Medical intelligent infusion management system based on visual identification

Abstract

The invention belongs to the technical field of visual medical treatment, and particularly relates to a medical intelligent infusion management system based on visual identification, which comprises a monitoring module and a transmission judgment module, the monitoring module is composed of a distributed monitoring unit and a transmission enhancement unit, can perform synchronous image acquisition and user information recording on a plurality of infusion bags, and realizes accurate labeling of the residual liquid amount and time of the infusion bags by analyzing the image number, the network state and the infusion rate and optimizing the image set transmission efficiency by using a transmission correction model; secondly, the system evaluates the risk probability of infusion bag replacement delay according to the corrected image set in combination with user data; the risk probability is fed back to the transmission enhancement unit, so that the image enhancement sequence is adjusted in real time, the delay risk is marked, and finally the image subjected to enhancement marking is displayed on the PAD equipment, so that medical staff are helped to know the infusion state in time, and the medical risk caused by the fact that an infusion bag is not replaced in time is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of visual medicine, and particularly relates to a medical intelligent infusion management system based on visual recognition. Background Art

[0002] Intravenous infusion means injecting drugs directly and slowly into the patient's vein within a certain period of time to achieve the treatment of various diseases. In order to ensure the safety and effectiveness of the infusion process, medical staff need to determine the infusion drip rate according to the patient's condition, age and the nature of the drugs used, and timely monitor the position of the liquid level, replace or stop the infusion of the liquid medicine; currently, most of the infusion monitoring uses weighing infusion monitoring or infrared monitoring technology. Among them, weighing monitoring is affected by factors such as the material of the infusion container, liquid shaking, and drug addition, with poor measurement accuracy, and installing weighing equipment requires modifying the infusion device, increasing costs and possibly damaging the original facilities; infrared monitoring technology is prone to interference from infrared signals due to the complex liquid flow in the infusion tube, bubbles and impurities, resulting in misjudgment; existing technologies also use visual technology for monitoring, but in the actual image transmission process, affected by factors such as network fluctuations and the number of synchronous transmissions, the image may experience packet loss, distortion, etc., resulting in a decline in image quality; for example, a kind of infusion process monitoring system and its monitoring method based on visual recognition disclosed in the Chinese patent with the authorization announcement number CN114588402B has complex algorithm processing, high requirements for hardware performance, and is prone to data processing delays when the network traffic pressure is high, affecting the timeliness of the infusion process judgment, and the image recognition link requires a large amount of training data and complex algorithm optimization. A kind of intravenous infusion monitoring method and system based on visual measurement disclosed in the Chinese patent with the publication number CN114596521A involves a variety of complex algorithms when identifying the liquid level position and drip rate, and is prone to recognition errors due to interference from light, the shape and material of the infusion bottle, etc. At the same time, both of these two technical solutions need to adjust the operation process, it is difficult to achieve seamless upgrade, and the acceptance of medical staff may not be high during promotion. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention proposes a medical intelligent infusion management system based on visual recognition. The system includes a monitoring module and a transmission discrimination module; the monitoring module consists of a distributed monitoring unit and a transmission enhancement unit, which can synchronously collect images of multiple infusion bags and record user information. By analyzing the number of images, network status and infusion rate, and using a transmission correction model to optimize the transmission efficiency of the image set, it realizes accurate annotation of the remaining liquid volume and time of the infusion bag. Secondly, the system combines the corrected image set with user data to evaluate the risk probability of infusion bag replacement delay; this risk probability is fed back to the transmission enhancement unit to adjust the image enhancement order in real time and mark the delay risk. Finally, the enhanced annotated image is displayed on the PAD device to help medical staff timely understand the infusion status and reduce the medical risk caused by untimely replacement of the infusion bag.

[0004] To achieve the above object, the present invention provides the following technical solutions: A medical intelligent infusion management system based on visual recognition, comprising: a monitoring module and a transmission discrimination module; the monitoring module includes a distributed monitoring unit and a transmission enhancement unit; Through the distributed monitoring unit, at least one or more infusion bags are synchronously collected for images, user information, and network transmission status. Based on the number of synchronously collected images, network transmission status, infusion rate, and the transmission correction model built in the transmission enhancement unit, a set of infusion bag images with synchronous transmission correction is obtained and the remaining liquid volume and remaining infusion time are marked. Based on the corrected set of infusion bag images combined with user information, through the transmission discrimination module, the risk probability of infusion bag replacement delay is obtained. The risk probability of infusion bag replacement delay is fed back to the transmission enhancement unit to adjust the real-time order of transmission image enhancement and mark the delay risk probability, and the enhanced and marked images are displayed on the PAD in real time.

[0005] Specifically, the construction process of the transmission correction model includes: According to the synchronously collected infusion bag images, the background of the infusion bag images is segmented through the background segmentation layer to obtain a set of infusion bag segmentation regions and a set of background segmentation regions, and the liquid level line coordinates in the infusion bag segmentation regions are marked. Based on the shape attribute information of the infusion bag, the infusion bag segmentation region, the full liquid volume, and the initial liquid level height, through the three-dimensional modeling layer, a three-dimensional model of the infusion bag is obtained. Based on the three-dimensional model of the infusion bag combined with different infusion rates, the position function of the liquid separation line at different infusion rates is fitted through a simulation algorithm combined with the support vector machine kernel function to obtain the position function of the liquid separation line corresponding to different infusion rates.

[0006] Specifically, the construction process of the transmission correction model further includes: Mask the liquid separation line in the infusion bag segmentation region, and input each masked infusion bag segmentation region and the position function of the liquid separation line into the feature extraction layer of the transmission correction model to obtain the feature space of each infusion bag segmentation region and the liquid level line coordinate space predicted by the position function of the liquid separation line. According to the feature space of the infusion bag segmentation region and the liquid level line coordinate space predicted by the position function of the liquid separation line combined with the marked liquid level line coordinates, the prediction error loss of the liquid level line position is obtained.

[0007] Specifically, the construction process of the transmission correction model further includes: The continuous image sequence of the infusion bag after masking the segmented areas corresponding to consecutive time points is input into the optical flow layer, and the motion vectors of each pixel corresponding to the segmented areas of the infusion bag after masking at consecutive time points are analyzed to obtain the dense optical flow field corresponding to the continuous segmented areas of the infusion bag after masking; Based on the change values of adjacent pixels in the dense optical flow field corresponding to each segmented area of the infusion bag after masking, a dynamic clustering sub-region pixel band space corresponding to each segmented area of the infusion bag after masking is constructed by combining the clustering algorithm with the three-quarter quantile; According to the pixel change values in the dynamic clustering sub-region pixel band space and the average pixel change value corresponding to the infusion bag full of liquid volume, a pixel change anomaly threshold is constructed to obtain the abnormal dynamic clustering sub-region pixel band; The abnormal dynamic clustering sub-region pixel band is the dynamic clustering sub-region pixel band corresponding to the most abnormal pixel change coordinate points and the largest abnormal deviation value.

[0008] Specifically, the construction process of the transmission correction model further includes: The abnormal dynamic clustering sub-region pixel band is input into the Gaussian kernel convolution, and the abnormal pixel points in the abnormal dynamic clustering sub-region pixel band are diffused to obtain the continuous abnormal pixel point coordinate space; The continuous abnormal pixel point coordinate space is fed back to the feature extraction layer to perform secondary correction on the liquid level line position prediction error loss and the predicted liquid level line coordinate space, and the corrected liquid level line position prediction error loss and the predicted liquid level line coordinate space are obtained; Based on the corrected liquid level line position prediction error loss, the predicted liquid level line coordinate space, and the infusion bag segmentation region feature space, the space of the infusion bag segmentation region to be enhanced is constructed.

[0009] Specifically, the construction process of the transmission correction model further includes: According to the available bandwidth for transmission, the packet loss rate and transmission delay corresponding to the number of synchronously transmitted images in different time periods in the network transmission state, the network quality assessment layer constructs the network quality index corresponding to different time periods; Based on the network quality index corresponding to different time periods and the preset network quality level interval, the network transmission level corresponding to different time periods is obtained; The network transmission levels corresponding to different time periods and the preset image enhancement resolution table are input into the hierarchical correction layer to perform simulated enhanced transmission on the space of the infusion bag segmentation region to be enhanced, and the transmission delay, packet loss loss, and the corresponding network level-image quality mapping matrix of each resolution image at different network transmission levels are obtained.

[0010] Specifically, the construction process of the transmission correction model further includes: Based on the pixel space of different segmented regions of the infusion bag after masking and the predicted liquid level line position, combined with the real-time infusion rate, through the probability evaluation layer with the Bayesian function built-in, obtain the replacement delay risk probability value and probability evaluation loss corresponding to each infusion bag at the current moment; Take the replacement delay risk probability value corresponding to each infusion bag at the current moment as the processing priority for the space of the segmented region of the infusion bag to be enhanced corresponding to the acquired image of each infusion bag; Feed back the processing priority for the space of the segmented region of the infusion bag to be enhanced corresponding to the acquired image of each infusion bag to the simulated enhancement transmission process, simulate the optimal image quality corresponding to each image with the processing priority under different network levels, and use the transmission delay as a measurement index to construct a processing priority - network level - image quality - transmission delay mapping matrix space.

[0011] Specifically, the construction process of the transmission correction model further includes: Build the processing priority - network level - image quality - transmission delay mapping matrix space into the output layer, input the space of the segmented region of the infusion bag to be enhanced corresponding to the current moment, the network quality index, and the processing priority corresponding to each infusion bag into the output layer, and obtain the infusion bag image with the corresponding resolution image quality and the corresponding image enhancement loss; Construct a transmission correction model loss function with the corrected liquid level line position prediction error loss, probability evaluation loss, transmission delay and packet loss, and image enhancement loss, perform simulation training on the transmission correction model, and analyze the transmission delay corresponding to each resolution image quality under different network qualities and processing priorities; When the transmission delay corresponding to each resolution image quality is less than the preset transmission delay, obtain the trained transmission correction model, and output the enhanced infusion bag image and the corresponding enhanced resolution under the corresponding network level.

[0012] Specifically, the process of masking the liquid separation line in the segmented region of the infusion bag includes: Based on the segmented region of the infusion bag, generate a multi-level pyramid feature map corresponding to the segmented region of the infusion bag through the Gaussian pyramid decomposition algorithm combined with the preset number of layers; Based on the multi-level pyramid feature map, perform horizontal gradient and vertical gradient convolution on each level of the pyramid feature map through the Sobel operator convolution, and obtain the horizontal and vertical gradient amplitudes and gradient direction angles corresponding to each level of the pyramid feature map; Based on the gradient direction angle combined with the preset gradient direction angle filtering interval, filter the horizontal and vertical gradient amplitudes to obtain the filtered horizontal and vertical gradient amplitudes corresponding to each level of the pyramid feature map; The channel convolution attention algorithm is used to weighted attenuate the horizontal and vertical gradient amplitudes corresponding to the filtering of each level of the pyramid feature map, and each level of the pyramid feature map after weighted attenuation is upsampled and fused through bilinear interpolation to obtain a multi-level pyramid fusion gradient map.

[0013] Specifically, the process of masking the liquid separation line in the segmentation area of ​​the infusion bag also includes: Based on the multi-level pyramid fusion gradient map, the mean and variance corresponding to the fusion gradient map are obtained through statistical algorithms; The mask interval is set based on the mean and variance corresponding to the fused gradient map, and the initial binary mask corresponding to the liquid level line in the segmentation area of ​​the infusion bag is generated according to the set mask interval combined with non-maximum suppression; Based on the initial binary mask corresponding to the liquid level line in the infusion bag segmentation area, the binary mask corresponding to the liquid level line in the infusion bag segmentation area is obtained through the 3×3 rectangular structure element S combined with the connected domain filter, and the liquid level line marked in the input infusion bag segmentation area is binary masked based on the binary mask corresponding to the liquid level line in the liquid bag segmentation area.

[0014] Compared with the prior art, the present invention has the following beneficial effects: In view of the deficiencies of the prior art, the present invention solves the problems of image packet loss and distortion caused by factors such as network fluctuations and the number of synchronous transmissions in the infusion management of the existing visual recognition technology by introducing a distributed monitoring unit and a transmission enhancement unit. Specifically, the present invention can efficiently and synchronously acquire images and record user information for multiple infusion bags, and optimize the image transmission quality using a transmission correction model, ensuring that a clear and accurate set of infusion bag images can be obtained even when the network status is poor, and accurate marking of the remaining liquid volume and infusion time can be achieved; in addition, the transmission discrimination module is combined to evaluate and feedback the probability of delayed risk of infusion bag replacement, and the image enhancement sequence and marking delay risk are adjusted in real time, and finally the enhanced and marked image is displayed on the PAD, which effectively improves the work efficiency of medical staff and the treatment safety of patients; therefore, compared with the prior art, the present invention improves the stability and accuracy of image transmission and reduces medical risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a process architecture diagram of a medical intelligent infusion management system based on visual recognition according to Embodiment 1 of the present invention; Figure 2 This is a diagram of the construction architecture of the transmission correction model of Example 1 of the present invention. DETAILED DESCRIPTION

[0016] Example 1 Intravenous infusion, as a common and effective disease treatment method, is to directly inject drugs into the patient's vein at a slow speed within a specific time period. In the daily operation of a hospital, the nursing resources are relatively limited, but they need to deal with a large number of infusion patients. Moreover, most patients do not have a dedicated caregiver during the infusion. In this case, how to accurately and efficiently transmit the real-time status of the infusion bag, such as the remaining drug amount, infusion rate, etc., clearly and quickly to the nurse's end has become a key problem to be solved urgently. If these information cannot be obtained in time, it is very easy to cause a delay in changing the liquid, resulting in blood backflow in the infusion tube, which not only affects the treatment process but may also bring discomfort or even additional health risks to the patient. At the same time, the delay in removing the hanging needle may also cause the patient to maintain the same posture for a long time, causing physical fatigue and occupying limited medical resources; for this reason, please refer to Figure 1 , an embodiment provided by the present invention: a medical intelligent infusion management system based on visual recognition, including: a monitoring module and a transmission discrimination module; the monitoring module includes a distributed monitoring unit and a transmission enhancement unit; The monitoring module is used for the synchronous acquisition of infusion bag images and the transmission correction of the acquired images; the distributed monitoring unit, according to the configured synchronous monitoring model, synchronously acquires at least one or more infusion bag images, user information, and the network transmission status during the image transmission process in real time; The transmission enhancement unit, according to the number of synchronously acquired images, the network transmission status, the infusion rate, and in combination with the configured transmission correction model, corrects the synchronously transmitted infusion bag image set and marks the remaining liquid volume and the remaining infusion time; The transmission enhancement unit, based on the corrected infusion bag image set combined with user information, obtains the risk probability of infusion bag replacement delay through the Bayesian function, and feeds back the risk probability of infusion bag replacement delay to the transmission enhancement unit to adjust the real-time order of the transmitted image enhancement and mark the delay risk probability. At the same time, the enhanced and marked image is displayed on the PAD (portable electronic device) in real time.

[0017] Furthermore, the working processes corresponding to the distributed monitoring unit, the transmission enhancement unit, and the transmission discrimination module in this embodiment include: Through the distributed monitoring unit, at least one or more infusion bags are synchronously collected for images, user information, and network transmission status. Based on the number of synchronously acquired images, the network transmission status, the infusion rate, and in combination with the transmission correction model built in the transmission enhancement unit, a synchronously transmitted and corrected infusion bag image set is obtained and the remaining liquid volume and the remaining infusion time are marked; Based on the corrected infusion bag image set combined with user information, through the transmission discrimination module, the risk probability of infusion bag replacement delay is obtained; Feed the replacement delay risk probability of the infusion bag back to the transmission enhancement unit, adjust the real-time enhancement order of the transmission image and annotate the delay risk probability, and display the enhanced and annotated image on the PAD in real time.

[0018] Further, please refer to Figure 2 , the construction process of the transmission correction model in this embodiment includes: Based on the synchronously collected infusion bag images, perform background segmentation on the infusion bag images through the background segmentation layer to obtain the infusion bag segmentation region set and the background segmentation region set, and annotate the liquid level line coordinates in the infusion bag segmentation region; Further, the specific process of background segmentation and annotation in this embodiment includes: Use depthwise separable convolution to construct the encoder of U-Net, and at the same time use transposed convolution to construct the decoder of U-Net. Input the synchronously collected infusion bag images into the constructed encoder of U-Net and gradually downsample them to the original Figure 1 / 32 size through the convolutional layer with a stride of 2. At the same time, input the downsampled infusion bag image features into the decoder of U-Net, splice them with the feature maps of the corresponding levels of the encoder, and perform channel alignment using SE attention while splicing to obtain the decoded infusion bag image features; The decoded infusion bag image features are combined with Dice Loss and Focal Loss, and through the output sublayer of U-Net, obtain the infusion bag segmentation region set. In the segmented infusion bag region, use a semi-automatic annotation tool (such as LabelMe) to calibrate the liquid level line coordinates and generate an annotation dataset.

[0019] Based on the infusion bag shape attribute information, the infusion bag segmentation region, the full liquid volume, and the initial liquid level height, obtain the infusion bag three-dimensional model through the three-dimensional modeling layer; Further, the construction process of the infusion bag three-dimensional model in this embodiment includes: Calculate the surface depth map of the infusion bag through stereo matching or laser speckle displacement, convert the depth map into a three-dimensional point cloud, and optimize the deformation model in combination with the infusion bag shape attributes (such as material flexibility parameters); Construct the infusion bag three-dimensional model according to the full-load volume parameter of the infusion bag; Based on the infusion bag three-dimensional model and different infusion rates, perform function fitting on the liquid separation line position function at different infusion rates through the simulation algorithm combined with the support vector machine kernel function to obtain the liquid separation line position function corresponding to different infusion rates.

[0020] Further, the liquid separation line in this embodiment refers to the dividing line corresponding to the liquid surface in the infusion bag.

[0021] Mask the liquid separation line in the segmented area of the infusion bag, and input each masked segmented area of the infusion bag and the liquid separation line position function into the feature extraction layer of the transmission correction model to obtain the feature space of each segmented area of the infusion bag and the liquid level line coordinate space predicted by the liquid separation line position function; Further, the process of masking the liquid separation line in the segmented area of the infusion bag in this embodiment includes: Based on the segmented area of the infusion bag, use the Gaussian pyramid decomposition algorithm in combination with a preset number of layers to generate multi-level pyramid feature maps corresponding to the segmented area of the infusion bag; Based on the multi-level pyramid feature maps, perform horizontal and vertical gradient convolutions on each level of pyramid feature maps through Sobel operator convolution to obtain the horizontal and vertical gradient amplitudes and gradient direction angles corresponding to each level of pyramid feature maps; Based on the gradient direction angle and in combination with a preset gradient direction angle filtering interval, filter the horizontal and vertical gradient amplitudes to obtain the filtered horizontal and vertical gradient amplitudes corresponding to each level of pyramid feature maps; Perform weighted attenuation on the filtered horizontal and vertical gradient amplitudes corresponding to each level of pyramid feature maps through the channel convolution attention algorithm, and perform upsampling fusion on each level of pyramid feature maps after weighted attenuation through bilinear interpolation to obtain a multi-level pyramid fusion gradient map.

[0022] Further, the process of obtaining the multi-level pyramid feature maps in this embodiment includes: Use the Gaussian pyramid decomposition algorithm in combination with a preset decomposition number of 3 layers to perform successive downsampling to generate three-level pyramid feature maps with σ = 1, 2, and 3; Apply the Sobel operator on each layer of pyramid feature map to extract horizontal and vertical gradients, and retain the edge responses in the horizontal direction (±10°); Further, in this embodiment, σ is the standard deviation of the Gaussian Kernel; Further, the successive downsampling process in this embodiment includes: Each layer of pyramid image is obtained by performing Gaussian blur on the previous layer of image and then downsampling (such as halving the size); For example: Original image (σ = 1 blur) → Downsampling to obtain the first layer.

[0023] The first layer of image (σ = 2 blur) → Downsampling to obtain the second layer.

[0024] The second layer of image (σ = 3 blur) → Downsampling to obtain the third layer.

[0025] Among them, the blurred layer with σ = 1: For the precise positioning of the liquid level line and suppressing the interference of small bubbles or slight reflections.

[0026] Fuzzy layer with σ = 2: Analyze the overall deformation of the label area of the infusion bag to avoid the influence of local noise.

[0027] Fuzzy layer with σ = 3: Extract the global contour of the infusion bag to assist in 3D modeling and liquid volume estimation.

[0028] Restore the multi-scale gradient map to the original size through bilinear interpolation to obtain a multi-level pyramid fusion gradient map; further, in this embodiment, each level of the pyramid feature map is superimposed according to weights (0.6, 0.3, 0.1) to enhance the continuity of the liquid level line.

[0029] Based on the multi-level pyramid fusion gradient map, obtain the mean and variance corresponding to the fusion gradient map through a statistical algorithm; Set a mask interval based on the mean and variance corresponding to the fusion gradient map, and generate an initial binary mask corresponding to the liquid level line in the segmented area of the infusion bag by combining the set mask interval with non-maximum suppression; Based on the initial binary mask corresponding to the liquid level line in the segmented area of the infusion bag, obtain a binary mask corresponding to the liquid level line in the segmented area of the infusion bag through a 3×3 rectangular structural element S combined with a connected component filter, and perform binary masking on the liquid level line labeled in the input segmented area of the infusion bag based on the binary mask corresponding to the liquid level line in the segmented area of the liquid bag.

[0030] Further, the detailed process of obtaining the binary mask in this embodiment includes: Based on the mean μ and variance s of the fusion gradient map, set the high threshold (μ + 1.5s) and low threshold (μ - 0.5s) corresponding to the pixel filtering interval; Retain the pixels above the high threshold as seed points based on the pixel filtering interval, and connect the areas above the low threshold to form an initial mask; Perform a closing operation (first dilation and then erosion) on the initial mask using a 3×3 rectangle to fill in the breaks, and filter out the noise through connected component analysis (area > 20 pixels) to obtain a binary mask corresponding to the liquid level line in the segmented area of the infusion bag; Based on the feature space of the segmented area of the infusion bag and the coordinate space of the predicted liquid level line of the liquid separation line position function combined with the labeled liquid level line coordinates, obtain the liquid level line position prediction error loss.

[0031] Input the continuous infusion bag image sequence constructed from the segmented area of the infusion bag after masking at consecutive time points into the optical flow layer, analyze the motion vectors of each pixel corresponding to consecutive time points in the segmented area of the infusion bag after masking, and obtain the dense optical flow field corresponding to the continuous segmented area of the infusion bag after masking; Based on the change values of adjacent pixels in the dense optical flow field corresponding to the segmented area of the infusion bag after each mask, a dynamic clustering sub-region pixel band space corresponding to the segmented area of the infusion bag after each mask is constructed by combining the clustering algorithm with the three-quarter quantile. According to the pixel change values in the dynamic clustering sub-region pixel band space and the pixel change mean value corresponding to the full-load liquid volume infusion bag, a pixel change anomaly threshold is constructed to obtain an abnormal dynamic clustering sub-region pixel band. The abnormal dynamic clustering sub-region pixel band is the dynamic clustering sub-region pixel band corresponding to the most abnormal pixel change coordinate points and the largest abnormal deviation value.

[0032] The abnormal dynamic clustering sub-region pixel band is input into the Gaussian kernel convolution, and the abnormal pixel points in the abnormal dynamic clustering sub-region pixel band are diffused to obtain a continuous abnormal pixel point coordinate space. The continuous abnormal pixel point coordinate space is fed back to the feature extraction layer to perform secondary correction on the liquid surface line position prediction error loss and the predicted liquid surface line coordinate space, and the corrected liquid surface line position prediction error loss and the predicted liquid surface line coordinate space are obtained. Based on the corrected liquid surface line position prediction error loss, the predicted liquid surface line coordinate space, and the infusion bag segmentation region feature space, a to-be-enhanced infusion bag segmentation region space is constructed.

[0033] According to the available bandwidth for transmission, the packet loss rate, and the transmission delay corresponding to the number of synchronously transmitted images in different time periods in the network transmission state, a network quality index corresponding to different time periods is constructed through the network quality evaluation layer. Based on the network quality index corresponding to different time periods and in combination with the preset network quality level interval, the network transmission level corresponding to different time periods is obtained. Further, the preset network quality level interval in this embodiment is specifically: preset NQI threshold (excellent ≥ 0.8, good ≥ 0.6, medium ≥ 0.4, poor < 0.4), and a dynamic matching image resolution strategy; where NQI is the abbreviation of the network quality index. The network transmission level corresponding to different time periods and the preset image enhancement resolution table are input into the hierarchical correction layer to perform simulated enhanced transmission on the to-be-enhanced infusion bag segmentation region space, and the transmission delay, packet loss loss, and the corresponding network level-image quality mapping matrix of each resolution image at different network transmission levels are obtained.

[0034] Based on the pixel space of different segmented areas of the infusion bag after the mask, the predicted liquid surface line position, and the real-time infusion rate, through the probability evaluation layer with the Bayesian function built in, the replacement delay risk probability value and the probability evaluation loss corresponding to each infusion bag at the current moment are obtained. Use the replacement delay risk probability value corresponding to each infusion bag at the current moment as the processing priority for the space of the segmented area of the infusion bag to be enhanced corresponding to the collected image of each infusion bag; Further, in this embodiment, the processing priority is set corresponding to the replacement delay risk probability value. The greater the replacement delay risk probability value, the higher the corresponding processing priority. Further, the processing priority in this embodiment is the priority for image enhancement and transmission of the infusion bag image, that is, the greater the corresponding replacement delay risk probability value of the infusion bag, the image of the corresponding infusion bag is first enhanced and transmitted to the corresponding nurse PAD for early warning.

[0035] Feed back the processing priority of the space of the segmented area of the infusion bag to be enhanced corresponding to the collected image of each infusion bag to the simulated enhancement transmission process, simulate the optimal image quality corresponding to each processing priority image under different network levels, and use the transmission delay as a measurement index to construct a processing priority-network level-image quality-transmission delay mapping matrix space; Build the processing priority-network level-image quality-transmission delay mapping matrix space into the output layer, input the space of the segmented area of the infusion bag to be enhanced corresponding to the current moment, the network quality index, and the processing priority corresponding to each infusion bag into the output layer to obtain the infusion bag image with the corresponding resolution image quality and the corresponding image enhancement loss; Construct a transmission correction model loss function with the corrected liquid level line position prediction error loss, probability evaluation loss, transmission delay and packet loss, and image enhancement loss, simulate and train the transmission correction model, and analyze the transmission delay corresponding to each resolution image quality under different network qualities and processing priorities; When the transmission delay corresponding to each resolution image quality is less than the preset transmission delay, obtain the trained transmission correction model and output the enhanced infusion bag image and the corresponding enhanced resolution under the corresponding network level.

[0036] Through multi-dimensional technology collaborative innovation, significant efficiency improvements have been achieved in the aspects of real-time performance, accuracy, and resource optimization of infusion monitoring. First, the U-Net architecture based on depthwise separable convolution, combined with a multi-level feature fusion mechanism, uses a stride-2 convolution in the encoding stage to achieve progressive downsampling, effectively compressing redundant information while retaining the morphological features of the infusion bag. During the decoding process, through the SE attention mechanism and cross-level feature stitching, the feature misalignment problem of traditional segmentation models in complex backgrounds is solved, significantly improving the segmentation accuracy of the infusion bag area and laying a reliable foundation for the liquid level line positioning. Second, for the liquid level fluctuation interference in the dynamic infusion scenario, multi-scale pyramid gradient analysis and dynamic mask generation technology are adopted. Through three-level Gaussian kernel decomposition, multi-level blur processing with σ ranging from 1 to 3 is achieved, capturing local details and global deformation features at different scales respectively. Combining the gradient direction filtering and channel attention weighted attenuation mechanism, the influence of noise such as reflection and bubbles on the liquid level line detection is effectively suppressed. At the same time, the mask continuity is optimized through closing operation and connected component analysis, reducing the liquid level line coordinate prediction error by about 40%. Furthermore, through optical flow field analysis and dynamic clustering sub-region detection technology, a dense optical flow field is constructed and the interquartile range statistical model is used to identify abnormal motion pixel clusters. Combining with the Gaussian kernel diffusion algorithm, the dynamic correction of the deformation features of the infusion bag is realized, improving the robustness of the system to scenarios of sudden changes in infusion rate or external force interference by about 35%. At the level of network transmission optimization, an innovative network quality index (NQI) dynamic evaluation model is introduced, mapping bandwidth, packet loss rate, and delay parameters to network transmission levels. Combining the processing priority-network level-image quality mapping matrix, the adaptive adjustment of image resolution and enhancement strategy is realized, further improving the network resource utilization rate while ensuring the priority transmission of key infusion bag monitoring data. In addition, based on the replacement delay risk probability model of the Bayesian function, by integrating multi-source data such as liquid volume, flow rate, and network status in real time, a dynamic risk assessment framework is constructed, greatly improving the early warning accuracy. And through the processing priority feedback mechanism, the task scheduling logic is optimized to ensure that the enhanced annotation information of high-risk infusion bags is transmitted to the nursing terminal along the optimal path. The synergistic effect of the above technical features enables the system to achieve systematic optimization of liquid volume monitoring accuracy, abnormal response speed, and resource allocation efficiency in complex clinical environments, providing an extensible technical paradigm for improving the level of infusion safety management.

[0037] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make changes, modifications, substitutions, and variations to the above embodiments without departing from the purpose and scope protected by the claims of the present invention. These all fall within the protection scope of the present invention.

Claims

1. Medical intelligent infusion management system based on visual recognition, characterized in that: include: Monitoring module and transmission identification module; the monitoring module includes a distributed monitoring unit and a transmission enhancement unit; The distributed monitoring unit collects synchronous images, user information and network transmission status of at least one or more infusion bags, and based on the number of synchronously collected images, network transmission status, infusion rate and the transmission correction model built into the transmission enhancement unit, obtains an infusion bag image set with synchronous transmission correction and marks the remaining liquid volume and remaining infusion time; Based on the corrected infusion bag image set combined with user information, the transmission discrimination module is used to obtain the infusion bag replacement delay risk probability; The delay risk probability of replacing the infusion bag is fed back to the transmission enhancement unit, the transmission image enhancement sequence is adjusted in real time and the delay risk probability is annotated, and the enhanced and annotated image is displayed on the PAD in real time.

2. The medical intelligent infusion management system based on visual recognition according to claim 1, characterized in that: The construction process of the transmission correction model includes: According to the synchronously collected infusion bag image, the background of the infusion bag image is segmented through the background segmentation layer to obtain an infusion bag segmentation area set and a background segmentation area set, and the coordinates of the liquid surface line in the infusion bag segmentation area are marked; Based on the infusion bag shape attribute information, the infusion bag segmentation area, the full liquid volume and the initial liquid level, a three-dimensional model of the infusion bag is obtained through the three-dimensional modeling layer; Based on the three-dimensional model of the infusion bag combined with different infusion rates, the liquid surface separation line position function under different infusion rates was fitted through the simulation algorithm combined with the support vector machine kernel function to obtain the separation line position function corresponding to different infusion rates.

3. The medical intelligent infusion management system based on visual recognition as claimed in claim 2, characterized in that: The construction process of the transmission correction model also includes: Masking the liquid separation lines in the infusion bag segmentation area, inputting each masked infusion bag segmentation area and the liquid separation line position function into the feature extraction layer of the transmission correction model, and obtaining the feature space of each infusion bag segmentation area and the liquid level line coordinate space predicted by the liquid separation line position function; The liquid level line position prediction error loss is obtained based on the liquid level line coordinate space predicted by the feature space of the infusion bag segmentation area and the liquid separation line position function combined with the marked liquid level line coordinates.

4. The medical intelligent infusion management system based on visual recognition as claimed in claim 3, characterized in that: The construction process of the transmission correction model also includes: The continuous infusion bag image sequence constructed by the masked infusion bag segmentation area corresponding to the continuous time points is input into the optical flow layer, and the motion vector of each pixel corresponding to the continuous time points of the masked infusion bag segmentation area is analyzed to obtain the dense optical flow field corresponding to the continuous masked infusion bag segmentation area; Based on the change values ​​of adjacent pixels in the dense optical flow field corresponding to each masked infusion bag segmentation area, a dynamic clustering sub-region pixel band space corresponding to each masked infusion bag segmentation area is constructed by combining the clustering algorithm with the quarter-third digit; According to the pixel change value in the dynamic clustering sub-region pixel band space and the pixel change mean corresponding to the full-load liquid infusion bag, a pixel change abnormality threshold is constructed to obtain an abnormal dynamic clustering sub-region pixel band; The abnormal dynamic clustering sub-region pixel band is a dynamic clustering sub-region pixel band corresponding to the largest number of abnormal pixel change coordinate points and the largest abnormal deviation value.

5. The medical intelligent infusion management system based on visual recognition as claimed in claim 4, characterized in that: The construction process of the transmission correction model also includes: The pixel band of the abnormal dynamic clustering sub-region is input into the Gaussian kernel convolution, and the corresponding abnormal pixel points in the pixel band of the abnormal dynamic clustering sub-region are diffused to obtain the continuous abnormal pixel point coordinate space; Feeding back the continuous abnormal pixel point coordinate space to the feature extraction layer to perform secondary correction on the liquid level line position prediction error loss and the predicted liquid level line coordinate space, so as to obtain the corrected liquid level line position prediction error loss and the predicted liquid level line coordinate space; The infusion bag segmentation area space to be enhanced is constructed based on the corrected liquid level position prediction error loss, the predicted liquid level coordinate space, and the infusion bag segmentation area feature space.

6. The medical intelligent infusion management system based on visual recognition as claimed in claim 5, characterized in that: The construction process of the transmission correction model also includes: According to the available bandwidth in different time periods in the network transmission status, the packet loss rate and transmission delay corresponding to the number of images transmitted simultaneously, the network quality index corresponding to different time periods is constructed through the network quality evaluation layer; Based on the network quality index corresponding to different time periods and the preset network quality level interval, the network transmission level corresponding to different time periods is obtained; The network transmission levels corresponding to the different time periods and the preset image enhancement resolution table are input into the hierarchical correction layer to simulate the enhanced transmission of the segmented area space of the infusion bag to be enhanced, so as to obtain the transmission delay and packet loss of each resolution image at different network transmission levels and the corresponding network level-image quality mapping matrix.

7. The medical intelligent infusion management system based on visual recognition according to claim 6, characterized in that: The construction process of the transmission correction model also includes: Based on the different pixel spaces of the infusion bag segmentation areas after masking and the predicted liquid level position combined with the real-time infusion rate, the probability value of replacement delay risk and probability evaluation loss corresponding to each infusion bag at the current moment are obtained through the probability evaluation layer with built-in Bayesian function; The replacement delay risk probability value corresponding to each infusion bag at the current moment is used as the processing priority of the infusion bag segmentation region space to be enhanced corresponding to each infusion bag acquisition image; The processing priority of the space of the infusion bag segmentation area to be enhanced corresponding to each infusion bag captured image is fed back to the simulated enhancement transmission process, the optimal image quality corresponding to each processing priority image under different network levels is simulated, and the transmission delay is used as a measurement indicator to construct a processing priority-network level-image quality-transmission delay mapping matrix space.

8. The medical intelligent infusion management system based on visual recognition according to claim 7, characterized in that: The construction process of the transmission correction model also includes: The processing priority-network level-image quality-transmission delay mapping matrix space is built into the output layer, and the segmented area space of the infusion bag to be enhanced corresponding to the current moment, the network quality index, and the processing priority corresponding to each infusion bag are input into the output layer to obtain the infusion bag image of the corresponding resolution image quality and the corresponding image enhancement loss; The loss function of the transmission correction model is constructed based on the corrected liquid level position prediction error loss, probability assessment loss, transmission delay and packet loss loss, and image enhancement loss. The transmission correction model is simulated and trained, and the transmission delay corresponding to each resolution image quality under different network qualities and processing priorities is analyzed. When the transmission delay corresponding to each resolution image quality is less than the preset transmission delay, the trained transmission correction model is obtained, and the enhanced infusion bag image and the corresponding enhanced resolution at the corresponding network level are output.

9. The medical intelligent infusion management system based on visual recognition according to claim 6, characterized in that: The process of masking the liquid separation line in the segmentation area of ​​the infusion bag comprises: Based on the infusion bag segmentation area, a multi-level pyramid feature map corresponding to the infusion bag segmentation area is generated by combining a Gaussian pyramid decomposition algorithm with a preset number of layers; Based on the multi-level pyramid feature map, the horizontal gradient and vertical gradient convolution are performed on each level of the pyramid feature map through the Sobel operator convolution to obtain the horizontal and vertical gradient amplitudes and gradient direction angles corresponding to each level of the pyramid feature map; Based on the gradient direction angle combined with the preset gradient direction angle filtering interval, the horizontal and vertical gradient amplitudes are filtered to obtain the horizontal and vertical gradient amplitudes corresponding to the filtered feature map of each level of the pyramid; The channel convolution attention algorithm is used to weighted attenuate the horizontal and vertical gradient amplitudes corresponding to the filtering of each level of the pyramid feature map, and each level of the pyramid feature map after weighted attenuation is upsampled and fused through bilinear interpolation to obtain a multi-level pyramid fusion gradient map.

10. The medical intelligent infusion management system based on visual recognition according to claim 9, characterized in that: The process of masking the liquid separation line in the segmentation area of ​​the infusion bag further includes: Based on the multi-level pyramid fusion gradient map, the mean and variance of the fusion gradient map are obtained through statistical algorithms; The mask interval is set based on the mean and variance corresponding to the fused gradient map, and the initial binary mask corresponding to the liquid level line in the segmentation area of ​​the infusion bag is generated according to the set mask interval combined with non-maximum suppression; Based on the initial binary mask corresponding to the liquid level line in the infusion bag segmentation area, the binary mask corresponding to the liquid level line in the infusion bag segmentation area is obtained through the 3×3 rectangular structure element S combined with the connected domain filter, and the liquid level line marked in the input infusion bag segmentation area is binary masked based on the binary mask corresponding to the liquid level line in the liquid bag segmentation area.

Images