A method and system for monitoring bleeding after thyroid surgery

Abstract

The present invention discloses a method and system for monitoring postoperative bleeding of the thyroid gland, which relates to the field of medical monitoring technology. The present invention adopts a YOLO-Seg network model and combines the SRCNN layer and the Blind Dconv layer to identify the drainage tube area and the dressing area with high precision, improve the accuracy of regional recognition, and enhance the clarity and deblurring effect of the image; the accurate detection of the color, category and proportion information of the drainage fluid provides strong support for judging whether the patient is bleeding; through comprehensive analysis of physiological indicator data, the overall health status of the patient is more comprehensively evaluated, the physiological indicator score accurately reflects the health status of the patient, and the potential bleeding risk is discovered in time, which provides a comprehensive, accurate and timely solution for the monitoring of patients after thyroid surgery, and effectively ensures the safety and recovery of the patient.

Description

Technical Field

[0001] The present invention relates to the technical field of medical monitoring, and in particular to a method and system for monitoring post-thyroid surgery bleeding. Background Art

[0002] Bleeding after thyroid surgery is one of the most serious complications of thyroid surgery, which will have a serious impact on the patient's health and subsequent recovery. Postoperative bleeding can cause swelling and pain in the neck, affecting the patient's comfort and quality of life. A small amount of bleeding may form a hematoma, causing swelling and pain in the neck. If the amount of bleeding is large, it may compress the trachea, causing difficulty breathing, and even endangering life. This acute airway obstruction requires emergency treatment, otherwise it may cause the patient to suffocate and die. In addition, bleeding may also cause damage to the recurrent laryngeal nerve, affect the movement of the vocal cords, and cause hoarseness or aphonia. Postoperative bleeding will increase the difficulty and time of recovery, and patients will need longer hospitalization and more medical interventions, such as hematoma removal and endotracheal intubation.

[0003] If postoperative bleeding is not treated in time, the patient may suffocate due to airway obstruction or even die. Even if the patient survives, long-term sequelae may remain. Therefore, a monitoring method is needed to closely monitor patients after thyroid surgery in order to detect and treat any signs of bleeding in a timely manner. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for monitoring post-thyroid surgery bleeding to improve the above-mentioned technical problems.

[0005] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0006] A method for monitoring post-thyroid surgery bleeding, comprising:

[0007] S1. Based on the monitoring interval, real-time acquisition of patient monitoring image data and physiological index data;

[0008] S2, build region recognition model and image detection model;

[0009] S3, inputting the monitoring image data into the region recognition model, and outputting the region recognition result; wherein the region recognition result includes the drainage tube region image and the dressing region image;

[0010] S4, inputting the region recognition result into the image detection model, and outputting the drainage tube detection result and the dressing detection result; wherein the dressing detection result is the bloodstain area, and the drainage tube detection result includes the drainage category result, the drainage tube color classification result and the proportion information difference;

[0011] S5. Calculate and test the physiological index data to obtain a physiological index score;

[0012] S6. Obtain monitoring results based on the drainage tube test results, dressing test results and physiological index scores.

[0013] Furthermore, the region recognition model includes a serially connected edge segmentation extraction module, an SRCNN layer, and a BlindDconv layer; the edge segmentation extraction module includes a YOLO-Seg network model; the YOLO-Seg network model includes a serially connected Backbone module, a Neck module, a Detection Head module, and a Segmentation Head module;

[0014] The image detection model includes a drainage tube area image detection module and a dressing area image detection module in parallel; the drainage tube area image detection module includes a color conversion submodule, a first color segmentation submodule, a color proportion calculation submodule, a texture feature extraction submodule and a liquid category classification submodule; the dressing area image detection module includes a second color segmentation submodule, a bloodstain detection submodule, a bloodstain judgment submodule and a bloodstain area calculation submodule connected in series.

[0015] Furthermore, the training process of the region recognition model is as follows:

[0016] S3-1, acquiring training monitoring image data and performing preprocessing to obtain preprocessed training monitoring image data;

[0017] S3-2, inputting the preprocessed training monitoring image data into the Backbone module, and outputting multi-scale monitoring image features;

[0018] S3-3, inputting the multi-scale monitoring image features into the Neck module, and outputting the fused multi-scale monitoring image features;

[0019] S3-4, inputting the fused multi-scale surveillance image features into the Detection Head module, and outputting the corresponding target detection frame and its category probability value;

[0020] S3-5, inputting the target detection frame and its category probability value into the Segmentation Head module, and outputting the initial training area recognition result;

[0021] S3-6, inputting the initial training area recognition result into the SRCNN layer, and outputting the processed initial training area recognition result;

[0022] S3-7, inputting the processed initial training area recognition result into the Blind Dconv layer, and outputting a training area recognition result; wherein the training area recognition result includes a drainage tube area training image and a dressing area training image;

[0023] S3-8. Calculate a first loss function based on the training area recognition result;

[0024] S3-9. Based on the first loss function, adjust the weight parameters of the region recognition model.

[0025] Furthermore, the color conversion submodule is used to perform color space conversion on the drainage tube region image to obtain a converted drainage tube region image;

[0026] The first color segmentation submodule is used to set a first color threshold and extract a color region; classify the converted drainage tube region image using a color classification algorithm to obtain a corresponding drainage tube color classification result; wherein the drainage tube color classification result is dark red, bright red, light red or other colors;

[0027] The color proportion calculation submodule is used to calculate the proportion information of the color area in the converted drainage tube area image; obtain the proportion information corresponding to the previous monitoring, calculate the difference between the two proportion information, and obtain the corresponding proportion information difference;

[0028] The texture feature extraction submodule is used to extract texture feature data of the converted drainage tube region image using a texture analysis algorithm;

[0029] The liquid category classification submodule is used to extract features from the texture feature data using a neural network, determine the drainage liquid category, and obtain a drainage category result; wherein the drainage category result is a slurry liquid or a non-slurry liquid;

[0030] The drainage tube detection result includes the drainage category result, the drainage tube color classification result and the proportion information difference.

[0031] Furthermore, the second color segmentation submodule is used to perform color conversion on the dressing area image; set a second color threshold, and segment the converted dressing area image using color segmentation to generate a corresponding image mask;

[0032] The bloodstain detection submodule is used to extract features from the image mask using a deep learning model to generate a bloodstain detection result; wherein the bloodstain detection result is a bleeding area of ​​the dressing;

[0033] The bloodstain judgment submodule is used to judge whether the patient is bleeding based on the bloodstain detection result, and obtain a corresponding judgment result; wherein the judgment result is that the patient is bleeding or the patient is not bleeding;

[0034] The bloodstain area calculation submodule is used to calculate the bloodstain area based on the bloodstain detection result and the judgment result.

[0035] Furthermore, the physiological index data include the patient's temperature, blood sugar, blood pressure, blood oxygen saturation, heart rate, respiratory rate and sound data;

[0036] The S5 comprises the following steps:

[0037] S5-1, obtaining the temperature and the sound data through a temperature sensor and a sound collection device;

[0038] S5-2, using a monitoring device to obtain the blood sugar, the blood pressure, the blood oxygen saturation, the heart rate, and the respiratory rate;

[0039] S5-3, analyzing the sound data using a signal processing algorithm to obtain a corresponding sound analysis result;

[0040] S5-4. Calculate the physiological indicator score based on the temperature, the sound analysis result, the blood sugar, the blood pressure, the blood oxygen saturation, the heart rate, and the respiratory rate.

[0041] Furthermore, the formula corresponding to the physiological index score is:

[0042] ;

[0043] in, , Represent the temperature coefficient and temperature fraction respectively, , Respectively represent the sound coefficient and sound score, , Respectively represent the blood sugar coefficient and blood sugar score, , Respectively represent blood pressure coefficient and blood pressure score, , Respectively represent the blood oxygen saturation coefficient and blood oxygen saturation score, , Represents heart rate coefficient and heart rate score respectively. , Represent the respiratory rate coefficient and respiratory rate score respectively. represents the physiological index score, represents a natural constant, Represents the bias term.

[0044] Further, the S6 comprises the following steps:

[0045] S6-1, determine whether any one of the following conditions is met: the physiological index score is lower than the physiological threshold, the drainage category result is serous fluid, and the drainage color classification result is dark red / bright red; if so, proceed to S6-2; otherwise, the monitoring result is that the patient has massive postoperative bleeding and needs urgent treatment;

[0046] S6-2. Determine whether the conditions that the proportion information difference is greater than 0 and the bloodstain area is greater than 0 are met; if so, the monitoring result is that the patient has a small amount of bleeding after surgery and needs treatment; otherwise, the monitoring result is that the patient is stable after surgery and needs observation.

[0047] A thyroid surgery bleeding monitoring system comprises an image acquisition module, a region recognition module, an image detection module, a physiological index data acquisition module, a physiological index score calculation module, and a monitoring module; wherein:

[0048] An image acquisition module, used to acquire monitoring image data and physiological index data of the patient in real time based on the monitoring interval;

[0049] The region recognition module is used to input the monitoring image data into the region recognition model and output the region recognition result;

[0050] An image detection module is used to input the region recognition result into the image detection model, and output the drainage tube detection result and the dressing detection result;

[0051] A physiological index data acquisition module is used to acquire the patient's physiological index data;

[0052] A physiological index score calculation module is used to analyze and calculate the physiological index data to obtain the physiological index score;

[0053] The monitoring module is used to obtain monitoring results based on the drainage tube detection results, the dressing detection results and the physiological index scores.

[0054] Furthermore, the physiological index data acquisition module includes a sound collection device, a temperature sensor, a monitoring device and an index preprocessing submodule; the physiological index score calculation module includes a sound analysis submodule and a physiological index score calculation submodule; the postoperative thyroid bleeding monitoring system also includes a warning module; wherein:

[0055] A warning module is used to send corresponding warning signals to medical staff according to the monitoring results;

[0056] A sound collection device, used to obtain sound data of the patient;

[0057] A temperature sensor for obtaining the patient's temperature;

[0058] Monitoring devices to obtain blood sugar, blood pressure, blood oxygen saturation, heart rate, and respiratory rate;

[0059] The indicator preprocessing submodule is used to denoise and interpolate data of temperature, sound data, blood sugar, blood pressure, blood oxygen saturation, heart rate, and respiratory rate;

[0060] The sound analysis submodule is used to analyze the pre-processed sound data using a signal processing algorithm to obtain corresponding sound analysis results;

[0061] The physiological index score calculation submodule is used to calculate the physiological index score based on the preprocessed temperature, sound analysis results, preprocessed blood sugar, preprocessed blood pressure, preprocessed blood oxygen saturation, preprocessed heart rate, and preprocessed respiratory rate.

[0062] The beneficial effects of the present invention are:

[0063] This method uses the YOLO-Seg network model combined with the SRCNN layer and the Blind Dconv layer to identify the drainage tube area and the dressing area with high precision. Through multi-scale feature extraction and fusion, the accuracy of regional recognition is improved, the clarity and deblurring effect of the image are improved, and accurate recognition under different lighting conditions is ensured. The color features and texture features of the drainage tube area and the dressing area are extracted, and the color, category and proportion information of the drainage fluid are accurately detected, which provides strong support for judging whether the patient is bleeding. Through comprehensive analysis of physiological indicator data, the overall health status of the patient is more comprehensively evaluated, and the health status of the patient is accurately reflected through the physiological indicator score, so as to timely discover potential bleeding risks.

[0064] This system uses deep learning models and signal processing algorithms to perform intelligent analysis of image and sound data, improving the accuracy and reliability of monitoring. It can send warning signals of different levels based on the monitoring results, and the various modules collaborate with each other to ensure the efficient operation of the system. It has a rapid response mechanism to ensure that when a patient bleeds, medical staff can take immediate action to avoid serious consequences. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0066] Figure 1 A flow chart of a method in an embodiment of the present invention;

[0067] Figure 2 A structural diagram of a region recognition model in an embodiment of the present invention;

[0068] Figure 3 A structural diagram of an image detection model in an embodiment of the present invention;

[0069] Figure 4 A system structure diagram in an embodiment of the present invention;

[0070] Figure 5 4 is a structural diagram of a physiological indicator data module and a physiological indicator score calculation module in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0072] See also Figure 1 The present embodiment provides a method for monitoring postoperative bleeding in the thyroid gland, including:

[0073] S1. Based on the monitoring interval, the patient's monitoring image data and physiological indicator data are acquired in real time; the physiological indicator data include the patient's temperature, blood sugar, blood pressure, blood oxygen saturation, heart rate, respiratory rate and sound data; the patient is monitored for 24 hours or 72 hours, and the monitoring interval is set to half an hour.

[0074] S2, build region recognition model and image detection model;

[0075] S3, inputting the monitoring image data into the region recognition model, and outputting the region recognition result; wherein the region recognition result includes the drainage tube region image and the dressing region image;

[0076] like Figure 2As shown, the region recognition model includes a serially connected edge segmentation extraction module, an SRCNN layer, and a BlindDconv layer; the edge segmentation extraction module includes a YOLO-Seg network model; the YOLO-Seg network model includes a serially connected Backbone module, a Neck module, a Detection Head module, and a Segmentation Head module;

[0077] This method selects the YOLO-Seg network model, SRCNN layer and Blind Dconv layer for region recognition, which has the advantages of high precision, image clarity improvement, deblurring processing, etc. It can adapt to different image conditions and has strong robustness and practicality. Through the end-to-end processing flow, the steps of region recognition and image enhancement are simplified, and the efficiency is improved.

[0078] The training process of the region recognition model is as follows:

[0079] S3-1. Acquire training monitoring image data and perform preprocessing to obtain preprocessed training monitoring image data; the preprocessing includes denoising and contrast enhancement.

[0080] S3-2, inputting the preprocessed training monitoring image data into the Backbone module, and outputting multi-scale monitoring image features;

[0081] S3-3, inputting the multi-scale monitoring image features into the Neck module, and outputting the fused multi-scale monitoring image features;

[0082] S3-4, inputting the fused multi-scale surveillance image features into the Detection Head module, and outputting the corresponding target detection frame and its category probability value;

[0083] S3-5, inputting the target detection frame and its category probability value into the Segmentation Head module, and outputting the initial training area recognition result;

[0084] S3-6, inputting the initial training area recognition result into the SRCNN layer to further improve the clarity of the image, and outputting the processed initial training area recognition result;

[0085] S3-7, inputting the processed initial training area recognition result into the Blind Dconv layer, performing deblurring processing, improving image quality, and outputting a training area recognition result; wherein the training area recognition result includes a drainage tube area training image and a dressing area training image;

[0086] S3-8. Calculate a first loss function based on the training area recognition result;

[0087] The first loss function The corresponding formula is:

[0088] ;

[0089] ;

[0090] ;

[0091] ;

[0092] in, Indicates The loss coefficient, represents the summation function, represents the YOLO-Seg loss function, represents the lighting condition factor, , Respectively represent BCE loss weight and Dice loss weight, , Respectively represent the BCE loss function and the Dice loss weight function, Represents the loss function corresponding to the SRCNN layer (mean square error loss function), , They represent the first and second recognition results of the initial training area after processing. The pixel value of the label, the processed initial training area recognition result pixel values, Represents the total number of pixels of the processed initial training area recognition results, represents the loss function corresponding to Blind Dconv, , represents the weight coefficient, , They represent image reconstruction loss (mean square error loss function) and blur kernel regularization loss function (L1 regularization). The value is 1 during daytime and 2 during nighttime.

[0093] The first loss function combines the damage functions of each module of the region recognition model, introduces the illumination condition factor, and dynamically adjusts the corresponding weight parameters, which can further improve the recognition accuracy and robustness of the region recognition model and adapt to different illumination conditions.

[0094] S3-9. Based on the first loss function, adjust the weight parameters of the region recognition model.

[0095] S4, inputting the region recognition result into the image detection model, and outputting the drainage tube detection result and the dressing detection result; wherein the dressing detection result is the bloodstain area, and the drainage tube detection result includes the drainage category result, the drainage tube color classification result and the proportion information difference;

[0096] like Figure 3 As shown, the image detection model includes a parallel drainage tube area image detection module and a dressing area image detection module; the drainage tube area image detection module includes a color conversion submodule, a first color segmentation submodule, a color proportion calculation submodule, a texture feature extraction submodule and a liquid category classification submodule.

[0097] The color conversion submodule is used to perform color space conversion on the drainage tube region image into an HSV color space to obtain a converted drainage tube region image;

[0098] The first color segmentation submodule is used to set the corresponding first color threshold according to dark red, bright red, and light red to extract the color area; use the color classification algorithm to classify the converted drainage tube area image to obtain the corresponding drainage tube color classification result; wherein the drainage tube color classification result is dark red, bright red, light red or other colors; the color classification algorithm can use the K-Nearest Neighbors algorithm or the KNN network. For example, the first color threshold can be set to:

[0099] Dark red is: H (0-10), S (100-255), V (100-255); bright red is: H (0-10), S (150-255), V (150-255); light red is: H (0-10), S (50-150), V (150-255).

[0100] The color proportion calculation submodule is used to calculate the proportion information of the color area in the converted drainage tube area image; obtain the proportion information corresponding to the previous monitoring, calculate the difference between the two proportion information, that is, the proportion information of the current monitoring minus the proportion information of the previous monitoring, and obtain the corresponding proportion information difference;

[0101] The texture feature extraction submodule is used to extract texture feature data of the converted drainage tube area image using a texture analysis algorithm; since the texture of a slurry liquid is usually rougher than that of an ordinary liquid, the texture analysis algorithm may select a gray level co-occurrence matrix (GLCM) to analyze the texture features of the liquid.

[0102] The liquid category classification submodule is used to extract features using a neural network, determine the drainage liquid category, and obtain a drainage category result; wherein the drainage category result is a slurry liquid or a non-slurry liquid;

[0103] The neural network can select LeNet-5 convolutional neural network, and the corresponding process is:

[0104] The texture feature data is processed through the convolution layer, and multiple convolution kernels are used for filtering to generate a feature map. The feature map is downsampled through the pooling layer to reduce the data dimension and retain important features. After alternating processing of multiple convolution and pooling layers, higher-level features are further extracted. The extracted feature map is flattened into a one-dimensional vector and input into the fully connected layer for feature combination and classification. The probability of each category is output through the Softmax function to complete the classification task.

[0105] The LeNet-5 convolutional neural network has a simple structure and is suitable for small-scale image classification tasks. It can automatically extract the features of drainage tube fluid and has the advantages of strong robustness and efficient classification. It can efficiently complete the classification task of drainage fluid categories and provide support for determining whether the patient is bleeding.

[0106] The drainage tube detection result includes the drainage category result, the drainage tube color classification result and the proportion information difference.

[0107] like Figure 3 As shown, the dressing area image detection module includes a second color segmentation submodule, a bloodstain detection submodule, a bloodstain judgment submodule and a bloodstain area calculation submodule connected in series.

[0108] The second color segmentation submodule is used to perform color conversion on the dressing area image; set a second color threshold, segment the converted dressing area image using color segmentation, and generate a corresponding image mask;

[0109] The bloodstain detection submodule is used to extract features from the image mask using a deep learning model to generate a bloodstain detection result; wherein the bloodstain detection result is a bleeding area of ​​the dressing;

[0110] The bloodstain judgment submodule is used to judge whether the patient is bleeding based on the bloodstain detection result, and obtain a corresponding judgment result; wherein the judgment result is that the patient is bleeding or the patient is not bleeding;

[0111] The bloodstain area calculation submodule is used to calculate the bloodstain area based on the bloodstain detection result and the judgment result. When the judgment result is that the patient is not bleeding, the bloodstain area is 0.

[0112] S5. Calculate and test the physiological index data to obtain a physiological index score;

[0113] The S5 comprises the following steps:

[0114] S5-1, obtaining the temperature and the sound data through a temperature sensor and a sound collection device;

[0115] S5-2, using a monitoring device to obtain the blood sugar, the blood pressure, the blood oxygen saturation, the heart rate, and the respiratory rate;

[0116] S5-3. Analyze the sound data using a signal processing algorithm to obtain a corresponding sound analysis result; the signal processing algorithm may use a multi-layer perceptron or a convolutional neural network, and the corresponding process is:

[0117] Feature extraction is performed on the sound data to obtain the corresponding power spectrum density; the power spectrum density is input into a multi-layer perceptron or a convolutional neural network to classify the breathing sounds to obtain the sound analysis results; the sound analysis results are dyspnea, rapid breathing or steady breathing.

[0118] S5-4. Calculate the physiological indicator score based on the temperature, the sound analysis result, the blood sugar, the blood pressure, the blood oxygen saturation, the heart rate, and the respiratory rate.

[0119] Based on the sound analysis results, the corresponding sound score is obtained, namely:

[0120] ;

[0121] Temperature fraction , blood sugar score , blood pressure score , blood oxygen saturation score , Heart Rate Score , respiratory rate score The corresponding formula is:

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] ;

[0127] ;

[0128] in, , Represent temperature, average temperature of patients, , Represents blood sugar and average blood sugar of patients, , Respectively represent the systolic blood pressure and the average systolic blood pressure of patients. , Respectively represent the diastolic blood pressure and the average diastolic blood pressure of the patient. , Respectively represent blood oxygen saturation and average blood oxygen saturation of patients. , Represent heart rate, average heart rate of patients, , Represent the respiratory rate and the average respiratory rate of patients respectively.

[0129] Thus, the formula corresponding to the physiological index score is:

[0130] ;

[0131] in, represents the temperature coefficient, Represents the sound coefficient, It represents the glycemic index. represents the blood pressure coefficient, Indicates the blood oxygen saturation coefficient, represents the heart rate coefficient, represents the respiratory rate coefficient, represents the physiological index score, represents a natural constant, Represents the bias term.

[0132] S6. Obtain monitoring results based on the drainage tube test results, dressing test results and physiological index scores.

[0133] The S6 comprises the following steps:

[0134] S6-1. Determine whether any one of the following conditions is met: the physiological index score is lower than the physiological threshold, the drainage category result is serous fluid, and the drainage color classification result is dark red / bright red; if so, proceed to S6-2; otherwise, the monitoring result is that the patient has massive postoperative bleeding and needs urgent treatment; the physiological threshold can be set to 0.3.

[0135] S6-2. Determine whether the conditions that the proportion information difference is greater than 0 and the bloodstain area is greater than 0 are met; if so, the monitoring result is that the patient has a small amount of bleeding after surgery and needs treatment; otherwise, the monitoring result is that the patient is stable after surgery and needs observation.

[0136] In summary, this method uses the YOLO-Seg network model combined with the SRCNN layer and the Blind Dconv layer to identify the drainage tube area and the dressing area with high precision. Through multi-scale feature extraction and fusion, the accuracy of regional recognition is improved, the clarity and deblurring effect of the image are improved, and accurate recognition under different lighting conditions is ensured. The color features and texture features of the drainage tube area and the dressing area are extracted to accurately detect the color, category and proportion of the drainage fluid, which provides strong support for judging whether the patient is bleeding. Through comprehensive analysis of physiological indicator data, the overall health status of the patient is more comprehensively evaluated, and the health status of the patient is accurately reflected through the physiological indicator scores, so as to timely discover potential bleeding risks.

[0137] like Figure 4 As shown, a thyroid surgery bleeding monitoring system includes an image acquisition module, a region recognition module, an image detection module, a physiological index data acquisition module, a physiological index score calculation module, and a monitoring module; wherein:

[0138] An image acquisition module, used to acquire monitoring image data and physiological index data of the patient in real time based on the monitoring interval;

[0139] The region recognition module is used to input the monitoring image data into the region recognition model and output the region recognition result;

[0140] An image detection module is used to input the region recognition result into the image detection model, and output the drainage tube detection result and the dressing detection result;

[0141] A physiological index data acquisition module is used to acquire the patient's physiological index data;

[0142] A physiological index score calculation module is used to analyze and calculate the physiological index data to obtain the physiological index score;

[0143] like Figure 5 As shown, the physiological index data acquisition module includes a sound collection device, a temperature sensor, and a monitoring device; the physiological index score calculation module includes a sound analysis submodule and a physiological index score calculation submodule; the thyroid postoperative bleeding monitoring system also includes a warning module; wherein:

[0144] The monitoring module is used to obtain monitoring results based on the drainage tube detection results, the dressing detection results and the physiological index score. The physiological index data acquisition module includes a sound collection device, a temperature sensor, a monitoring device and an index preprocessing submodule; the physiological index score calculation module includes a sound analysis submodule and a physiological index score calculation submodule; the thyroid postoperative bleeding monitoring system also includes a warning module; wherein:

[0145] A warning module is used to send corresponding warning signals to medical staff according to the monitoring results;

[0146] A sound collection device, used to obtain sound data of the patient;

[0147] A temperature sensor for obtaining the patient's temperature;

[0148] Monitoring devices to obtain blood sugar, blood pressure, blood oxygen saturation, heart rate, and respiratory rate;

[0149] The indicator preprocessing submodule is used to denoise and interpolate data of temperature, sound data, blood sugar, blood pressure, blood oxygen saturation, heart rate, and respiratory rate;

[0150] The sound analysis submodule is used to analyze the pre-processed sound data using a signal processing algorithm to obtain corresponding sound analysis results;

[0151] The physiological index score calculation submodule is used to calculate the physiological index score based on the preprocessed temperature, sound analysis results, preprocessed blood sugar, preprocessed blood pressure, preprocessed blood oxygen saturation, preprocessed heart rate, and preprocessed respiratory rate.

[0152] In the monitoring module, when the monitoring result shows that the patient has a lot of bleeding after surgery and needs to be treated urgently, or the patient has a small amount of bleeding after surgery and needs to be treated, the first-level alarm instruction is transmitted to the warning module, and the patient information and operation information are transmitted to the computer or mobile phone of the medical staff, and a warning is issued. When the monitoring result shows that the patient is stable after surgery and needs to be observed, the observation instruction is transmitted to the warning module. The warning device integrates the physiological index score, drainage tube test results, dressing test results and patient information into pictures and texts, and sends them to the computer or mobile phone of the medical staff, and reminds the medical staff to observe the patient. Providing an intuitive visual interface, medical staff can easily view the patient's monitoring images, physiological index data and monitoring results. The graphic and text display method enables medical staff to quickly understand and deal with the patient's condition, improving work efficiency.

[0153] In summary, this system uses deep learning models and signal processing algorithms to perform intelligent analysis of image and sound data, thereby improving the accuracy and reliability of monitoring. It can send warning signals of different levels based on the monitoring results, and the various modules collaborate with each other to ensure the efficient operation of the system. It has a rapid response mechanism to ensure that when a patient bleeds, medical staff can take measures in the first place to avoid serious consequences.

[0154] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for monitoring bleeding after thyroid surgery, characterized in that: include: S1. Based on the monitoring interval, real-time acquisition of patient monitoring image data and physiological index data; S2, build region recognition model and image detection model; S3, inputting the monitoring image data into the region recognition model, and outputting the region recognition result; wherein the region recognition result includes the drainage tube region image and the dressing region image; S4, inputting the region recognition result into the image detection model, and outputting the drainage tube detection result and the dressing detection result; wherein the dressing detection result is the bloodstain area, and the drainage tube detection result includes the drainage category result, the drainage tube color classification result and the proportion information difference; S5. Calculate and test the physiological index data to obtain a physiological index score; S6. Obtain monitoring results based on the drainage tube test results, dressing test results and physiological index scores; The region recognition model includes a serially connected edge segmentation extraction module, an SRCNN layer, and a Blind Dconv layer; the edge segmentation extraction module includes a YOLO-Seg network model; the YOLO-Seg network model includes a serially connected Backbone module, a Neck module, a Detection Head module, and a Segmentation Head module; The image detection model includes a drainage tube area image detection module and a dressing area image detection module in parallel; the drainage tube area image detection module includes a color conversion submodule, a first color segmentation submodule, a color proportion calculation submodule, a texture feature extraction submodule and a liquid category classification submodule; the dressing area image detection module includes a second color segmentation submodule, a bloodstain detection submodule, a bloodstain judgment submodule and a bloodstain area calculation submodule connected in series.

2. A method for monitoring postoperative bleeding in thyroid surgery according to claim 1, characterized in that: The training process of the region recognition model is as follows: S3-1, acquiring training monitoring image data and performing preprocessing to obtain preprocessed training monitoring image data; S3-2, inputting the preprocessed training monitoring image data into the Backbone module, and outputting multi-scale monitoring image features; S3-3, inputting the multi-scale monitoring image features into the Neck module, and outputting the fused multi-scale monitoring image features; S3-4, inputting the fused multi-scale surveillance image features into the Detection Head module, and outputting the corresponding target detection frame and its category probability value; S3-5, inputting the target detection frame and its category probability value into the Segmentation Head module, and outputting the initial training area recognition result; S3-6, inputting the initial training area recognition result into the SRCNN layer, and outputting the processed initial training area recognition result; S3-7, inputting the processed initial training area recognition result into the Blind Dconv layer, and outputting a training area recognition result; wherein the training area recognition result includes a drainage tube area training image and a dressing area training image; S3-8. Calculate a first loss function based on the training area recognition result; S3-9. Based on the first loss function, adjust the weight parameters of the region recognition model.

3. A method for monitoring postoperative bleeding after thyroid surgery according to claim 1, characterized in that: The color conversion submodule is used to perform color space conversion on the drainage tube region image to obtain a converted drainage tube region image; The first color segmentation submodule is used to set a first color threshold and extract a color region; classify the converted drainage tube region image using a color classification algorithm to obtain a corresponding drainage tube color classification result; wherein the drainage tube color classification result is dark red, bright red, light red or other colors; The color proportion calculation submodule is used to calculate the proportion information of the color area in the converted drainage tube area image; obtain the proportion information corresponding to the previous monitoring, calculate the difference between the two proportion information, and obtain the corresponding proportion information difference; The texture feature extraction submodule is used to extract texture feature data of the converted drainage tube region image using a texture analysis algorithm; The liquid category classification submodule is used to extract features from the texture feature data using a neural network, determine the drainage liquid category, and obtain a drainage category result; wherein the drainage category result is a slurry liquid or a non-slurry liquid; The drainage tube detection result includes the drainage category result, the drainage tube color classification result and the proportion information difference.

4. A method for monitoring post-thyroid surgery bleeding according to claim 1, characterized in that: The second color segmentation submodule is used to perform color conversion on the dressing area image; Setting a second color threshold, segmenting the converted dressing area image using color segmentation, and generating a corresponding image mask; The bloodstain detection submodule is used to extract features from the image mask using a deep learning model to generate a bloodstain detection result; wherein the bloodstain detection result is a bleeding area of ​​the dressing; The bloodstain judgment submodule is used to judge whether the patient is bleeding based on the bloodstain detection result, and obtain a corresponding judgment result; wherein the judgment result is that the patient is bleeding or the patient is not bleeding; The bloodstain area calculation submodule is used to calculate the bloodstain area based on the bloodstain detection result and the judgment result.

5. The method for monitoring post-thyroid surgery bleeding according to claim 1, characterized in that: The physiological index data include the patient's temperature, blood sugar, blood pressure, blood oxygen saturation, heart rate, respiratory rate and voice data; The S5 comprises the following steps: S5-1, obtaining the temperature and the sound data through a temperature sensor and a sound collection device; S5-2, using a monitoring device to obtain the blood sugar, the blood pressure, the blood oxygen saturation, the heart rate, and the respiratory rate; S5-3, analyzing the sound data using a signal processing algorithm to obtain a corresponding sound analysis result; S5-4. Calculate the physiological indicator score based on the temperature, the sound analysis result, the blood sugar, the blood pressure, the blood oxygen saturation, the heart rate, and the respiratory rate.

6. A method for monitoring post-thyroid surgery bleeding according to claim 5, characterized in that: The formula corresponding to the physiological index score is: ； in, , Represent the temperature coefficient and temperature fraction respectively, , Respectively represent the sound coefficient and sound score, , Respectively represent the blood sugar coefficient and blood sugar score, , Respectively represent blood pressure coefficient and blood pressure score, , Respectively represent the blood oxygen saturation coefficient and blood oxygen saturation score, , Represents heart rate coefficient and heart rate score respectively. , Represent the respiratory rate coefficient and respiratory rate score respectively. represents the physiological index score, represents a natural constant, Represents the bias term.

7. A method for monitoring post-thyroid surgery bleeding according to claim 3 or 4, characterized in that: The S6 comprises the following steps: S6-1, determine whether any one of the following conditions is met: the physiological index score is lower than the physiological threshold, the drainage category result is serous fluid, and the drainage color classification result is dark red / bright red; if so, proceed to S6-2; otherwise, the monitoring result is that the patient has massive postoperative bleeding and needs urgent treatment; S6-2. Determine whether the conditions that the proportion information difference is greater than 0 and the bloodstain area is greater than 0 are met; if so, the monitoring result is that the patient has a small amount of bleeding after surgery and needs treatment; otherwise, the monitoring result is that the patient is stable after surgery and needs observation.

8. A system for monitoring bleeding after thyroid surgery, used to implement a method for monitoring bleeding after thyroid surgery as claimed in any one of claims 1 to 7, characterized in that: It includes an image acquisition module, a region recognition module, an image detection module, a physiological index data acquisition module, a physiological index score calculation module, and a monitoring module; wherein: An image acquisition module, used to acquire monitoring image data and physiological index data of the patient in real time based on the monitoring interval; The region recognition module is used to input the monitoring image data into the region recognition model and output the region recognition result; An image detection module is used to input the region recognition result into the image detection model, and output the drainage tube detection result and the dressing detection result; A physiological index data acquisition module is used to acquire the patient's physiological index data; A physiological index score calculation module is used to analyze and calculate the physiological index data to obtain the physiological index score; The monitoring module is used to obtain monitoring results based on the drainage tube detection results, the dressing detection results and the physiological index scores.

9. A thyroid surgery bleeding monitoring system according to claim 8, characterized in that: The physiological index data acquisition module includes a sound collection device, a temperature sensor, a monitoring device and an index preprocessing submodule; The physiological index score calculation module includes a sound analysis submodule and a physiological index score calculation submodule; The thyroid surgery bleeding monitoring system further comprises a warning module; wherein: A warning module is used to send corresponding warning signals to medical staff according to the monitoring results; A sound collection device, used to obtain sound data of the patient; A temperature sensor for obtaining the patient's temperature; Monitoring devices to obtain blood sugar, blood pressure, blood oxygen saturation, heart rate, and respiratory rate; The indicator preprocessing submodule is used to denoise and interpolate data of temperature, sound data, blood sugar, blood pressure, blood oxygen saturation, heart rate, and respiratory rate; The sound analysis submodule is used to analyze the pre-processed sound data using a signal processing algorithm to obtain corresponding sound analysis results; The physiological index score calculation submodule is used to calculate the physiological index score based on the preprocessed temperature, sound analysis results, preprocessed blood sugar, preprocessed blood pressure, preprocessed blood oxygen saturation, preprocessed heart rate, and preprocessed respiratory rate.

Images