Wireless alarm general-purpose ECMO tube bubble detection device and method of machine vision

Abstract

The application discloses a wireless alarm general-purpose ECMO pipeline bubble detection equipment and method of machine vision, and the wireless alarm general-purpose ECMO pipeline bubble detection equipment of machine vision comprises a wireless WIFI transmission module, a microcontroller, a camera module, an annular light source, a background base plate, a target detection pipeline, a support, a touch support data panel and a power supply; the camera module, the touch support data panel, the wireless WIFI transmission module and the power supply are all connected to the microcontroller, the support is connected to the microcontroller, the annular light source and the background base plate respectively, the power supply is arranged on the top of the microcontroller, the target detection pipeline is arranged above the background base plate, the wireless WIFI transmission module is arranged directly above the touch support data panel, the camera module is arranged below the touch support data panel, and the annular light source is arranged directly below the camera module; the problems of low universality and low intelligence in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of pipeline air bubbles, and in particular to a machine vision-based wireless alarm general-purpose ECMO pipeline air bubble detection device and method. Background Technology

[0002] This patent utilizes machine vision-based binarized image differential collaborative generalized Pareto distribution (GPD) probability prediction threshold setting technology to achieve real-time monitoring and alarm of air bubbles in extracorporeal membrane oxygenation (ECMO) tubing. The system can operate independently without relying on a specific ECMO brand. It uses a microcontroller to control a camera module to capture real-time images of the target tubing. After binarizing and differentially analyzing each frame of the image, the system counts the number of white pixels in the differential image. If this value exceeds a preset threshold, the system assumes air bubbles are present in the tubing and initiates an alarm; otherwise, the system assumes no air bubbles are present and continues monitoring.

[0003] The system alarm operations include local sound and light alarms as well as remote mobile terminal software alarms via Wi-Fi. It can also instantly store information such as real-time image frames and differential image frames that trigger alarm operations to record alarm event details, so that relevant data can be used for subsequent event inspection or research.

[0004] The system's alarm thresholds and monitoring environment (such as the setting or modification of different target monitoring pipeline environments) can be set by the user through the data panel, and the user can export the differential / real-time image frame data that is stored when the alarm is triggered.

[0005] During system operation, the wireless network connection with the mobile terminal software enables the present invention to have complete remote service functions, including remote alarm, remote operation status monitoring, remote data reading and remote parameter setting, and other remote access and control functions.

[0006] The threshold selection of the system is achieved by fitting the extreme values ​​of the number of white pixels in the differential image obtained by sampling under the normal bubble-free state of the pipeline during ECMO operation in the field to the GPD distribution, and selecting those with a prediction probability of at least less than 95% confidence interval.

[0007] Problems with existing technology

[0008] Currently, investigations have revealed the following two problems with existing ECMO-related bubble detection devices (hereinafter referred to as bubble detectors):

[0009] Low versatility:

[0010] Bubble detectors from different manufacturers may employ different technical specifications and parameters. A specific brand of bubble detector often has dependencies on its compatible ECMO operating platform in terms of usage requirements, transmission interfaces, and data formats, resulting in a lack of cross-brand compatibility. This means that, given that the ECMO operating platform brand has already been selected and configured, the available bubble detectors are often limited to that brand or a limited number of supported third-party manufacturers, leading to significant limitations.

[0011] In addition, current clinical bubble detectors based on traditional infrared or ultrasound detection technologies suffer from high air attenuation, which inevitably leads to system structural constraints under the requirement of close contact with tubing. This means that these products typically only monitor one type of tubing, such as 3 / 8 inch (adult) tubing or 1 / 4 inch (children's) tubing, and cannot simultaneously monitor both. If monitoring different tubing sizes is required, additional bubble detectors must be replaced or purchased, resulting in low flexibility and significantly increasing equipment purchase costs.

[0012] Low level of intelligence

[0013] Current clinical bubble detectors have limited functionality in terms of alarms and parameter settings. Firstly, current devices only offer local audible and visual alarms and do not support network connectivity, remote monitoring and querying of operational status, remote alarms, or alarm data storage and export. Some devices also lack the ability to set detection parameters (such as the size of the target bubble). In short, current devices offer limited functionality and have a low level of intelligence.

[0014] The present invention addresses these pain points by using a machine vision-based, wirelessly alarm-enabled, general-purpose ECMO pipeline bubble detection device and method. Summary of the Invention

[0015] To address the problems existing in the prior art, this invention provides a machine vision-based wireless alarm general-purpose ECMO pipeline bubble detection device and method.

[0016] The technical solution of the present invention is as follows:

[0017] The machine vision-based wireless alarm general-purpose ECMO tubing bubble detection device includes a wireless WIFI transmission module, a microcontroller, a camera module, a ring light source, a background plate, a target detection tubing, a bracket, a touch-enabled data panel, and a power supply.

[0018] The camera module, touch support data panel, wireless WIFI transmission module, and power supply are all connected to the microcontroller. The bracket is connected to the microcontroller, the ring light source, and the background plate. The power supply is located on top of the microcontroller, the target detection pipeline is located above the background plate, the wireless WIFI transmission module is located directly above the touch support data panel, the camera module is located below the touch support data panel, and the ring light source is located directly below the camera module.

[0019] Preferably, the lens of the camera module is used to pass through the central ring light source, thereby aligning with the target detection pipeline for monitoring.

[0020] Preferably, the machine vision-based wireless alarm general-purpose ECMO tubing bubble detection method includes the following steps:

[0021] Step S1: After the system is set up and powered on, the user interacts with the CNC data panel to set the target monitoring pipeline environment mode and alarm bubble size parameters.

[0022] Step S2: Within the system, the interaction control logic between the user and the data panel is implemented through the internal software logic of the microcontroller;

[0023] Step S3: The settings for the detection pipeline environment and the alarm bubble size are both completed by changing the alarm decision threshold inside the microcontroller;

[0024] Step S4: Based on the two consecutive black and white images obtained after binarizing the two real-time image frames captured in succession, traverse all pixels on the two black and white images, and use bitwise logic XOR operation on the binary form of the pixel values ​​located at the same coordinate position to obtain the pixel value of the differential frame image at this coordinate position.

[0025] Step S5: GPD distribution fitting completes the periodic linear envelope endpoints of the line graph drawn from the white pixels in the continuously obtained differential frame images;

[0026] Step S6: After the settings are completed, the system will take real-time pictures of the pipeline through the camera module, perform binarization and difference operations on the image frames, and calculate the number of white pixels in the difference image in real time.

[0027] Step S7: The system will determine whether there are air bubbles in the pipeline based on this value. If the value does not exceed the judgment threshold, the system will determine that no air bubble exceeding the preset target size has been detected, and will not execute the alarm procedure, and will continue to detect. Otherwise, if the value exceeds the judgment threshold, the system will determine that an air bubble exceeding the preset target size has been detected and will execute the alarm procedure.

[0028] Step S8: The system's local sound program will be completed by controlling the connected sound-generating unit and light-emitting unit through the microcontroller;

[0029] Step S9: Under the condition that the wireless transmission Wi-Fi module connected to the microcontroller is bridged with the mobile terminal software, the remote alarm program will send specific data content to trigger the mobile terminal software to call pop-up notification and vibration to complete the process.

[0030] Step S10: The specific data content can be arbitrarily specified, but the mobile software must be able to recognize the alarm data content in this alarm program and push alarm information to the mobile user.

[0031] Step S11: The system has the function of storing alarm event information;

[0032] Step S12: The alarm event number starts from 0 and increments sequentially as new alarm events occur after the system starts up;

[0033] Step S13: Alarm event information will be stored sequentially in a MicroSD memory card connected to the microcontroller to support subsequent operations such as data transfer, export, and clearing of alarm event data by the user;

[0034] Step S14: After an alarm is triggered, the system must be manually reset on-site before it can return to normal and continue monitoring.

[0035] Step S15: In non-alarm state, the system supports remote users to query system operating status information or set / change detection parameters via Wi-Fi wireless network through mobile terminal software;

[0036] Step S16: The system's operating status information is dynamically updated periodically by the microcontroller after the system starts up and stored in the storage module connected to the microcontroller;

[0037] Step S17: After the microcontroller is bridged with the remote mobile software through the wireless transmission Wifi module connected to it, it updates the operating status information to the remote user through the wireless transmission Wifi module network and displays the dynamic updates of system information in the mobile software.

[0038] Preferably, the alarm bubble size parameter in step S1 will be used to change the decision threshold inside the system;

[0039] The target monitoring tubing environment mode setting includes two tubing type options for ECMO: 1 / 4 inch for children and 3 / 8 inch for adults.

[0040] The alarm bubble size parameters include: small and medium; small is less than 2mm in diameter, and medium is greater than 2mm in diameter.

[0041] Preferably, the alarm decision threshold in S3 is set by binarizing and performing continuous real-time frame difference operations on the image captured in real time by the camera module under the condition that no target size bubble appears in the target detection pipeline environment, and then fitting the GPD distribution, and selecting the value corresponding to the GPD probability density function obtained by fitting the GPD probability density function to predict the probability that is at least outside the 95% confidence interval.

[0042] Binarization involves mapping image pixel values ​​to 0 if the pixel value is less than a preset threshold; otherwise, it maps the pixel value to the maximum allowed value in the camera pixel storage format.

[0043] Preferably, in step S5, the number of white pixels represents the maximum allowed value of the pixel storage format content;

[0044] In GPD distribution fitting, the selection method for the parameters of the GPD distribution is not specified.

[0045] Preferably, the alarm procedure in step S7 includes local sound and light alarms, as well as remote alarms via mobile software based on a Wi-Fi wireless network.

[0046] Preferably, the alarm event information in step S11 includes: alarm event number, alarm trigger time, cumulative running time since power-on, differential image frame that triggered the alarm, and two original real-time image frames that participated in constructing the differential image frame.

[0047] Preferably, the system operating status information in step S15 includes the system number, system startup time, system cumulative running time, and system cumulative historical alarm count; the set / changed detection parameters include the target detection pipeline environment and alarm bubble size.

[0048] Preferably, the system information dynamic update frequency in step S17 should be at least once every 8 seconds.

[0049] The beneficial effects of the machine vision-based wireless alarm general-purpose ECMO pipeline bubble detection device and method of the present invention are as follows:

[0050] 1. This invention represents a significant leap forward in terms of product flexibility and intelligent functionality.

[0051] 2. Based on the technical characteristics of visual inspection, the present invention enables the system to be free from the unavoidable system structure constraints caused by the high air attenuation characteristics of traditional infrared or ultrasonic inspection technologies when high fitting requirements are required in pipelines.

[0052] 3. In terms of functional flexibility and versatility, this invention allows users to interact with a micro controller via a touch-sensitive data panel. The micro controller will adapt to different target monitoring pipeline environments and set the size of the bubble to be inspected based on user operations and software logic.

[0053] 4. The pre-stored logical decision threshold that is changed during the user parameter setting process is derived from the high goodness fit of the GPD distribution. This invention provides a strong and reliable statistical interpretation for threshold selection, making this invention highly theoretically credible in scenarios with high detection accuracy requirements.

[0054] 5. The present invention enables the system to provide data value that differs from traditional infrared or ultrasonic detection technologies by allowing the read-out storage of alarm data from real-time image frames and differential image frames that trigger alarms. This data value allows users to perform post-event tracing and event callback after an alarm event occurs, and can be further used for the analysis and research of alarm events.

[0055] 6. Due to the characteristics of the system's independent working mode, this invention is free from dependence on brands, transmission interfaces, data formats, etc., compared with similar products, and has strong cross-brand versatility.

[0056] 7. Connecting with the accompanying mobile application gives the system intelligent networking capabilities, enabling the system to provide remote services.

[0057] 8. In terms of alarms, in addition to having the same local sound and light alarm function as similar products, this invention can also send wireless alarm messages to mobile software via a wireless Wi-Fi network; in the absence of an alarm, the remote service function will also enable users to remotely query the status of the operating system, read data, and adjust monitoring parameters in real time, which will alleviate the density and intensity of on-site real-time monitoring work for medical staff to a certain extent. Attached Figure Description

[0058] Figure 1 This is a flowchart of the system monitoring status of the present invention.

[0059] Figure 2 This is a schematic diagram of the component connections of the present invention.

[0060] Figure 3 This is a flowchart of the remote query program of the present invention.

[0061] Reference numerals: 1-Wireless WIFI transmission module, 2-Micro controller, 3-Camera module, 4-Ring light source, 5-Background plate, 6-Target detection pipeline, 7-Bracket, 8-Touch support data panel, 9-Power supply. Detailed Implementation

[0062] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0063] A machine vision-based wireless alarm general-purpose ECMO tubing bubble detection device includes: a wireless WIFI transmission module 1, a microcontroller 2, a camera module 3, a ring light source 4, a background plate 5, a target detection tubing 6, a bracket 7, a touch-enabled data panel 8, and a power supply 9.

[0064] The camera module 3, touch support data panel 8, wireless WIFI transmission module 1, and power supply 9 are all connected to the microcontroller 2. The bracket 7 is connected to the microcontroller 2, the ring light source 4, and the background plate 6. The power supply 9 is located on top of the microcontroller 2. The target detection pipe 6 is located above the background plate 5. The wireless WIFI transmission module 1 is located directly above the touch support data panel 8. The camera module 3 is located below the touch support data panel 8. The ring light source 4 is located directly below the camera module 3.

[0065] In this embodiment, the lens of the camera module 3 is used to pass through the ring light source 4 in the center, thereby aiming at the target detection pipeline 6 for monitoring.

[0066] A machine vision-based wireless alarm method for detecting air bubbles in ECMO tubing includes the following steps:

[0067] Step S1: After the system is set up and powered on, the user interacts with the CNC data panel to set the target monitoring pipeline environment mode and alarm bubble size parameters.

[0068] Step S2: Within the system, the interaction control logic between the user and the data panel is implemented through the internal software logic of the microcontroller;

[0069] Step S3: The settings for the detection pipeline environment and the alarm bubble size are both completed by changing the alarm decision threshold inside the microcontroller;

[0070] Step S4: Based on the two consecutive black and white images obtained after binarizing the two real-time image frames captured in succession, traverse all pixels on the two black and white images, and use bitwise logic XOR operation on the binary form of the pixel values ​​located at the same coordinate position to obtain the pixel value of the differential frame image at this coordinate position.

[0071] Step S5: GPD distribution fitting completes the periodic linear envelope endpoints of the line graph drawn from the white pixels in the continuously obtained differential frame images;

[0072] Step S6: After the settings are completed, the system will take real-time pictures of the pipeline through the camera module, perform binarization and difference operations on the image frames, and calculate the number of white pixels in the difference image in real time.

[0073] Step S7: The system will determine whether there are air bubbles in the pipeline based on this value. If the value does not exceed the judgment threshold, the system will determine that no air bubble exceeding the preset target size has been detected, and will not execute the alarm procedure, and will continue to detect. Otherwise, if the value exceeds the judgment threshold, the system will determine that an air bubble exceeding the preset target size has been detected and will execute the alarm procedure.

[0074] Step S8: The system's local sound program will be completed by controlling the connected sound-generating unit and light-emitting unit through the microcontroller;

[0075] Step S9: Under the condition that the wireless transmission Wi-Fi module connected to the microcontroller is bridged with the mobile terminal software, the remote alarm program will send specific data content to trigger the mobile terminal software to call pop-up notification and vibration to complete the process.

[0076] Step S10: The specific data content can be arbitrarily specified, but the mobile software must be able to recognize the alarm data content in this alarm program and push alarm information to the mobile user.

[0077] Step S11: The system has the function of storing alarm event information;

[0078] Step S12: The alarm event number starts from 0 and increments sequentially as new alarm events occur after the system starts up;

[0079] Step S13: Alarm event information will be stored sequentially in a MicroSD memory card connected to the microcontroller to support subsequent operations such as data transfer, export, and clearing of alarm event data by the user;

[0080] Step S14: After an alarm is triggered, the system must be manually reset on-site before it can return to normal and continue monitoring.

[0081] Step S15: In non-alarm state, the system supports remote users to query system operating status information or set / change detection parameters via Wi-Fi wireless network through mobile terminal software;

[0082] Step S16: The system's operating status information is dynamically updated periodically by the microcontroller after the system starts up and stored in the storage module connected to the microcontroller;

[0083] Step S17: After the microcontroller is bridged with the remote mobile software through the wireless transmission Wifi module connected to it, it updates the operating status information to the remote user through the wireless transmission Wifi module network and displays the dynamic updates of system information in the mobile software.

[0084] In step S1 of this implementation scheme, the alarm bubble size parameter will be used to change the decision threshold inside the system;

[0085] The target monitoring tubing environment mode setting includes two tubing type options for ECMO: 1 / 4 inch for children and 3 / 8 inch for adults.

[0086] The alarm bubble size parameters include: small and medium; small is less than 2mm in diameter, and medium is greater than 2mm in diameter.

[0087] In step S3 of this implementation scheme, the alarm decision threshold is set by binarizing and performing continuous real-time frame difference operations on the image captured in real time by the camera module under the condition that no target size bubble appears in the target detection pipeline environment, and then fitting the GPD distribution. The predicted probability of the GPD probability density function obtained by the numerical corresponding to the fitted value is at least outside the 95% confidence interval.

[0088] Binarization involves mapping image pixel values ​​to 0 if the pixel value is less than a preset threshold; otherwise, it maps the pixel value to the maximum allowed value in the camera pixel storage format.

[0089] In step S5 of this implementation scheme, the number of white pixels represents the maximum allowed value of the pixel storage format content.

[0090] In GPD distribution fitting, the selection method for the parameters of the GPD distribution is not specified.

[0091] In step S7 of this implementation scheme, the alarm procedures include local sound and light alarms, as well as remote alarms via mobile software based on Wi-Fi wireless network.

[0092] The alarm event information in step S11 of this implementation scheme includes: alarm event number, alarm trigger time, cumulative running time since power-on, differential image frame that triggered the alarm, and two original real-time image frames that participated in constructing the differential image frame.

[0093] In step S15 of this implementation plan, the system operating status information includes the system number, system startup time, system cumulative running time, and system cumulative historical alarm count; the set / changed detection parameters include the target detection pipeline environment and alarm bubble size.

[0094] The system information dynamic update frequency in step S17 of this implementation plan should be at least once every 8 seconds.

[0095] When this implementation plan is implemented, the system's operation plan is as follows: Figure 1 As shown.

[0096] After the system is deployed and powered on, the user interacts with the CNC data panel to set the target monitoring pipeline environment mode and alarm bubble size parameters.

[0097] The alarm bubble size parameter will be used to change the internal decision threshold of the system;

[0098] The target monitoring tubing environment mode setting includes two tubing type options for ECMO: 1 / 4 inch for children and 3 / 8 inch for adults.

[0099] Alarm bubble size parameters include: small (diameter less than 2mm) and medium (diameter range greater than 2mm).

[0100] Internally, the interaction control logic between the user and the data panel is implemented through the internal software logic of the microcontroller.

[0101] The settings for the detection pipeline environment and the alarm bubble size are both accomplished by modifying the alarm decision threshold within the microcontroller. The alarm decision threshold is set by binarizing and performing continuous real-time frame differencing on the image captured in real-time by the camera module, fitting a GPD distribution, and selecting values ​​whose predicted probability from the fitted GPD probability density function is at least outside the 95% confidence interval, provided that the target size bubble does not appear in the detection pipeline environment. Specifically, binarization maps image pixel values ​​less than a preset threshold to 0; otherwise, it maps to the maximum permissible value within the camera pixel storage format.

[0102] No specific criteria are specified for selecting this preset threshold.

[0103] Based on the two consecutive black and white images obtained by binarizing the two real-time image frames captured in succession, all pixels in the two black and white images are traversed, and bitwise XOR operation is performed on the binary form corresponding to the pixel values ​​of the pixels located at the same coordinate position to obtain the pixel value of the differential frame image at this coordinate position.

[0104] GPD distribution fitting completes the periodic linear envelope endpoints of the line graph drawn from the white pixels in the continuously obtained differential frame images. The number of white pixels is the maximum allowed value of the pixel storage format; the selection method of the parameters of GPD distribution in GPD distribution fitting is not specified.

[0105] After the settings are completed, the system will take real-time pictures of the pipeline through the camera module, perform binarization and differential operations on the image frames, and calculate the number of white pixels in the differential image in real time.

[0106] The system will determine whether an air bubble appears in the pipeline based on this value. If the value does not exceed the judgment threshold, the system will determine that no air bubble exceeding the preset target size has been detected, and no alarm will be executed, and the detection will continue. Otherwise, if the value exceeds the judgment threshold, the system will determine that an air bubble exceeding the preset target size has been detected and will execute the alarm procedure. The alarm procedure includes local sound and light alarms, as well as remote alarms via mobile software based on the Wi-Fi wireless network.

[0107] The system's local sound and alarm programs will be completed by controlling the connected sound-emitting unit and light-emitting unit through a microcontroller, respectively.

[0108] The remote alarm program will complete the process by bridging the wireless transmission Wi-Fi module connected to the microcontroller with the mobile software and sending specific data content.

[0109] The specific data content can be arbitrarily specified, but the mobile software must be able to recognize this alarm data content and push alarm information to the mobile user.

[0110] The system has the function of storing alarm event information, which includes: alarm event number, alarm trigger time, cumulative running time since power-on, differential image frame that triggered the alarm, and two original real-time image frames that participated in the construction of the differential image frame.

[0111] Alarm event numbers start from 0 and increment sequentially with each new alarm event after system startup. Alarm event information will be stored sequentially on a MicroSD memory card connected to the microcontroller to support subsequent operations such as data transfer, export, and clearing by the user.

[0112] After an alarm is triggered, the system must be manually reset on-site before it can return to normal and continue monitoring.

[0113] In non-alarm mode, the system allows remote users to query system operating status information or set / change detection parameters via Wi-Fi using mobile software. System operating status information includes system number, system startup time, system cumulative running time, and system cumulative historical alarm count. Setting / changing detection parameters includes target detection pipeline environment and alarm bubble size.

[0114] The system's operating status information is dynamically updated periodically by the microcontroller after the system starts up and stored in the storage module connected to the microcontroller.

[0115] After the microcontroller is bridged with remote mobile software via a connected wireless Wi-Fi module, it updates the operating status information to the remote user through the wireless Wi-Fi network and displays the dynamically updated system information in the mobile software. The system information update frequency should be at least once every 8 seconds.

[0116] The present invention relates to the following components: microcontroller 2, camera module 3, touch support data panel, ring light source 4, bracket 7, wireless transmission Wifi module 1 and power supply 9.

[0117] The system connection method is as follows: Figure 2As shown: Camera module 3, touch-enabled data panel, wireless transmission WIFI module 1, and power supply 9 are directly connected to the microcontroller chip 2; a dedicated bracket is used to fix the monitoring system and the ring light source 4. During use, the lens of camera module 3 passes through the center of the ring light source 4 and is aimed at the target detection pipeline 6 for monitoring.

[0118] This invention utilizes machine vision-based binarized image differential collaborative GPD threshold probability prediction technology to achieve real-time monitoring and alarm of air bubbles in ECMO tubing. Its key technical aspects include the following:

[0119] Feature extraction based on image binarization difference

[0120] Binarizing and differencing the real-time captured images from the camera allows the system to significantly highlight the difference between the captured bubbles in the pipe and the background environment. Image binarization can be implemented using adaptive or non-adaptive algorithms. Continuously differencing the pixel values ​​between consecutive black-and-white binarized image frames quickly maximizes the highlighting of the difference information between two consecutive frames while minimizing the common information. Finally, by counting the number of white pixels in the resulting image after the differencing operation, the parameter values ​​used for real-time determination of the presence of bubbles in the pipe can be obtained.

[0121] Threshold setting for predicted probability based on GPD fitting

[0122] The threshold for the number of white pixels in the differential frame image used to determine the presence of air bubbles in the pipeline is selected by fitting a GPD distribution to the endpoints of the periodic linear envelope of the number of white pixels in the differential frame image under normal operating conditions of the pipeline being monitored. The selected value corresponds to a range within at least a 95% confidence interval of the predicted probability of the fitted GPD probability density function. The algorithm for selecting the parameters for GPD fitting and the definition of the extreme values ​​of the data points used for fitting can be accomplished using different algorithms.

[0123] Wide range of applicable scenarios support

[0124] Vision-based pipeline bubble detection avoids the system structural constraints inherent in traditional infrared or ultrasonic detection technologies, which require high pipe fit due to high air attenuation. This allows users to customize the target pipeline environment based on actual needs. Furthermore, users can interact with the microcontroller via a touch-enabled data panel to set the size of the target bubble. This support for both types of functions significantly broadens the applicability of this invention and improves system flexibility.

[0125] Supports intelligent networked remote service functions

[0126] The wireless network connection provided by the accompanying mobile software enables this invention to possess comprehensive remote service functions, including remote alarms, remote operational status monitoring, remote data reading, and remote parameter setting—all remote access and control functions. This functionality will significantly reduce the workload and intensity of on-site care for medical personnel.

[0127] Real-time storage function for alarm-triggered event data is supported.

[0128] The system will automatically store real-time and differential image frames that trigger alarm procedures, so that users can perform post-event tracing and event callback after the alarm event occurs, providing important research value for event analysis.

[0129] In response to the shortcomings of existing products, such as limited versatility and low intelligence, the improved technology of this invention achieves a significant leap forward in terms of product flexibility and intelligent functionality.

[0130] Based on the technical characteristics of visual inspection, this invention enables the system to avoid the unavoidable system structure constraints caused by the high air attenuation characteristics of traditional infrared or ultrasonic inspection technologies when high-fitting requirements are required in pipelines.

[0131] In terms of functional flexibility and versatility, this invention allows users to interact with a micro controller via a touch-sensitive data panel. The micro controller will adapt to different target monitoring pipeline environments and set the size of the bubble to be inspected based on the user's operation and through software logic.

[0132] The pre-stored logic decision threshold that is changed during the user parameter setting process is derived from the high goodness fit of the GPD distribution. The latter provides a strong and reliable statistical interpretation for threshold selection, making the present invention highly theoretically credible in scenarios requiring high detection accuracy.

[0133] In addition, the ability to read and store alarm data from real-time and differential image frames that trigger alarms enables the system to provide data value that differs from traditional infrared or ultrasonic detection technologies. This allows users to perform post-event tracing and event callback after an alarm event occurs, and further to analyze and study the alarm event.

[0134] Meanwhile, due to the characteristics of the system's independent working mode, this invention is free from dependence on brands, transmission interfaces, data formats, etc., compared to similar products, and has strong cross-brand versatility.

[0135] The integration with the accompanying mobile application grants the system intelligent networking capabilities, enabling remote service functionality. Regarding alarms, in addition to the local audible and visual alarm functions consistent with similar products, this invention can also send wireless alarm messages to the mobile application via Wi-Fi. Even when no alarm is triggered, the remote service function allows users to remotely query the system's status, read data, and adjust monitoring parameters in real time, thus alleviating the workload and intensity of on-site monitoring for medical staff to some extent. Figure 3 As shown.

Claims

1. A machine vision-based wireless alarm general-purpose ECMO tubing bubble detection device, characterized in that, Includes a wireless WIFI transmission module (1), a micro controller (2), a camera module (3), a ring light source (4), a background plate (5), a target detection pipeline (6), a bracket (7), a touch support data panel (8), and a power supply (9). The camera module (3), touch support data panel (8), wireless WIFI transmission module (1) and power supply (9) are all connected to the microcontroller (2). The bracket (7) is connected to the microcontroller (2), the ring light source (4) and the background plate (5) respectively. The power supply (9) is located on the top of the microcontroller (2). The target detection pipe (6) is located above the background plate (5). The wireless WIFI transmission module (1) is located directly above the touch support data panel (8). The camera module (3) is located below the touch support data panel (8). The ring light source (4) is located directly below the camera module (3). The lens of the camera module (3) is used to pass through the ring light source (4) in the center, thereby aiming at the target detection pipeline (6) for monitoring; The machine vision-based wireless alarm general-purpose ECMO tubing bubble detection method includes the following steps: Step S1: After the system is set up and powered on, the user interacts with the CNC data panel to set the target monitoring pipeline environment mode and alarm bubble size parameters. Step S2: Within the system, the interaction control logic between the user and the touch-enabled data panel is implemented through the internal software logic of the microcontroller; Step S3: The settings for the detection pipeline environment and the alarm bubble size are both completed by changing the alarm decision threshold inside the microcontroller; Step S4: Based on the two consecutive black and white images obtained after binarizing the two real-time image frames captured in succession, traverse all pixels on the two black and white images, and use bitwise logic XOR operation on the binary form of the pixel values ​​located at the same coordinate position to obtain the pixel value of the differential frame image at this coordinate position. Step S5: GPD distribution fitting completes the periodic linear envelope endpoints of the line graph drawn from the white pixels in the continuously obtained differential frame images; Step S6: After the settings are completed, the system will take real-time pictures of the pipeline through the camera module, perform binarization and difference operations on the image frames, and calculate the number of white pixels in the difference image in real time. Step S7: The system will determine whether there are air bubbles in the pipeline based on this value. If the value does not exceed the judgment threshold, the system will determine that no air bubble exceeding the preset target size has been detected, and will not execute the alarm procedure, and will continue to detect. Otherwise, if the value exceeds the judgment threshold, the system will determine that an air bubble exceeding the preset target size has been detected and will execute the alarm procedure. Step S8: The system's local audible and visual alarm program will be completed by controlling the connected sound and light-emitting units through the microcontroller; Step S9: Under the condition that the wireless transmission Wi-Fi module connected to the microcontroller is bridged with the mobile terminal software, the remote alarm program will send specific data content to trigger the mobile terminal software to call pop-up notification and vibration to complete the process. Step S10: The specific data content can be arbitrarily specified, but the mobile software must be able to recognize the alarm data content in this alarm program and push alarm information to the mobile user. Step S11: The system has the function of storing alarm event information; Step S12: The alarm event number starts from 0 and increments sequentially as new alarm events occur after the system starts up; Step S13: Alarm event information will be stored sequentially in a MicroSD memory card connected to the microcontroller to support subsequent operations such as data transfer, export, and clearing of alarm event data by the user; Step S14: After an alarm is triggered, the system must be manually reset on-site before it can return to normal and continue monitoring. Step S15: In non-alarm state, the system supports remote users to query system operating status information or set / change detection parameters via Wi-Fi wireless network through mobile terminal software; Step S16: The system's operating status information is dynamically updated periodically by the microcontroller after the system starts up and stored in the storage module connected to the microcontroller; Step S17: After the microcontroller is bridged with the remote mobile terminal software through the wireless transmission Wifi module connected to it, it updates the operating status information to the remote user through the wireless transmission Wifi module network and displays the dynamic update of system information in the mobile terminal software. In step S1, the alarm bubble size parameter will be used to change the internal decision threshold of the system. The target monitoring tubing environment mode setting includes two tubing type options for ECMO: 1 / 4 child inch and 3 / 8 adult inch. The alarm bubble size parameters include: small and medium; small means the diameter is less than 2mm, and medium means the diameter is greater than 2mm; The alarm decision threshold in S3 is set by binarizing and performing continuous real-time frame difference operations on the image captured in real time by the camera module under the condition that no target size bubble appears in the target detection pipeline environment, and then fitting the GPD distribution after selecting the value corresponding to the GPD probability density function obtained by fitting the fit, and the predicted probability is at least outside the 95% confidence interval. The binarization is performed such that if the image pixel value is less than a certain preset threshold, the image pixel value is mapped to 0; otherwise, it is mapped to the maximum value allowed by the camera pixel storage format. In step S5, the number of white pixels refers to the maximum number of pixel values ​​allowed by the pixel storage format content. The alarm program in step S7 includes local sound and light alarms, as well as remote alarms via mobile software based on Wi-Fi wireless network. The alarm event information in step S11 includes: alarm event number, alarm trigger time, cumulative running time since power-on, differential image frame that triggered the alarm, and two original real-time image frames that participated in constructing the differential image frame; The system operating status information in step S15 includes system number, system startup time, system cumulative running time, and system cumulative historical alarm count; the set / changed detection parameters include target detection pipeline environment and alarm bubble size; The system information dynamic update frequency in step S17 should be at least once every 8 seconds.

Images