvm （1，ho） ~uvm （60，ho） ]; the interval of the longitudinal expansion and contraction velocity of the chest cavity, denoted as [vvl （1，ho） ~vvl （60，ho） , vvm （1，ho） ~vvm （60，ho） ]; where ho represents the number of healthy people; Step S272: Assume that the transverse expansion and contraction speed of the chest cavity of the patient (or healthy person) at the uth second within 1 minute is uvr （u） , the longitudinal velocity of the chest cavity is vvr （u） , the respiratory frequency is ff; the value range of u is: 1~60; Define the relation S: ; Among them, uvr （s） and vvr （s） , respectively represent the transverse expansion and contraction velocity and longitudinal velocity of the chest cavity of the patient (or healthy human body) at the sth second within 1 minute, and the value range of s is: 1~u; ε represents the (error) judgment value; (the value of ε is 0.1, and the user or relevant technical personnel can adjust the value of ε according to actual needs) j represents the imaginary unit, and e represents the base of the natural logarithm; Step S273: [uvl （1，ho） ~uvl （60，ho） , uvm （1，ho） ~uvm （60，ho） ] and [vvl （1，ho） ~vvl （60，ho） , vvm （1，ho） ~vvm （60，ho） ]Substitute into the relation S (uvr （u） and vvr （u） and uvr （s） and vvr （s） ), calculate the breathing frequency range of a healthy person and get [fh （1） , fh （ho） ]; The uv （1） ~uv （60） and vv （1） ~vv （60） Substituting into the relation S (uvr （u） and vvr （u） and uvr （s） and vvr （s） ), calculate the patient's respiratory rate, recorded as hf; Step S274: Determine whether hf is in [fh （1） , fh （ho） ] interval; If yes, it means the patient is breathing normally and no treatment is required (and step S3 is skipped); If not, an alarm is immediately sounded, and the medical records of all target hospitals are obtained, the patient's pathological condition is analyzed, and the process proceeds to step S3.

[0022] Step S3: extract the medical record information with the same name as the patient's disease name as pathological information; construct a lesion index change equation based on the pathological information; predict the patient's condition change based on the lesion index change equation, obtain diagnostic information, and feedback; the specific steps of step S3 are as follows: Step S31: extracting the medical record information with the same disease name as the patient from all the medical record information of the target hospital as pathological information; counting the number of patients in the pathological information, recorded as mu; The first patient in the pathological information is taken as patient C; the treatment time of patient C is obtained and recorded as td (unit: day); The pathogen content of patient C from day 1 to day td is recorded as: sy （1） ,sy （2） ~sy （td） ; Immune cell content is recorded as: im （1）、im （2） ~im （td） ; Non-immune cell content is recorded as: ni （1） 、ni （2） ~ni （td） ; The content of liver and kidney secretion is recorded as: lk （1） lk （2） ~lk （td） ; Obtain the content of disease markers in healthy people as syy, the content of immune cells as imm, the content of non-immune cells as nii, the content of liver and kidney secretions as lkk, and construct the pathological equation SS （1） ; Step S311: Analyze the changes in the lesion index of patient C from the first to the second day, and calculate the Kalman gain Kdk （1-2） ; Step S3111: Construct the lesion index matrix of patient C on the first day, denoted as B （1） , the lesion index matrix on the second day is recorded as B （2） ; Matrix B （1） The mathematical representation of is: ; Matrix B （2） The mathematical representation of is: ; Step S3112: Calculate matrix B （1） Becomes matrix B （2） The state transfer matrix of （1-2） ; Matrix B （1-2） The calculation formula is: ; Where × represents matrix multiplication, T represents the transpose of the matrix, and -1 represents the inverse of the matrix; Step S3113: Calculate matrix B （1） The difference matrix of （1） ; Matrix RB （1） The calculation formula is: ; Where - represents matrix subtraction; Calculate the matrix B （2） The difference matrix of （2） ; Matrix RB （2） The calculation formula is: ; Calculate the matrix B （1） Becomes matrix B （2） The error transition matrix is ​​denoted as matrix QB （1-2） ; Matrix QB （1-2）The mathematical formula is: QB （1-2） =RB （2） -RB （1） ; Step S3114: Calculate sy （1） 、im （1） 、ni （1） andlk （1） The average value is denoted as ah; Calculate the matrix B （1） The decentralized matrix of （1） ;APd （1） The calculation formula is: APd （1） =B （1） -[ah×(I （4） ) T ]; where - represents matrix subtraction, T represents the transpose of the matrix, and I （4） represents a (4×1) all-1 matrix; Calculate the matrix B （1） The error covariance matrix of （1） ; Matrix Pd （1） The calculation formula is: Pd （1） = (1 / 4) × (APd （1） T ×APd （1） ); Step S3115: Calculate matrix B （2） The pre-error covariance matrix of （2） ;Pk （2） The calculation formula is: ; Among them, × represents matrix multiplication, T represents matrix transpose, and + represents matrix addition; Calculate the Kalman gain Kdk （1-2） ,Kdk （1-2） The calculation formula is: ; Step S312: Repeat the calculation of the Kalman gain Kdk （1-2） The same process as above was used to analyze the changes in the lesion indexes of patient C from the 2nd to the 3rd day to the (td-1) to the td day, and calculate the Kalman gain Kdk （2-3） ~Kdk （（td-1）-td） ; Calculate Kdk （1-2） ~Kdk （（td-1）-td） The average of , denoted as Kkk; Step S313: Assume that the disease marker content corresponding to patient C on day k is sy（k） , the immune cell content is im （k） , non-immune cell content is ni （k） , liver and kidney secretion content is lk （k） ; Assume that the symptom marker content of patient C on day (k+1) is sy （k+1） , the immune cell content is im （k+1） , non-immune cell content is ni （k+1） , liver and kidney secretion content is lk （k+1） ; Construct equation SS （1） The initial mathematical expression is: ; Wherein, QC represents the error matrix of (4×1), and all initial parameters in the matrix QC are 0; Step S314: Set sy （1） ~sy （td） 、im （1） ~im （td） 、ni （1） ~ni （td） andlk （1） ~lk （td） Substitute into equation SS (1,1) and calculate the value of QC using MATLAB software to obtain CC; Substituting CC into equation SS （1） From the initial mathematical expression, we get the equation SS(1,1): ; Process A315: Repeated construction of equation SS （1） The same process as above is used to construct the lesion equations corresponding to the 2nd to muth patients in the pathological information, and the equation SS is obtained. （2） ~ Equation SS （mu） ; Using MATLAB software, equation SS （1） ~ Equation SS （mu） Fitting is performed to obtain the equation for the change of lesion index; Process A32: Obtain the current pathogen content in the patient's medical examination report, recorded as ym, immune cell content as mu, non-immune cell content as nu, and liver and kidney secretion content as li; Substitute ym, mu, nu and li into the lesion index change equation to calculate the patient's lesion index change on the second day as diagnostic information.

[0023] Step S4: Continuously update and analyze the monitoring video until the patient recovers.

[0024] Embodiment 2 See also Figure 2 , a respiratory motion-based ICU image processing system comprises: Image acquisition module: used to obtain the patient's disease name and medical examination report, and obtain the patient's monitoring video in the ICU; Image analysis module: used to enhance the monitoring video and mark the patient's chest contour through background modeling technology (such as Gaussian mixture model); detect the subtle movement of the patient's chest contour in the monitoring video to determine the patient's respiratory movement; perform frequency domain analysis on the patient's respiratory movement (using Fourier transform) to obtain the patient's breathing periodicity information and determine whether the patient's respiratory frequency is normal; if so, no processing is performed (and step S3 is skipped); if not, an alarm is immediately sounded, and the medical records of all target hospitals are obtained to analyze the patient's pathological conditions; Patient monitoring module: used to extract medical record information with the same name as the patient's disease as pathological information; construct a lesion index change equation based on the pathological information; predict the patient's condition change based on the lesion index change equation, obtain diagnostic information, and provide feedback; Continuous monitoring module: used to continuously update and analyze monitoring videos until the patient recovers.

[0025] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technical personnel in this field according to actual conditions. For example, if there are weight coefficients and proportional coefficients, their set sizes are to quantify each parameter to obtain a specific value, which is convenient for subsequent comparison. Regarding the size of the weight coefficient and the proportional coefficient, it is sufficient as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0026] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for processing images in a ward based on respiratory motion, characterized in that: The method comprises: Step S1: Obtaining the patient's monitoring video; Step S2: Enhance the monitoring video and mark the patient's chest contour through background modeling technology; detect the subtle movement of the patient's chest contour in the monitoring video to determine the patient's respiratory movement; perform frequency domain analysis on the patient's respiratory movement to obtain the patient's breathing periodicity information and determine whether the patient's respiratory frequency is normal; if so, do not process it; if not, analyze the patient's pathological condition; Step S3: construct a lesion index change equation; predict the patient's condition change and obtain diagnostic information; Step S4: Continuously update and analyze the monitoring video according to the lesion condition and diagnosis information until the patient recovers.

2. The method for processing images in a ICU based on respiratory motion according to claim 1, characterized in that: The specific steps of step S2 are as follows: Step S21: define a processing step for marking the chest contour of a patient; obtain a resolution of the monitoring video Lb×Hb; and use the image at time t in the monitoring video as a monitoring image; Define triple 2-1: (r （x，y） , g （x，y） , b （x，y） ); (r （x，y） , g （x，y） , b （x，y） ) represents the RGB value of the pixel at (x, y) on the monitoring image; Step S22: The monitoring image is taken as image A; the RGB values ​​of all pixels on image A are recorded as (r （1，1） , g （1，1） , b （1，1） )~(r （Lb，Hb） , g （Lb，Hb） , b （Lb，Hb） ); calculate r （1，1） ~r （Lb，Hb） The average value of ar,g （1，1） ~g （Lb，Hb） The average value of ag,b （1，1） ~b （Lb，Hb） The average value ab; The number of Gaussian filters is recorded as N, and the scale of the 1st to Nth Gaussian filters is recorded as d （1） ~d （N） , the standard deviation is denoted as σ （1） ~σ （N） ; Calculate the weights of the 1st to Nth Gaussian filters and get ω （1） ~ω （N） ; Step S23: Transform the monitoring image once to obtain the gray value gr （1，1） ~gr （Lb，Hb） .

3. A method for processing images in a ICU based on respiratory motion according to claim 2, characterized in that: The step S2 further comprises: Step S24: Count the number of pixels with gray value 0 and record it as ga （0） ; Similarly, the number of pixels with a grayscale value of 255 is recorded as ga （255） ; Let pj （z） Indicates the frequency of pixels whose gray value is less than or equal to z; ga （i） Represents the number of pixels whose grayscale value is i and whose grayscale value is less than or equal to z; Define calculation formula 2-6: ; Calculate the pixel frequency pj whose gray value is less than or equal to 0 to 255 （0） ~pj （255） ; Step S25: Define calculation formula 2-7: ； Among them, gr （x，y） Represents the gray value, Indicates that the gray value is less than or equal to gr （x，y） The pixel frequency, xf （x，y） Indicates gr （x，y） The transformation value of Adaptively enhance the contrast of the monitoring image to obtain xf （1，1） ~xf （Lb，Hb） ; xf （1，1） ~xf （Lb，Hb） The corresponding image is used as a quasi-image; the background of the aligned image is eliminated, the aligned image is segmented by adaptive threshold, and the chest contour of the patient is marked; Step S26: obtaining the current time dh hours and dm minutes; extracting the monitoring video of the patient within dh hours (dm-1) minutes as the target video; taking the picture per second in the target video as the target picture; calculating the chest movement information of the patient in 1 minute; Step S27: Based on the patient's chest transverse expansion and contraction velocity uv within 1 minute （1） ~uv （60） and the thoracic longitudinal velocity vv （1） ~vv （60） , analyzing the patient's respiratory movements.

4. The method for processing images in a ICU based on respiratory motion according to claim 2, characterized in that: The specific steps of step S22 are as follows: Step S221: Calculate r （1，1） ~r （Lb，Hb） The mean square error Sr;g （1，1） ~g （Lb，Hb） The mean square error Sg; b （1，1） ~b （Lb，Hb） The mean square error Sb is as follows: ； ； ； Step S222: back up image A to obtain a duplicate image; extract d （1） ~d （N） The minimum value dl in Define calculation formula 2-2: ； Among them, Fr （k）（x，y） , Fr （k）（x，y） , Fr （k）（x，y） , respectively represent the color intensity of red, green and blue at the monitoring image (x, y) after being processed by the k-th Gaussian filter; d （k） and σ （k） , respectively represent the scale and standard deviation of the kth Gaussian filter; use the 1st to Nth Gaussian filters to process the heavy image and obtain image A （1） ~Image A （N） ; Step S223: Calculate image A （1） The mean square error of the color intensities of red, green, and blue is sr （1） ,sg （1） and sb （1） ; Image A （1） The RGB values ​​of all pixels are recorded as (r （1）（1，1） , g （1）（1，1） , b （1）（1，1） )~(r （1）（Lb，Hb） , g （1）（Lb，Hb） , b （1）（Lb，Hb） ); Calculate r （1）（1，1） ~r （1）（Lb，Hb） The average value of （1） , g （1）（1，1） ~g （1）（Lb，Hb） The average value of ag （1） , b （1）（1，1） ~b （1）（Lb，Hb） The average value of ab （1） .

5. The method for processing images in a ICU based on respiratory motion according to claim 4, characterized in that: The subsequent steps of step S223 are as follows: Step S224: Calculate r （1）（1，1） ~r （1）（Lb，Hb） The mean square error sr （1） ; g （1）（1，1） ~g （1）（Lb，Hb） The mean square error sg （1） ; b （1）（1，1） ~b （1）（Lb，Hb） The mean square error sb （1） ; ； ； ； Step S225: Calculate image A （2） To Image A （N） The mean square error of the color intensities of red, green, and blue is sr （2） ~sr （N） , sg （2） ~sg （N） and sb （2） ~sb （N） ; Let image A （k） The mean square error of the color intensity of red, green, and blue is sr （k） ,sg （k） and sb （k） ; The weight coefficient of the kth Gaussian filter is zq （k） ; Define calculation formula 2-3: ; Calculate the weight coefficients zq of the 1st to Nth Gaussian filters （1） ~zq （N） ; Step S226: Calculate zq （1） ~zq （N） The sum is recorded as aq; the weight coefficient of the kth Gaussian filter is recorded as zq （k） , the weight is denoted as ω （1） , define calculation formula 2-4: ω （k） =(zq （k） / aq); Calculate the weights ω of the 1st to Nth Gaussian filters （1） ~ω （N） .

6. The method for processing images in a ICU based on respiratory motion according to claim 2, characterized in that: The specific steps of step S23 are as follows: Step S231: define calculation formula 2-5: ； Among them, Rr （x，y） , Rg （x，y） and Rb （x，y） , respectively represent the color intensity of the pixel point (x, y) on the monitoring image after red, green and blue transformation; Fr （k）（x，y） , Fg （k）（x，y） , Fb （k）（x，y） , respectively represent the color intensity of red, green and blue at the monitoring image (x, y) after being processed by the k-th Gaussian filter; Step S232: Transform the monitoring image once to obtain (Rr （1，1） , Rg （1，1） , Rb （1，1） )~(Rr （Lb，Hb） , Rg （Lb，Hb） , Rb （Lb，Hb） ), converted into grayscale value, and get gr （1，1） ~gr （Lb，Hb） .

7. The method for processing images in a ICU based on respiratory motion according to claim 3, characterized in that: The specific steps of step S26 are as follows: Step S261: Set the picture at the dtth second in the target video as picture Pr （dt） , the (dt+1) second picture is picture Pr （dt+1） ; Step S262: Count the pictures Pr （dt） The number of pixels of the patient's chest contour in st; Put the picture Pr （dt） The pixel points of the chest contour of the patient in the figure are taken as punctuation points; the positions of the 1st to the stth punctuation points are recorded as (sx （1） ,sy （1） )~(sx （st） ,sy （st） ); in the picture Pr （dt） In the above example, the grayscale values ​​of the 1st to the stth punctuation points are extracted and recorded as Ii (sx （1） ,sy （1） )~Ii(sx （st） ,sy （st） ); Step S263: In the picture Pr （dt+1） Extract the positions corresponding to the 1st to the st punctuation marks (tx （1） ,ty （1） )~(tx （st） ,ty （st） ); extract(tx （1） ,ty （1） )~(tx （st） ,ty （st） ) is the gray value of the pixel point Io (tx （1） ,ty （1） )~Io(tx （st） ,ty （st） ).

8. The method for processing images in a ICU based on respiratory motion according to claim 7, characterized in that: The subsequent steps of step S263 are as follows: Step S264: Picture Pr （dt） In the example, the position of the rth punctuation mark (sx （r） ,sy （r） ), grayscale Ii (sx （r） ,sy （r） ); Picture Pr （dt+1） In the example, the position of the rth punctuation mark (tx （r） ,ty （r） ), Grayscale Io (tx （r） ,ty （r） ); Define calculation formula 2-8: ； Among them, Ix （r） ,Iy （r） and Iio （r） , respectively represent the horizontal gradient, vertical gradient and brightness gradient of the rth punctuation point; calculate the horizontal gradient Ix from the 1st to the stth punctuation point （1） ~Ix （st） , longitudinal gradient Iy （1） ~Iy （st） , brightness gradient Iio （1） ~Iio （st） ; Step S265: Record the patient's chest transverse expansion and contraction velocity at the dt second as uv （dt） , the longitudinal velocity of the thorax is recorded as vv （dt） ; Define calculation formula 2-9: ； Step S266: Extract the monitoring image of the patient at dh hours (dm-2) minutes and 59 seconds as the reference image; set the value of dt to 1, and calculate the transverse expansion and contraction velocity uv of the patient's chest cavity within dh hours (dm-1) minutes （1） ~uv （60） and the thoracic longitudinal velocity vv （1） ~vv （60） As chest movement information.

9. The method for processing images in a ICU based on respiratory motion according to claim 3, characterized in that: The specific steps of step S27 are as follows: Step S271: Assume that the patient's chest transverse expansion and contraction speed at the qth second is uv （q） , the longitudinal velocity of the chest cavity is vv （q） ; Get the chest transverse expansion and contraction velocity interval of the first to the hoth healthy person within 1 minute, recorded as [uvl （1，ho） ~uvl （60，ho） , uvm （1，ho） ~uvm （60，ho） ]; the interval of the longitudinal expansion and contraction velocity of the chest cavity, denoted as [vvl （1，ho） ~vvl （60，ho） , vvm （1，ho） ~vvm （60，ho） ]; Step S272: Assume that the patient's chest transverse expansion and contraction speed at the uth second within 1 minute is uvr （u） , the longitudinal velocity of the chest cavity is vvr （u） , the respiratory frequency is ff; define the relationship S: ； Among them, uvr （s） and vvr （s） , respectively represent the transverse expansion and contraction velocity and longitudinal velocity of the chest cavity of the patient at the sth second within 1 minute; ε represents the judgment value; j represents the imaginary unit; Step S273: [uvl （1，ho） ~uvl （60，ho） , uvm （1，ho） ~uvm （60，ho） ] and [vvl （1，ho） ~vvl （60，ho） , vvm （1，ho） ~vvm （60，ho） ] into the relation S, and calculate the breathing frequency range of a healthy person, and we get [fh （1） , fh （ho） ]; The uv （1） ~uv （60） and vv （1） ~vv （60） Substituting into the relation S, calculate the patient's respiratory rate, recorded as hf; Step S274: Determine whether hf is in [fh （1） , fh （ho） ] interval; If the patient is present, no action will be taken; if the patient is not present, an alarm will be immediately issued and the patient's pathological condition will be analyzed.

10. A respiratory motion-based ICU image processing system, applicable to a respiratory motion-based ICU image processing method according to any one of claims 1 to 9, characterized in that: The system comprises: Image acquisition module: used to obtain patient monitoring video; Image analysis module: used to enhance the monitoring video and mark the patient's chest contour through background modeling technology; detect the subtle movement of the patient's chest contour in the monitoring video to determine the patient's respiratory movement; perform frequency domain analysis on the patient's respiratory movement to obtain the patient's breathing periodicity information and determine whether the patient's respiratory frequency is normal; if so, no processing will be done; if not, an alarm will be immediately sounded and the medical records of all target hospitals will be obtained to analyze the patient's pathological conditions; Patient monitoring module: used to construct the lesion index change equation; predict the patient's condition change according to the lesion index change equation and obtain diagnostic information; Continuous monitoring module: used to continuously update and analyze monitoring videos based on lesion conditions and diagnostic information until the patient recovers.

Images