A visual combined multi-parameter monitoring lung isolation catheter

Abstract

The present invention discloses a visual combined multi-parameter monitoring lung isolation catheter, which relates to the technical field of medical devices. The lung isolation catheter includes: an outer catheter, an inner catheter and an endotracheal intubation endoscope detection system; wherein the inner catheter cuff and the outer catheter cuff are both embedded with pressure sensors; the top of the outer catheter is connected to a four-way connector, which includes an outer catheter interface, an inner catheter interface, a suction interface, and a ventilation interface; the lower end of the outer catheter is embedded with a flushing pipe, a camera, a temperature sensor, and a pH sensor; the endotracheal intubation endoscope detection system includes a data processing module for locating the glottis using a trained glottis recognition model based on an improved Yolov5s network for the collected tracheal image; and is also used to compare the temperature value, pH value, and pressure value with a preset threshold value, and to alarm when the preset threshold value is exceeded. The present invention integrates the multi-parameter monitoring function into one, which can reduce the need for multiple intubations.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and in particular to a visual combined multi-parameter monitoring lung isolation catheter. Background Art

[0002] With the increasing incidence of thoracic and lung diseases and the increase in the number of thoracic and lung surgeries, the anesthesia operation technology and anesthesia management of thoracic and lung surgeries have also been significantly improved. When double-lung isolation and single-lung ventilation are required during thoracic and lung surgeries, a double-lumen bronchial tube must be used to achieve this without affecting the surgical operation and ensuring the patient's normal oxygenation and ventilation.

[0003] Despite numerous advances in endotracheal intubation technology, most are still based on a single endoscopic imaging method for identification. During intubation, the endoscopic image has a relatively small viewing angle, and image contrast, target distance, and target size all change, making it difficult for doctors to quickly lock onto the target. Sputum and airway secretions obscure targets such as the tracheal and esophageal openings, interfering with the visual image. Furthermore, existing endotracheal tubes are unable to monitor values ​​such as cuff compression, gastric acid reflux, and gas temperature in real time, resulting in delayed risk warnings. They are also unable to detect cuff pressure in real time. Excessive pressure can easily damage the airway mucosa, while low pressure can lead to air leaks and ventilator-associated pneumonia. Furthermore, they are unable to monitor the airway in real time, making it impossible to quickly reposition the tube after displacement. Summary of the Invention

[0004] To this end, the present invention provides a visual combined multi-parameter monitoring lung isolation catheter to try to solve or alleviate one or more of the above problems.

[0005] A visual combined multi-parameter monitoring lung isolation catheter, comprising: an outer catheter 6, an inner catheter 3, and an endotracheal intubation endoscopic detection system 9; wherein the inner catheter 3 is inserted into the outer catheter 6, an inner catheter cuff 1 is provided at the bottom of the inner catheter 3; an inner tube pressure sensor 2 is embedded in the inner catheter cuff 1;

[0006] An outer catheter cuff 4 is provided at the bottom of the outer catheter 6, and an outer tube pressure sensor 5 is embedded in the outer catheter cuff 4; a four-way connector 16 is connected to the top of the outer catheter 6, and the four-way connector 16 includes an outer catheter interface 161, an inner catheter interface 162, a sputum suction interface 163, and a ventilation interface 164; a flushing pipe, a camera 19, a temperature sensor 20, and a pH sensor 21 are embedded on the end surface of the lower end of the outer catheter 6, and the camera 19, the temperature sensor 20, and the pH sensor 21 are connected to the endotracheal intubation detection system 9 via a wired or wireless manner;

[0007] The inner tube pressure sensor 2 and the outer tube pressure sensor 5 are respectively connected to the endotracheal intubation endoscopy detection system 9 via wired or wireless means;

[0008] The endotracheal intubation detection system 9 includes a data processing module and a display module; wherein the data processing module is configured to perform glottis positioning on the tracheal image captured by the camera 19 using a trained glottis recognition model based on the improved Yolov5s network; and compare the temperature value captured by the temperature sensor 20, the pH value captured by the pH sensor 21, and the pressure value captured by the inner tube pressure sensor 2 or the outer tube pressure sensor 5 with a preset threshold value, and issue an alarm when the temperature value, pH value or pressure value exceeds the preset threshold value; the display module is configured to display the glottis position, temperature value, pH value and pressure value.

[0009] Furthermore, the outer catheter 6 is connected to the four-way joint 16 through the outer catheter interface 161, and the inner catheter 3 is inserted into the outer catheter 6 through the inner catheter interface 162, and the suction interface 163 and the ventilation interface 164 are both connected to the interior of the outer catheter 6; the suction interface 163 is connected to the perforated semi-sealed cap 15, the perforated semi-sealed cap 15 is connected to the interface at the top of the branch joint 17, the tops of the inner catheter 3 and the ventilation interface 164 are both connected to the interface at the bottom of the branch joint 17, and the branch joint 17 is a three-way joint; a ventilation switch is also provided on the branch joint 17, and the ventilation switch controls the communication between the ventilator and the inner catheter 3 or the ventilation interface 164.

[0010] Furthermore, the side end of the branch connector 17 is connected to the Y-shaped connector assembly 18, and the Y-shaped connector assembly 18 includes a three-way connector 181 and a connecting hose 182. The three-way connector 181 includes a ventilator connection port 1811, an internal catheter connection port 1812, and a ventilation connection port 1813. The ventilator is connected to the ventilator connection port 1811. There are two connecting hoses 182, and the two connecting hoses 182 are respectively connected to the internal catheter connection port 1812 and the ventilation connection port 1813. The other end of the connecting hose 182 is connected to the suction interface 163 or the ventilation interface 164.

[0011] Furthermore, a locking nut 14 is provided on the inner catheter interface 162 , and the locking nut 14 is threadedly connected to the inner catheter interface 162 . The outer circumference of the locking nut 14 is circular, and a plurality of anti-slip protrusions 141 are provided along the circumference of the outer circumference of the locking nut 14 .

[0012] Furthermore, the perforated semi-sealed cap 15 includes a connecting head 151 , a connecting cover 152 , and a connecting belt 153 . The connecting head 151 is connected to the connecting cover 152 via the connecting belt 153 , and the connecting cover 152 can be buckled onto the connecting head 151 .

[0013] Furthermore, the improvements of the improved Yolov5s network in the data processing module include: introducing a multi-scale attention mechanism for cross-space learning in the backbone network part of the original Yolov5s network; the operating mechanism of the multi-scale attention mechanism is: first, the input feature map is divided into multiple groups, and each group of feature maps is processed in parallel by convolution kernels of different sizes; then, inter-channel interaction is performed through 1×1 convolution blocks; then, spatial context information is captured through 3×3 convolution blocks; finally, spatial attention maps of different scales are generated through global average pooling and feature aggregation.

[0014] Furthermore, the improvements of the improved Yolov5s network in the data processing module include: introducing a context enhancement module in the neck network part of the original Yolov5s network, and the context enhancement module adopts an adaptive fusion method, including: obtaining adaptive weights through convolution, splicing and Softmax function; aggregating context information to the output by calculating the weighted sum.

[0015] Furthermore, the improvements of the improved Yolov5s network in the data processing module include: the loss function adopts the improved IoU loss function, and the loss value calculation formula is as follows:

[0016]

[0017] Where, express Loss function; represents the adjustment factor; P represents the penalty factor.

[0018] Furthermore, when the pressure value exceeds the preset threshold, an alarm is issued and a dynamic airbag pressure adjustment program is started to adjust the pressure of the inner catheter cuff 1 or the outer catheter cuff 4; in the dynamic airbag pressure adjustment program, when the pressure value exceeds the upper limit of the preset threshold, the pressure is released to reduce the pressure; when the pressure value exceeds the lower limit of the preset threshold, the pressure is increased.

[0019] Furthermore, the preset threshold of the temperature value is set to 38 degrees Celsius; the preset threshold of the pH value is set to 4.5; the upper limit of the preset threshold of the pressure value is set to 30cm , the preset lower threshold is set to 20cm .

[0020] The present invention has the following technical effects:

[0021] The present invention proposes a visual combined multi-parameter monitoring lung isolation catheter. First, a four-way connector is provided on the outer catheter, integrating the functions of "suctioning, drug administration, and monitoring" into one, which can reduce the need for multiple intubations; a locking nut is added to the four-way connector to lock the inner catheter to prevent the inner tube from loosening; secondly, values ​​such as airbag compression, gastric acid reflux, and gas temperature can be monitored in real time, and risk warnings can be issued in real time; thirdly, the improved YOLOv5s network is used to achieve precise positioning of the glottis, and the improved YOLOv5s network has higher accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 This is an overall structural diagram of a visual combined multi-parameter monitoring lung isolation catheter provided by an embodiment of the present invention.

[0024] Figure 2 Schematic diagram of the end face structure of the lower end of the outer conduit in an embodiment of the present invention.

[0025] Figure 3 2 is a schematic structural diagram of a four-way joint in an embodiment of the present invention.

[0026] Figure 4 Schematic diagram of the structure of the Y-shaped connector assembly in an embodiment of the present invention.

[0027] Figure 5 1 is a schematic structural diagram of a locking nut in an embodiment of the present invention.

[0028] Figure 6 2 is a schematic structural diagram of a semi-sealed cap with a hole in an embodiment of the present invention.

[0029] Figure 7 This is the network structure diagram of YOLOv5s.

[0030] Figure 8 4 is an example diagram of an input module of a tracheal intubation endoscopy detection system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.

[0032] The embodiment of the present invention provides a visual combined multi-parameter monitoring lung isolation catheter, such as Figures 1-3 As shown, the lung isolation catheter includes an outer catheter 6, an inner catheter 3 and an endotracheal intubation detection system 9; wherein, the inner catheter 3 is inserted into the outer catheter 6, and an inner catheter cuff 1 is provided at the bottom of the inner catheter 3; an inner tube pressure sensor 2 is embedded in the inner catheter cuff 1; an outer catheter cuff 4 is provided at the bottom of the outer catheter 6, and an outer tube pressure sensor 5 is embedded in the outer catheter cuff 4; the top of the outer catheter 6 is connected to a four-way connector 16, and the four-way connector 16 includes an outer catheter interface 161, an inner catheter interface 162, a suction interface 163, and a ventilation interface 164; a flushing pipe, a camera 19, a temperature sensor 20 and a pH sensor 21 are embedded on the end face of the lower end of the outer catheter 6, and the camera 19, the temperature sensor 20 and the pH sensor 21 are connected to the endotracheal intubation detection system 9 by wired or wireless means; the inner tube pressure sensor 2 and the outer tube pressure sensor 5 are respectively connected to the endotracheal intubation detection system 9 by wired or wireless means;

[0033] The endotracheal intubation detection system 9 includes a data processing module and a display module; wherein the data processing module is configured to perform glottis positioning on the tracheal image captured by the camera 19 using a trained glottis recognition model based on the improved Yolov5s network; and compare the temperature value captured by the temperature sensor 20, the pH value captured by the pH sensor 21, and the pressure value captured by the inner tube pressure sensor 2 or the outer tube pressure sensor 5 with a preset threshold value, and issue an alarm when the temperature value, pH value or pressure value exceeds the preset threshold value; the display module is configured to display the glottis position, temperature value, pH value and pressure value.

[0034] In this embodiment, the flushing conduit extends upward from the wall of the outer catheter 6 to its upper end, where it is connected to the flushing tube 22, which is then connected to the flushing connector 7. The outer catheter 6 has an outer inflation conduit and a pressure sensor line conduit embedded within its wall. The bottom of the outer inflation conduit is connected to the outer catheter cuff 4, and the top of the outer inflation conduit is connected to the outer inflation conduit 11. The other end of the outer inflation conduit 11 is connected to the endotracheal tube detection system 9. The outer pressure sensor wire 10 passes through the pressure sensor line conduit and is connected to the endotracheal tube detection system 9. The inner catheter 3 has an inner inflation conduit and a pressure sensor line conduit embedded within its wall. The bottom of the inner inflation conduit is connected to the inner catheter cuff 1, and the top of the inner inflation conduit is connected to the inner inflation conduit 13. The other end of the inner inflation conduit 13 is connected to the endotracheal tube detection system 9. The inner pressure sensor wire 12 passes through the pressure sensor line conduit and is connected to the endotracheal tube detection system 9.

[0035] In this embodiment, if the camera 19, temperature sensor 20, pH sensor 21 are connected to the endotracheal tube detection system 9 via a wired connection, the connecting wires 8 are wound together and extended upward from the wall of the outer tube 6 until they pass through and are connected to the endotracheal tube detection system 9.

[0036] In this embodiment, the external catheter 6 is connected to the four-way connector 16 through the external catheter interface 161, and the internal catheter 3 is inserted into the external catheter 6 through the internal catheter interface 162. The suction interface 163 and the ventilation interface 164 are both connected to the inside of the external catheter 6. A suction tube can be inserted through the suction interface to suction the external catheter 6, or a bronchoscope can be inserted to assist in intubation of the external catheter 6; or a suction tube can be inserted through the internal catheter interface 162 to suction the internal catheter 3, or a bronchoscope can be inserted to assist in intubation of the internal catheter 3. The tops of the internal catheter 3 and the ventilation interface 164 are both connected to a branch connector 17, and the ventilator is connected to the side ends of the two branch connectors 17 through a Y-shaped connector assembly 18.

[0037] The shunt connector 17 is a three-way connector. The top interface of the shunt connector 17 is connected to the perforated semi-sealed cap 15, the bottom interface of the shunt connector 17 is connected to the inner tube 3 or the ventilation interface 164, and the sputum suction interface 163 is connected to the perforated semi-sealed cap 15. The shunt connector 17 is also provided with a ventilation switch, which controls the communication between the ventilator and the inner tube 3 or the ventilation interface 164. The ventilation switch can be a valve provided at the side interface of the shunt connector 17, or other device that can control the flow of gas. By controlling the ventilation switch, the communication between the ventilator and the inner tube 3 or the ventilation interface 164 can be controlled, thereby achieving subsequent single-lung ventilation.

[0038] like Figure 4As shown, the Y-shaped connector assembly 18 includes a three-way connector 181 and a connecting hose 182. The three-way connector 181 includes a ventilator connection port 1811, an internal catheter connection port 1812, and a ventilation connection port 1813. The ventilator is connected to the ventilator connection port 1811. There are two connecting hoses 182. The two connecting hoses 182 are respectively connected to the internal catheter connection port 1812 and the ventilation connection port 1813. The other end of the connecting hose 182 is connected to the suction interface 163 or the ventilation interface 164.

[0039] The inner conduit interface 162 is provided with a locking nut 14, which is threadedly connected to the inner conduit interface 162. Figure 5 As shown, the outer circumference of the locking nut 14 is circular, and the outer circumference of the locking nut 14 is provided with multiple anti-slip protrusions 141 along its circumference, which can effectively increase the friction of the locking nut 14, making it easier to tighten or unscrew it toward the outside of the inner catheter interface 162; the inner wall of the locking nut 14 is provided with an internal thread 142, and the outer circumference of the inner catheter interface 162 is provided with an external thread compatible with the locking nut 14, and the inner wall of the inner catheter interface 162 gradually becomes smaller from top to bottom, and a circular through hole 143 is provided on the upper part of the interior of the locking nut 14, and the circular through hole 143 extends downward along the inner wall of the locking nut, and the inner catheter 3 enters the inner catheter interface 162 through the circular through hole 143. The outer diameter of the circular through hole 143 is smaller than the inner diameter of the top of the inner catheter interface 162, and larger than the inner diameter of the bottom of the inner catheter interface 162. When the inner catheter 3 needs to be fixed, the locking nut 14 is rotated so that the locking nut 14 is screwed down on the outside of the inner catheter interface 162 through the thread. At this time, the circular through hole 143 enters the inner catheter interface 162, and the bottom of the circular through hole 143 enters the bottom of the inner catheter interface 162 and is squeezed, causing the circular through hole 143 to shrink inward and squeeze the outer wall of the inner catheter 3, thereby completing the fixation of the inner catheter 3.

[0040] The top of the inclined pipe of the four-way joint 16 is provided with a semi-sealed cap 15, which can be opened to insert instruments such as suction tubes and bronchoscopes to achieve various functions, such as Figure 6 As shown, the perforated semi-sealed cap 15 includes a connector 151, a connecting cover 152, and a connecting strap 153. The connector 151 is connected to the connecting cover 152 via the connecting strap 153, and the connecting cover 152 can be snapped onto the connector 151. The connecting cover 152 on the connector 151 can be removed or tightened as needed. For example, when suctioning of the external catheter 6 or the internal catheter 3 is not required, the perforated semi-sealed cap 15 needs to remain tightened to ensure that the external catheter 6 does not communicate with the outside world through the suction port 163 and the branch joint 17, and the internal catheter 3 does not communicate with the outside world through the branch joint 17, thereby preventing the gas in the ventilator from escaping.

[0041] In this embodiment, the camera 19 is a 2-megapixel CMOS camera with a waterproof coating and a 5mm LED cold light source.

[0042] During use, completely deflate the air within the outer and inner catheter cuffs 4 and 1, lubricate the inner catheter 3, and then insert the inner catheter 3 into the inner catheter port 162 in the four-way connector 16. Insert the lower end of the inner catheter 3 through the small hole until the inner catheter cuff 1 is completely exposed from the lower end of the outer catheter 6. Once the glottis is clearly exposed, insert the outer and inner catheters 6 and 3 into the trachea. Connect the camera 19 to a dedicated tablet computer via the USB Type-C port, which has an image capture function. If the lens is blurry, connect a 20ml syringe to the flushing connector 7. Low-pressure, low-volume air can be injected through the syringe to flush the camera 19 lens and clear it. Under video surveillance, position the intubation tube. Once the inner catheter 3 is properly inserted and aligned, tighten the locking nut 14 above the inner catheter port 162 to secure the inner catheter 3.

[0043] If sputum is attached to the outer catheter 6 and the inner catheter 3 during the insertion process, or the line of sight of the camera 19 is blocked when the outer catheter 6 is inserted, the perforated semi-sealed cap 15 on the suction interface 163 can be opened, and the suction tube can be inserted into the suction interface 163 to suction the outer catheter 6, or a bronchoscope can be inserted to assist in intubation of the outer catheter 6; if sputum is attached to the inner catheter 3 during the insertion process, or the line of sight of the camera 19 is blocked when the inner catheter 3 is inserted, the perforated semi-sealed cap 15 on the branch joint 17 above the inner catheter 3 can be opened, and the suction tube can be inserted into the branch joint 17 above the inner catheter 3 to suction the inner catheter 4, or a bronchoscope can be inserted to assist in intubation of the outer catheter 6.

[0044] Connect the Y-shaped connector assembly 18 to the two branch connectors 17 (blue and transparent respectively), connect the blue branch connector 17 to the inner catheter 3, connect the transparent branch connector 17 to the ventilation interface 164, so that it is connected to the inside of the outer catheter 6, and connect the other end of the Y-shaped connector assembly 18 to the ventilation system (ventilator). After the inner catheter 3 is inserted into the right bronchus, turn on the ventilation switch on the blue branch connector 17 and turn off the ventilation switch on the transparent branch connector 17. At this time, the gas in the ventilator enters the blue branch connector 17 through the Y-shaped connector assembly 18, and continues to flow into the inner catheter 3, and enters the right bronchus with the inner catheter 3, which can achieve right-side single lung ventilation and left-side collapsed lung (if the inner catheter 3 is inserted into the left bronchus, repeat the above steps for left-side single lung ventilation and right-side collapsed lung); after the inner catheter 3 is inserted into the right bronchus, turn off the ventilation switch on the blue branch connector 17 and turn on the ventilation switch on the transparent branch connector 17. At this time, the gas in the ventilator passes through the Y-shaped connector The head assembly 18 enters the transparent branch connector 17 and continues to flow into the outer catheter 6 and into the main trachea along with the outer catheter 6. Since the inner catheter cuff 1 is inflated and pressed against the trachea, the gas in the ventilator can only enter the left bronchus and not the right bronchus, thereby achieving left-side single-lung ventilation and right-side collapsed lung (if the inner catheter 3 is inserted into the left bronchus, the above steps are repeated for right-side single-lung ventilation and left-side collapsed lung); after the inner catheter 3 is inserted into the right (left) lateral bronchus, the ventilation switches on the two branch connectors 17 are turned on. At this time, the gas in the ventilator enters the two branch connectors 17 through the Y-shaped connector assembly 18 and continues to flow into the outer catheter 6 and the inner catheter 3, thereby achieving double-lung ventilation.

[0045] In this embodiment, the improvements of the improved Yolov5s network in the data processing module include: introducing an efficient multi-scale attention mechanism for cross-space learning into the backbone network part of the original Yolov5s network; introducing a context enhancement module into the neck network part; and improving the loss function from IoU to PIoU loss function.

[0046] like Figure 7As shown in the figure, the YOLOv5s network structure mainly consists of three parts: the backbone network (Backbone), the neck network (Neck), and the head network (Head). The backbone is mainly composed of the following components: convolutional layers (Conv), C3 modules, and fast spatial pyramid pooling (SPPF) modules. The convolutional layers are combined with batch normalization layers and Silu activation functions to extract target features. The C3 module enhances feature learning through residual learning and concatenation operations. The SPPF module improves processing speed and optimizes feature capture of multi-scale targets by using multiple small-sized pooling kernels. The neck uses the feature pyramid network (FPN) structure, supplemented by the PANet structure, to enhance feature fusion and localization information transmission. The head is responsible for generating detection boxes and performing classification, localization, and confidence scoring on them.

[0047] The embodiment of the present invention introduces an efficient multi-scale attention mechanism for cross-space learning - the EMA attention mechanism - into the backbone network part, which increases the model's ability to extract features of the target to be measured, and also makes the model pay more attention to the detection of small targets. EMA first performs feature grouping, dividing the input feature map X into multiple groups to facilitate its learning of different semantic information; then a multi-scale parallel sub-network is designed. While performing channel reshaping, it extracts the attention weights of the grouped feature maps through three parallel sub-paths. A 1×1 convolution kernel is used to process the interaction between channels, which helps to capture local cross-channel information, and a 3×3 convolution kernel is used to capture a wider range of spatial contextual information; then cross-spatial and cross-channel feature information aggregation is performed, and the output information of the 1×1 convolution kernel is encoded using 2D global average pooling. It is converted into the corresponding dimensional shape before the channel feature activation mechanism, and the processed features are subjected to dot product operation to construct the first spatial attention map that can capture spatial information of different scales; then the same dot product operation is applied in the 3×3 convolution branch to obtain the second spatial attention map; finally, the sigmoid function is used to aggregate the feature map of each group output and the two generated spatial attention weights to enhance the contextual information of all pixels.

[0048] The EMA module achieves cross-spatial learning through feature grouping and multi-scale parallel sub-networks. It first divides the input feature map into multiple groups, each of which is processed in parallel using convolution kernels of different sizes to capture spatial information at different scales. 1×1 convolution blocks are then used to interact between channels and capture local cross-channel information. 3×3 convolution blocks are used to capture spatial context. Finally, global average pooling and feature aggregation are used to generate spatial attention maps at different scales. The EMA module boasts a flexible design. By using parallel convolution blocks and avoiding dimensionality reduction, it reduces the model's parameter count and computational overhead, improving computational efficiency. Its ease of integration and lightweight nature make it easy to integrate into the existing YOLO network architecture, enhancing the model's feature extraction capabilities and multi-scale fusion performance.

[0049] The embodiment of the present invention introduces a context enhancement module - CAM module in the Neck network part, which balances the training process, reduces feature conflicts, improves the detection accuracy of small targets, and enhances the robustness of the model while maintaining computational efficiency. It expands the scale of the shared convolution kernel and increases the receptive field of the convolution kernel without changing the scale of the feature map. The CAM module applies dilated convolution with different dilated convolution rates to obtain contextual information of different receptive fields, and injects it into the FPN network in YOLOv5 from top to bottom. It processes the features through dilated convolution of 3×3 convolution kernels with different dilation ratios of 1, 3, and 5, so that the module can achieve the purpose of multi-scale extraction of feature maps.

[0050] The CAM module includes three fusion mechanisms: weighted fusion, cascade operation, and adaptive operation. Weighted fusion and cascade operation both adjust the number of feature map passes through three 1×1 convolutions and then fuse features in the spatial and channel dimensions. Adaptive operation uses convolution, concatenation, and the Softmax function to obtain adaptive weights. By calculating the weighted sum, it aggregates contextual information into the output. The CAM connection method in this embodiment is the connection method of adaptive operation.

[0051] The accuracy of object detection depends on the design of the loss function, and the bounding box loss function is crucial to improving model performance. The traditional IoU loss function comprehensively considers various spatial position information and can achieve good convergence results for boundary regression problems. However, during model training, high and low IoU samples contribute unbalancedly to the loss value, and the penalty factors selected for different samples may be suboptimal. This embodiment of the present invention improves the loss function of YOLOv5s to the PIoU loss function, effectively improving the detection box distortion caused by large sample differences and enhancing the robustness of the model.

[0052] Loss value The calculation formula is as follows:

[0053] IoU

[0054] Where IoU represents the IoU loss function; P represents the penalty factor, , 、 、 、 Represents the absolute value of the distance between the edge of the predicted box and the edge of the real box, 、 is the length and width of the real frame.

[0055] Furthermore, considering that different data sets may have different requirements for quality samples, the embodiment of the present invention adjusts the factor Introducing the PIoU function, the specific loss value As shown below:

[0056] IoU

[0057] Adjustment Factor The value of is greater than 1. When the adjustment factor When the value is small, the loss function tends to take the maximum gradient in low-quality samples; as the adjustment factor As increases, the model gradually focuses on higher quality samples.

[0058] The improved Yolov5s network is trained using multiple tracheal images to obtain a trained glottis recognition model; the glottis is located using the trained glottis recognition model based on the improved Yolov5s network for the tracheal images captured by the camera 19, and the catheter position is then calculated.

[0059] The improved YOLOv5s network proposed in the embodiment of the present invention first introduces the EMA module in the Backbone part, thereby enhancing the model's multi-scale fusion and feature extraction capabilities; secondly, the CAM context enhancement module is introduced in the Neck part, thereby balancing the training process, reducing feature conflicts, improving the detection accuracy of small objects, and enhancing the robustness of the model while maintaining computational efficiency; finally, replacing the new loss function improves the convergence of the loss function and balances the contribution of each quality sample to the loss.

[0060] In this embodiment, the endotracheal intubation detection system 9 may include an input module in addition to a data processing module and a display module. Figure 8 As shown, the input module includes an on / off key, function keys, a charging key, a decompression key, etc.

[0061] In this embodiment, the collected temperature value, pH value, and pressure value are compared with preset thresholds, and an alarm is issued when the temperature value, pH value, and pressure value exceed the preset thresholds, including:

[0062] 1) Airway gas temperature approaches core body temperature (more accurate than axillary or rectal temperature), providing real-time feedback on temperature changes in critically ill patients (such as those with sepsis or intraoperative hypothermia), reducing the risk of infection associated with frequent invasive temperature measurements. Prolonged high-flow oxygen therapy can lead to airway dryness and localized hyperthermia (>40°C), and sudden increases in airway temperature during laser surgery (requiring immediate interruption to avoid burns). Exceeding the temperature limit triggers an alarm, prompting adjustments to ventilation parameters or suspension of the procedure. The temperature sensor displays the tracheal temperature in real time on the endotracheal intubation monitoring system. This temperature sensor can be a PT1000, triggering an audible and visual alarm when the local temperature exceeds 38°C (preset temperature threshold).

[0063] 2) A sudden decrease in airway pH (e.g. pH < 4) may indicate gastric acid reflux, which is the main cause of ventilator-associated pneumonia (VAP); real-time alarms can prompt medical staff to suction sputum or adjust body positions in time to reduce the risk of aspiration. The pH sensor can display the endotracheal pH value in real time in the endotracheal intubation detection system 9. Among them, the pH sensor can be used Solid-state electrode, the surface is covered with a 0.2μm hydrophobic breathable membrane; and when the pH value is less than 4.5 (preset pH threshold), an audible and visual alarm is triggered.

[0064] 3) Pressure sensors can optimize airbag pressure and prevent airway necrosis. Local tissue ischemia (such as excessive compression of the intubation airbag) can lead to anaerobic metabolism, lactic acid accumulation, and a decrease in pH (normal airway pH ≈ 7.0–7.5). The airbag inflation volume is dynamically adjusted according to airway pressure changes to prevent leakage or excessive compression. Among them, the pressure sensor can be MEMS piezoresistive, ±1cm Accuracy. When the pressure exceeds a preset threshold, an alarm is triggered and a dynamic airbag pressure adjustment program is activated to adjust the pressure of inner catheter cuff 1 or outer catheter cuff 4. In the dynamic airbag pressure adjustment program, when the pressure exceeds the upper limit of the preset threshold, the pressure is released to reduce the pressure; when the pressure exceeds the lower limit of the preset threshold, the pressure is increased.

[0065] For example, if the monitored pressure is greater than the preset upper limit (i.e. 30cm ), the air release valve is activated (0.5 ml / s); if the monitored pressure is less than the preset lower limit (i.e. 20 cm ), the micro air pump is started; if the monitored pressure is greater than the preset lower limit and less than the preset lower limit (i.e. the pressure is within 20cm and 30cm If the error is greater than 10%, the airbag volume is fine-tuned (PID parameters: Kp=0.8, Ki=0.2, Kd=0.1). Safety mechanism: Dual redundant pressure sensors cross-validate, switching to manual mode when the error is greater than 10%.

[0066] The structural design of the diversion four-way connector 16 of the present invention integrates "sputum suction, drug administration, monitoring (such as The device integrates "pulmonary embolism monitoring" and "pulmonary intubation monitoring" functions, reducing the need for multiple intubations. A locking nut 14 is added to the four-way connector 16 to lock the inner tube 3 to prevent it from loosening. The outer tube 6 is located above the carina, reducing stimulation to the nerves in the tracheal carina area. The flexible combination of inner and outer tubes allows for flexible switching of left and right lung isolation, reducing operational difficulty. The device has a flexible and adjustable field of view, a clear field of view, a high-definition anti-fog camera, and a side flushing system to flush blood and secretions, providing the operator with a clear field of view. The small cuff lung isolation design is suitable for different bronchial diameters and has a wide range of applications.

[0067] Applications in thoracic surgery are as follows:

[0068] Intubation preparation: camera calibration (white balance adjustment), preset airbag pressure 25cm ;

[0069] Insertion process: After the AI ​​system identifies the glottis, the AR interface guides the rotating catheter to the left mainstem bronchus (right lung isolation); the insertion depth is indicated (28 ± 2 cm from the incisors) with vibration feedback;

[0070] Intraoperative monitoring: When surgical traction causes airway pressure fluctuations, the airbag automatically inflates 2ml of air to maintain the seal; the pH sensor detects gastric acid reflux (pH=3.9) and immediately triggers the suction pump to start.

[0071] The applications of long-term ventilation in the ICU are as follows:

[0072] Automatically execute the mucosal protection program every 4 hours: the airbag pressure is periodically reduced to 15cm (Last for 2 minutes); the temperature sensor detected a local temperature rise to 39.2°C, warning of possible early infection.

[0073] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A visual combined multi-parameter monitoring lung isolation catheter, characterized in that: It comprises an outer catheter (6), an inner catheter (3) and an endotracheal tube endoscopy detection system (9); wherein, The inner catheter (3) is inserted into the outer catheter (6); an inner catheter cuff (1) is provided at the bottom of the inner catheter (3); an inner tube pressure sensor (2) is embedded inside the inner catheter cuff (1); The outer catheter (6) is provided with an outer catheter cuff (4) at the bottom, and an outer tube pressure sensor (5) is embedded inside the outer catheter cuff (4); the top of the outer catheter (6) is connected to a four-way connector (16), and the four-way connector (16) includes an outer catheter interface (161), an inner catheter interface (162), a sputum suction interface (163), and a ventilation interface (164); a flushing pipe, a camera (19), a temperature sensor (20), and a pH sensor (21) are embedded on the end face of the lower end of the outer catheter (6), and the camera (19), the temperature sensor (20), and the pH sensor (21) are connected to the endotracheal intubation detection system (9) by wired or wireless means; the outer The catheter (6) is connected to the four-way connector (16) through the outer catheter interface (161), the inner catheter (3) is inserted into the outer catheter (6) through the inner catheter interface (162), and the sputum suction interface (163) and the ventilation interface (164) are both connected to the inside of the outer catheter (6); the sputum suction interface (163) is connected to the perforated semi-sealed cap (15), the perforated semi-sealed cap (15) is connected to the interface at the top of the shunt connector (17), the inner catheter (3) and the ventilation interface (164) are both connected to the interface at the bottom of the shunt connector (17), and the shunt connector (17) is a three-way connector; a ventilation switch is also provided on the shunt connector (17), and the ventilation switch controls the communication between the ventilator and the inner catheter (3) or the ventilation interface (164); A locking nut (14) is provided on the inner catheter interface (162), and the locking nut (14) is threadedly connected to the inner catheter interface (162); The inner wall of the locking nut (14) is provided with an internal thread (142), the outer periphery of the inner catheter interface (162) is provided with an external thread adapted to the locking nut (14), and the inner wall of the inner catheter interface (162) gradually becomes smaller from top to bottom, a circular through hole (143) is provided above the inner portion of the locking nut (14), and the circular through hole (143) extends downward along the inner wall of the locking nut, and the inner catheter (3) enters the inner catheter interface (162) through the circular through hole (143), and the outer diameter of the circular through hole (143) is smaller than the inner diameter of the top of the inner catheter interface (162) and larger than the inner diameter of the bottom of the inner catheter interface (162); The inner tube pressure sensor (2) and the outer tube pressure sensor (5) are respectively connected to the endotracheal intubation endoscopy detection system (9) via a wired or wireless method; The endotracheal intubation detection system (9) includes a data processing module and a display module; wherein the data processing module is configured to perform glottis positioning on the tracheal image captured by the camera (19) using a trained glottis recognition model based on an improved Yolov5s network, and compare the temperature value captured by the temperature sensor (20), the pH value captured by the pH sensor (21), and the pressure value captured by the inner tube pressure sensor (2) or the outer tube pressure sensor (5) with a preset threshold value, and to issue an alarm when the temperature value, pH value or pressure value exceeds the preset threshold value, and at the same time, to start a dynamic air bag pressure adjustment program to adjust the pressure of the inner catheter cuff (1) or the outer catheter cuff (4); in the dynamic air bag pressure adjustment program, when the pressure value exceeds the upper limit of the preset threshold value, the pressure is released to reduce the pressure; when the pressure value exceeds the lower limit of the preset threshold value, the pressure is increased; and the display module is configured to display the glottis position, temperature value, pH value and pressure value.

2. A visual combined multi-parameter monitoring lung isolation catheter according to claim 1, characterized in that: The side end of the branch connector (17) is connected to a Y-shaped connector assembly (18), and the Y-shaped connector assembly (18) includes a three-way connector (181) and a connecting hose (182). The three-way connector (181) includes a ventilator connection port (1811), an internal catheter connection port (1812), and a ventilation connection port (1813). The ventilator is connected to the ventilator connection port (1811). There are two connecting hoses (182), and the two connecting hoses (182) are connected to the internal catheter connection port (1812) and the ventilation connection port (1813), respectively. The other end of the connecting hose (182) is connected to the sputum suction interface (163) or the ventilation interface (164).

3. The visual combined multi-parameter monitoring lung isolation catheter according to claim 2, characterized in that: The outer periphery of the locking nut (14) is circular, and a plurality of anti-slip protrusions (141) are provided on the outer periphery of the locking nut (14) along its circumference.

4. The visual combined multi-parameter monitoring lung isolation catheter according to claim 3, characterized in that: The perforated semi-sealed cap (15) comprises a connecting head (151), a connecting cover (152), and a connecting belt (153); the connecting head (151) is connected to the connecting cover (152) via the connecting belt (153); and the connecting cover (152) can be buckled onto the connecting head (151).

5. The visual combined multi-parameter monitoring lung isolation catheter according to claim 1, characterized in that: The improvements of the improved Yolov5s network in the data processing module include: introducing a multi-scale attention mechanism for cross-space learning in the backbone network part of the original Yolov5s network; the operating mechanism of the multi-scale attention mechanism is: first, the input feature map is divided into multiple groups, and each group of feature maps is processed in parallel by convolution kernels of different sizes; then, inter-channel interaction is performed through 1×1 convolution blocks; then, spatial context information is captured through 3×3 convolution blocks; finally, spatial attention maps of different scales are generated through global average pooling and feature aggregation.

6. The visual combined multi-parameter monitoring lung isolation catheter according to claim 1, characterized in that: The improvements of the improved Yolov5s network in the data processing module include: introducing a context enhancement module in the neck network part of the original Yolov5s network, and the context enhancement module adopts an adaptive fusion method, including: obtaining adaptive weights through convolution, splicing and Softmax function; aggregating context information to the output by calculating the weighted sum.

7. The visual combined multi-parameter monitoring lung isolation catheter according to claim 1, characterized in that: The improvements of the improved Yolov5s network in the data processing module include: the loss function adopts the improved IoU loss function, and the loss value calculation formula is as follows: ； Where, express Loss function; represents the adjustment factor; P represents the penalty factor, , 、 、 、 Represents the absolute value of the distance between the edge of the predicted box and the edge of the real box, 、 is the length and width of the real frame.

8. The visual combined multi-parameter monitoring lung isolation catheter according to claim 7, characterized in that: The preset threshold of the temperature value is set to 38 degrees Celsius; the preset threshold of the pH value is set to 4.5; the upper limit of the preset threshold of the pressure value is set to 30cm , the preset lower threshold is set to 20cm .

Images