Transfusion exosmosis optical monitoring and early warning system based on machine learning
Through multi-channel optical reflectivity detection and machine learning, a tissue state index tensor is constructed, which solves the problem of unstable infusion extravasation monitoring in existing technologies, realizes adaptive discrimination and detailed early warning of extravasation risks, and improves infusion safety.
Patent Information
- Application Number
- CN202510793108.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-13
- Publication Date
- 2025-09-16
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the existing technology, the infusion extravasation monitoring method mainly relies on single-channel optical signals and lacks multi-dimensional feature extraction, resulting in instability in different bands and individual conditions. There is a lack of a unified and generalizable analysis model, making it difficult to achieve early identification and graded warning.
Multi-channel optical reflectivity detection combined with machine learning was used to construct a multi-channel optical reflectivity time series. The tissue state index tensor was constructed through the reflectivity derivative and band difference function. Dynamic feature stability analysis and response rate measurement were performed, an adaptive label set was generated, and an extravasation risk level prediction model was trained. Multi-factor fusion early warning was also performed.
It significantly improves the ability to identify extravasation risks, can adapt to different individual and tissue conditions, achieves detailed and accurate comprehensive graded warnings, and improves the ability to identify extravasation events early and respond in real time.
Smart Images
Figure CN120656754A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical devices, and in particular to an optical monitoring and early warning system for infusion extravasation based on machine learning. Background Art
[0002] During clinical intravenous infusion, extravasation is a common but serious complication, especially in the use of irritating drugs, hypertonic fluids or in special populations such as the elderly and infants. If extravasation is not identified and intervened in time, it may lead to adverse consequences such as tissue necrosis, infection, and impaired limb function. Therefore, the construction of an extravasation monitoring system that can identify in real time, provide intelligent warnings, and have the ability to respond and make decisions has become an important research direction in the current field of intelligent infusion safety assurance. The solution proposed in the present invention is an infusion extravasation monitoring and warning system that integrates multi-channel optical reflectivity detection technology and a classification prediction method based on machine learning. The various functional modules and data processing processes of the system are designed around actual tissue changes and classification response modeling, which can achieve early identification and graded warning of extravasation events. In the existing technology, single-channel optical signals are mostly used to judge infusion extravasation. The method is relatively crude and mainly relies on simple thresholds or trend fitting for judgment. It fails to extract tissue change characteristics from multiple dimensions and shows obvious instability in different bands and different individual conditions. There is a lack of a unified and generalizable analysis model. Summary of the Invention
[0003] In response to the shortcomings of the existing technology, the present invention provides an optical monitoring and early warning system for infusion extravasation based on machine learning to solve the problems raised in the above background technology.
[0004] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides an optical monitoring and early warning system for infusion extravasation based on machine learning, comprising the following steps: S1. Construct a multi-channel optical monitoring model with a high response rate matching algorithm to obtain a multi-channel optical reflectivity time series. S2. Use multi-channel optical reflectance time series to construct reflectance derivative and band difference function to obtain tissue state index tensor; S3. Use the tissue state index tensor to perform dynamic feature stability analysis and response rate measurement to obtain an adaptive label set; S4. Using tissue status indicators and adaptive labels to train a multi-class classification model to obtain an extravasation risk level prediction model; S5. Use the extravasation risk level prediction model to classify and map the real-time indicator data to obtain a visual warning level signal; S6. Use the classification output and warning level signal to perform multi-factor fusion to obtain a comprehensive graded warning result.
[0005] To further optimize this technical solution, step S1 first collects the original signal, then performs optical signal filtering and denoising, then performs inter-channel alignment and calibration, and finally performs multi-channel high-response matching modeling. The final multi-channel optical reflectivity time series expression is: ; in, :Indicates at time A multi-channel optical reflectivity vector sequence; :Indicates the wavelength is The corresponding Channels at time reflectivity; : The number of selected main channels.
[0006] To further optimize the technical solution, step S2 includes a first-order reflectivity derivative function, whose function formula is: ; in, :Indicates the wavelength is The channel at time The time derivative of the reflectivity at the moment, that is, the rate at which the reflectivity changes with time; :Indicates the wavelength is The channel at time Optical reflectance value at the moment; : Time variable, the domain is a continuous time period , corresponding to the continuous monitoring period of the optical sensor; : No. The central wavelength corresponding to each channel satisfies .
[0007] To further optimize this technical solution, step S2 further includes a dual-band optical difference function, the function formula of which is: ; in, :Indicates wavelength pair In time The value of the joint spectral difference enhancement function at the moment; : Indicates the reflectivity difference between the two channels, which is used to measure the instantaneous difference between the bands; : represents the time derivative of the sum of the reflectivities of the two channels, which is used to enhance the dynamic response caused by band coupling; :The two selected wavelength channels satisfy , is the total number of channels.
[0008] To further optimize this technical solution, step S2 constructs a tissue state index tensor based on the first-order reflectivity derivative function and the dual-band optical difference function. The corresponding formula model is: ; in, : Organizational state indicator tensor, representing the time The multi-dimensional feature set at each moment is composed of the time derivative features of all channels and the joint difference features between any channel pairs; :All satisfied channel combinations to ensure that symmetric channel pairs are not counted repeatedly.
[0009] To further optimize this technical solution, the above formula in step S2 includes the following process during the execution of the calculation: Obtain multi-channel reflectivity time series ; For each channel Calculate the first derivative ; Calculate the band difference function for each pair of channels ; Constructing organizational status indicator tensor .
[0010] To further optimize this technical solution, the step S3 first performs indicator sequence window division and local statistics extraction, and extracts the sliding window from the continuous time indicator tensor , construct the mean and variation rate of each indicator in the window, tensor Middle The indicators are , for the time window Conduct point statistics; The mean value within the window is: ; The variance within the window is: ; in, : The time span of the sliding window for analysis, the unit is consistent; : integral variable, representing the continuous time points within the window; :The first in the organizational status indicator set indicator numbers, total indicators.
[0011] To further optimize this technical solution, step S3 then constructs a dynamic response rate function, whose formula model is: ; in, Is a very small positive number, used to avoid the denominator being zero, here set .
[0012] To further optimize this technical solution, the step S3 finally performs label generation rule formulation and cluster discrimination, and calculates the feature set using the above formula After that, the state of each time window is classified by the label discriminant function, and its function formula is: ; in, : The judgment threshold of the variance within the window, used to identify whether the volatility is significant; : The lower threshold of the response rate, identifying weak but non-risk state changes; : Upper threshold of response rate, identifying rapid changes that may lead to tissue leakage.
[0013] To further optimize this technical solution, the optical monitoring and early warning system for infusion extravasation based on machine learning includes: a multi-channel optical monitoring and reflectivity modeling module, a tissue state feature construction module, a label adaptive generation module, a multi-category risk level model training module, a real-time risk classification and early warning mapping module and a multi-factor fusion grading output module.
[0014] In a second aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of an optical monitoring and early warning system for infusion extravasation based on machine learning as described in the first aspect of the present invention are implemented.
[0015] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of an optical monitoring and early warning system for infusion extravasation based on machine learning as described in the first aspect of the present invention are implemented.
[0016] Compared with the existing technology, the present invention provides an optical monitoring and early warning system for infusion extravasation based on machine learning, which has the following beneficial effects: The optical monitoring and early warning system for infusion extravasation based on machine learning sets a combined indicator model of "reflectivity derivative function" and "band difference function" constructed based on a multi-channel optical reflectivity time series. The present invention can dynamically characterize the tissue state from two dimensions: the rate of change of the optical signal and the difference in response between bands, and extract stable state characteristic indicators with classification and differentiation capabilities. On this basis, by constructing an adaptive label set and training a classification model, the model can adapt to different individuals and tissue conditions, and enhance the ability to distinguish the risk level of extravasation. This combined indicator model can not only significantly improve the feature expression of the training samples, but also quickly generate warning level signals through state indicator mapping in the real-time prediction stage, and further perform multi-factor fusion analysis with the classification output, thereby forming a more detailed and accurate comprehensive graded warning result. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0018] Figure 1 This is a flow chart of an optical monitoring and early warning system for infusion extravasation based on machine learning proposed by the present invention; Figure 2 This is a schematic diagram of the process of constructing the reflectivity derivative and band difference function of the optical monitoring and early warning system for infusion extravasation based on machine learning proposed by the present invention. DETAILED DESCRIPTION
[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0021] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.
[0022] Example 1: Reference Figures 1 and 2 , which is the first embodiment of the present invention, provides an optical monitoring and early warning system for infusion extravasation based on machine learning, comprising the following steps: S1. Construct a multi-channel optical monitoring model with a high response rate matching algorithm to obtain a multi-channel optical reflectivity time series. In step S1, raw signal acquisition is first performed. A multi-band optical sensor module is used to illuminate local tissue using LED excitation light with wavelengths of 730nm, 850nm, and 940nm. A photodiode array is used to receive reflected light signals. No less than five sets of full-band light intensity values are collected per second to form raw optical reflectance data. Then, the optical signal is filtered and denoised using the mature wavelet threshold denoising technology to suppress high-frequency noise caused by environmental interference, circuit noise and epidermal capillary movement. Then, channel alignment and calibration are performed. Cross-correlation technology is used to align the time series of channels in different bands to solve the problem of asynchronous response between channels. Amplitude normalization is performed using a preset reference tissue model to eliminate baseline offset caused by individual differences in skin absorption characteristics. Finally, multi-channel high-response matching modeling was performed, and a principal component matching strategy was introduced to screen out two to three channel combinations with the largest response amplitude changes and the strongest stability from the principal component signals of multiple channels. These combinations were used as the main channels for subsequent analysis to ensure that early tissue changes in extravasation could be quickly reflected in key channel signals.
[0023] The multi-channel optical reflectivity time series expression finally obtained in step S1 is: ; in, :Indicates at time A multi-channel optical reflectivity vector sequence; :Indicates the wavelength is The corresponding Channels at time reflectivity; : is the number of selected main channels; Each It is essentially a one-dimensional time series function sampled at equal time intervals, which has been processed by wavelet denoising and cross-correlation calibration.
[0024] S2. Use multi-channel optical reflectance time series to construct reflectance derivative and band difference function to obtain tissue state index tensor; Step S2 constructs a reflectivity derivative function and a band difference function based on the multi-channel optical reflectivity time series obtained in step S1, and further obtains a tissue state index set.
[0025] Step S2 includes a first-order reflectivity derivative function, whose function formula is: ; in, :Indicates the wavelength is The channel at time The time derivative of the reflectivity at the moment, that is, the rate at which the reflectivity changes with time; :Indicates the wavelength is The channel at time Optical reflectance value at the moment; : Time variable, the domain is a continuous time period , corresponding to the continuous monitoring period of the optical sensor; : No. The central wavelength corresponding to each channel satisfies ; This function represents the The reflectivity change rate per unit time is used to measure the sudden or gradual change of tissue state over time, especially with high responsiveness to the initial fluid accumulation of extravasation.
[0026] Step S2 also includes a dual-band optical difference function, the function formula of which is: ; in, :Indicates wavelength pair In time The value of the joint spectral difference enhancement function at the moment; : Indicates the reflectivity difference between the two channels, which is used to measure the instantaneous difference between the bands; : represents the time derivative of the sum of the reflectivities of the two channels, which is used to enhance the dynamic response caused by band coupling; :The two selected wavelength channels satisfy , is the total number of channels.
[0027] The formula introduces a derivative term of the sum of the dual-band optical reflectance and weights the band difference by its derivative amplitude, thereby emphasizing the joint response intensity caused by tissue fluid diffusion in different bands.
[0028] In step S2, a tissue state index tensor is constructed based on the first-order reflectivity derivative function and the dual-band optical difference function. The corresponding formula model is: ; in, : Organizational state indicator tensor, representing the time The multi-dimensional feature set at each moment is composed of the time derivative features of all channels and the joint difference features between any channel pairs; :All satisfied channel combinations to ensure that symmetric channel pairs are not counted repeatedly.
[0029] The above formula in step S2 includes the following process during the execution of the calculation: Obtain multi-channel reflectivity time series ; For each channel Calculate the first derivative ; Calculate the band difference function for each pair of channels ; Constructing organizational status indicator tensor .
[0030] Step S2 constructs the reflectivity derivative function Band difference variation function , jointly generate a set of organizational status indicators with temporal dynamics Compared with the static spectral difference analysis or single-channel feature-based monitoring methods commonly used in existing mature technologies, the difference is that step S2 introduces the coupled expression of cross-band dynamic change rates, so that tissue state perception not only has instantaneous reflection characteristics, but can also sensitively capture the direction and intensity of changes in the response rate of tissues under different channel combinations, thereby improving the dynamic discrimination ability of complex tissue state transition processes (such as early extravasation), and breaking through the limitations of traditional methods in multi-channel interactive expression and real-time response feature extraction.
[0031] S3. Use the tissue state index tensor to perform dynamic feature stability analysis and response rate measurement to obtain an adaptive label set; Step S3 is based on the tissue status index tensor obtained in step S2 , performing time-domain statistics and variability analysis on the derivative information and band difference functions contained therein, thereby measuring the dynamic stability and response rate characteristics of the indicator within a specific window. Through this analysis process, continuous time series indicators are converted into distinguishable behavioral pattern labels, enabling the dynamic generation of adaptive label sets.
[0032] Step S3 first performs indicator sequence window division and local statistics extraction, extracting the sliding window from the continuous time indicator tensor , construct the mean and variation rate of each indicator in the window, tensor Middle The indicators are , for the time window Conduct point statistics; The mean value within the window is: ; The variance within the window is: ; in, : The time span of the sliding window for analysis, the unit is consistent; : integral variable, representing the continuous time points within the window; : The kth indicator number in the organizational status indicator set, there are indicators; This integral statistic is used to quantify the stability and intensity of each state indicator within a local time segment.
[0033] Then the dynamic response rate function is constructed, and its formula model is: ; in, Is a very small positive number, used to avoid the denominator being zero, here set ; Finally, the label generation rules and clustering discrimination are formulated, and the feature set is calculated by the above formula. After that, the state of each time window is classified by the label discriminant function, and its function formula is: ; in, : The judgment threshold of the variance within the window, used to identify whether the volatility is significant; : The lower threshold of the response rate, identifying weak but non-risk state changes; : Upper threshold of response rate, identifying rapid changes that may lead to tissue leakage.
[0034] At the time of generation, the variance is calculated from the collected multi-channel optical monitoring data. and response rate Perform K-Means clustering to analyze data distribution, divide the data into three categories according to the cluster center, distinguish the low-volatility area from other areas, and obtain , distinguishing between medium-speed and fast responses, and obtaining and .
[0035] Tag Set This constitutes the final output of step S3: the adaptive label set.
[0036] Step S3 is different from existing methods that use static thresholds or expert experience to predefine labels (such as "extravasation" and "suspicious changes"). Instead, it is based on the dynamic behavioral characteristics of reflectance derivatives and band coupling indicators. Through window analysis, it extracts stability and response rate and generates label sets in real time. It can adapt to the dynamic evolution of tissue optical properties in different individuals and different band combinations, significantly improving the generalization and discrimination of labels.
[0037] S4. Using tissue status indicators and adaptive labels to train a multi-class classification model to obtain an extravasation risk level prediction model; Step S4 is to and Perform window sliding aggregation to generate sample pairs , to ensure that the same label corresponds to a state sequence, and use min-max standardization to normalize each indicator Perform numerical normalization to ensure consistent scales between model input dimensions; The sliding window mechanism is then used to expand the state indicator set by time period. Each sample consists of: an indicator vector of fixed length and a corresponding single label ; constitutes the standard training pair in supervised learning ; Step S4 uses random forest classifier to construct multiple decision trees, automatically learn the nonlinear relationship between state indicators and labels, and finally the trained model outputs the state indicators at any time point. The corresponding risk level prediction results.
[0038] S5. Use the extravasation risk level prediction model to classify and map the real-time indicator data to obtain a visual warning level signal; Step S5 first collects and preprocesses real-time status indicator data, using the same window strategy as step S4 to ensure consistency with the training model input structure; uses the same window strategy in step S4 to construct a fixed-length vector to ensure that the prediction model input scale matches; and uses the training period normalization parameters saved in step S4 to normalize each dimension of the indicator to ensure that the prediction model input scale matches.
[0039] The standardized indicator vector is then input into the model trained in step S4, and classification reasoning is performed in real time to determine the risk level.
[0040] Finally, establish clear risk level-signal color mapping rules: Low risk ( ) → green; Medium risk ( ) → yellow; High risk ( ) → red; The system presents the color signal after the current level mapping through a graphical interface, LED indicator light or other means; Through a mature state mapping mechanism, the classification results are converted into state signals with different visual perception intensities.
[0041] S6. Use the classification output and warning level signal to perform multi-factor fusion to obtain a comprehensive graded warning result; Step S6 integrates the warning basis of each separated dimension into a unique and interpretable comprehensive graded warning result based on the risk level classification result outputted in step S5 and the real-time visual warning signal, so as to guide the system to trigger a higher level intervention strategy or alarm mechanism.
[0042] Step S6 first constructs a multi-factor fusion vector, using the following variables as fusion inputs to construct a state vector: : Risk level prediction probability from the S5 classification model, corresponding to the level 、 、 ; : corresponds to the visual signal level, derived from S5 output (green = 1, yellow = 2, red = 3); : The duration of the consistency between the current risk level and the previous level; : auxiliary intervention trigger mark; Construct the fusion state vector: .
[0043] Then calculate the comprehensive risk score function and define a fusion function , the vector Mapped to a real-valued risk score , which is used to quantify the current overall extravasation risk level: ; in: It is the fusion weight parameter preset by the system; Emphasize the weighting of high risk level probabilities; Reflects the visual signal level; Introducing nonlinear penalties for duration; Grants direct risk bonuses to special triggering events.
[0044] Finally, a comprehensive graded warning result is determined based on the risk score. , construct segmentation and grading judgment rules: ; in, The threshold value of the comprehensive graded warning score is The distribution of is obtained by quantile division; : Comprehensive graded warning level label, including Represents no risk, Same as above, representing low to high risk; The final output result of step S6 is: , that is, the comprehensive graded warning results.
[0045] Traditional medical early warning systems are mostly based on single-factor trigger logic, such as: When an indicator exceeds the threshold, an early warning is triggered; Or directly use the classification model output as the basis for alarm without considering its uncertainty (such as probability); No fusion of visual cues or persistent behavioral parameters.
[0046] This step, by constructing a multi-factor fusion vector + risk scoring function + threshold classification mechanism, is significantly different from traditional methods and has the following differences: Integrate multiple factors (such as probability distribution, signal level, and duration) to improve risk identification accuracy; The scoring function mapping mechanism makes risk expression more continuous and adjustable; The fusion results support graded response, not single-point triggering, and are more in line with clinical graded intervention needs.
[0047] Example 2: Reference Figures 1 and 2 , which is the second embodiment of the present invention, provides an optical monitoring and early warning system for infusion extravasation based on machine learning, which includes: a multi-channel optical monitoring and reflectivity modeling module, a tissue state feature construction module, a label adaptive generation module, a multi-category risk level model training module, a real-time risk classification and early warning mapping module and a multi-factor fusion classification output module.
[0048] The multi-channel optical monitoring and reflectivity modeling module is based on the settings in step S1. It constructs a multi-channel reflectivity monitoring model through a high response rate matching algorithm, collects the reflectivity signal of the puncture site in real time, forms a standardized time series, and provides basic data input for subsequent analysis. The tissue state feature construction module is based on the settings in step S2, performs derivative analysis and band difference modeling on the reflectivity time series, and constructs a tissue state indicator set to characterize the state change characteristics of tissue pressure, leakage, etc. The label adaptive generation module is set up in step S3, dynamically identifies feature stability and response rate based on tissue status indicators, and generates an adaptive label set with time period and type labels as the target label source for model training; The multi-category risk level model training module is based on the setting of step S4, combines the indicator set outputted by step S2 and the adaptive label outputted by step S3, and uses non-weighted modeling methods such as random forest and multi-layer perceptron to complete the multi-category classification model training and establish an extravasation risk prediction model; The real-time risk classification and warning mapping module is set up based on step S5. During the model deployment phase, the trained risk classification model is used to predict the risk level of the indicator data collected in real time, and is converted into a visual warning level signal through multi-level threshold mapping; The multi-factor fusion grading output module is set up based on step S6, integrates the model classification results and warning signals, performs quantile level division through the multi-factor scoring model, and outputs the final comprehensive graded warning results for clinical operation recommendation output.
[0049] Example 3: This embodiment also provides a computer device, which is suitable for an optical monitoring and early warning system for infusion extravasation based on machine learning, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement an optical monitoring and early warning system for infusion extravasation based on machine learning as proposed in the above embodiment.
[0050] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, an optical monitoring and early warning system for infusion extravasation based on machine learning as proposed in the above embodiment is implemented.
[0051] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.
[0052] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.
[0053] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0054] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.
[0055] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.
[0056] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. An optical monitoring and early warning system for infusion extravasation based on machine learning, characterized in that: The following steps are involved: S1. Construct a multi-channel optical monitoring model with a high response rate matching algorithm to obtain a multi-channel optical reflectivity time series. S2. Use multi-channel optical reflectance time series to construct reflectance derivative and band difference function to obtain tissue state index tensor; S3. Use the tissue state index tensor to perform dynamic feature stability analysis and response rate measurement to obtain an adaptive label set; S4. Using tissue status indicators and adaptive labels to train a multi-class classification model to obtain an extravasation risk level prediction model; S5. Use the extravasation risk level prediction model to classify and map the real-time indicator data to obtain a visual warning level signal; S6. Use the classification output and warning level signal to perform multi-factor fusion to obtain a comprehensive graded warning result.
2. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 1 is characterized in that: In step S1, the original signal is first collected, and then the optical signal is filtered and denoised, followed by inter-channel alignment and calibration, and finally multi-channel high-response matching modeling. The final multi-channel optical reflectivity time series expression is: ； in, :Indicates at time A multi-channel optical reflectivity vector sequence; :Indicates the wavelength is The corresponding Channels at time reflectivity; : The number of selected main channels.
3. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 1 is characterized in that: The step S2 includes a first-order reflectivity derivative function, the function formula of which is: ； in, :Indicates the wavelength is The channel at time The time derivative of the reflectivity at the moment, that is, the rate at which the reflectivity changes with time; :Indicates the wavelength is The channel at time Optical reflectance value at the moment; : Time variable, the domain is a continuous time period , corresponding to the continuous monitoring period of the optical sensor; : No. The central wavelength corresponding to each channel satisfies .
4. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 3 is characterized in that: The step S2 also includes a dual-band optical difference function, the function formula of which is: ； in, :Indicates wavelength pair In time The value of the joint spectral difference enhancement function at the moment; : Indicates the reflectivity difference between the two channels, which is used to measure the instantaneous difference between the bands; : represents the time derivative of the sum of the reflectivities of the two channels, which is used to enhance the dynamic response caused by band coupling; :The two selected wavelength channels satisfy , is the total number of channels.
5. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 3 is characterized in that: In step S2, a tissue state index tensor is constructed based on the first-order reflectivity derivative function and the dual-band optical difference function. The corresponding formula model is: ； in, : Organizational state indicator tensor, representing the time The multi-dimensional feature set at each moment is composed of the time derivative features of all channels and the joint difference features between any channel pairs; :All satisfied channel combinations to ensure that symmetric channel pairs are not counted repeatedly.
6. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 1, characterized in that: The above formula in step S2 includes the following process during the execution of the calculation: Obtain multi-channel reflectivity time series ; For each channel Calculate the first derivative ; Calculate the band difference function for each pair of channels ; Constructing organizational status indicator tensor .
7. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 1 is characterized in that: The step S3 first performs indicator sequence window division and local statistics extraction, and extracts the sliding window from the continuous time indicator tensor. , construct the mean and variation rate of each indicator in the window, tensor Middle The indicators are , for the time window Conduct point statistics; The mean value within the window is: ； The variance within the window is: ； in, : The time span of the sliding window for analysis, the unit is consistent; : integral variable, representing the continuous time points within the window; :The first in the organizational status indicator set indicator numbers, total indicators.
8. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 7 is characterized in that: The step S3 then constructs a dynamic response rate function, the formula model of which is: ； in, Is a very small positive number, used to avoid the denominator being zero, here set .
9. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 7, characterized in that: The step S3 finally performs label generation rule formulation and cluster discrimination, and calculates the feature set by the above formula. After that, the state of each time window is classified by the label discriminant function, and its function formula is: ； in, : The judgment threshold of the variance within the window, used to identify whether the volatility is significant; : The lower threshold of the response rate, identifying weak but non-risk state changes; : Upper threshold of response rate, identifying rapid changes that may lead to tissue leakage.
10. The optical monitoring and early warning system for infusion extravasation based on machine learning according to claim 1, characterized in that: The functional modules of the system include multi-channel optical monitoring and reflectivity modeling module, tissue state feature construction module, label adaptive generation module, multi-category risk level model training module, real-time risk classification and warning mapping module and multi-factor fusion classification output module.