A classification method for monitoring images in respiratory intensive care units

Abstract

The present invention discloses a classification method for monitoring images in a respiratory intensive care unit, relates to the medical field, and solves the problem of poor application effect of existing monitoring image classification methods for respiratory intensive care units. The method comprises the following steps: S1: continuously monitoring medical indicators of monitored patients in the intensive care unit, obtaining respiratory index images and oxygenation index images, and obtaining continuous medical image data; S2: performing intermittent medical indicator monitoring on the monitored patients in the intensive care unit to obtain multiple medical image analysis images, and obtaining intermittent medical image data; S3: obtaining a medical image data set, creating a monitoring image classification model to classify the medical image data set, and obtaining monitoring image classification data; and S4: evaluating the monitoring status of the monitored patients according to the monitoring image classification data. The present invention improves the application value of monitoring images in the respiratory intensive care unit.

Description

Technical Field

[0001] The present invention belongs to the medical field and relates to image analysis technology, in particular to a classification method for monitoring images in a respiratory intensive care unit. Background Art

[0002] The existing classification method for monitoring images in respiratory intensive care units has the following defects when monitoring patient status:

[0003] Existing classification methods for monitoring images in respiratory intensive care units are unable to combine and analyze the continuous monitoring data acquired in real time by medical monitoring equipment with the intermittently acquired medical imaging data when monitoring patient status, resulting in a lack of comprehensiveness in the monitoring process.

[0004] Existing classification methods for respiratory intensive care unit monitoring images cannot visualize the collected numerical indicators and do not combine large model technology for automated image classification. This results in low image classification efficiency and makes it difficult to fully realize the medical value of monitoring images.

[0005] To this end, we propose a classification method for respiratory intensive care unit monitoring images. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a classification method for monitoring images in the respiratory intensive care unit. The present invention aims to improve the classification efficiency of monitoring images and improve the monitoring quality of the intensive care unit.

[0007] In order to achieve the above object, the present invention adopts the following technical solution: a classification method for monitoring images of respiratory intensive care units, comprising the following specific steps:

[0008] Step S1: Continuously monitor medical indicators of patients in the intensive care unit, and obtain respiratory index images and oxygenation index images by analyzing the monitoring results to obtain continuous medical image data;

[0009] Step S2: performing interval medical index monitoring on the monitored patient in the intensive care unit, and obtaining medical imaging analysis images by analyzing the monitoring results to obtain interval medical image data;

[0010] Step S3: obtaining a medical image dataset corresponding to each monitored patient based on the continuous medical image data and the intermittent medical image data, creating a monitoring image classification model to classify the medical image dataset, and obtaining monitoring image classification data;

[0011] Step S4: Evaluate the monitoring status of the monitored patient based on the monitoring image classification data.

[0012] Furthermore, the step S1 further includes the following specific steps:

[0013] Step S11: selecting a critical care patient from a plurality of monitored patients in the critical care unit as a sample monitored patient;

[0014] Step S12: During the process of monitoring the medical indicators of the sample monitoring patients, the time value corresponding to the current moment is used as the end time of the period to mark a medical monitoring period of fixed duration, and the medical monitoring period is divided into a number of monitoring sub-periods of equal duration, obtaining monitoring sub-periods H1 to Ha;

[0015] Step S13: monitoring the respiratory index of the sample monitoring patient in the medical monitoring cycle, and obtaining a respiratory index image by analyzing the monitoring results;

[0016] Step S14: monitoring the oxygenation index of the sample monitoring patient in the medical monitoring cycle, and obtaining an oxygenation index image by analyzing the monitoring results;

[0017] Step S15: Acquire the oxygenation index image and respiratory index image corresponding to each monitored patient to obtain continuous medical image data.

[0018] Furthermore, the step S13 further includes the following specific steps:

[0019] Step S131: obtaining the gas volume of a single breath of the patient during the monitoring sub-cycle H1 to the monitoring sub-cycle Ha, and obtaining the tidal volume of the H1 cycle to the Ha cycle;

[0020] Step S132: Obtain the cyclic respiratory frequency of the sample monitoring patient from the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, and obtain the H1 cycle respiratory frequency to the Ha cycle respiratory frequency;

[0021] Step S133: During the monitoring sub-cycle H1 to the monitoring sub-cycle Ha, the mean arterial carbon dioxide partial pressure of the sample monitoring patient in each monitoring sub-cycle is obtained, and the arterial carbon dioxide partial pressure of the H1 cycle to the Ha cycle is obtained;

[0022] Step S134: The periodic tidal volume VT, periodic respiratory frequency RR and periodic arterial carbon dioxide partial pressure PaCO2 corresponding to the same monitoring sub-cycle are calculated to obtain the periodic respiratory index RFA. The specific formula is: ;

[0023] Step S135: naming the periodic respiratory indices corresponding to the H1 monitoring sub-cycle to the Ha monitoring sub-cycle as H1 periodic respiratory index to Ha periodic respiratory index in sequence;

[0024] Step S136: In the existing broken line statistical graph template, the monitoring sub-cycle is used as the horizontal axis and the cycle respiratory index is used as the vertical axis. The H1 monitoring sub-cycle to the Ha monitoring sub-cycle and the H1 cycle respiratory index to the Ha cycle respiratory index are plotted in the broken line statistical graph template to obtain a respiratory index image.

[0025] Furthermore, the step S14 further includes the following specific steps:

[0026] Step S141: obtaining the mean airway pressure corresponding to each breathing process of the sample monitoring patient during the monitoring sub-cycle H1 to the monitoring sub-cycle Ha, and calculating the mean airway pressure corresponding to multiple breathing processes in the same monitoring sub-cycle to obtain the airway pressure value of the H1 cycle and the airway pressure of the Ha cycle;

[0027] Step S142: obtaining the oxygen inhalation concentration corresponding to each breathing process of the sample monitoring patient during the monitoring sub-cycle H1 to the monitoring sub-cycle Ha, and calculating the average of the oxygen inhalation concentrations corresponding to multiple breathing processes in the same monitoring sub-cycle to obtain the oxygen inhalation concentration of the H1 cycle and the oxygen inhalation concentration of the Ha cycle;

[0028] Step S143: During the monitoring sub-cycle H1 to the monitoring sub-cycle Ha, respectively obtain the average arterial oxygen partial pressure of the sample monitoring patient in each monitoring sub-cycle, and obtain the arterial oxygen partial pressure of the H1 cycle to the Ha cycle;

[0029] Step S144: Calculate the periodic oxygen inhalation concentration FiO2, periodic airway pressure MAP, and periodic arterial oxygen partial pressure PaO2 corresponding to the same monitoring sub-cycle to obtain the periodic oxygenation index OI. The specific formula is:

[0030] ;

[0031] Step S145: naming the cycle oxygenation indexes corresponding to the H1 monitoring sub-cycle to the Ha monitoring sub-cycle as H1 cycle oxygenation index to Ha cycle oxygenation index in sequence;

[0032] Step S146: In the existing broken line statistical graph template, the monitoring sub-cycle is used as the horizontal axis and the cycle oxygenation index is used as the vertical axis. The H1 monitoring sub-cycle to the Ha monitoring sub-cycle and the H1 cycle oxygenation index to the Ha cycle oxygenation index are plotted in the broken line statistical graph template to obtain an oxygenation index image.

[0033] Furthermore, the step S2 further includes the following specific steps:

[0034] Step S21: obtaining a plurality of medical examination images of the sample monitoring patient during the medical monitoring cycle, and naming the plurality of medical examination images in chronological order as examination images Y1 to Yb;

[0035] Step S22: performing image analysis on the Y1 examination image to obtain the Y1 image diagnosis coefficient;

[0036] Step S23: acquiring the image diagnostic coefficients corresponding to the Y2 examination image to the Yb examination image respectively, to obtain the Y2 image diagnostic coefficient to the Yb image diagnostic coefficient;

[0037] Step S24: In the existing broken line statistical graph template, the examination image is used as the horizontal coordinate, and the image diagnostic coefficient is used as the vertical coordinate. The Y1 examination image to the Yb examination image and the Y1 image diagnostic coefficient to the Yb image diagnostic coefficient are plotted in the broken line statistical graph template to obtain a diagnostic coefficient image;

[0038] Step S25: acquiring the diagnostic coefficient image corresponding to each monitored patient to obtain interval medical image data;

[0039] The step S22 further includes the following specific steps:

[0040] Step S221: using an image segmentation algorithm to segment the patient's lung organ tissue in the Y1 examination image to obtain an image lung tissue region, and using an image segmentation algorithm to segment the lung lesion region in the Y1 examination image to obtain an image lesion region;

[0041] Step S222: defining a plurality of pixel filling blocks of equal area in the Y1 inspection image, using the pixel filling blocks to fill the area of ​​the lung tissue region of the image, and after the filling is completed, counting the number of pixel filling blocks filled in the lung tissue region of the image to obtain the number of lung filling blocks, and using the pixel filling blocks to fill the area of ​​the lesion region of the image, and after the filling is completed, counting the number of pixel filling blocks filled in the lesion region of the image to obtain the number of lesion filling blocks;

[0042] Step S223: Calculate the ratio of the number of lesion filling blocks to the number of lung filling blocks to obtain the lesion area filling ratio;

[0043] Step S224: performing edge smoothness analysis on the lesion area of ​​the image to obtain a quantified value of the edge smoothness of the lesion area;

[0044] Step S225: Calculate the product of the quantified value of the lesion edge smoothness and the filling area ratio of the lesion area to obtain the Y1 imaging diagnosis coefficient.

[0045] Furthermore, the step S224 further includes the following specific steps:

[0046] Step S2241: using an edge extraction algorithm to extract the edge of the image lesion area in the Y1 inspection image to obtain the edge line of the lesion area;

[0047] Step S2242: setting a number of feature rectangles of equal area to cover the edge line of the lesion area. During the covering process, it is necessary to ensure that the edge line of the lesion area passes through any two opposite sides of the feature rectangles;

[0048] Step S2243: naming the characteristic rectangles covering the edge line of the lesion area as the first characteristic rectangle to the cth characteristic rectangle respectively;

[0049] Step S2244: performing a flatness analysis on the edge line of the lesion area within the first characteristic rectangle to obtain a first edge flatness quantization value;

[0050] Step S2245: performing flatness analysis on the edge lines of the lesion area from the second characteristic rectangle to the cth characteristic rectangle respectively, and obtaining the second edge flatness quantization value to the cth edge flatness quantization value;

[0051] Step S2246: Calculate the average of the first edge smoothness quantization value to the cth edge smoothness quantization value to obtain the lesion area edge smoothness quantization value.

[0052] Furthermore, the step S2244 further includes the following specific steps:

[0053] Mark the two opposite sides of the feature rectangle where the edge line of the lesion area passes as the first feature side and the second feature side, and mark any rectangular side between the first feature side and the second feature side as the third feature side;

[0054] Mark the intersection of the edge line of the lesion area and the first characteristic edge as the first characteristic point, mark the intersection of the edge line of the lesion area and the second characteristic edge as the second characteristic point, and mark the midpoint of the third characteristic edge as the third characteristic point;

[0055] Mark the circle determined by the first characteristic point, the second characteristic point, and the third characteristic point as the first characteristic circle, and mark the center of the first characteristic circle as the local characteristic point of the lesion, mark several lesion area points on the edge line of the lesion area, and name the marked lesion area points as the first lesion area point to the dth lesion area point respectively;

[0056] In the first characteristic circle, the first characteristic point and the second characteristic point are connected to obtain a lesion characteristic chord, and a length value of the lesion characteristic chord is obtained to obtain a first preset leveling distance;

[0057] Connect the first lesion area point to the dth lesion area point and the lesion local feature point to obtain the first lesion edge line to the dth lesion edge line, and obtain the length value of the first lesion edge line to the dth lesion edge line to obtain the length value of the first edge line to the dth edge line;

[0058] Obtaining the radius of the first characteristic circle to obtain a second preset leveling distance;

[0059] A first edge smoothness quantization value is obtained by calculating the first edge connection length value to the dth edge connection length value, the first preset smoothing distance, and the second preset smoothing distance;

[0060] Calculate the quantized value of the first edge flatness. The specific formula is as follows:

[0061] ;

[0062] Among them, Pad1 is the first edge smoothing quantization value, Blxi is the length value of the i-th edge connection line, Ysj1 is the first preset smoothing distance, Ysj2 is the second preset average distance, and d is the number value corresponding to the lesion area point.

[0063] Furthermore, the step S3 further includes the following specific steps:

[0064] Step S31: Acquire continuous medical image data, and acquire an oxygenation index image and a respiratory index image corresponding to each monitored patient based on the continuous medical image data;

[0065] Step S32: obtaining interval medical image data, and obtaining a diagnostic coefficient image corresponding to each monitored patient based on the interval medical image data;

[0066] Step S33: merging the oxygenation index image, respiratory index image, and diagnostic coefficient image corresponding to the same monitored patient into a monitoring image set;

[0067] Step S34: acquiring monitoring image sets corresponding to multiple historical patients based on historical monitoring data of the intensive care unit to obtain multiple historical monitoring image sets;

[0068] Step S35: Divide the historical monitoring image set into a first type image set and a second type image set to obtain a historical monitoring image annotation set;

[0069] Step S36: dividing the historical monitoring image annotation set into a monitoring image training set and a monitoring image test set according to the image training and testing ratio;

[0070] Step S37: creating a monitoring image classification model using the monitoring image training set and the monitoring image test set;

[0071] Step S38: using the monitoring image classification model to identify the plurality of monitoring image sets as a first type of image set and a second type of image set, and obtaining monitoring image classification data.

[0072] Furthermore, the step S35 further includes the following specific steps:

[0073] Step S351: selecting a sample monitoring image set from the historical monitoring image annotation set, and obtaining an oxygenation index image, a respiratory index image, and a diagnostic coefficient image based on the sample monitoring image set;

[0074] Step S352: In the respiratory index image, the coordinate points corresponding to the H1 cycle respiratory index to the Ha cycle respiratory index are marked as H1 respiratory coordinate points to Ha respiratory coordinate points, and the slope between each two consecutive respiratory coordinate points between the H1 respiratory coordinate points and the Ha respiratory coordinate points is numerically acquired to obtain a-1 respiratory coordinate slope values, and the obtained a-1 respiratory coordinate slope values ​​are averaged to obtain the sample respiratory image change rate;

[0075] Step S353: In the oxygenation index image, the coordinate points corresponding to the H1 cycle oxygenation index to the Ha cycle oxygenation index are marked as H1 oxygenation coordinate points to Ha oxygenation coordinate points, and the slope between each two consecutive coordinate points between the H1 oxygenation coordinate point and the Ha oxygenation coordinate point is numerically obtained to obtain a-1 oxygenation coordinate slope values, and the obtained a-1 oxygenation coordinate slope values ​​are averaged to obtain the sample oxygenation image change rate;

[0076] Step S354: In the diagnostic coefficient image, the coordinate points corresponding to the Y1 image diagnostic coefficient to the Yb image diagnostic coefficient are marked as Y1 image coordinate points to Yb image coordinate points, and the slope between each two consecutive coordinate points between the Y1 image coordinate point and the Yb image coordinate point is numerically obtained, b-1 image coordinate slope values, and the obtained b-1 image coordinate slope values ​​are averaged to obtain the sample diagnostic image change rate;

[0077] Step S355: Calculate the average of the sample respiratory image change rate, the sample oxygenation image change rate, and the sample diagnostic image change rate to obtain a monitoring state comprehensive change rate, obtain a monitoring state comprehensive change rate threshold, and compare the monitoring state comprehensive change rate with the monitoring state comprehensive change rate threshold. If the monitoring state comprehensive change rate is greater than or equal to the monitoring state comprehensive change rate threshold, mark the sample monitoring image set as a first type image set; if the monitoring state comprehensive change rate is less than the monitoring state comprehensive change rate threshold, mark the sample monitoring image set as a second type image set.

[0078] Step S356: Mark each historical monitoring image set separately.

[0079] Furthermore, the step S4 further includes the following specific steps:

[0080] Step S41: Acquire monitoring image classification data, and acquire a first type image set and a second type image set according to the monitoring image classification data;

[0081] Step S42: when the monitoring image set corresponding to the monitored patient is a first type of image set, it is assessed that the monitored patient is in a monitoring fluctuation state;

[0082] Step S43: When the monitored image set corresponding to the monitored patient is the second type of image set, it is assessed that the monitored patient is in a stable monitoring state.

[0083] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0084] 1. The present invention combines and analyzes the continuous monitoring data acquired in real time by medical monitoring equipment with the medical imaging data acquired intermittently, thereby improving the comprehensiveness of the monitoring process and the accuracy of the monitoring results;

[0085] 2. The present invention performs image visualization processing on the collected numerical indicators and combines large model technology to perform automatic image classification, thereby improving the efficiency of image classification and fully realizing the medical value of monitoring images. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0087] Figure 1 It is a diagram of the implementation steps of the present invention;

[0088] Figure 2 This is a schematic diagram of the characteristic rectangular coverage of the present invention;

[0089] Figure 3 Schematic diagram of the edge line of the lesion area of ​​the present invention. DETAILED DESCRIPTION

[0090] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0091] Example 1

[0092] See also Figure 1The present invention provides a technical solution: a classification method for monitoring images in a respiratory intensive care unit, comprising the following specific steps:

[0093] Step S1: Continuously monitor medical indicators of patients in the intensive care unit, and obtain respiratory index images and oxygenation index images by analyzing the monitoring results to obtain continuous medical image data;

[0094] The step S1 further includes the following specific steps:

[0095] One intensive care patient is selected from multiple monitored patients in the intensive care unit as a sample monitored patient;

[0096] It should be noted here that:

[0097] The monitored patients referred to herein are patients in the intensive care unit who are connected to condition monitoring equipment. The condition monitoring equipment involved herein includes but is not limited to a ventilator, a blood oxygen saturation monitor, and a gas analyzer;

[0098] In the process of monitoring the medical indicators of the sample monitoring patients, the time value corresponding to the current moment is used as the end time of the period to mark a medical monitoring period of fixed duration, and the medical monitoring period is divided into several monitoring sub-periods of equal duration, obtaining monitoring sub-periods H1 to Ha;

[0099] It should be noted here that:

[0100] In this application, as the time value corresponding to the current moment changes, the duration corresponding to the medical monitoring cycle is fixed, so that the start time point of the medical monitoring cycle also changes with the time value corresponding to the current moment, thereby realizing dynamic updating of the medical monitoring cycle;

[0101] In this application, H mentioned here is the identifier corresponding to the monitoring sub-cycle, and a is the quantity value corresponding to the monitoring sub-cycle.

[0102] Monitor the respiratory index of sample monitoring patients in the medical monitoring cycle, and obtain respiratory index images by analyzing the monitoring results;

[0103] The details are as follows:

[0104] During the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, the gas volume of a single breath of the sample monitoring patient is obtained to obtain the tidal volume of the H1 cycle to the tidal volume of the Ha cycle;

[0105] Obtain the periodic respiratory frequency of the sample monitoring patient from the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, and obtain the H1 periodic respiratory frequency to the Ha periodic respiratory frequency;

[0106] It should be noted here that:

[0107] In this application, the H1 cycle respiratory rate referred to herein is the average number of breaths per minute of the sample monitored patient during the H1 monitoring sub-cycle;

[0108] In the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, the average arterial carbon dioxide partial pressure of the sample monitoring patient in each monitoring sub-cycle is obtained, and the arterial carbon dioxide partial pressure of the H1 cycle to the arterial carbon dioxide partial pressure of the Ha cycle is obtained;

[0109] The periodic respiratory index RFA is calculated by calculating the periodic tidal volume VT, periodic respiratory frequency RR and periodic arterial carbon dioxide partial pressure PaCO2 corresponding to the same monitoring sub-cycle. The specific formula is: ;

[0110] The periodic respiratory indices corresponding to the H1 monitoring sub-cycle to the Ha monitoring sub-cycle are named H1 periodic respiratory index to Ha periodic respiratory index respectively;

[0111] In specific implementation, VT×RR represents the minute ventilation of the monitored patient. The higher the ventilation, the stronger the breathing ability, and the lower the PaCO2, the more effective the excretion of carbon dioxide. In clinical studies, the higher the ventilation, the more effective the excretion of carbon dioxide, which can indicate that the patient's breathing is more efficient. In the experiment, there are VT=0.5 L, RR=20 times / minute, PaCO2=45mmHg, and respiratory index = 45mmHg(10L / min)≈0.222L / (min⋅mmHg);

[0112] In the existing broken line statistical graph template, the monitoring sub-cycle is used as the horizontal axis and the cycle respiratory index is used as the vertical axis. The H1 monitoring sub-cycle to the Ha monitoring sub-cycle and the H1 cycle respiratory index to the Ha cycle respiratory index are plotted in the broken line statistical graph template to obtain a respiratory index image;

[0113] Monitor the oxygenation index of sample patients in the medical monitoring cycle and obtain oxygenation index images by analyzing the monitoring results;

[0114] From the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, the mean airway pressure corresponding to each breathing process of the sample monitoring patient is obtained, and the mean airway pressure corresponding to multiple breathing processes in the same monitoring sub-cycle is averaged to obtain the H1 cycle airway pressure value and the Ha cycle airway pressure;

[0115] From the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, the oxygen inhalation concentration corresponding to each breathing process of the sample monitoring patient is obtained, and the oxygen inhalation concentration corresponding to multiple breathing processes in the same monitoring sub-cycle is averaged to obtain the oxygen inhalation concentration of the H1 cycle and the oxygen inhalation concentration of the Ha cycle;

[0116] In the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, the average arterial oxygen partial pressure of the sample monitoring patient in each monitoring sub-cycle is obtained, and the arterial oxygen partial pressure of the H1 cycle to the arterial oxygen partial pressure of the Ha cycle is obtained;

[0117] The periodic oxygen inhalation concentration FiO2, periodic airway pressure MAP, and periodic arterial oxygen partial pressure PaO2 corresponding to the same monitoring sub-cycle are calculated to obtain the periodic oxygenation index OI. The specific formula is:

[0118] ;

[0119] It should be noted here that:

[0120] In this application, the cyclic oxygenation index involved here is an indicator of the patient's pulmonary oxygenation capacity. The higher the cyclic oxygenation index OI, the worse the patient's pulmonary oxygenation function. The cyclic oxygen inhalation concentration FiO2 and the cyclic arterial oxygen partial pressure PaO2 are directly proportional to the cyclic oxygenation index OI, and the cyclic airway pressure MAP is inversely proportional to the cyclic oxygenation index OI.

[0121] In actual operation, if PaO2=70mmHg, FiO2=0.5, and MAP=15mmHg, the periodic respiratory index OI can be calculated as 2.33;

[0122] The cycle oxygenation indexes corresponding to the H1 monitoring sub-cycle to the Ha monitoring sub-cycle are named H1 cycle oxygenation index to Ha cycle oxygenation index respectively;

[0123] In the existing broken line statistical graph template, the monitoring sub-cycle is used as the horizontal axis and the cycle oxygenation index is used as the vertical axis. The H1 monitoring sub-cycle to the Ha monitoring sub-cycle and the H1 cycle oxygenation index to the Ha cycle oxygenation index are plotted in the broken line statistical graph template to obtain an oxygenation index image;

[0124] The oxygenation index image and respiratory index image corresponding to each monitored patient are acquired respectively to obtain continuous medical image data.

[0125] Step S2: performing interval medical index monitoring on the monitored patient in the intensive care unit, and obtaining medical imaging analysis images by analyzing the monitoring results to obtain interval medical image data;

[0126] The step S2 further includes the following specific steps:

[0127] Acquire several medical examination images of the sample monitoring patient during the medical monitoring cycle, and name the several medical examination images in chronological order as Y1 examination image to Yb examination image;

[0128] It should be noted here that:

[0129] In this application, Y referred to herein is an identifier corresponding to a medical examination image, b is a quantity value corresponding to the medical examination image, and b is an integer greater than 0;

[0130] In this application, the medical examination images involved herein are specifically tomographic images, and the tomographic scan site is the patient's lungs;

[0131] In the present application, the lung tissue area, lung tissue position, and lung tissue scanning angle scanned by the Y1 examination image to the Yb examination image are all the same.

[0132] Perform image analysis on the Y1 examination image to obtain the Y1 imaging diagnostic coefficient;

[0133] The details are as follows:

[0134] Using an image segmentation algorithm to segment the patient's lung organ tissue in the Y1 examination image to obtain an image lung tissue area, and using an image segmentation algorithm to segment the lung lesion area in the Y1 examination image to obtain an image lesion area;

[0135] In the Y1 inspection image, several pixel filling blocks of equal area are defined, and the pixel filling blocks are used to fill the area of ​​the lung tissue region of the image. After the filling is completed, the number of pixel filling blocks filled in the lung tissue region of the image is counted to obtain the number of lung filling blocks. The pixel filling blocks are used to fill the area of ​​the lesion region of the image. After the filling is completed, the number of pixel filling blocks filled in the lesion region of the image is counted to obtain the number of lesion filling blocks.

[0136] The ratio of the number of lesion filling blocks to the number of lung filling blocks was calculated to obtain the filling area ratio of the lesion area;

[0137] It should be noted here that:

[0138] In this application, the pixel filling block is 0.001mm in area. 2 The area error caused by the overflow of pixel filling blocks in the lung tissue area and the edge area of ​​the image lesion area is negligible;

[0139] Perform edge smoothness analysis on the lesion area of ​​the image to obtain the quantitative value of the edge smoothness of the lesion area;

[0140] The details are as follows:

[0141] Use edge extraction algorithm to extract the edge of the image lesion area in the Y1 inspection image to obtain the edge line of the lesion area;

[0142] See also Figure 2 , set several feature rectangles of equal area to cover the edge line of the lesion area. During the covering process, it is necessary to ensure that the edge line of the lesion area passes through any two opposite sides of the feature rectangle;

[0143] The feature rectangles covering the edge line of the lesion area are named as the first feature rectangle to the cth feature rectangle;

[0144] It should be noted here that:

[0145] In this application, c is the number value corresponding to the characteristic rectangle, and c is an integer greater than 0;

[0146] Performing a flatness analysis on the edge line of the lesion area within the first characteristic rectangle to obtain a first edge flatness quantization value;

[0147] The details are as follows:

[0148] See also Figure 3 , mark the two opposite sides of the feature rectangle where the edge line of the lesion area passes as the first feature side and the second feature side, and mark any rectangular side between the first feature side and the second feature side as the third feature side;

[0149] Mark the intersection of the edge line of the lesion area and the first characteristic edge as the first characteristic point, mark the intersection of the edge line of the lesion area and the second characteristic edge as the second characteristic point, and mark the midpoint of the third characteristic edge as the third characteristic point;

[0150] Mark the circle determined by the first characteristic point, the second characteristic point, and the third characteristic point as the first characteristic circle, and mark the center of the first characteristic circle as the local characteristic point of the lesion, mark several lesion area points on the edge line of the lesion area, and name the marked lesion area points as the first lesion area point to the dth lesion area point respectively;

[0151] It should be noted here that:

[0152] In the present application, d referred to here is the quantity value corresponding to the lesion area point, and d is an integer greater than 0.

[0153] In the first characteristic circle, the first characteristic point and the second characteristic point are connected to obtain a lesion characteristic chord, and a length value of the lesion characteristic chord is obtained to obtain a first preset leveling distance;

[0154] Connect the first lesion area point to the dth lesion area point and the lesion local feature point to obtain the first lesion edge line to the dth lesion edge line, and obtain the length value of the first lesion edge line to the dth lesion edge line to obtain the length value of the first edge line to the dth edge line;

[0155] Obtaining the radius of the first characteristic circle to obtain a second preset leveling distance;

[0156] A first edge smoothness quantization value is obtained by calculating the first edge connection length value to the dth edge connection length value, the first preset smoothing distance, and the second preset smoothing distance;

[0157] Calculate the quantized value of the first edge flatness. The specific formula is as follows:

[0158] ;

[0159] Wherein, Pad1 is the first edge smoothing quantization value, Blxi is the length of the i-th edge line, Ysj1 is the first preset smoothing distance, Ysj2 is the second preset average distance, and d is the number value corresponding to the lesion area point;

[0160] It should be noted here that:

[0161] In the present application, the i-th edge connection length value involved here may be any edge connection length value from the first edge connection length value to the d-th edge connection length value;

[0162] In this application, if the edge line of the lesion area in the first characteristic rectangle is in an extremely flat state, the edge line of the lesion area should be in a coincidence state with the lesion characteristic chord, that is, the sum of the values ​​of the length of the first edge line to the length of the dth edge line should be the same as The values ​​are equal, and the first edge smoothness quantization value calculated is 0. It can be seen that the larger the first edge smoothness quantization value is, the worse the smoothness of the edge line of the lesion area in the first characteristic rectangle is.

[0163] Repeat the process of obtaining the first edge smoothness quantization value, and perform flatness analysis on the edge lines of the lesion area from the second characteristic rectangle to the cth characteristic rectangle, to obtain the second edge smoothness quantization value to the cth edge smoothness quantization value;

[0164] Calculate the average of the first edge smoothness quantization value to the cth edge smoothness quantization value to obtain the ward edge smoothness quantization value;

[0165] The product of the quantified value of the lesion edge smoothness and the filling area ratio of the lesion area is calculated to obtain the Y1 imaging diagnostic coefficient;

[0166] It should be noted here that:

[0167] In this application, the Y1 imaging diagnostic coefficient referred to herein is an indicator for measuring the severity of the patient's condition as reflected by the Y1 examination image feedback. The Y1 imaging diagnostic coefficient indicates that the patient's condition is more serious;

[0168] It is understandable that: under the same other external conditions, the larger the quantitative value of the smoothness of the edge of the lesion area (the worse the smoothness of the edge of the lesion area), the larger the ratio of the filling area to the lesion area, and the more serious the patient's condition.

[0169] Repeat the process of obtaining the image diagnostic coefficient of Y1, and obtain the image diagnostic coefficients corresponding to the examination images Y2 to Yb respectively, to obtain the image diagnostic coefficients of Y2 to Yb;

[0170] In the existing broken line statistical graph template, the examination image is used as the horizontal coordinate, the image diagnostic coefficient is used as the vertical coordinate, and the Y1 examination image to the Yb examination image and the Y1 image diagnostic coefficient to the Yb image diagnostic coefficient are plotted in the broken line statistical graph template to obtain a diagnostic coefficient image;

[0171] The diagnostic coefficient image corresponding to each monitored patient is acquired respectively to obtain interval medical image data.

[0172] Step S3: obtaining a medical image dataset corresponding to each monitored patient based on the continuous medical image data and the intermittent medical image data, creating a monitoring image classification model to classify the medical image dataset, and obtaining monitoring image classification data;

[0173] The step S3 further includes the following specific steps:

[0174] Acquire continuous medical image data, and acquire an oxygenation index image and a respiratory index image corresponding to each monitored patient based on the continuous medical image data;

[0175] Acquire interval medical image data, and acquire a diagnostic coefficient image corresponding to each monitored patient based on the interval medical image data;

[0176] Merging the oxygenation index image, respiratory index image, and diagnostic coefficient image corresponding to the same monitored patient into a monitoring image set;

[0177] Create a surveillance image classification model;

[0178] The details are as follows:

[0179] Acquire monitoring image sets corresponding to multiple historical patients according to historical monitoring data of the intensive care unit to obtain multiple historical monitoring image sets;

[0180] It should be noted here that:

[0181] In this application, the historical monitoring image set and the image in the monitoring image set have the same image format.

[0182] Dividing the historical monitoring image set into a first type image set and a second type image set to obtain a historical monitoring image annotation set;

[0183] The details are as follows:

[0184] Selecting a sample monitoring image set from a historical monitoring image annotation set, and obtaining an oxygenation index image, a respiratory index image, and a diagnostic coefficient image based on the sample monitoring image set;

[0185] In the respiratory index image, the coordinate points corresponding to the H1 cycle respiratory index to the Ha cycle respiratory index are marked as H1 respiratory coordinate point to Ha respiratory coordinate point, and the slope between each two consecutive respiratory coordinate points between the H1 respiratory coordinate point and the Ha respiratory coordinate point is numerically obtained to obtain a-1 respiratory coordinate slope values, and the obtained a-1 respiratory coordinate slope values ​​are averaged to obtain the sample respiratory image change rate;

[0186] In the oxygenation index image, the coordinate points corresponding to the H1 cycle oxygenation index to the Ha cycle oxygenation index are marked as H1 oxygenation coordinate point to Ha oxygenation coordinate point, and the slope between each two consecutive coordinate points between the H1 oxygenation coordinate point and the Ha oxygenation coordinate point is numerically obtained to obtain a-1 oxygenation coordinate slope values, and the obtained a-1 oxygenation coordinate slope values ​​are averaged to obtain the sample oxygenation image change rate;

[0187] In the diagnostic coefficient image, the coordinate points corresponding to the Y1 image diagnostic coefficient to the Yb image diagnostic coefficient are marked as Y1 image coordinate points to Yb image coordinate points, and the slope between each two consecutive coordinate points between the Y1 image coordinate point and the Yb image coordinate point is numerically obtained, b-1 image coordinate slope values ​​are obtained, and the average of the obtained b-1 image coordinate slope values ​​is calculated to obtain the sample diagnostic image change rate;

[0188] Calculate the average of the sample respiratory image change rate, the sample oxygenation image change rate, and the sample diagnostic image change rate to obtain the comprehensive monitoring state change rate, obtain the comprehensive monitoring state change rate threshold, and compare the comprehensive monitoring state change rate with the comprehensive monitoring state change rate threshold. If the comprehensive monitoring state change rate is greater than or equal to the comprehensive monitoring state change rate threshold, mark the sample monitoring image set as a first type of image set; if the comprehensive monitoring state change rate is less than the comprehensive monitoring state change rate threshold, mark the sample monitoring image set as a second type of image set;

[0189] It should be noted here that:

[0190] The image stability in the first type of image set involved here is less than that in the second type of image set, that is, the monitoring stability of the patient corresponding to the first type of image set is less than that of the patient corresponding to the second type of image set.

[0191] Label each historical surveillance image set separately;

[0192] The historical monitoring image annotation set is divided into a monitoring image training set and a monitoring image test set according to the image training and testing ratio;

[0193] It should be noted here that:

[0194] In this application, the image training-test ratio is specifically set to 7:3, that is, the ratio of the number of medical monitoring images in the monitoring image training set and the monitoring image test set is 7:3;

[0195] Create an image recognition model using an existing artificial intelligence platform and train it using a surveillance image training set;

[0196] The image recognition model is tested using the monitoring image test set, and the recognition accuracy is obtained. When the recognition accuracy is greater than or equal to the target recognition accuracy, the image recognition model training is completed and the monitoring image classification model is obtained. When the recognition accuracy is less than the target recognition accuracy, the image recognition model is continued to be trained using the monitoring image training set until the recognition accuracy is greater than or equal to the target recognition accuracy.

[0197] Using a monitoring image classification model, the plurality of monitoring image sets are identified as a first type image set and a second type image set, thereby obtaining monitoring image classification data;

[0198] Step S4: Evaluate the monitoring status of the monitored patient based on the monitoring image classification data;

[0199] The step S4 further includes the following specific steps:

[0200] Acquire monitoring image classification data, and acquire a first type image set and a second type image set respectively according to the monitoring image classification data;

[0201] When the monitored image set corresponding to the monitored patient is a first type of image set, it is assessed that the monitored patient is in a monitoring fluctuation state;

[0202] When the monitored image set corresponding to the monitored patient is the second type of image set, it is assessed that the monitored patient is in a stable monitoring state.

[0203] In this application, if a corresponding calculation formula appears, the above calculation formula is dimensionless and its numerical calculation is performed. The weight coefficient, proportional coefficient and other coefficients in the formula are set to a result value obtained by quantifying each parameter. Regarding the size of the weight coefficient and the proportional coefficient, as long as it does not affect the proportional relationship between the parameter and the result value, it is acceptable.

[0204] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A classification method for monitoring images in a respiratory intensive care unit, characterized in that: include: Step S1: Continuously monitoring medical indicators of a patient in an intensive care unit, acquiring a respiratory index image and an oxygenation index image, and obtaining continuous medical image data; Step S2: performing interval medical index monitoring on the monitored patient in the intensive care unit, acquiring multiple medical imaging analysis images, and obtaining interval medical image data; The step S2 further includes the following specific steps: Step S21: Acquire inspection images Y1 to Yb; Step S22: performing image analysis on the Y1 examination image to obtain the Y1 image diagnosis coefficient; Step S23: acquiring the Y2 image diagnostic coefficient to the Yb image diagnostic coefficient respectively; Step S24: Using the examination image as the abscissa and the image diagnostic coefficient as the ordinate, the Y1 examination image to the Yb examination image and the Y1 image diagnostic coefficient to the Yb image diagnostic coefficient are plotted in a broken line statistical graph template to obtain a diagnostic coefficient image; Step S25: acquiring the diagnostic coefficient image corresponding to each monitored patient to obtain interval medical image data; The step S22 further includes the following specific steps: Step S221: segmenting the patient's lung organ tissue in the Y1 examination image to obtain an image lung tissue region, and segmenting the lung lesion region in the Y1 examination image using an image segmentation algorithm to obtain an image lesion region; Step S222: defining a plurality of pixel filling blocks of equal area in the Y1 examination image, using the pixel filling blocks to fill the lung tissue area of ​​the image and counting the number of lung filling blocks, and using the pixel filling blocks to fill the lesion area of ​​the image and counting the number of lesion filling blocks; Step S223: Calculate the ratio of the number of lesion filling blocks to the number of lung filling blocks to obtain the lesion area filling ratio; Step S224: performing edge smoothness analysis on the lesion area of ​​the image to obtain a quantified value of the edge smoothness of the lesion area; Step S225: Calculate the product of the quantified value of the lesion edge smoothness and the filling area ratio of the lesion area to obtain the Y1 imaging diagnostic coefficient; Step S3: Obtain a medical image dataset corresponding to each monitored patient, create a monitoring image classification model to classify the medical image dataset, and obtain monitoring image classification data; Step S4: Evaluate the monitoring status of the monitored patient based on the monitoring image classification data.

2. A classification method for monitoring images in a respiratory intensive care unit according to claim 1, characterized in that: The step S1 further includes the following specific steps: Step S11: selecting a critical care patient from multiple monitored patients as a sample monitored patient; Step S12: Mark a medical monitoring cycle and divide the medical monitoring cycle into H1 monitoring sub-cycles to Ha monitoring sub-cycles; Step S13: monitoring the respiratory index of the sample monitoring patient in the medical monitoring cycle to obtain a respiratory index image; Step S14: monitoring the oxygenation index of the sample monitoring patient in the medical monitoring cycle to obtain an oxygenation index image; Step S15: Acquire the oxygenation index image and respiratory index image corresponding to each monitored patient to obtain continuous medical image data.

3. A classification method for monitoring images in a respiratory intensive care unit according to claim 2, characterized in that: The step S13 further includes the following specific steps: Step S131: obtaining the gas volume of a single breath of the patient during the monitoring sub-cycle H1 to the monitoring sub-cycle Ha, and obtaining the tidal volume of the H1 cycle to the Ha cycle; Step S132: Obtain the cyclic respiratory frequency of the sample monitoring patient from the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, and obtain the H1 cycle respiratory frequency to the Ha cycle respiratory frequency; Step S133: obtaining the mean arterial carbon dioxide partial pressure of the sample monitoring patient during the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, and obtaining the arterial carbon dioxide partial pressure of the H1 cycle to the Ha cycle; Step S134: The periodic tidal volume VT, periodic respiratory frequency RR and periodic arterial carbon dioxide partial pressure PaCO2 corresponding to the same monitoring sub-cycle are calculated to obtain the periodic respiratory index RFA. The specific formula is: ; Step S135: naming the periodic respiratory indices corresponding to the H1 monitoring sub-cycle to the Ha monitoring sub-cycle as H1 periodic respiratory index to Ha periodic respiratory index; Step S136: Using the monitoring sub-cycle as the abscissa and the cycle respiration index as the ordinate, plot the H1 monitoring sub-cycle to the Ha monitoring sub-cycle and the H1 cycle respiration index to the Ha cycle respiration index in the broken line statistical graph template to obtain a respiration index image.

4. The classification method for monitoring images in a respiratory intensive care unit according to claim 2, characterized in that: The step S14 further includes the following specific steps: Step S141: obtaining the average airway pressure of the sample monitored patients during the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, and obtaining the airway pressure from the H1 cycle to the Ha cycle; Step S142: obtaining the oxygen inhalation concentration of each breathing process of the sample monitoring patient during the H1 monitoring sub-cycle to the Ha monitoring sub-cycle, and performing mean calculation to obtain the oxygen inhalation concentration of the H1 cycle to the Ha cycle; Step S143: obtaining the mean arterial oxygen partial pressure of the patient during the monitoring sub-cycle H1 to the monitoring sub-cycle Ha, and obtaining the arterial oxygen partial pressure of the H1 cycle to the Ha cycle; Step S144: Calculate the periodic oxygen inhalation concentration FiO2, periodic airway pressure MAP, and periodic arterial oxygen partial pressure PaO2 corresponding to the same monitoring sub-cycle to obtain the periodic oxygenation index OI. The specific formula is: ； Step S145: naming the cycle oxygenation indexes corresponding to the H1 monitoring sub-cycle to the Ha monitoring sub-cycle as H1 cycle oxygenation index to Ha cycle oxygenation index in sequence; Step S146: Using the monitoring sub-cycle as the horizontal axis and the cycle oxygenation index as the vertical axis, the H1 monitoring sub-cycle to the Ha monitoring sub-cycle and the H1 cycle oxygenation index to the Ha cycle oxygenation index are plotted in the broken line statistical graph template to obtain an oxygenation index image.

5. The classification method for monitoring images in a respiratory intensive care unit according to claim 1, characterized in that: The step S224 further includes the following specific steps: Perform edge extraction on the image lesion area in the Y1 examination image to obtain the edge line of the lesion area; Set a number of feature rectangles with equal areas to cover the edge line of the lesion area; The feature rectangles covering the edge line of the lesion area are named the first feature rectangle to the cth feature rectangle respectively; Performing a flatness analysis on the edge line of the lesion area within the first characteristic rectangle to obtain a first edge flatness quantization value; Performing flatness analysis on the edge lines of the lesion area from the second characteristic rectangle to the cth characteristic rectangle respectively, and obtaining the second edge flatness quantization value to the cth edge flatness quantization value; The first edge smoothness quantization value to the cth edge smoothness quantization value are averaged to obtain the ward edge smoothness quantization value.

6. The classification method for monitoring images in a respiratory intensive care unit according to claim 5, characterized in that: Get the first edge smoothness quantization value as follows: Mark the two opposite sides of the feature rectangle where the edge line of the lesion area passes as the first feature side and the second feature side, and mark any rectangular side between the first feature side and the second feature side as the third feature side; Mark the intersection of the edge line of the lesion area and the first characteristic edge as the first characteristic point, mark the intersection of the edge line of the lesion area and the second characteristic edge as the second characteristic point, and mark the midpoint of the third characteristic edge as the third characteristic point; Mark the circle determined by the first characteristic point, the second characteristic point, and the third characteristic point as the first characteristic circle, mark the center of the first characteristic circle as the local characteristic point of the lesion, and mark the first lesion area point to the dth lesion area point on the lesion area edge line; In the first characteristic circle, the first characteristic point and the second characteristic point are connected to obtain a lesion characteristic chord, and a length value of the lesion characteristic chord is obtained to obtain a first preset leveling distance; Connecting the first lesion area point to the dth lesion area point and the lesion local feature point, and obtaining length values ​​of the obtained multiple lines to obtain the length value of the first edge line to the dth edge line; Obtaining the radius of the first characteristic circle to obtain a second preset leveling distance; The first edge smoothness quantization value Pzd1 is obtained by calculating the first edge connection length value Blx1 to the dth edge connection length value Blxd, the first preset smoothing distance Ysj1, and the second preset smoothing distance Ysj2. The specific formula is as follows: 。 7. The classification method for monitoring images in a respiratory intensive care unit according to claim 1, characterized in that: The step S3 further includes the following specific steps: Step S31: Acquire continuous medical image data, and acquire an oxygenation index image and a respiratory index image corresponding to each monitored patient based on the continuous medical image data; Step S32: obtaining interval medical image data, and obtaining a diagnostic coefficient image corresponding to each monitored patient based on the interval medical image data; Step S33: merging the oxygenation index image, respiratory index image, and diagnostic coefficient image corresponding to the same monitored patient into a monitoring image set; Step S34: acquiring monitoring image sets corresponding to multiple historical patients based on historical monitoring data of the intensive care unit to obtain multiple historical monitoring image sets; Step S35: Divide the historical monitoring image set into a first type image set and a second type image set to obtain a historical monitoring image annotation set; Step S36: dividing the historical monitoring image annotation set into a monitoring image training set and a monitoring image test set according to the image training and testing ratio; Step S37: creating a monitoring image classification model using the monitoring image training set and the monitoring image test set; Step S38: using the monitoring image classification model to identify the plurality of monitoring image sets as a first type of image set and a second type of image set, and obtaining monitoring image classification data.

8. The classification method for monitoring images in a respiratory intensive care unit according to claim 7, characterized in that: The step S35 further includes the following specific steps: Step S351: selecting a sample monitoring image set from the historical monitoring image annotation set, and obtaining an oxygenation index image, a respiratory index image, and a diagnostic coefficient image based on the sample monitoring image set; Step S352: In the respiratory index image, the coordinate points corresponding to the H1 cycle respiratory index to the Ha cycle respiratory index are marked as H1 respiratory coordinate points to Ha respiratory coordinate points, and the slope between each two consecutive respiratory coordinate points is numerically obtained and averaged to obtain the sample respiratory image change rate; Step S353: In the oxygenation index image, the coordinate points corresponding to the H1 cycle oxygenation index to the Ha cycle oxygenation index are marked as H1 oxygenation coordinate points to Ha oxygenation coordinate points, and the slope between each two consecutive coordinate points is numerically obtained and averaged to obtain the sample oxygenation image change rate; Step S354: In the diagnostic coefficient image, the coordinate points corresponding to the Y1 image diagnostic coefficient to the Yb image diagnostic coefficient are marked as Y1 image coordinate points to Yb image coordinate points, and the slope between each two consecutive coordinate points is numerically obtained and averaged to obtain the sample diagnostic image change rate; Step S355: Calculate the average of the sample respiratory image change rate, the sample oxygenation image change rate, and the sample diagnostic image change rate to obtain a monitoring state comprehensive change rate, and obtain a monitoring state comprehensive change rate threshold. If the monitoring state comprehensive change rate is greater than or equal to the monitoring state comprehensive change rate threshold, mark the sample monitoring image set as a first type image set; if the monitoring state comprehensive change rate is less than the monitoring state comprehensive change rate threshold, mark the sample monitoring image set as a second type image set. Step S356: Mark each historical monitoring image set separately.

9. The classification method for monitoring images in a respiratory intensive care unit according to claim 1, characterized in that: The step S4 further includes the following specific steps: Step S41: Acquire monitoring image classification data, and acquire a first type image set and a second type image set according to the monitoring image classification data; Step S42: when the monitoring image set corresponding to the monitored patient is a first type of image set, it is assessed that the monitored patient is in a monitoring fluctuation state; Step S43: When the monitored image set corresponding to the monitored patient is the second type of image set, it is assessed that the monitored patient is in a stable monitoring state.

Images