Drainage tube bleeding early warning device

Abstract

The invention discloses a drainage tube bleeding early-warning device, which comprises a drainage tube body which is retained in the body of a patient, and a part of the drainage tube body is a transparent end; the flowmeter is connected with the transparent end; the optical sensing module and the color sensor set monitoring frequency; the transmission module is respectively connected with the flowmeter, the optical sensing module and the color sensor; the microprocessor is connected with the transmission module and predicts the bleeding probability through a preset machine learning model; the cloud module is connected with the microprocessor; the patient terminal and the medical care terminal are both connected with the cloud module, and probability threshold values, wavelength information threshold values and liquid color information are preset in the patient terminal and the medical care terminal to judge whether bleeding occurs or not. According to the drainage tube bleeding early warning device, direct judgment and machine learning model prediction are combined through flow data, wavelength information and liquid color information of liquid in a drainage tube, early warning can be given to patients and medical staff in time, and the drainage tube bleeding early warning device belongs to the technical field of medical equipment.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to a bleeding warning device for drainage tubes. Background Technology

[0002] Patients who have undergone kidney, ureter, or bladder surgery often require indwelling drainage tubes to drain fluid. During home care after discharge, the characteristics of the drainage fluid (especially bleeding) are a key indicator for assessing recovery. Currently, monitoring is mainly done through subjective observation and regular follow-up visits, which carries risks of delayed feedback, strong subjectivity, and untimely warnings, potentially leading to worsening complications and increased readmission rates. Summary of the Invention

[0003] To address the above shortcomings, this invention provides a drainage tube bleeding early warning device. By directly judging the flow rate, wavelength information, and color information of the fluid in the drainage tube and combining this with machine learning model prediction, it can promptly issue early warnings to patients and medical staff.

[0004] The specific technical solution is as follows: A drainage tube bleeding early warning device, comprising: The drainage tube body is placed in the patient's kidney, ureter, and bladder. The outlet end of the drainage tube body extends out of the patient's urethra, and the part of the drainage tube body extending out of the patient's urethra is transparent. The monitoring module includes a flow meter, an optical sensing module, and a color sensor. The flow meter is connected to the transparent end, and the optical sensing module and color sensor are both clamped to the transparent end via clips, or the optical sensing module and color sensor are connected to the transparent end via a transparent tee tube. The optical sensing module and color sensor are set to monitor the frequency and collect the flow rate data, wavelength information, and liquid color information of the liquid in the drainage tube. The transmission module connects to the flow meter, the optical sensing module, and the color sensor. The microprocessor is connected to the transmission module and is used to receive and process the flow data, wavelength information and liquid color information transmitted by the transmission module, and predict the probability of bleeding through a preset machine learning model. The cloud module connects to the microprocessor and is used to receive data processed by the microprocessor. The terminal module includes a patient terminal and a medical care terminal, both of which are connected to the cloud module. Both the patient terminal and the medical care terminal are preset with probability thresholds, wavelength information thresholds, and fluid color information to determine whether bleeding has occurred.

[0005] Preferably, the drainage tube body includes a left renal duct segment, a right renal duct segment, and a urethral segment, with the left and right renal duct segments connected to the urethral segment via a three-way tube; The drainage tube body may include a left renal duct segment and a urethral segment, which are connected by a tube connector; The drainage tube body may include a right renal duct segment and a urethral segment, which are connected by a tube connector; The urethral segment ends in a transparent part, extending out of the patient's urethra. The left and right renal duct segments remain in the patient's left and right ureters, respectively. The connection between the left and right renal duct segments and the urethral segment is in the bladder, as is the connection between the right and right renal duct segments and the urethral segment.

[0006] Preferably, it also includes a power supply, with the microprocessor, flow meter, optical sensing module and color sensor all connected to the power supply.

[0007] Preferably, the microprocessor processes the flow data, wavelength information, and liquid color information through a prediction model. The prediction model includes a data collection module, a data processing module, and a machine learning model. The data collection module is used to continuously collect the patient's test data, which is distributed in a time series. The data processing module is used to clean, align, and normalize the data. The machine learning model uses the processed data to make predictions and determines whether bleeding will occur based on a preset threshold.

[0008] Preferably, the patient's test data includes bleeding risk index, real-time blood flow, and time series of cumulative blood loss.

[0009] Preferably, the specific method for cleaning the data is to process all outliers and missing values ​​detected by the sensors, wherein missing values ​​are processed by interpolation.

[0010] Preferably, the specific method for aligning data is to align data of different frequencies on the timeline to form a unified time segment.

[0011] Preferably, the specific method for normalizing the data is to scale the data to the same scale.

[0012] Preferably, before the machine learning model makes predictions on the data, the prepared labeled dataset is divided into a training set, a validation set, and a test set. The machine learning model is trained using the training set data, allowing it to adjust its internal parameters and learn the mapping relationship from input features to output results. The hyperparameters are adjusted using the validation set to prevent overfitting. Finally, the generalization ability of the model is evaluated using the test set. The trained machine learning model is then used to continuously make predictions on the detected data.

[0013] Preferably, the patient terminal and the medical staff terminal compare the predicted value output by the machine learning model with P, and determine whether to trigger an alert based on the interval of P in which the predicted value output by the machine learning model falls.

[0014] Compared with the prior art, the beneficial effects of the present invention are: to realize real-time, objective and remote monitoring of the drainage fluid status of patients after discharge, to automatically identify bleeding risks through data modeling, and to issue timely warnings to patients and medical staff, thereby achieving early intervention and improving patient safety and medical quality. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0016] Figure 1 This is a schematic diagram of a drainage tube bleeding early warning device; Figure 2 This is a schematic diagram of the drainage tube body.

[0017] 1 is the main body of the drainage tube, 11 is the left renal duct segment, 12 is the right renal duct segment, 13 is the three-way stopcock, 14 is the urethral segment, and 15 is the clamp. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of this invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0020] In the description of this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. Where the terms "first," "second," and "third" are used for descriptive purposes and to distinguish technical features, they should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features.

[0021] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. Furthermore, the technical features involved in the different embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0022] like Figures 1-2 As shown, this embodiment provides a drainage tube bleeding early warning device, comprising: The drainage tube body 1 is placed in the patient's kidney, ureter, and bladder. The outlet end of the drainage tube body 1 extends out from the patient's urethra, and the part of the drainage tube body 1 that extends out from the patient's urethra is a transparent end. The monitoring module includes a flow meter, an optical sensing module, and a color sensor. The flow meter is connected to the transparent end. Both the optical sensing module and the color sensor are clamped to the transparent end via clip 15, or connected to the transparent end via a transparent three-way tube 13. The optical sensing module and color sensor are set to monitor frequencies to collect flow rate data, wavelength information, and liquid color information of the fluid in the drainage tube. The optical sensing module uses a multi-wavelength (e.g., near-infrared, visible light) photoelectric sensor attached to the transparent end of the drainage tube near the outlet. By analyzing the absorption and scattering spectra of light passing through the drainage fluid, the fluid components can be identified non-invasively and in real time, with a focus on monitoring changes in hemoglobin concentration to quantify the amount of bleeding. The flow meter accurately measures the drainage volume per unit time. The color sensor, as an auxiliary to the optical sensing, makes a preliminary judgment on the color of the drainage fluid (from pale yellow to pink to bright red). The flow meter, optical sensing module, and color sensor automatically collect optical data, flow rate data, and color data of the drainage fluid at a set frequency (e.g., once every 5 minutes).

[0023] The transmission module connects to the flow meter, optical sensor module, and color sensor. For short-range connections, Bluetooth 5.0 BLE allows for communication between the device and patient / medical terminals via an app, balancing low power consumption and stability. For longer distance connections, the patient / medical terminals synchronize data with the cloud module.

[0024] The microprocessor, connected to the transmission module, receives and processes the flow data, wavelength information, and liquid color information transmitted by the transmission module, and predicts the bleeding probability through a preset machine learning model. The microprocessor (MCU) is responsible for controlling the sensor, performing preliminary data calculations and packaging. Specifically, the MCU performs preliminary processing and fusion of the raw data to calculate the "bleeding risk index" and "real-time drainage volume".

[0025] The cloud module connects to the microprocessor to receive data processed by the microprocessor. It stores basic patient information, surgical records, and medical history, and receives and stores sensor data (bleeding index, blood flow) uploaded from a large number of patient apps. It also stores patient-uploaded symptom records and examination report images. Using OCR and NLP technologies, key indicators such as "urine occult blood +++", "red blood cell count", and "white blood cell count" are automatically extracted from the report images and converted into structured data. This data is then correlated with real-time sensor data, patient subjective symptoms, and extracted examination results to form a complete patient health timeline.

[0026] The terminal module includes a patient terminal and a medical staff terminal, both connected to a cloud module. Both terminals are preset with probability thresholds, wavelength information thresholds, and fluid color information to determine the presence of bleeding. The patient and medical staff terminals are smartphones. They receive data from the cloud module and visually display drainage volume trends and changes in the bleeding risk index in chart form. Patients can manually enter subjective symptoms, such as pain level (VAS score), body temperature, activity level, and urine color. Patients can also upload photos of follow-up laboratory reports (such as urinalysis, blood routine tests), and imaging reports. The patient terminal receives alerts from the cloud module and alerts the patient via sound and vibration; the medical staff terminal also receives these alerts simultaneously, facilitating direct contact between the attending physician or nurse and the patient. After synchronizing data from the cloud module, the patient and medical staff terminals determine the presence of bleeding based on preset safety thresholds, such as "bleeding risk index readings > X threshold for 3 consecutive times" or "drainage volume > Y ml and blood color concentration > Z within 1 hour"; this is done using statistical process control. For example, "The bleeding risk index shows a continuous and significant upward trend within 6 hours." Finally, a pre-set machine learning model is used to predict the probability of bleeding. The machine learning model uses historical patient data (including data of patients who were eventually diagnosed with bleeding complications) for supervised learning. The fused multi-source time series data is input into the machine learning model, which predicts the probability of a clinically intervention-required bleeding event occurring within a future period (e.g., 24 hours).

[0027] The terminal module can set different levels of early warning mechanisms within different threshold ranges. Level 1 warning (low risk): only reminds the patient and suggests strengthening observation. Level 2 warning (medium risk): reminds both the patient and the nurse. Level 3 warning (high risk): reminds the patient, family members, follow-up nurses and the attending physician at the hospital. The notification can be sent through multiple channels: through App push, SMS, automatic outbound telephone calls, etc., to ensure that the warning is received in a timely manner.

[0028] It should be noted that the optical sensing module is the core sensor of the device, responsible for directly detecting the composition of the drainage fluid, especially the concentration of hemoglobin. It emits light of a specific wavelength (such as near-infrared or red light) through the liquid in the drainage tube and receives the transmitted or reflected light signals. By analyzing the degree of absorption of different wavelengths of light, the concentration of blood can be quantitatively analyzed, enabling early and objective bleeding warnings. The optical sensing module is directly connected to the power supply to obtain the required operating voltage and connects to the microprocessor via a digital bus such as I2C or SPI. The MCU controls its emission and sampling timing and reads the raw spectral data it generates.

[0029] The optical sensing module can be connected to the drainage tube body 1 by integrating the optical sensing module, color sensor, etc., into a lightweight, hinged clip. One side of the clip has a light source (such as an LED), and the other side has a light signal receiver (such as a photodiode). In use, simply clamp the transparent section of the drainage tube between these clips 15, ensuring good contact between the tube wall and the sensor without air bubbles.

[0030] The tee tube 13 is a short, transparent, standard-interface rigid tee tube segment. Both ends of this segment connect to the drainage system, with one side port sealed and internally integrating all optical sensors. This segment becomes a fixed part of the drainage system. This approach provides more stable measurements but requires modifications to the existing piping.

[0031] The optical sensing module and color sensor are integrated side-by-side inside the main structure (clamp 15 or tube segment), with precise optical path alignment to ensure that light can pass perpendicularly through the drainage tube and liquid. They are solidified and encapsulated in a waterproof, biocompatible material, contacting only the outer wall of the drainage tube and not the bodily fluids.

[0032] The color sensor serves as an auxiliary and verification tool for the optical sensing module. It uses an RGB filter to perceive the overall color of the liquid, thus determining whether the drainage fluid changes from pale yellow and pink to bright red. It is relatively inexpensive and primarily used for rapid color determination and cross-validation with data from the optical sensing module, improving system reliability. The color sensor is connected to a power source and to a microprocessor via an I2C bus; the MCU reads the RGB color values ​​it outputs.

[0033] Flow meters accurately measure the volume (flow rate) of drainage fluid flowing per unit time and the cumulative drainage volume. Bleeding risk is related not only to concentration but also to the total volume. Combining hemoglobin concentration and flow rate data allows for the calculation of blood loss per unit time, a more accurate risk assessment indicator. Flow meters are connected to a power supply, and their output may be an analog signal (voltage) or a digital pulse (one pulse per unit volume flowed). Therefore, it connects to the analog input pin or digital interrupt pin of the MCU, which performs counting and calculations. Flow meters typically employ clamp-on ultrasonic flow meters or miniature turbine flow meters. Clamp-on ultrasonic flow meters have two ultrasonic transducers mounted on the main structure, located on either side of the pipe, calculating the flow rate by measuring the time difference of ultrasonic wave propagation in the forward and reverse flows. This method is also non-invasive. Miniature turbine flow meters, if integrated into a pipe section, can be used as a separate miniature module connected in series in the drainage line, with the impeller driven by the liquid.

[0034] The microprocessor coordinates the timing of all sensor operations, such as commanding optical sensors to emit light and reading data from color sensors. It performs preliminary processing and data fusion on the raw data received from the sensors. For example, it converts optical data into a "bleeding risk index," calculates the blood loss rate by combining it with flow rate, and packages the processed key data (such as risk index, flow rate, timestamp, and device ID) into a data packet. This packet is then uploaded to a cloud module or sent directly to a terminal module, which is powered by a power supply. It connects to all sensors through different bus interfaces (I2C, SPI, ADC, GPIO) to receive their data and connects to a Bluetooth module (for direct data transmission at close range) via a UART serial port to send commands and data packets.

[0035] The microprocessor, transmission module, and power supply—these core electronic components—are integrated into a main control box, which is connected to the aforementioned "sensor clip 15" or "sensor segment" via a flexible ribbon cable. The main control box can be attached to the patient's abdominal skin or secured with a bandage to reduce the load on the drainage tube outlet. The battery, typically a rechargeable button cell battery or a small polymer lithium battery, is located inside the main control box.

[0036] The drainage tube body 1 includes a left renal duct segment 11, a right renal duct segment 12, and a urethral segment 14. The left renal duct segment 11 and the right renal duct segment 12 are connected to the urethral segment 14 through a three-way tube. The drainage tube body 1 may include a left renal duct segment 11 and a urethral segment 14, which are connected by a tube connector; The drainage tube body 1 may include a right renal duct segment 12 and a urethral segment 14, which are connected by a tube connector; The distal end of urethral segment 14 is transparent and extends out of the patient's urethra. Left renal duct segment 11 and right renal duct segment 12 are placed in the patient's left and right ureters, respectively. The connection between left renal duct segment 11 and urethral segment 14 is in the bladder, and the connection between right renal duct segment 12 and urethral segment 14 is in the bladder.

[0037] The device also includes a power supply, to which the microprocessor, flow meter, optical sensing module, and color sensor are all connected. The power supply is a rechargeable lithium battery, providing at least a week of battery life. It supplies power to all components requiring electricity, including the microprocessor, transmission module, optical sensing module, color sensor, and flow meter, via power traces on the circuit board.

[0038] The microprocessor processes flow data, wavelength information, and liquid color information through a prediction model, which includes a data collection module, a data processing module, and a machine learning model. The data collection module continuously collects patient test data, which is distributed in a time series. The data processing module cleans, aligns, and normalizes the data. The machine learning model uses the processed data to make predictions and determines whether bleeding will occur based on a preset threshold.

[0039] The patient data collected includes bleeding risk index, real-time blood flow, and time series of cumulative blood loss.

[0040] The specific method for cleaning the data is to process all outliers and missing values ​​detected by the sensors, with missing values ​​being processed using interpolation.

[0041] The specific method for aligning data is to align data of different frequencies on the timeline to form a unified time segment.

[0042] The specific method for normalizing data is to scale the data to the same scale.

[0043] Before making predictions on data, the machine learning model divides the prepared labeled dataset into a training set, a validation set, and a test set. The machine learning model is trained using the training set data, allowing it to adjust its internal parameters and learn the mapping relationship from input features to output results. The hyperparameters are adjusted using the validation set to prevent overfitting. Finally, the generalization ability of the model is evaluated using the test set. The trained machine learning model is then used to continuously make predictions on the detected data.

[0044] The patient terminal and the medical staff terminal compare the predicted value output by the machine learning model with the value of P, and determine whether to trigger an alert based on the interval of the predicted value output by the machine learning model in P.

[0045] It should be noted that the machine learning model includes a multimodal input encoding layer, a feature fusion layer, a pattern recognition layer, and an output layer, including a sensor data encoder, a patient report encoder, and a clinical event encoder. High-frequency acquired sensor time-series data is input into the sensor data encoder, which uses a bidirectional LSTM network. The sensor data encoder analyzes the sensor data stream from two directions (from front to back and from back to front), capturing the long-term temporal dependencies and contextual correlations of bleeding risk indicators. The sensor data encoder outputs a high-level feature representation for each time point, containing comprehensive risk information for that moment and the time periods before and after it.

[0046] By inputting low-frequency, irregular patient-reported data (pain scores, body temperature, etc.) into the patient-report encoder, the patient-report encoder uses a bidirectional GRU network (gated cyclic unit) to extract the temporal variation patterns of the patient's subjective symptoms and align them with the sensor data on the time axis.

[0047] Sparse discrete event labels (such as "upload lab report", "contact doctor" etc.) are input into the clinical event encoder. The clinical event encoder uses a multilayer perceptron to transform discrete events into dense feature vectors, representing the degree of impact of these events on the overall risk.

[0048] The feature fusion layer calculates the "importance score" of the feature vector at each time point, assigns higher weights to important time points and lower weights to less important time points, forming a weighted temporal feature representation.

[0049] The feature fusion layer concatenates the outputs of the three encoders along the feature dimension, and then performs deep interaction through a multi-layer neural network to establish non-linear correlations between different data sources, thereby identifying complex risk patterns that cannot be discovered by a single data type.

[0050] The pattern recognition layer includes a convolutional trend extractor and a multi-scale statistical feature calculator. The convolutional trend extractor employs a one-dimensional convolutional neural network, which scans temporal features using "time windows" of varying widths (e.g., 3 hours, 6 hours, 12 hours). The multi-scale statistical feature calculator calculates the mean, standard deviation, first / second difference, and cumulative effect (total blood loss) within the moving window. The pattern recognition layer transforms raw fluctuations into interpretable clinical indicators.

[0051] The output layer includes a main prediction network, an auxiliary prediction network, and an uncertainty estimation module. By inputting the fused high-level features and trend features into the main prediction network, which adopts a multi-layer fully connected network and uses the Sigmoid activation function at the end, the main prediction network outputs a bleeding probability value between 0 and 1, representing the likelihood of a bleeding event requiring intervention occurring within the next 24 hours.

[0052] The auxiliary prediction network is used to output risk levels in parallel, making it easier for users and medical staff to understand the risk level.

[0053] The uncertainty estimation module outputs a confidence score for each prediction, telling doctors "how reliable this prediction is," thus avoiding over-reliance on prediction results when the model is uncertain.

[0054] It should be noted that the predicted probability P output by the machine learning model does not directly trigger an alarm. Instead, it is combined with the aforementioned preset threshold range. For example, P < 0.3 indicates low risk, 0.3 ≤ P < 0.7 indicates medium risk, and P ≥ 0.7 indicates high risk. If the machine learning predicted probability P ≥ 0.7 AND (optical sensor bleeding index > YOR for 2 consecutive hours) or the patient reports a sudden increase in pain, then a "high-risk predictive warning" is triggered. This method of warning can greatly improve the accuracy and clinical reliability of the warning.

[0055] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A bleeding warning device for a drainage tube, characterized in that, include: The drainage tube body is placed in the patient's kidney, ureter, and bladder. The outlet end of the drainage tube body extends out of the patient's urethra, and the part of the drainage tube body extending out of the patient's urethra is transparent. The monitoring module includes a flow meter, an optical sensing module, and a color sensor. The flow meter is connected to the transparent end, and the optical sensing module and color sensor are both clamped to the transparent end via clips, or the optical sensing module and color sensor are connected to the transparent end via a transparent tee tube. The optical sensing module and color sensor are set to monitor the frequency and collect the flow rate data, wavelength information, and liquid color information of the liquid in the drainage tube. The transmission module connects to the flow meter, the optical sensing module, and the color sensor. The microprocessor is connected to the transmission module and is used to receive and process the flow data, wavelength information and liquid color information transmitted by the transmission module, and predict the probability of bleeding through a preset machine learning model. The cloud module connects to the microprocessor and is used to receive data processed by the microprocessor. The terminal module includes a patient terminal and a medical care terminal, both of which are connected to the cloud module. Both the patient terminal and the medical care terminal are preset with probability thresholds, wavelength information thresholds, and fluid color information to determine whether bleeding has occurred.

2. The drainage tube bleeding early warning device according to claim 1, characterized in that, The drainage tube body includes a left renal duct segment, a right renal duct segment, and a urethral segment. The left and right renal duct segments are connected to the urethral segment via a three-way tube. The drainage tube body may include a left renal duct segment and a urethral segment, which are connected by a tube connector; The drainage tube body may include a right renal duct segment and a urethral segment, which are connected by a tube connector; The urethral segment ends in a transparent part, extending out of the patient's urethra. The left and right renal duct segments remain in the patient's left and right ureters, respectively. The connection between the left and right renal duct segments and the urethral segment is in the bladder, as is the connection between the right and right renal duct segments and the urethral segment.

3. The drainage tube bleeding early warning device according to claim 1, characterized in that, It also includes a power supply, and the microprocessor, flow meter, optical sensing module and color sensor are all connected to the power supply.

4. The drainage tube bleeding early warning device according to claim 1, characterized in that, The microprocessor processes flow data, wavelength information, and liquid color information through a prediction model. The prediction model includes a data collection module, a data processing module, and a machine learning model. The data collection module is used to continuously collect patient test data, which is distributed in time series. The data processing module is used to clean, align, and normalize the data. The machine learning model uses the processed data to make predictions and determines whether bleeding will occur based on a preset threshold.

5. A drainage tube bleeding early warning device according to claim 4, characterized in that, The patient data collected includes bleeding risk index, real-time blood flow, and time series of cumulative blood loss.

6. The drainage tube bleeding early warning device according to claim 4, characterized in that, The specific method for cleaning the data is to process all outliers and missing values ​​detected by the sensors, with missing values ​​being processed using interpolation.

7. A drainage tube bleeding early warning device according to claim 4, characterized in that, The specific method for aligning data is to align data of different frequencies on the timeline to form a unified time segment.

8. A drainage tube bleeding early warning device according to claim 4, characterized in that, The specific method for normalizing data is to scale the data to the same scale.

9. A drainage tube bleeding early warning device according to claim 1, characterized in that, Before making predictions on data, the machine learning model divides the prepared labeled dataset into a training set, a validation set, and a test set. The machine learning model is trained using the training set data, allowing it to adjust its internal parameters and learn the mapping relationship from input features to output results. The hyperparameters are adjusted using the validation set to prevent overfitting. Finally, the generalization ability of the model is evaluated using the test set. The trained machine learning model is then used to continuously make predictions on the detected data.

10. A drainage tube bleeding early warning device according to claim 9, characterized in that, The patient terminal and the medical staff terminal compare the predicted value output by the machine learning model with the value of P, and determine whether to trigger an alert based on the interval of the predicted value output by the machine learning model in P.

Images