Mode recognition system for operating state of drainage tube in cardiothoracic surgery department

Abstract

The invention relates to the technical field of medical state monitoring, in particular to a cardiothoracic surgery drainage tube running state mode recognition system which comprises a drainage imaging positioning module, a leading edge rhythm judgment module, a liquid column direction judgment module, a state collaborative recognition module and a running state labeling module. According to the method, the visual range of the drainage liquid in the transparent section is positioned through the continuous images, the leading edge boundary and the starting and ending positions of the liquid column are extracted, the quantitative tracking capability of the dynamic characteristics of the liquid is enhanced, and the recognition accuracy of stagnation and flowing states is improved through leading edge displacement change and rhythmicity analysis; the judgment of the length change direction of the liquid column defines the flow direction of the liquid, reduces single-frame visual misjudgment, cooperatively judges the leading edge state and the direction of the liquid column, screens out a synchronous section, improves the stability and consistency of state recognition, completes state labeling in combination with an image timestamp, constructs a sequential and continuous operation state recognition system, and improves the accuracy of state recognition. And the intelligence and accuracy of drainage state monitoring are enhanced.

Description

Technical Field

[0001] This invention relates to the field of medical condition monitoring technology, and in particular to a pattern recognition system for the operational status of a cardiothoracic surgical drainage tube. Background Technology

[0002] The field of medical status monitoring technology refers to a set of technologies for the continuous observation and judgment of patients' physiological status and the operational status of medical devices during the medical process. Its core aspects include the acquisition, discrimination, and recording of the status of target objects in the clinical environment. This technological field typically involves collecting observable information in the medical setting and identifying the status using predetermined criteria to support monitoring and management during the medical process. Among these, the pattern recognition system for the operational status of traditional cardiothoracic surgery drainage tubes refers to the technology for identifying the operational status of drainage tubes in patients after cardiothoracic surgery. Its focus is on determining whether the drainage tube is patent, blocked, dislodged, or operating abnormally. Traditional methods typically involve setting up cameras in the bedside area to acquire continuous images of the drainage tube and its connections. Based on visual features such as the position, shape, outline, and fluid flow appearance of the drainage tube in the images, the acquired features are matched with pre-defined operational status feature samples to identify and differentiate the operational status of the drainage tube.

[0003] Traditional monitoring methods rely on cameras to capture images of drainage tubes and their connections. Identification is primarily based on comparing the location, shape, and fluid appearance of the drainage tubes within the images, which suffers from insufficient capture of dynamic fluid changes. Since image analysis is largely based on feature matching of static frames, it struggles to accurately reflect the continuous movement of the fluid over time, leading to ambiguous results when determining whether the fluid is stagnant or flowing smoothly. Key behaviors such as the direction and rhythm of fluid movement are difficult to accurately identify from a single image, often resulting in confusion between flow direction and visual angle, and misjudgments of fluid surface oscillations. Feature sample matching strategies rely on manually set feature templates, which are difficult to adapt to the actual manifestations of different drainage fluid states, leading to a high misjudgment rate. For example, when the fluid exhibits intermittent microflow or no obvious morphological changes, the system may misidentify it as stagnant, affecting the timeliness of postoperative intervention. Furthermore, existing systems lack systematic integration of drainage state change trends, failing to form cross-time period identification and annotation, resulting in mostly point-like state identification lacking the ability to present continuity and overall trends. These shortcomings limit the applicability and reliability of traditional drainage monitoring methods in actual clinical settings, and can easily lead to diagnostic errors and management delays. Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention provides a pattern recognition system for the operational status of cardiothoracic surgical drainage tubes. The technical solution is as follows: On the one hand, a pattern recognition system for the operational status of cardiothoracic surgical drainage tubes is provided, the system comprising: The drainage imaging and positioning module acquires continuous images of the transparent segment of the drainage tube in the postoperative bed area, locates the imaging range of the transparent segment, extracts the visible liquid surface area of ​​the drainage fluid, determines the position of the leading edge boundary, as well as the start and end positions of the liquid column, and generates a time series of the visible position of the drainage fluid. The leading edge rhythm discrimination module extracts the changes in the leading edge boundary position at adjacent moments based on the time series of the visible position of the drainage fluid, judges the trend of the leading edge displacement state within a continuous time window, and marks the same rhythm state if the change trend is consistent, and generates a leading edge advancement rhythm state sequence in chronological order. The liquid column direction discrimination module extracts the positions of both ends of the liquid column in the visible position sequence of the drainage liquid through the corresponding timestamp, and calculates the length difference between adjacent frames to construct the liquid column length change trend. Based on the liquid column length change trend, the change direction is determined, and the changes are recorded in chronological order to generate a liquid column change direction label sequence. The state coordination recognition module makes corresponding judgments within a time window based on the sequence of liquid column change directions, filters out the intervals in which the leading edge propulsion rhythm and the liquid column change direction have a preset corresponding relationship, and generates a synchronous interval for the drainage operation status. The operation status labeling module matches the image timestamp range of the synchronization interval of the drainage operation status with the preset classification labels of the operation status of cardiothoracic surgery drainage to obtain the drainage operation status labeling results.

[0005] As a further aspect of the present invention, the time series of the visible position of the drainage fluid includes an image acquisition timestamp sequence, a leading edge boundary position sequence, a liquid column start position sequence, and a liquid column end position sequence; the leading edge advancement rhythm state sequence includes a leading edge advancement state segment, a leading edge stagnation state segment, and a sampling period consistency judgment mark; the liquid column change direction sequence includes a liquid column growth direction mark, a liquid column shortening direction mark, and a liquid column length maintenance mark; the drainage operation state synchronization interval includes a synchronization interval start timestamp, a synchronization interval end timestamp, and a consistent change relationship type; and the drainage operation state labeling result includes a state classification label, a labeling segment start timestamp, a labeling segment end timestamp, and a labeling segment number.

[0006] As a further aspect of the present invention, the drainage imaging positioning module includes: The image acquisition submodule acquires images of the transparent segment of the drainage tube in the postoperative ward area of ​​cardiothoracic surgery, identifies the area of ​​difference between the image and the background boundary, locates the position where the edge of the transparent segment appears continuously in consecutive image frames, determines the spatial boundary of the imaging range, and generates the coordinate values ​​of the imaging area of ​​the transparent segment. The liquid surface recognition submodule extracts the light and dark boundary features of the liquid in the image based on the coordinate values ​​of the transparent segment imaging area, identifies the position of the liquid front boundary within the image frame, and determines the start and end of the liquid column by combining the range of pixel grayscale changes within the region, generating the position interval values ​​of the two ends of the liquid column. The time series processing submodule arranges image frames according to the image acquisition timestamp order based on the position interval values ​​at both ends of the liquid column, associates the corresponding liquid column position change trend in the image, integrates them into continuously changing sequence data, and generates a time series of visible liquid column positions.

[0007] As a further aspect of the present invention, the leading-edge rhythm discrimination module includes: The displacement extraction submodule acquires the time series of the visible position of the drainage fluid, extracts the position values ​​of the leading edge boundary corresponding to adjacent image frames in continuous sampling time, calculates the displacement difference of the leading edge boundary between time periods, and sorts out the displacement change sequence in combination with the image acquisition time sequence to generate the leading edge displacement change trend value. The state recognition submodule identifies the movement continuity of the leading edge boundary position within a continuous time period based on the trend value of the leading edge displacement, filters data segments with the same displacement direction in adjacent time periods, judges the distribution status of the continuous displacement interval and the stationary interval, and obtains the leading edge continuity judgment result. The rhythm classification submodule calls the frontal continuity judgment result, sorts the state results within the time period in order, marks the displacement interval as the advancing state, marks the stationary interval as the stagnant state, arranges all classification results in the order of image timestamps, and generates the frontal advancing rhythm state sequence.

[0008] As a further aspect of the present invention, the liquid column direction determination module includes: The state extraction submodule obtains the state sequence of the leading-edge propulsion rhythm, filters the sampling period timestamps corresponding to the propulsion state, calls the time sequence of the visible position of the drainage fluid under the timestamp, extracts the positions of the start and end of the liquid column in the transparent segment of the corresponding image frame, and generates the liquid column position interval value. The length monitoring submodule calculates the change in liquid column length in adjacent image frames within the propulsion state cycle based on the liquid column position interval value, organizes the continuous liquid column length sequence in the propulsion and stagnation states, and obtains the liquid column change trend. The direction recognition submodule calls the liquid column change trend to determine the direction of liquid column length change within a continuous sampling period, filters time periods of rising, falling or stable states, arranges the time periods corresponding to the change types according to the image acquisition time sequence, and generates a sequence of liquid column change directions.

[0009] As a further aspect of the present invention, the state coordination recognition module includes: The segment division submodule obtains the state sequence of the leading-edge propulsion rhythm and the change direction sequence of the liquid column, establishes a unified time axis based on a common time range, sets a fixed time window to divide continuous time segments, extracts the corresponding state values ​​of the two sequences within the time segment, and generates a set of state-aligned segments. The state comparison submodule determines whether the state of the advancing front and the direction of liquid column change are consistent within each segment based on the set of state alignment segments. It then filters segments in which the advancing front and the rising liquid column occur simultaneously or the stagnant front and the stable liquid column occur simultaneously, thus obtaining a list of state consistency segments. The synchronous extraction submodule calls the state consistency segment list, extracts the start and end time values ​​of the corresponding time period in the list, integrates the consecutively occurring consistency segments and removes the discontinuous segments, establishes a continuous state maintenance interval, and generates a traffic diversion operation state synchronization interval.

[0010] As a further aspect of the present invention, the running status labeling module includes: The time matching submodule obtains the synchronization interval of the traffic diversion operation status, extracts the start and end image acquisition timestamps corresponding to the interval, formats all time periods according to a fixed time format, establishes a standardized set of time intervals, and generates a synchronization interval time index value. The tag comparison submodule calls the preset cardiothoracic surgery drainage operation status classification tags according to the synchronization interval time index value, compares the correspondence between the status type and the tag classification content within the time period, filters the intervals that meet the classification tag conditions, and obtains the status classification tag matching value. The segment labeling submodule calls the status classification label matching value, writes the label type into the corresponding synchronization interval, constructs a unified segment labeling sequence, arranges and summarizes all labeling information in chronological order, and generates the traffic diversion operation status labeling result.

[0011] As a further aspect of the present invention, the postoperative ward area is the bedside space area where the drainage tube is fixed and continuously monitored when the patient is in a bedridden monitoring state after completing thoracic and cardiovascular surgery. The transparent section of the drainage tube is a section of tube made of transparent material that is installed in the chest cavity or pericardial drainage tube and is used to observe the flow of drainage fluid. The visible liquid surface area of ​​the drainage fluid refers to the area formed by the drainage fluid in the image of the transparent section of the drainage tube, which is distinguished from the background of the tube wall in terms of image brightness or color. The starting and ending positions of the liquid column refer to the starting and ending boundary positions of the drainage liquid column along the length of the pipe within the transparent section of the drainage tube. The time series of visible drainage fluid positions is a set of the leading edge boundary positions of the drainage fluid and the starting and ending positions of the fluid column, arranged in the order of image acquisition timestamps.

[0012] As a further aspect of the present invention, the leading edge displacement state refers to the state determination result formed by whether the position of the leading edge boundary of the drainage fluid changes within adjacent sampling periods; The aforementioned leading-edge advance rhythm state sequence is a state sequence formed by leading-edge advance states and leading-edge stagnation states arranged in the order of image acquisition time.

[0013] As a further embodiment of the present invention, the fixed time window is a continuous sampling period interval set according to the image acquisition timestamp, which is used to make corresponding judgments on the state within multiple sampling periods. The sequence of liquid column change directions is a set of liquid column length extension or contraction changes arranged in chronological order of image acquisition time. The synchronous interval of the drainage operation state refers to the time segment within a continuous time range in which the rhythm of the leading edge propulsion and the direction of change of the liquid column maintain a consistent relationship. The classification labels for the operation status of cardiothoracic drainage are pre-set category identification information used to identify different drainage operation statuses based on the cardiothoracic drainage monitoring scenario.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the visible range of the drainage fluid in the transparent segment is located by continuous image positioning, and the leading edge boundary and the start and end positions of the liquid column are extracted, which enhances the quantitative tracking ability of the dynamic characteristics of the liquid. The analysis of the displacement change and rhythm of the leading edge improves the accuracy of identifying stagnation and flow states. The judgment of the direction of liquid column length change clarifies the liquid flow direction and reduces single-frame visual misjudgment. The leading edge state and liquid column direction are judged in a coordinated manner to screen out synchronous segments, which improves the stability and consistency of state recognition. The state is marked by combining image timestamps, and a time-sequential and continuous operation state recognition system is constructed, which enhances the intelligence and accuracy of drainage state monitoring. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a system block diagram of the present invention; Figure 3 This is a flowchart of the drainage imaging positioning module in this invention; Figure 4This is a flowchart of the leading-edge rhythm discrimination module in this invention; Figure 5 This is a flowchart of the liquid column direction determination module in this invention; Figure 6 This is a flowchart of the state coordination identification module in this invention; Figure 7 This is a flowchart of the running status labeling module in this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides a pattern recognition system for the operational status of cardiothoracic surgical drainage tubes, such as... Figure 1-2 The diagram shows a pattern recognition system for the operational status of a cardiothoracic surgical drainage tube. The system includes: The drainage imaging and positioning module acquires continuous images of the transparent segment of the pleural or pericardial drainage tube within the postoperative bed area of ​​cardiothoracic surgery, locates the imaging range of the transparent segment of the drainage tube, extracts the visible fluid surface area, determines the position of the leading edge boundary of the drainage fluid and the positions of the beginning and end of the fluid column, and organizes them according to the image acquisition timestamp order to generate a time series of the visible position of the drainage fluid. The postoperative ward area for cardiothoracic surgery is the bedside space where drainage tubes are fixed and continuously monitored when a patient is in bedridden monitoring after completing cardiothoracic surgery. The transparent section of a drainage tube is a section of a tube that is placed in a chest or pericardial drainage tube, made of transparent material, and used to observe the flow of drainage fluid. The visible fluid surface area refers to the area formed by the drainage fluid in the transparent section of the drainage tube and distinguished from the background of the tube wall in terms of image brightness or color. The position of the leading edge boundary of the drainage fluid refers to the position of the frontmost liquid surface boundary of the drainage fluid in the image along the flow direction within the transparent section of the drainage tube; The starting and ending positions of the liquid column refer to the starting and ending boundary positions of the drainage liquid column along the length of the pipe within the transparent section of the drainage tube. The time series of visible drainage fluid positions is a collection of the leading edge boundary positions of the drainage fluid and the starting and ending positions of the fluid column, arranged in the order of image acquisition timestamps. The leading edge rhythm discrimination module extracts the changes in the boundary position of the leading edge of the drainage fluid at adjacent sampling times based on the time series of the visible position of the drainage fluid, calculates the leading edge displacement change state, and combines it with the image acquisition timestamp to form a continuous time sequence. It makes a consistency judgment on the leading edge displacement change state within the continuous sampling period, forms a distinction between leading edge advancement and leading edge stagnation, and generates a leading edge advancement rhythm state sequence. The state of frontal displacement change refers to the state determination result formed by whether the position of the frontal boundary of the drainage fluid changes within adjacent sampling periods; The advancing state and the stagnant state are two types of leading motion states distinguished by the change in the leading displacement. The advancing state indicates that the position of the leading boundary changes, while the stagnant state indicates that the position of the leading boundary remains unchanged. The liquid column direction discrimination module is based on the leading edge advance rhythm state sequence. It extracts the positions of the two ends of the liquid column in the visible position sequence of the drainage liquid through the corresponding timestamp, and calculates the length difference between adjacent frames to construct the liquid column length change trend. Based on the liquid column length change trend, the change direction is determined, and the changes are recorded in time order to generate a liquid column change direction label sequence. The leading-edge advance rhythm state sequence is a state sequence formed by leading-edge advance states and leading-edge stagnation states arranged in the order of image acquisition time; The state coordination identification module is based on the state sequence of the leading edge propulsion rhythm and the liquid column change direction sequence. Within a fixed time window, it makes corresponding judgments on the state of the leading edge propulsion and the change direction of the liquid column, and filters the intervals in which the two maintain a consistent change relationship within the same time period to generate the synchronous interval of the drainage operation state. A fixed time window is a continuous sampling period interval set based on the image acquisition timestamp, used to make corresponding judgments on the state within multiple sampling periods; The liquid column length maintenance state refers to the liquid column length state in which the positions of the start and end ends of the liquid column do not change within adjacent sampling periods; The sequence of liquid column change directions is a set of liquid column length extension or contraction directions arranged in chronological order of image acquisition time. The consistent change relationship refers to the state correspondence that occurs simultaneously when the forward propulsion state and the liquid column change direction are both established within the same time interval. The operation status labeling module matches and judges the synchronization interval with the preset cardiothoracic surgery drainage operation status classification labels based on the image acquisition timestamp range corresponding to the synchronization interval of the drainage operation status, and completes the segment labeling of the drainage tube operation status to obtain the drainage operation status labeling results. The synchronous interval of the drainage operation state refers to the time segment within a continuous time range in which the rhythm of the leading edge propulsion and the direction of change of the liquid column maintain a consistent relationship. The classification labels for the operation status of cardiothoracic drainage are pre-set category identification information used to identify different drainage operation statuses based on the cardiothoracic drainage monitoring scenario.

[0023] The time series of the visible location of the drainage fluid includes the image acquisition timestamp sequence, the leading edge boundary position sequence, the liquid column start position sequence, and the liquid column end position sequence. The leading edge advancement rhythm state sequence includes the leading edge advancement state segment, the leading edge stagnation state segment, and the sampling period consistency judgment mark. The liquid column change direction sequence includes the liquid column growth direction mark, the liquid column shortening direction mark, and the liquid column length maintenance mark. The drainage operation status synchronization interval includes the synchronization interval start timestamp, the synchronization interval end timestamp, and the consistent change relationship type. The drainage operation status annotation results include the status classification label, the annotation segment start timestamp, the annotation segment end timestamp, and the annotation segment number.

[0024] Specifically, such as Figure 2 , 3 As shown, the drainage imaging positioning module includes: The image acquisition submodule acquires images of the transparent segment of the drainage tube in the postoperative ward area of ​​cardiothoracic surgery, identifies areas in the image that are significantly different from the background boundary, locates the position where the edge of the transparent segment appears continuously in consecutive image frames, determines the spatial boundary of the imaging range, and generates the coordinate values ​​of the imaging area of ​​the transparent segment. A high-definition camera fixed to the side of the hospital bed was used, with a frame rate of 10 frames per second to ensure image continuity. The acquisition area covered approximately 30cm from the patient's side to encompass the entire transparent segment of the drainage tube. After the images entered the processing flow, the region of interest was first defined to narrow down the analysis scope. Then, edge detection was performed, using algorithms such as Canny to identify boundaries with significant grayscale changes in the image. Thresholds were set with an upper limit of 100 and a lower limit of 50 to ensure that only clear boundaries were retained. After edge extraction, morphological closing operations were used to eliminate breaks and small noise contours. Through grayscale analysis, areas with significant grayscale fluctuations were selected as candidate regions for the transparent segment. Then, each... The coordinates of candidate region edges in a frame of an image are statistically analyzed, and the frequency of each edge point in the image sequence is recorded. When the proportion of an edge point in consecutive frames exceeds 60%, it is determined to be a stable boundary point. All stable points are summarized, and the maximum and minimum values ​​of their horizontal and vertical coordinates are extracted to form the upper left and lower right corner coordinates of a rectangular region, which are used as the boundary values ​​of the transparent segment imaging region. For example, if the frequently occurring horizontal coordinate range is 125 to 275 and the vertical coordinate range is 80 to 210, then the coordinates of the transparent segment region are upper left corner (125, 80) and lower right corner (275, 210), thus clarifying the spatial region range required for image analysis.

[0025] The liquid surface recognition submodule extracts the light and dark boundary features of the liquid in the image based on the coordinate values ​​of the transparent segment imaging area, identifies the position of the liquid front boundary within the image frame, and determines the start and end of the liquid column by combining the range of pixel grayscale changes within the region, generating the position interval values ​​of the two ends of the liquid column. The system receives the coordinates of the transparent segment imaging region, crops the image of this region for subsequent analysis, traverses the image column data within the region frame by frame, calculates the vertical difference of gray values ​​by column, identifies gray value abrupt change points as candidate locations of the liquid-gas interface, sets a gray value change threshold of 30, and records any point with a change greater than this value as a candidate boundary. Among all candidate points, density clustering is further used to identify continuously concentrated boundary points, with clustering parameters set to a neighborhood radius of 3 pixels and a minimum cluster number of 5, to filter out the true boundary locations. Then, the system analyzes the average gray value of the entire column within the transparent region, comparing it column by column from the top of the image downwards with the average gray value of the transparent gas region, setting a baseline offset value, for example, adding 20, and finding the first row exceeding this baseline value as the starting point of the liquid column. The analysis continues downwards, finding the lowest boundary continuously below the baseline value as the ending point of the liquid column. If the starting position in the Y direction of the image is 130 and the ending position is 210, then the liquid column interval is 130 to 210, and this position interval will be used for subsequent location analysis.

[0026] The time series processing submodule arranges image frames according to the image acquisition timestamp based on the position interval values ​​at both ends of the liquid column, associates the corresponding liquid column position change trend in the image, integrates them into continuously changing sequence data, and generates a time series of visible liquid column positions. The system receives the upper and lower boundary values ​​of the liquid column and timestamp information from image frames. It then arranges the start and end coordinates of each image frame in chronological order, recording them as three data items: timestamp, start position, and end position. For example, 10:00:00 corresponds to start coordinate 130 and end coordinate 210, calculating the liquid column length in that frame as 80 pixels. Similarly, at 10:00:01, coordinates range from 130 to 215, corresponding to a liquid column length of 85 pixels, and so on, forming a continuously changing time series. To facilitate subsequent trend analysis, a sliding window is applied to this series, with a window width of 5 frames. The average liquid column length is calculated for every 5 frames, generating a smooth change curve. Data storage uses a three-field format: time, upper edge position of the liquid column, lower edge position of the liquid column, and the calculated liquid column length. Finally, the system outputs the dynamic change data of the liquid column in the image sequence, forming a time series of the visible liquid column position.

[0027] Specifically, such as Figure 2 , 4 As shown, the leading edge rhythm discrimination module includes: The displacement extraction submodule acquires the time series of the visible position of the drainage fluid, extracts the position values ​​of the leading edge boundary corresponding to adjacent image frames in continuous sampling time, calculates the displacement difference of the leading edge boundary between time periods, and sorts out the displacement change sequence in combination with the image acquisition time sequence to generate the leading edge displacement change trend value. To obtain the time series of the visible location of the drainage fluid, the vertical coordinates of the leading edge boundary in each frame of the image sequence are first extracted. The image sampling frequency is set to 10 frames per second, so the time interval between adjacent frames is 0.1 seconds. During the extraction process, the leading edge position in each frame is marked. For example, the leading edge position in frame 1 is Y1=140, and in frame 2 it is Y2=143. The displacement difference between the two frames is 3 pixels. When converting pixel units to actual physical units, an image resolution scaling factor needs to be used. For example, after camera calibration, each pixel corresponds to an actual length of 0.2mm, so the displacement difference is 0.6mm. This process is repeated for the entire image sequence. Frame displacement calculation involves subtracting each displacement value from the previous frame and recording it as a difference sequence. All calculated displacement data are then sorted in ascending order by image timestamp, forming a continuous displacement difference sequence with time as the main axis. Each record is formatted as {timestamp t_i, leading edge position Y_i, displacement ΔY_i}. For example, if Y=140 at 10:00:00.0, Y=147 at 10:00:00.1, the displacement ΔY is 7 pixels, or 1.4mm. Continuing to obtain the value at 10:00:02, Y=147, and ΔY is 0 pixels. After arranging the above sequence by time, the complete leading edge displacement trend value is output.

[0028] The state recognition submodule identifies the continuity of movement of the leading edge boundary position within a continuous time period based on the trend value of the leading edge displacement, filters data segments with the same displacement direction in adjacent time periods, judges the distribution status of the continuous displacement interval and the stationary interval, and obtains the leading edge continuity judgment result. Based on the trend value of the displacement change of the leading edge, the motion state of the leading edge boundary on the time axis is analyzed. First, the direction of the continuous displacement difference sequence is determined. The judgment logic is set as follows: when the ΔY value is greater than the set positive threshold, it is positive motion; when it is less than the set negative threshold, it is negative motion. If the absolute value of the ΔY value is less than the stationary threshold, it is considered that there is no obvious displacement. The stationary threshold is set to 1 pixel, corresponding to a physical length of 0.2 mm. Data segments with the same ΔY direction at adjacent time points are filtered. For example, if ΔY is positive or negative in 5 consecutive frames, it is determined to be a continuous displacement segment. Conversely, if ΔY is 0 in consecutive frames or fluctuates within the stationary threshold, it is determined to be a stationary segment. The start and end times of each segment are determined according to the timestamp. For example, if ΔY is 2, 3, 4, and 3 pixels respectively from 10:00:00 to 10:00:03, and the direction is consistent and positive, it is a continuous displacement segment. Then, the distribution of all continuous motion segments and stationary segments in the whole time period is statistically analyzed, and a data structure is generated to record the time range, displacement direction, and state type of each segment, thereby obtaining the leading edge continuity judgment result.

[0029] The rhythm classification submodule calls the frontal continuity judgment result, sorts the state results within the time period in order, marks the displacement interval as the advancing state, marks the stationary interval as the stagnant state, arranges all classification results in the order of image timestamps, and generates the frontal advancing rhythm state sequence. The motion states in the time series are classified by calling the results of the frontal continuity judgment. The identified continuous motion segments are uniformly marked as advancing states, and the corresponding time range is consistent with the original data. The continuous stationary segments are uniformly marked as stationary states. The entire sequence is organized in the order of timestamps to form a complete state flow. For example, 10:00:00 to 10:00:03 is marked as advancing, 10:00:03 to 10:00:06 is marked as stationary, 10:00:06 to 10:00:10 is marked as advancing again, and so on. All classification results are recorded. Each record includes start time, end time, and state label. All records are arranged in ascending order of time in list form. The final sequence is the frontal advance rhythm state sequence.

[0030] Specifically, such as Figure 2 , 5 As shown, the liquid column direction determination module includes: The state extraction submodule obtains the state sequence of the leading-edge propulsion rhythm, filters the sampling period timestamps corresponding to the propulsion state, calls the time sequence of the visible position of the drainage fluid under the timestamp, extracts the positions of the start and end of the liquid column in the transparent segment of the corresponding image frame, and generates the liquid column position interval value. After obtaining the leading-edge propulsion rhythm state sequence, the time periods marked as propulsion states are extracted one by one. The sampling timestamps corresponding to all propulsion states are obtained, and the image frame numbers corresponding to the corresponding times are filtered out in chronological order. The parsed time series information of the visible liquid column position in these image frames is then used to extract the start and end positions of the liquid column within the transparent segment of the corresponding image. The leading-edge propulsion state is strongly correlated with the liquid column movement process; therefore, it is necessary to ensure that the propulsion state timestamps are perfectly aligned with the image acquisition time. For example, when the propulsion state time period is from 10:00:00 to 10:00:04, the corresponding image frame number is... From 100 to 140, the image sampling frequency is 10 frames per second. Each frame of the image has a clear timestamp. The upper and lower boundary positions are extracted from the liquid column position information frame by frame. For example, if the starting position of the first frame is Y1=130 and the ending position is Y2=210, the liquid column interval is 80 pixels. If the image resolution is 1 pixel corresponds to 0.2 mm, then the corresponding liquid column length is 16 mm. The upper and lower boundary coordinates of the liquid column in all propulsion frames are extracted in sequence and organized into a structured data list. Each record contains three items: timestamp, starting position, and ending position. Finally, the liquid column position interval value corresponding to each image frame in the propulsion state is generated.

[0031] The length monitoring submodule calculates the change in liquid column length in adjacent image frames within the propulsion state cycle based on the liquid column position interval value, organizes the continuous liquid column length sequence in the propulsion and stagnation states, and obtains the liquid column change trend. The change in liquid column length is calculated based on the interval values ​​of the liquid column position. First, the length of the liquid column in each frame is defined as the difference between the upper and lower boundary positions. For example, if the starting coordinate in an image frame is 130 and the ending coordinate is 210, then the liquid column length is 80 pixels. The length difference between adjacent frames is calculated. For example, if the liquid column length in frame 1 is 80 pixels and in frame 2 is 84 pixels, then the change is 4 pixels. Multiplying this by the pixel conversion ratio of 0.2mm, the change is 0.8mm. For each image frame in the advancing state, the change between adjacent frames needs to be calculated to form a length change sequence. At the same time, the same process is performed on the image frames in the stationary state time period. The sample extraction process is as follows, but in this state, only the stability of the liquid column length is recorded. The difference between adjacent frames is not calculated. Instead, the change range is checked to see if it is within the set threshold range. For example, the allowable fluctuation in the static state is no more than 2 pixels, i.e. 0.4mm. If the length difference between consecutive frames does not exceed this threshold, it is considered to remain unchanged. Finally, the liquid column length data of all advancing and stagnant segments are recorded as two sets of time series. The advancing segment sequence includes three items: time, liquid column length, and length difference. The stagnant segment sequence includes two items: time and liquid column length. By sorting and arranging the two types of data, the continuous change trend of the liquid column over time is obtained.

[0032] The direction recognition submodule calls the liquid column change trend to determine the direction of liquid column length change within a continuous sampling period, filters the time periods of rising, falling or stable state, arranges the time periods corresponding to the change types according to the image acquisition time sequence, and generates a sequence of liquid column change directions. After retrieving the liquid column change trend, the direction of the liquid column length change values ​​within all continuous sampling periods is determined. The liquid column length sequence is read in the order of image timestamps. For each pair of adjacent image frames, the length difference ΔL is calculated. A positive judgment condition is set as ΔL greater than a positive threshold, and a negative judgment condition is set as ΔL less than a negative threshold. Both positive and negative thresholds are set to 2 pixels, corresponding to 0.4 mm. If ΔL is greater than 2 pixels, it is marked as rising; if it is less than -2 pixels, it is marked as falling; if it fluctuates between plus or minus 2 pixels, it is marked as a stable state. Each time period is classified into three categories: rising, falling, and stable. For example, if the lengths of the liquid column between 10:00:00 and 10:00:03 are 80, 85, 90, and 92 pixels respectively, then ΔL is 5, 5, and 2 pixels respectively, all in a positive direction. Therefore, this period is judged as an upward state. If the lengths of the liquid column between 10:00:03 and 10:00:06 are 92, 91, and 92 pixels respectively, then ΔL is -1 and 1 pixel respectively, which is within the set fluctuation range and is judged as a stable state. These classification results are arranged sequentially along the entire time axis. Each result contains three fields: start time, end time, and change type, ultimately forming a sequence of liquid column change direction.

[0033] Specifically, such as Figure 2 , 6 As shown, the state coordination recognition module includes: The segment division submodule obtains the state sequence of the leading-edge propulsion rhythm and the liquid column change direction sequence, establishes a unified time axis based on a common time range, sets a fixed time window to divide continuous time segments, extracts the corresponding state values ​​of the two sequences within the time segment, and generates a set of state-aligned segments. To obtain the propulsion rhythm state sequence and the liquid column change direction sequence, firstly, extract the timestamp range of each sequence. Determine the intersection of the two timestamps as the unified analysis time interval. For example, if the time range of the propulsion rhythm state sequence is 10:00:00 to 10:10:00, and the time range of the liquid column change direction sequence is 10:01:00 to 10:09:00, then the unified time axis range is set to 10:01:00 to 10:09:00. Then, a fixed time window is set for segmentation. A commonly used window length is 30 seconds, meaning each 30-second interval is considered an analysis unit. The calculation method is to divide the unified time range into equal parts according to a set interval. If the starting time is 10:01:00, then the segments are divided sequentially. The time intervals are divided sequentially, from 10:01:00 to 10:01:30, 10:01:30 to 10:02:00, etc., until 10:09:00. Within each time interval, the advancing or stagnant state is extracted from the advancing rhythm sequence, and the rising, falling, or stable state is extracted from the liquid column direction sequence, forming a two-dimensional state combination item. The data structure can be defined as {start of time interval, end of time interval, advancing state, liquid column state}. For example, if the advancing state is "advancing" and the liquid column state is "rising" in the interval from 10:01:00 to 10:01:30, then the corresponding state combination for this interval is {advancing, rising}. The state combinations in all intervals are extracted in this way, and finally, a set of state-aligned intervals is generated.

[0034] The state comparison submodule determines whether the state of the forward propulsion and the direction of liquid column change are consistent within each segment based on the set of state alignment segments. It then filters segments where the forward propulsion and liquid column rise occur simultaneously or where the forward stagnation and liquid column stabilization occur simultaneously, thus obtaining a list of state consistency segments. Based on the set of state alignment segments, state consistency is determined. For each segment, the advancing state and the liquid column state are logically judged. The consistency condition is set to two combinations: the advancing state and the rising liquid column occur simultaneously, or the stagnant state and the stable liquid column occur simultaneously. Other combinations, such as advancing and descending, advancing and stable, stagnant and rising, or stagnant and descending, are considered inconsistent states. The judgment process matches the state combinations within each segment. For example, if the advancing state is advancing and the liquid column state is rising within the segment from 10:02:00 to 10:02:30, it meets one of the consistency conditions and is recorded as a consistent segment. If the next segment is advancing and stable, it is excluded. The program is set to... The consistency judgment logic is as follows: if the propulsion state value = 1 and the liquid column state value = 1, or the propulsion state value = 0 and the liquid column state value = 0, then it is recorded as a consistent segment. The state values ​​are defined as: propulsion = 1, stagnation = 0, liquid column rise = 1, stability = 0, and fall = -1. This logic matching operation is performed on all segments to filter out the time periods that meet the conditions. The start and end times and the corresponding state combinations are recorded in sequence, and the output is a list of consistent segments. For example, the list content can be {10:02:00 to 10:02:30, propulsion-rise}, {10:04:30 to 10:05:00, stagnation-stability}, etc., which finally form a list of state consistent segments.

[0035] Synchronously extract the list of consistent segments in the submodule call status, extract the start and end time values ​​of the corresponding time period in the list, integrate the consecutively occurring consistent segments and remove the intermittent segments, establish a continuous status maintenance interval, and generate a synchronous interval for the traffic diversion operation status. The system retrieves the state consistency segment list, extracts the start and end times for each record, and arranges all records in chronological order. It then determines whether adjacent segments are continuous, defining continuity as the end time of the current segment equals the start time of the next segment. For example, segment 1 (10:02:00 to 10:02:30) and segment 2 (10:02:30 to 10:03:00) are considered continuous. If there is a time interval between two segments, such as segment 3 (10:03:30 to 10:04:00) with a 30-second gap, it is considered a non-continuous segment. All continuous segments are then merged into a single complete segment. For continuous state intervals, the start time of the first segment and the end time of the last segment are taken as the boundary of the merged interval. The minimum interval threshold for interruption judgment is set to be equal to the window length. That is, as long as the interval is greater than or equal to 30 seconds, the discontinuous segment is removed. The final output is several continuous and consistent interval segments. Each record includes the start time, end time and total duration, forming a synchronous interval structure for the traffic diversion operation status, such as {10:02:00 to 10:03:00, duration 60 seconds}, {10:05:00 to 10:06:30, duration 90 seconds}, etc., which finally generate the synchronous interval for the traffic diversion operation status.

[0036] Specifically, such as Figure 2 , 7 As shown, the running status labeling module includes: The time matching submodule obtains the synchronization interval of the traffic acquisition operation status, extracts the start and end image acquisition timestamps corresponding to the interval, formats all time periods according to a fixed time format, establishes a standardized set of time intervals, and generates the synchronization interval time index value. After obtaining the synchronization interval of the traffic acquisition operation status, the start and end times of each interval are first extracted. The corresponding time values ​​are derived from the timestamp information attached to each frame of the image acquisition system. Each timestamp is uniformly formatted as "yyyy-MM-dd HH:mm:ss". To ensure format consistency, all time values ​​need to be standardized. Time standardization uses a time format function to unify all time fields into string format and rounds down to the nearest second. For example, if the timestamp of a certain frame is "2022-12-22 10:15:34.527", then the standardized time is "2022-12-22". "10:15:34", and then, taking the synchronization interval as a unit, the start and end times of each interval are standardized. The extracted results are summarized into a unified set of time intervals. Each item contains two fields: standard start time and standard end time. Then, based on the time sampling frequency (e.g., 16.67 frames per second), the range of image frame numbers corresponding to the interval is derived. The time is mapped to the image sequence index by looking up a table or converting it with a formula. If the start time is 10:15:00, the corresponding frame number is 15000, and the end time is 10:16:00, the corresponding frame number is 16000. Then, the image frames corresponding to this time interval are frames 15000 to 16000. This frame index is the synchronization interval time index value.

[0037] The tag comparison submodule calls the preset thoracic and cardiothoracic drainage operation status classification tags based on the synchronization interval time index value, compares the correspondence between the status type and the tag classification content within the time period, filters the intervals that meet the classification tag conditions, and obtains the status classification tag matching value. Based on the synchronization interval time index value, a status label comparison is performed for each synchronization time period. First, a pre-set label library for classifying the operational status of cardiothoracic drainage is established. This label library is composed of expert experience data, with typical labels including "stable drainage period," "intermittent peristaltic period," and "flow stagnation period." The status combination characteristics corresponding to each label are defined. For example, "stable drainage period" corresponds to a propulsive state lasting longer than 60 seconds and a fluid column rising state proportion exceeding 80%; "intermittent peristaltic period" requires frequent state transitions with each transition lasting no more than 30 seconds; and "flow stagnation period" corresponds to a stagnant and stable fluid column state lasting longer than 90 seconds. These judgment rules are then applied to assess the status within each synchronization time period. The system matches the state structure. First, it calculates the duration of the advancing and stagnant states and the proportion of rising and stable liquid columns within each time period. Then, it calculates the cumulative number of seconds for each state and the percentage of each state using a simple statistical formula. For example, if a certain time period lasts for 60 seconds, with the advancing state accounting for 40 seconds and the rising state accounting for 45 seconds, then the advancing state accounts for 66% and the rising state accounts for 75%. This segment does not meet any label threshold and is not considered a successful classification. If another time period has an advancing state of 50 seconds and a rising state of 55 seconds, with percentages of 83% and 86% respectively, it meets the "stable drainage period" condition, and the label for that time period is successfully matched. Finally, the same process is performed on all segments to generate the corresponding state classification label matching value.

[0038] The segment labeling submodule calls the status classification label matching value, writes the label type into the corresponding synchronization interval, constructs a unified segment labeling sequence, arranges and summarizes all labeling information in chronological order, and generates the traffic diversion operation status labeling result. The system calls the status classification label matching value and establishes a mapping relationship between each matched label result and the corresponding time segment. First, it iterates through all successfully matched time segments and writes the label value into the original synchronization interval data structure. A new field "label type" is added, which contains the name of the successfully matched label. For example, the segment 10:15:00 to 10:16:30 is labeled as "stable drainage period". If a segment does not match any label, it is labeled as null or set to "undefined". After all segments are labeled, the segments are reordered in chronological order to ensure that the output sequence is strictly arranged in chronological order. Then, all labeling results are output in a structured format. The structure contains three items: start time, end time, and status label. It is stored as a label sequence by segment, such as {10:15:00 to 10:16:30, stable drainage period}, {10:16:30 to 10:17:00, undefined}, {10:17:00 to 10:18:00, flow stagnation period}, etc. Finally, a complete drainage operation status labeling result is generated.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.

Claims

1. A pattern recognition system for the operating state of a cardiothoracic surgery drainage tube, characterized by The system comprises: The drainage imaging positioning module acquires continuous images of the transparent section of the drainage tube in the postoperative bed area, positions the imaging range of the transparent section, extracts the visible liquid level area of the drainage fluid, judges the position of the leading edge boundary, and the starting position and ending position of the liquid column, and generates a time sequence of the visible position of the drainage fluid; The leading edge rhythm discrimination module extracts the change of the leading edge boundary position at adjacent time points based on the time sequence of the visible position of the drainage fluid, judges the trend of the leading edge displacement state in the continuous time window, and if the change trend remains consistent, marks it as the same rhythm state, and generates a sequence of the leading edge advancing rhythm state in time sequence; The liquid column direction discrimination module extracts the positions of the two ends of the liquid column in the visible position sequence of the drainage fluid through the corresponding time stamp based on the sequence of the leading edge advancing rhythm state, calculates the length difference between adjacent frames, constructs the change trend of the length of the liquid column, judges the change direction according to the change trend of the length of the liquid column, records in time sequence, and generates a sequence of the change direction label of the liquid column; The state cooperative recognition module judges and filters the interval where the leading edge advancing rhythm and the change direction of the liquid column present a preset corresponding relationship based on the sequence of the change direction of the liquid column in a time window, and generates a synchronous interval of the drainage running state; The running state labeling module matches the image time stamp range of the synchronous interval of the drainage running state with the preset thoracic and cardiovascular surgery drainage running state classification label to obtain a drainage running state labeling result.

2. The pattern recognition system for the running state of a drainage tube in cardiothoracic surgery according to claim 1, characterized in that: The time sequence of the visible position of the drainage fluid includes an image acquisition time stamp sequence, a leading edge boundary position sequence, a liquid column starting end position sequence, and a liquid column ending end position sequence, the sequence of the leading edge advancing rhythm state includes a leading edge advancing state section, a leading edge stagnation state section, and a sampling cycle consistency judgment mark, the sequence of the change direction of the liquid column includes a liquid column growth direction mark, a liquid column shortening direction mark, and a liquid column length keeping mark, the synchronous interval of the drainage running state includes a synchronous interval starting time stamp, a synchronous interval ending time stamp, and a consistent change relationship type, and the drainage running state labeling result includes a state classification label, a labeled section starting time stamp, a labeled section ending time stamp, and a labeled section number.

3. The pattern recognition system for the running state of a drainage tube in cardiothoracic surgery according to claim 1, characterized in that, The drainage imaging positioning module comprises: The image acquisition sub-module acquires the image of the transparent section of the drainage tube in the postoperative bed area of thoracic and cardiovascular surgery, identifies the area with boundary difference in the image, locates the position where the edge of the continuous image frame appears continuously, judges the spatial boundary of the imaging range, and generates the coordinate value of the imaging area of the transparent section; The liquid level recognition sub-module extracts the liquid light-dark boundary feature in the image according to the coordinate value of the imaging area of the transparent section, identifies the leading edge boundary position in the image frame, judges the starting end and ending end of the liquid column in combination with the gray scale change range of the area, and generates the interval value of the position of the liquid column at two ends; The time sequence arrangement sub-module arranges the image frames in the order of the image acquisition time stamp based on the interval value of the position of the liquid column at two ends, associates the change trend of the position of the liquid column in the corresponding image, integrates it into sequence data with continuous change, and generates the time sequence of the visible position of the liquid column.

4. The system according to claim 3, wherein the system is characterized by The leading edge rhythm discrimination module comprises: The displacement extraction submodule obtains the visible position time sequence of the drainage liquid, extracts the position values of the front boundary corresponding to adjacent image frames in continuous sampling time points, calculates the displacement difference value of the front boundary between the front and rear time points, arranges the displacement change sequence in combination with the image acquisition time sequence, and generates the front displacement change trend value; The state recognition submodule recognizes the movement continuity of the front boundary position in the continuous time period according to the front displacement change trend value, screens the data segments with consistent displacement directions in adjacent time periods, judges the distribution state of the displacement continuous interval and the static interval, and obtains the front continuity judgment result; The rhythm classification submodule calls the front continuity judgment result, sequentially classifies the state results in the time period, marks the displacement interval as a pushing state and the static interval as a stagnation state, arranges all classification results in the image timestamp sequence, and generates the front pushing rhythm state sequence.

5. A pattern recognition system for the running state of a drainage tube in cardiothoracic surgery according to claim 4, characterized in that The liquid column direction discrimination module comprises: The state extraction submodule obtains the front pushing rhythm state sequence, screens the sampling period time stamps corresponding to the pushing state, calls the visible position time sequence of the drainage liquid at the time stamp, extracts the starting end and ending end positions of the liquid column in the transparent section in the corresponding image frame, and generates the liquid column position interval value; The length monitoring submodule calculates the liquid column length change value in the adjacent image frames in the pushing state period according to the liquid column position interval value, arranges the continuous liquid column length sequence in the pushing and stagnation states, and obtains the liquid column change trend; The direction recognition submodule calls the liquid column change trend, judges the liquid column length change direction in the continuous sampling period, screens the time period in the rising, falling or stable state, arranges the time period corresponding change type according to the image acquisition time sequence, and generates the liquid column change direction sequence.

6. A pattern recognition system for the running state of a drainage tube in cardiothoracic surgery according to claim 5, characterized in that The state cooperative recognition module comprises: The section division submodule obtains the front pushing rhythm state sequence and the liquid column change direction sequence, establishes a unified time axis based on the common time range, sets a fixed time window to divide the continuous time section, respectively extracts the corresponding state values of the two sequences in the time section, and generates a state alignment section set; The state comparison submodule judges whether there is consistency state between the front pushing state and the liquid column change direction in each section according to the state alignment section set, screens the sections in which the front pushing and the liquid column rising appear at the same time or the front stagnation and the liquid column stable appear at the same time, and obtains a state consistency section list; The synchronization extraction submodule calls the state consistency section list, extracts the start and end time values of the corresponding time period in the list, integrates the continuously appearing consistency sections and eliminates the intermittent sections, establishes a continuous state maintenance interval, and generates a drainage running state synchronization interval.

7. A pattern recognition system for the running state of a drainage tube in cardiothoracic surgery according to claim 6, characterized in that The running state labeling module comprises: The time matching submodule obtains the drainage running state synchronization interval, extracts the start and end image acquisition time stamps corresponding to the interval, formats all time periods in a fixed time format, establishes a standardized time interval set, and generates a synchronization interval time index value; The label comparison submodule calls a preset thoracic and cardiac surgery drainage running state classification label according to the synchronization interval time index value, compares the corresponding relationship between the state type in the time period and the label classification content in sequence, filters the interval that meets the classification label condition, and obtains a state classification label matching value; The section labeling submodule calls the state classification label matching value, writes the label type into the corresponding synchronization interval, constructs a unified section labeling sequence, arranges and summarizes all labeling information in time sequence, and generates a drainage running state labeling result.

8. The pattern recognition system for the running state of a drainage tube for cardiothoracic surgery according to claim 1, characterized by The postoperative bed area is a bedside space area where the drainage tube is fixed and continuously monitored when the patient is in a bedridden monitoring state after completing a thoracic and cardiac surgery; the transparent section of the drainage tube is a section of the tube made of transparent material and used for observing the flow state of the drainage fluid; the visible fluid level area of the drainage fluid refers to an area formed by the drainage fluid in the image of the transparent section of the drainage tube and distinguished from the tube wall background in image brightness or color; the start and end positions of the liquid column refer to the corresponding start boundary position and end boundary position of the liquid column in the length direction of the pipeline in the transparent section of the drainage tube; the visible position time sequence of the drainage fluid is a collection of the leading edge boundary position of the drainage fluid and the start and end positions of the liquid column arranged in the order of image acquisition time stamp.

9. The pattern recognition system for the running state of a drainage tube for cardiothoracic surgery according to claim 1, characterized by The front edge displacement state refers to the state determination result formed by whether the leading edge boundary position of the drainage fluid changes in adjacent sampling periods; The front edge advancing rhythm state sequence is a state sequence formed by the front edge advancing state and the front edge stagnation state arranged in the order of image acquisition time.

10. The pattern recognition system for the running state of a drainage tube for cardiothoracic surgery according to claim 6, characterized by The fixed time window is a continuous sampling period interval set according to the image acquisition time stamp, used for corresponding judgment of the state in multiple sampling periods; The liquid column change direction sequence is a direction state collection of the extension change or retraction change of the liquid column length arranged in the order of image acquisition time; The drainage running state synchronization interval refers to a time section in which the front edge advancing rhythm state and the liquid column change direction maintain a consistent change relationship within a continuous time range; The thoracic and cardiac surgery drainage running state classification label is a category identification information used to identify different drainage running states preset according to the thoracic and cardiac surgery drainage monitoring scene.

Images