Artificial intelligence-based replantation finger blood vessel crisis monitoring and early warning method and system

Abstract

The application provides an artificial intelligence-based monitoring and early warning method and system for vascular crisis of replanted fingers, comprising: acquiring skin temperature monitoring data, blood flow velocity signals and pulse wave transmission time of the affected side and the healthy side of the replanted fingers; mapping the skin temperature monitoring data and the blood flow velocity signals to the same coordinate system, taking the R wave peak of the pulse wave signal as the time reference point, interpolating and resampling the skin temperature monitoring data and filtering the skin temperature monitoring data to generate continuous skin temperature monitoring data flow; to solve the problem that the time and space resolution of the skin temperature data obtained by the traditional monitoring method is low, the monitoring deviation of the skin temperature makes it difficult for medical staff to accurately capture the dynamic change information of the microcirculation, and further cannot timely and accurately judge whether the vascular crisis occurs, increases the risk of necrosis of the replanted fingers, and seriously affects the success rate of the replanted fingers.

Description

Technical Field

[0001] This invention relates to the field of monitoring and early warning technology, and in particular to a method and system for monitoring and early warning of vascular crises in replanted fingers based on artificial intelligence. Background Technology

[0002] Replanted finger refers to a finger that has been completely or partially severed due to trauma or surgery and has been surgically reattached to its original position to restore blood supply.

[0003] However, in the postoperative monitoring of replanted fingers, traditional monitoring methods yield skin temperature data with low spatiotemporal resolution. Temporally, the sampling intervals are too long, failing to capture rapid dynamic changes in skin temperature over time. Spatially, the monitoring points are sparsely distributed, making it difficult to comprehensively reflect skin temperature differences in different areas of the finger. Furthermore, the dynamic changes in the microcirculation of replanted fingers are extremely complex and rapid; even minute fluctuations in skin temperature can indicate vascular crisis. Therefore, the bias in skin temperature monitoring makes it difficult for medical staff to accurately capture these dynamic changes in microcirculation, thus hindering timely and accurate judgment of whether vascular crisis has occurred. This increases the risk of replanted finger necrosis and severely impacts the success rate of finger replantation.

[0004] Based on this, this application proposes an artificial intelligence-based method and system for monitoring and early warning of vascular crises in replanted fingers. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that the spatiotemporal resolution of skin temperature data obtained by traditional monitoring methods is low. The monitoring deviation of skin temperature makes it difficult for medical staff to accurately capture the dynamic changes in these microcirculations, thus making it impossible to timely and accurately determine whether vascular crisis has occurred, increasing the risk of necrosis of replanted fingers, and seriously affecting the success rate of finger replantation.

[0006] To achieve the above objectives, the present invention provides a method and system for monitoring and early warning of vascular crises in replanted fingers based on artificial intelligence.

[0007] According to a first aspect of the present invention, an artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers is provided, comprising: Acquire skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger; Skin temperature monitoring data and blood flow velocity signals are mapped to the same coordinate system, and the skin temperature monitoring data are interpolated, resampled and filtered using the R-wave peak of the pulse wave signal as the time reference point to generate a continuous skin temperature monitoring data stream. Based on continuous skin temperature monitoring data stream, periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points are identified to extract and form a skin temperature fluctuation feature set. Based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave conduction time, the local vascular resistance index is calculated. An individualized baseline model was constructed based on the skin temperature monitoring data and blood flow velocity signal of the healthy side, and the deviation between the skin temperature value and the blood flow velocity signal of the affected side was dynamically calculated based on the individualized baseline model. Skin color images of the affected finger are collected periodically, and the skin color images are compared with a predefined skin colorimetric card. The current skin color grade of the affected finger is determined based on the comparison results. The composite early warning index of the affected finger is composed of local vascular resistance index, deviation degree and current skin color grade. A dynamic early warning threshold range is generated according to the patient's monitoring stage. When the composite early warning index is not within the dynamic early warning threshold range, a vascular crisis of the replanted finger is determined and an early warning signal is output to the medical terminal.

[0008] Optionally, skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time of the replanted finger on both the affected and unaffected sides can be obtained, specifically including: Skin temperature monitors were used to monitor the affected and healthy sides of the replanted finger to obtain skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger.

[0009] Optionally, the skin temperature monitor includes a monitoring chip and a flexible skin temperature sensor array and a photoplethysmography (PPG) sensor, both electrically connected to the monitoring chip. The flexible skin temperature sensor array is used to monitor and acquire skin temperature data on the affected and healthy sides, and the PPG sensor is used to monitor and acquire blood flow velocity signals and pulse wave conduction time on the affected and healthy sides.

[0010] Optionally, skin temperature monitoring data and blood flow velocity signals are mapped to the same coordinate system, and the skin temperature monitoring data is interpolated, resampled, and filtered using the R-wave peak of the pulse wave signal as the time reference point to generate a continuous skin temperature monitoring data stream, specifically including: Record the physical coordinates of each node in the flexible skin temperature sensor array and the light emission and light receiving points of the photoplethysmography (PPG) sensor to generate a set of mapping points; Obtain the calibration point set of the three-dimensional calibration board, and establish the mapping relationship between the physical coordinates of each mapping point in the mapping point set and the pixel coordinates of each calibration point in the calibration point set through the three-dimensional calibration board and the least squares method. Generate a spatial transformation matrix to map skin temperature monitoring data and blood flow velocity signals to the same coordinate system based on the spatial transformation matrix. Acquire electrocardiogram (ECG) signals and extract their corresponding R-wave peaks. Use the Pan-Tompkins algorithm to extract the R-wave peaks corresponding to the blood flow velocity signals and compare them with the R-wave peaks corresponding to the ECG signals by time difference. Use a dynamic time warping algorithm to align the R-wave peaks of the ECG signals and the R-wave peaks of the blood flow velocity signals to form the R-wave peaks of the pulse wave signals and define them as time reference points. At the time reference point, the skin temperature monitoring data is subjected to cubic spline interpolation and filtering at a sampling rate of 100Hz to generate a continuous skin temperature monitoring data stream.

[0011] Optionally, a calibration point set of a 3D calibration board is obtained, and a mapping relationship between the physical coordinates of each mapping point in the mapping point set and the pixel coordinates of each calibration point in the calibration point set is established using the 3D calibration board and the least squares method. A spatial transformation matrix is ​​generated to map skin temperature monitoring data and blood flow velocity signals to the same coordinate system based on the spatial transformation matrix. Specifically, this includes: Based on the design spacing of the 3D calibration board, the pixel coordinates of each calibration point in the 3D calibration board are extracted by a corner detection algorithm to generate a calibration point set; Simulated affected and healthy finger bodies with the same shape as the affected and healthy finger bodies were constructed respectively. A three-dimensional calibration plate was attached to the surface of the simulated affected and healthy finger bodies. The mapping relationship between the physical coordinates of each mapping point and each calibration point was established by the least squares method, and a spatial transformation matrix was generated. Skin temperature monitoring data and blood flow velocity signals are mapped to the same coordinate system using a spatial transformation matrix.

[0012] Optionally, based on continuous skin temperature monitoring data stream, periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points are identified to extract and form a skin temperature fluctuation feature set. Then, based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave propagation time, the local vascular resistance index is calculated, specifically including: The power spectral density of the skin temperature monitoring data stream in each frequency band of the continuous skin temperature monitoring data stream is calculated based on the wavelet transform algorithm, and the frequency of the frequency band corresponding to the maximum power spectral density is selected as the dominant frequency, so as to identify the periodic fluctuation segment based on the dominant frequency. Piecewise linear regression was used to fit the continuous skin temperature monitoring data stream to identify the slope of frequency bands with an upward or downward trend, and the frequency bands whose slopes exceeded the clinical experience threshold were marked as trend fluctuation segments. By calculating the difference between each sampling point in the continuous skin temperature monitoring data stream and the median value of the previous 5 sampling points, the current sampling point is determined to be a sudden change point when the difference is greater than the preset change threshold. Extract the fluctuation features of periodic fluctuation segments, trend fluctuation segments, and abrupt fluctuation points to form a skin temperature fluctuation feature set; Based on skin temperature fluctuation feature set, blood flow velocity signal and pulse wave conduction time, the local vascular resistance index is calculated using a support vector regression model.

[0013] Optionally, an individualized baseline model can be constructed based on skin temperature monitoring data and blood flow velocity signals from the healthy side, and the deviation between the skin temperature value and blood flow velocity signal on the affected side can be dynamically calculated based on the individualized baseline model, specifically including: Using skin temperature monitoring data and blood flow velocity signals from the healthy side as input, and combining machine learning algorithms, an individualized baseline model that reflects the correlation between skin temperature monitoring data and blood flow velocity signals is constructed. The skin temperature and blood flow velocity signals of the affected side are input into the individualized baseline model to obtain the theoretical skin temperature and theoretical blood flow velocity signals corresponding to the affected side. The deviation between the affected side's skin temperature and the theoretical skin temperature, as well as the difference between the affected side's blood flow velocity signal and the theoretical blood flow velocity signal, is obtained by calculating the absolute value of these differences.

[0014] Optionally, skin color images of the affected finger are periodically collected, and the skin color images are compared against a predefined skin colorimetric chart. The current skin color grade of the affected finger is determined based on the comparison results, specifically including: Using a camera mounted on a skin temperature monitor, images of the skin color on the affected side of the fingers are periodically collected at preset time intervals. The skin color image of the affected side finger is geometrically corrected and grayscale normalized using the ROI localization algorithm in order to correct the skin color image of the affected side finger. The color gamut of the corrected affected finger skin image is matched with a predefined four-order skin color chart. The Euclidean distance between each pixel in the corrected affected finger skin image in LAB space and the center of each order of the four-order skin color chart is calculated, and the order with the smallest Euclidean distance is extracted and defined as the skin color level of the corresponding pixel. The skin color level of each pixel in the corrected skin color image of the affected side is statistically analyzed, and the level with the highest percentage is determined as the current skin color level of the affected side of the finger.

[0015] Optionally, a composite early warning index for the affected finger can be constructed using local vascular resistance index, deviation, and current skin color grade. A dynamic early warning threshold range can be generated based on the patient's monitoring stage. When the composite early warning index falls outside the dynamic early warning threshold range, a vascular crisis in the replanted finger is identified, and an early warning signal is output to the medical terminal. Specifically, this includes: A composite early warning indicator for the affected side of the finger is formed by combining local vascular resistance index, deviation, and current skin color grade. The postoperative time of the patient is obtained, and the patient monitoring stage is divided into the edema stage and the healing stage according to the postoperative time. The current monitoring stage of the patient is identified by the corresponding timestamp of the patient through the medical terminal. Acquire a dynamic threshold model library that includes the threshold monitoring range during the edema stage and the monitoring range during the healing stage, so as to extract the threshold parameter range of the corresponding stage from the dynamic threshold model library based on the current monitoring stage; The composite early warning index is compared with the threshold parameter range. When the composite early warning index is not within the dynamic early warning threshold range, a vascular crisis in the replanted finger body is determined and an early warning signal is output to the medical terminal.

[0016] According to a second aspect of the present invention, an artificial intelligence-based replanted finger vascular crisis monitoring and early warning system is also provided, for implementing the artificial intelligence-based replanted finger vascular crisis monitoring and early warning method of any of the above claims, the replanted finger vascular crisis monitoring and early warning system comprising: The data acquisition module is used to acquire skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger. The data stream generation module is used to map skin temperature monitoring data and blood flow velocity signals to the same coordinate system, and uses the R-wave peak of the pulse wave signal as the time reference point to interpolate, resample, and filter the skin temperature monitoring data to generate a continuous skin temperature monitoring data stream. The local vascular resistance index calculation module can identify periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points based on continuous skin temperature monitoring data streams, extract and form a skin temperature fluctuation feature set, and calculate the local vascular resistance index based on the blood flow velocity signal and pulse wave conduction time in the skin temperature fluctuation feature set. The deviation calculation module can construct an individualized baseline model based on the skin temperature monitoring data and blood flow velocity signal of the healthy side, and dynamically calculate the deviation between the skin temperature value and blood flow velocity of the affected side based on the individualized baseline model. The current skin color level determination module is used to periodically collect skin color images of the affected side of the finger, compare the skin color images with a predefined skin color matching card, and determine the current skin color level of the affected side of the finger based on the comparison results. The warning signal output module can construct a composite warning index for the affected finger by combining the local vascular resistance index, deviation, and current skin color level. It can also generate a dynamic warning threshold range based on the patient's monitoring stage. When the composite warning index is not within the dynamic warning threshold range, it can determine the vascular crisis of the replanted finger and output a warning signal to the medical terminal.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes an artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers. By mapping skin temperature monitoring data and blood flow velocity signals to the same coordinate system, and using the R-wave peak of the pulse wave signal as a time reference for interpolation, resampling, and filtering, a high spatiotemporal resolution continuous data stream is constructed, solving the spatiotemporal discreteness problem of traditional single-parameter monitoring. Furthermore, by identifying periodic, trend, and sudden fluctuation characteristics based on the continuous skin temperature monitoring data stream, and combining the blood flow velocity signal with the pulse wave propagation time to calculate the local vascular resistance index, quantitative analysis of hemodynamic abnormalities is achieved, breaking through the limitations of traditional methods that rely solely on... The limitations of the skin temperature threshold are addressed by constructing an individualized baseline model based on data from the healthy side and dynamically calculating the deviation of the affected side. This is combined with adaptive adjustment of the judgment threshold at the postoperative stage to solve the baseline drift problem caused by individual differences and postoperative physiological changes. Furthermore, by fusing local vascular resistance index, deviation, and skin color grade to generate a composite early warning index, and generating a dynamic early warning threshold range based on the patient monitoring stage, multi-parameter spatiotemporal alignment fusion and clinical context awareness are achieved to provide accurate early warning. This significantly improves the sensitivity of vascular crisis identification and reduces the false alarm rate, thereby reducing the risk of necrosis of replanted fingers and improving the success rate of finger replantation.

[0018] The artificial intelligence-based replanted finger vascular crisis monitoring and early warning system proposed in this invention belongs to the same general inventive concept as the artificial intelligence-based replanted finger vascular crisis monitoring and early warning method of this invention. It should at least have the same technical effects as the artificial intelligence-based replanted finger vascular crisis monitoring and early warning method of this invention. Therefore, this invention will not elaborate further here.

[0019] As can be seen from the above, the technical solution of the present invention can effectively solve the problem that the spatiotemporal resolution of skin temperature data obtained by traditional monitoring methods is low. The monitoring deviation of skin temperature makes it difficult for medical staff to accurately capture the dynamic changes of these microcirculations, thus making it impossible to timely and accurately judge whether vascular crisis has occurred, increasing the risk of necrosis of replanted fingers, and seriously affecting the success rate of finger replantation.

[0020] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0021] The present invention can be better understood by referring to the following description taken in conjunction with the accompanying drawings, in which the same or similar reference numerals are used throughout the drawings to denote the same or similar parts.

[0022] Figure 1 This is a flowchart illustrating an artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers, as an embodiment of the present invention. Figure 2 This is a schematic diagram of the principle of the artificial intelligence-based replanted finger vascular crisis monitoring and early warning system according to an embodiment of the present invention. Detailed Implementation To enable those skilled in the art to more fully understand the technical solutions of the present invention, exemplary embodiments of the present invention will be described more comprehensively and in detail below with reference to the accompanying drawings. Obviously, the one or more embodiments of the present invention described below are merely one or more specific ways to implement the technical solutions of the present invention, and are not exhaustive. It should be understood that other ways belonging to a general inventive concept can be used to implement the technical solutions of the present invention, and should not be limited to the embodiments described exemplary. Based on one or more embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] Reference Figure 1 The embodiments of the present invention provide an artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers, including: Step S1: Obtain skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time from the affected and healthy sides of the replanted finger. Step S2: Map the skin temperature monitoring data and blood flow velocity signal to the same coordinate system, and use the R wave peak of the pulse wave signal as the time reference point to interpolate, resample, and filter the skin temperature monitoring data to generate a continuous skin temperature monitoring data stream. Step S3: Identify periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points based on continuous skin temperature monitoring data stream, so as to extract and form a skin temperature fluctuation feature set, and calculate the local vascular resistance index based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave conduction time; Step S4: Construct an individualized baseline model based on the skin temperature monitoring data and blood flow velocity signal of the healthy side, and dynamically calculate the deviation between the skin temperature value and the blood flow velocity signal of the affected side based on the individualized baseline model; Step S5: Periodically collect skin color images of the affected side's fingers, compare the skin color images with a predefined skin colorimetric card, and determine the current skin color grade of the affected side's fingers based on the comparison results; Step S6: The local vascular resistance index, deviation, and current skin color grade are used to construct a composite early warning index for the affected finger. A dynamic early warning threshold range is generated based on the patient's monitoring stage. When the composite early warning index is not within the dynamic early warning threshold range, a vascular crisis in the replanted finger is determined and an early warning signal is output to the medical terminal.

[0024] In one embodiment, step S1 involves acquiring skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger, specifically including: Skin temperature monitors were used to monitor the affected and healthy sides of the replanted finger to obtain skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger.

[0025] In one specific embodiment, the skin temperature monitor includes a monitoring chip and a flexible skin temperature sensor array and a photoplethysmography (PPG) sensor, both electrically connected to the monitoring chip. The flexible skin temperature sensor array is used to monitor and acquire skin temperature monitoring data of the affected and healthy sides, and the PPG sensor is used to monitor and acquire blood flow velocity signals and pulse wave conduction time of the affected and healthy sides.

[0026] In one specific embodiment, the monitoring chip is electrically connected to the medical terminal.

[0027] Specifically, embodiments of the present invention utilize a flexible skin temperature sensor array that can conform to the curved surface of the replanted finger to monitor the temperature distribution in various regions of the affected and healthy sides of the finger in real time, and convert the temperature signal into an electrical signal to be output to the monitoring chip. Furthermore, a photoplethysmography (PPG) sensor is used to detect the difference in light absorption caused by changes in blood volume by emitting 660nm red light and 940nm infrared light through the skin, thereby generating a pulse wave signal that reflects the blood flow velocity, and simultaneously acquiring the pulse wave conduction time.

[0028] This breakthrough overcomes the limitations of traditional methods that rely solely on skin temperature thresholds. By using a monitoring chip, it achieves spatiotemporal alignment and real-time analysis of skin temperature and blood flow velocity. Combined with the dynamic early warning function of medical terminals, it forms a closed-loop system from data collection to clinical decision-making. This effectively solves the technical problems of low spatiotemporal resolution and inability to capture dynamic changes in microcirculation in traditional monitoring methods, and significantly reduces the risk of necrosis of replanted fingers.

[0029] Furthermore, the monitoring chip is a biosignal processing chip, which can realize real-time acquisition, filtering and multimodal fusion analysis of skin temperature and pulse wave signals through hardware acceleration. Its specific working principle is existing technology, so this invention will not elaborate further here.

[0030] It is worth noting that both the flexible skin temperature sensor array and the photoplethysmography (PPG) sensor can be mounted on the housing of the skin temperature monitor and made in contact with the affected and healthy sides by means of adhesive bonding.

[0031] In one embodiment, in step S2, skin temperature monitoring data and blood flow velocity signals are mapped to the same coordinate system, and the skin temperature monitoring data is interpolated, resampled, and filtered using the R-wave peak of the pulse wave signal as the time reference point to generate a continuous skin temperature monitoring data stream. Specifically, this includes: Step S21: Record the physical coordinates of each node in the flexible skin temperature sensor array and the light emission and light receiving points of the photoplethysmography (PPG) sensor to generate a mapping point set; Step S22: Obtain the calibration point set of the three-dimensional calibration board, and establish the mapping relationship between the physical coordinates of each mapping point in the mapping point set and the pixel coordinates of each calibration point in the calibration point set through the three-dimensional calibration board and the least squares method, and generate a spatial transformation matrix to map the skin temperature monitoring data and blood flow velocity signal to the same coordinate system based on the spatial transformation matrix; Step S23: Acquire electrocardiogram (ECG) signals and extract their corresponding R-wave peaks. Extract the R-wave peaks corresponding to the blood flow velocity signals using the Pan-Tompkins algorithm, and compare them with the R-wave peaks corresponding to the ECG signals by time difference. Align the R-wave peaks of the ECG signals and the R-wave peaks of the blood flow velocity signals using a dynamic time warping algorithm to form the R-wave peaks of the pulse wave signals and define them as time reference points. Step S24: At the time reference point, the skin temperature monitoring data is subjected to cubic spline interpolation and filtering at a sampling rate of 100Hz to generate a continuous skin temperature monitoring data stream.

[0032] Specifically, embodiments of the present invention utilize a three-dimensional calibration plate and a spatial transformation matrix constructed by the least squares method to unify the distributed data of the flexible skin temperature sensor array and the point data of the photoplethysmography pulse wave sensor into the same coordinate system, so as to eliminate spatial misalignment caused by differences in sensor deployment positions and ensure the spatial correspondence accuracy of skin temperature and blood flow velocity signals.

[0033] In the time dimension, the Pan-Tompkins algorithm and the dynamic time warping algorithm are combined to achieve time synchronization of the R-wave peaks of the electrocardiogram signal and the blood flow velocity signal. At the time reference point of the R-wave peak of the formed pulse wave signal, the skin temperature time series is reconstructed by cubic spline interpolation with a high sampling rate of 100Hz. This solves the time series discreteness problem caused by the original sensor's 10Hz sampling rate, so that the generated continuous skin temperature monitoring data stream has both high spatial resolution and high temporal resolution. This provides a reliable data foundation for subsequent fluctuation feature extraction, local vascular resistance index calculation and dynamic early warning threshold management, significantly improving the accuracy and clinical applicability of vascular crisis identification.

[0034] It is worth noting that the 100Hz sampling rate was chosen because it meets the Nyquist sampling theorem requirements for skin temperature and blood flow velocity signals, while balancing data volume and computational efficiency, and avoiding high-frequency noise interference.

[0035] In one specific embodiment, in step S22, the calibration point set of the three-dimensional calibration board is obtained, and the mapping relationship between the physical coordinates of each mapping point in the mapping point set and the pixel coordinates of each calibration point in the calibration point set is established through the three-dimensional calibration board and the least squares method, generating a spatial transformation matrix to map the skin temperature monitoring data and blood flow velocity signal to the same coordinate system based on the spatial transformation matrix, specifically including: Step S221: Based on the design spacing of the 3D calibration board, extract the pixel coordinates of each calibration point in the 3D calibration board using a corner detection algorithm to generate a calibration point set; Step S222: Construct simulated affected and healthy finger bodies with the same shape as the affected and healthy finger bodies, respectively, and attach the three-dimensional calibration plate to the surface of the simulated affected and healthy finger bodies to establish the mapping relationship between the physical coordinates of each mapping point and the pixel coordinates of each calibration point through the least squares method, and generate a spatial transformation matrix; Step S223: Map the skin temperature monitoring data and blood flow velocity signal to the same coordinate system using a spatial transformation matrix.

[0036] In one specific embodiment, the three-dimensional calibration plate is a flexible silicone plate with a known design spacing.

[0037] Specifically, the construction of simulated affected and healthy finger bodies involves acquiring surface point cloud data of the real affected and healthy finger bodies using a laser scanner, eliminating noise and optimizing geometry using point cloud processing algorithms, and then using CAD software to reconstruct the surface of the point cloud to generate a digital model that is completely consistent with the shape and size of the real finger body. Based on 3D printing technology, flexible materials such as silicone are used to create a physical simulated finger body, ensuring that its biomechanical properties are close to those of real tissue.

[0038] In embodiments of the present invention, a three-dimensional calibration plate is attached to the surface of a simulated finger, and the pixel coordinates of each calibration point in the three-dimensional calibration plate are extracted based on a corner detection algorithm to generate a calibration point set. Then, the least squares method is used to fit each mapping point and each calibration point to calculate the rigid transformation matrix from pixel coordinates to physical coordinates to generate a spatial transformation matrix.

[0039] Furthermore, since both the simulated affected and unaffected fingers are rigid models, the stability of the calibration process can be ensured, thus resolving calibration errors caused by the dynamic deformation of the actual affected finger due to postoperative swelling and movement. In addition, the simulated affected and unaffected fingers are reusable, providing a consistent reference benchmark for different monitoring stages, thereby significantly improving the reliability of vascular crisis early warning.

[0040] In one specific embodiment, in step S23, an electrocardiogram (ECG) signal is acquired and its corresponding R-wave peak is extracted. The R-wave peak corresponding to the blood flow velocity signal is extracted using the Pan-Tompkins algorithm, and its time difference is compared with that of the R-wave peak corresponding to the ECG signal. A dynamic time warping algorithm is then used to align the R-wave peaks of the ECG signal and the blood flow velocity signal, forming the R-wave peak of the pulse wave signal and defining it as a time reference point. Specifically, this includes: Step S231: Remove baseline drift and high-frequency noise from the blood flow velocity signal by using a 0.5-17Hz bandpass filter to form a purified blood flow velocity signal; Step S232: Calculate the first derivative and squared difference of the purified blood flow velocity signal, and integrate the first derivative and squared difference by moving the window to obtain continuous first derivative integral signal and continuous squared difference integral signal; Step S233: Fuse the continuous first derivative integral signal and the continuous square difference integral signal in a ratio of 1:7 to generate a feature enhancement signal; Step S234: Extract the R-wave peak corresponding to the feature enhancement signal using the Pan-Tompkins algorithm, and compare it with the R-wave peak corresponding to the acquired electrocardiogram signal by time difference. Align the R-wave peak of the electrocardiogram signal and the R-wave peak of the blood flow velocity signal using the dynamic time warping algorithm, and define the confirmed R-wave peak timestamp as the time reference point.

[0041] In one specific embodiment, in step S232, the first derivative and squared difference of the purified blood flow velocity signal are calculated, and the first derivative and squared difference are integrated through a moving window of 50-100ms to obtain continuous first derivative integral signals and continuous squared difference integral signals. This also includes: The first derivative and the squared difference are integrated through a moving window of 50-100ms to obtain the first derivative integral signal and the squared difference integral signal, both of which are discrete. By using low-pass filtering, the discrete first-derivative integral signal and square difference integral signal are converted into continuous curves reflecting the slope and amplitude changes of the purified blood flow velocity, respectively, to obtain continuous first-derivative integral signals and continuous square difference integral signals.

[0042] Specifically, the formation of purified blood flow velocity signals can suppress electromyographic interference and sensor thermal noise in blood flow velocity signals, retain the frequency band related to heart pulsation, provide a clean signal for the extraction of subsequent feature enhancement signals, and avoid false detection of R-wave peaks due to noise interference.

[0043] By fusing continuous first-derivative integral signals and continuous square difference integral signals in a 1:7 ratio, a feature enhancement signal with prominent amplitude variation characteristics can be generated. Based on the prominent amplitude variation characteristics, the peak energy of the pulse wave rising branch in the feature enhancement signal is significantly enhanced, thereby improving the anti-interference ability of the R-wave peak detection of the blood flow velocity signal. This allows for more accurate alignment with the R-wave peak of the electrocardiogram signal to establish a unified time reference.

[0044] It is worth noting that blood flow velocity signals reflect changes in blood flow in the peripheral vessels of the affected and healthy fingers, while electrocardiogram (ECG) signals directly record cardiac electrical activity. Due to physiological conduction delays and differences in hardware acquisition, there is a time shift between the two. By extracting the R-wave peaks from both and aligning them using a dynamic time warping algorithm, a unified time reference can be established. This ensures the spatiotemporal alignment of subsequent skin temperature and blood flow velocity signals, providing a reliable temporal dimension reference for accurate early warning of vascular crises, and significantly improving the anti-interference capability and clinical applicability of vascular crisis monitoring.

[0045] Furthermore, the acquisition of electrocardiogram signals is achieved by attaching the silver chloride electrodes of a medical-grade electrocardiograph to specific lead positions on the chest, and the extraction of the R wave peak is also based on the Pan-Tompkins algorithm.

[0046] The specific principles of the Pan-Tompkins algorithm are existing technologies, therefore, this invention will not elaborate further on them here.

[0047] In one embodiment, in step S3, periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points are identified based on the continuous skin temperature monitoring data stream to extract and form a skin temperature fluctuation feature set. Based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave propagation time, the local vascular resistance index is calculated, specifically including: Step S31: Calculate the power spectral density of the skin temperature monitoring data stream for each frequency band in the continuous skin temperature monitoring data stream based on the wavelet transform algorithm, and select the frequency of the frequency band corresponding to the maximum power spectral density as the dominant frequency, so as to identify the periodic fluctuation segment based on the dominant frequency. Step S32: Perform piecewise linear regression algorithm fitting on the continuous skin temperature monitoring data stream to identify the frequency band slope with an upward or downward trend, and extract the frequency bands whose slope exceeds the clinical experience threshold as trend fluctuation segments; Step S33: Calculate the difference between each sampling point in the continuous skin temperature monitoring data stream and the median value of its previous 5 sampling points, so that when the difference is greater than the preset mutation threshold, the current sampling point is determined to be a mutation point. Step S34: Extract the fluctuation features of periodic fluctuation segments, trend fluctuation segments, and abrupt change points to form a skin temperature fluctuation feature set; Step S35: Calculate the local vascular resistance index using a support vector regression model based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave conduction time.

[0048] Specifically, the clinical experience threshold is determined based on the 90-99% distribution range of normal skin temperature fluctuation rates in large-scale clinical data, and the clinical experience threshold corresponding to this distribution range is 0.02-0.05. This is to ensure that trend fluctuation segment identification captures only pathology-related changes.

[0049] The preset mutation threshold ranges from 0.1 to 0.3. It uses statistical analysis of the standard deviation of instantaneous skin temperature fluctuations in healthy individuals, and sets the standard deviation to be 2-4 times the standard deviation of instantaneous skin temperature fluctuations, in order to balance the sensitivity of abrupt change detection and the resistance to noise interference.

[0050] The embodiments of the present invention capture instantaneous changes in skin temperature in real time by using median differential mutation detection. By comparing the difference between the median value of the current sampling point and the median value of the previous 5 sampling points with a preset threshold, the abrupt change point is determined. This enables rapid capture of instantaneous temperature changes caused by crises such as vasospasm or embolism, providing a key time window for early intervention and significantly improving the timeliness of crisis warning.

[0051] In one specific embodiment, in step S31, the power spectral density of the skin temperature monitoring data stream for each frequency band in the continuous skin temperature monitoring data stream is calculated based on the wavelet transform algorithm, and the frequency of the frequency band corresponding to the maximum power spectral density is selected as the dominant frequency, so as to identify the periodic fluctuation segment based on the dominant frequency, specifically including: Step S311: Using a sliding window with a window length of 5-10 minutes, the power spectral density of the skin temperature monitoring data stream in each frequency band of the continuous skin temperature monitoring data stream is calculated by combining the wavelet transform algorithm, and the frequency of the frequency band corresponding to the maximum power spectral density is defined as the dominant frequency. Step S312: Process the time-domain signal segment in the skin temperature monitoring data stream corresponding to the dominant frequency using a bandpass filter to extract the purified periodic fluctuation signal; Step S313: Calculate the period duration of the periodic fluctuation signal using the autocorrelation function analysis algorithm, and determine the period start point based on the first peak point in the period duration. Based on the period duration and the time difference between the period start point and the adjacent peak point, obtain and identify the periodic fluctuation segments in the periodic fluctuation signal based on the period duration characteristics and phase offset characteristics.

[0052] Specifically, the range of the 5-10 minute sliding window is determined based on time periods that can balance time-frequency resolution. Furthermore, the extraction of purified periodic signals can eliminate high-frequency noise and low-frequency drift interference, thus improving signal quality.

[0053] By calculating the cycle duration using the autocorrelation function analysis algorithm and determining the cycle start point using the first peak point, and extracting phase shift features by combining the time difference between adjacent peaks, it is possible to identify complete fluctuation segments based on cycle duration stability and phase consistency.

[0054] It is worth noting that the multi-scale analysis of the wavelet transform algorithm can capture the transient characteristics of non-stationary skin temperature signals, so as to focus on key fluctuation components by the dominant frequency, avoid misjudgment caused by multi-frequency interference, and significantly improve the sensitivity and robustness of vascular crisis early warning.

[0055] In one specific embodiment, in step S32, a piecewise linear regression algorithm is used to fit the continuous skin temperature monitoring data stream to identify frequency band slopes with upward or downward trends, and frequency bands whose slopes exceed clinical experience thresholds are marked as trend fluctuation segments. Specifically, this includes: Step S321: Perform exponentially weighted moving average noise reduction processing on the skin temperature monitoring data stream of each frequency band to obtain smooth skin temperature monitoring data for each frequency band; Step S322: Fit the smoothed skin temperature monitoring data for each frequency band using a piecewise linear regression algorithm to identify the slope of frequency bands with an upward or downward trend, and extract the frequency bands with slopes exceeding 0.05. The frequency band corresponding to the clinical experience threshold is marked as the trend fluctuation segment.

[0056] Specifically, embodiments of the present invention utilize an exponentially weighted moving average to smooth the skin temperature data stream for each frequency band, suppressing short-term noise while preserving physiological trends and ensuring the stability of trend analysis. Furthermore, a piecewise linear regression algorithm is used to fit the smoothed skin temperature monitoring data into multiple linear frequency bands, precisely quantifying the slope and duration of each band, and particularly identifying trend fluctuations with slopes exceeding the clinically experienced threshold of 0.05℃ / min.

[0057] In one specific embodiment, step S34 involves extracting the fluctuation features of periodic fluctuation segments, trend fluctuation segments, and abrupt change points to form a skin temperature fluctuation feature set, specifically including: Step S341: Based on the period duration characteristics and phase offset characteristics, generate the fluctuation characteristics of the periodic fluctuation segment; Step S342: Calculate the slope direction of the frequency band slope corresponding to the trend fluctuation segment based on the linear regression equation, and obtain the slope change amplitude feature based on the absolute value of the frequency band slope, so as to generate the fluctuation feature of the trend fluctuation segment through the slope direction and slope change amplitude feature. Step S343: Based on the phase difference between the abrupt wave point and the adjacent periodic wave segment, obtain the abrupt amplitude of the abrupt wave point and generate the wave characteristics of the abrupt wave point. Step S344: Extract the fluctuation features of periodic fluctuation segments, trend fluctuation segments, and abrupt change points, and combine them with blood flow velocity signals and pulse wave conduction time to form a skin temperature fluctuation feature set.

[0058] Specifically, the fluctuation characteristics of periodic fluctuation segments generated based on cycle duration and phase shift can quantify the physiological rhythm of skin temperature, providing a benchmark for distinguishing normal fluctuations from pathological cycle disorders. By extracting the trend slope direction and slope change amplitude features through linear regression equations, the dynamic changes of local blood flow can be intuitively reflected, providing quantitative indicators to distinguish between physiological fluctuations and pathological abnormalities, and significantly improving the specificity and timeliness of vascular crisis early warning.

[0059] The fluctuation characteristics of abrupt wave points obtained by utilizing the phase difference between abrupt wave points and adjacent periodic wave segments can be used to distinguish between thermoregulation and sudden events.

[0060] Therefore, the resulting skin temperature fluctuation feature set can provide comprehensive data support for subsequent support vector regression models, thereby improving the sensitivity of vascular crisis early warning and effectively solving the technical problem that single features are easily affected by noise interference.

[0061] In one embodiment, in step S35, based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave conduction time, the local vascular resistance index is calculated using a support vector regression model, specifically including: Step S351: Perform Kalman filtering on the blood flow velocity signal to obtain a denoised blood flow velocity signal; Step S352: Obtain the historical resistance vascular index set, and construct a support vector machine regression model by combining it with the support vector machine regression algorithm; Step S353: Input the skin temperature fluctuation feature set, the denoised blood flow velocity signal, and the pulse wave conduction time into the support vector machine model to calculate and output the local vascular resistance index.

[0062] Specifically, the embodiments of the present invention integrate multi-dimensional parameters such as skin temperature fluctuation characteristics, blood flow velocity signals, and pulse wave conduction time to achieve accurate quantification of the local vascular resistance index. This effectively solves the technical problem that a single parameter is easily affected by individual differences and noise interference, and provides a reliable basis for early warning of vascular crisis.

[0063] It is worth noting that the historical resistance vascular index set was obtained by simultaneously collecting skin temperature, blood flow velocity signals and pulse wave conduction time data from clinical patients, combined with the real vascular resistance values ​​obtained by Doppler ultrasound or catheterization as labels.

[0064] In one embodiment, in step S4, an individualized baseline model is constructed based on the skin temperature monitoring data and blood flow velocity signal of the healthy side, and the deviation between the skin temperature value and the blood flow velocity signal of the affected side is dynamically calculated based on the individualized baseline model, specifically including: Step S41: Using skin temperature monitoring data and blood flow velocity signals from the healthy side as input, and combining machine learning algorithms, construct an individualized baseline model that reflects the correlation between skin temperature monitoring data and blood flow velocity signals. Step S42: Input the skin temperature and blood flow velocity signals of the affected side into the individualized baseline model to obtain the theoretical skin temperature and theoretical blood flow velocity signals corresponding to the affected side; Step S43: The deviation between the affected side skin temperature value and the theoretical skin temperature value, as well as the blood flow velocity signal on the affected side and the theoretical blood flow velocity signal, are calculated by the absolute value of the difference between the affected side skin temperature value and the theoretical skin temperature value, to obtain the deviation between the affected side skin temperature value and the blood flow velocity signal.

[0065] Specifically, embodiments of the present invention obtain theoretical skin temperature and theoretical blood flow velocity signals by constructing an individualized baseline model, which can accurately reflect the relationship pattern between normal skin temperature and blood flow velocity signals on the healthy side.

[0066] The calculated deviation quantifies the difference between the actual measured value on the affected side and the theoretical value derived from the normal pattern on the healthy side. This helps medical staff to clearly and intuitively understand the abnormal condition on the affected side. For example, a large deviation may indicate circulatory disorders, tissue damage, or other pathological changes on the affected side, providing crucial evidence for timely diagnosis and the development of targeted treatment plans. This improves the accuracy of disease diagnosis and the effectiveness of treatment, and ultimately better addresses the technical problem of disease diagnosis and treatment being affected by the inability to accurately quantify the degree of abnormality on the affected side.

[0067] In one embodiment, in step S5, skin color images of the affected finger are periodically acquired, and the skin color images are compared with a predefined skin colorimetric card. The current skin color level of the affected finger is determined based on the comparison results. Specifically, this includes: Step S51: Using the camera mounted on the skin temperature monitor, periodically collect skin color images of the affected side of the fingers at preset time intervals; Step S52: Perform geometric correction and grayscale normalization on the skin color image of the affected side's fingers using the ROI localization algorithm to correct the skin color image of the affected side's fingers. Step S53: Perform color gamut matching between the corrected affected side finger skin color image and the predefined four-order skin color chart. By calculating the Euclidean distance between each pixel in the corrected affected side finger skin color image in LAB space and the center of each order of the four-order skin color chart, extract the order with the smallest Euclidean distance and define it as the skin color level of the corresponding pixel. Step S54: Statistically analyze the skin color level of each pixel in the corrected skin color image of the affected side finger, and determine the level with the highest percentage as the current skin color level of the affected side finger.

[0068] In one specific embodiment, the four-level skin color chart includes a black range, a dark brown range, a rosy range, and a pale white range.

[0069] Specifically, the preset time interval is based on the real-time needs of monitoring the condition of patients with replanted fingers. If the condition changes rapidly, the time interval needs to be shortened in order to capture changes in skin color in a timely manner. The preset time interval ranges from 5 to 60 minutes.

[0070] The center of each colorimetric chart is defined in the LAB color space: LAB is a color model that closely approximates human visual perception, where L represents lightness, A represents the range from green to red, and B represents the range from blue to yellow. The four-level skin tone colorimetric chart is transformed within the LAB color space. By calculating the average coordinates of each colorimetric chart in LAB space, the center coordinates of each colorimetric chart in LAB space are obtained. These center coordinates represent the center of the corresponding colorimetric chart.

[0071] The embodiments of the present invention quantify the skin color level of each pixel by minimizing the Euclidean distance, avoiding subjective judgment errors. The skin color level with the highest proportion in the skin color image of the affected side is taken as the current state, which can reflect the fullness of the microcirculation of the finger in real time, providing an objective basis for the early identification of vascular crisis and further enhancing the reliability of vascular crisis early warning.

[0072] In one embodiment, in step S6, a composite early warning index for the affected finger is constructed by combining the local vascular resistance index, deviation, and current skin color grade. A dynamic early warning threshold range is generated based on the patient's monitoring stage. When the composite early warning index is not within the dynamic early warning threshold range, a vascular crisis in the replanted finger is determined, and an early warning signal is output to the medical terminal. Specifically, this includes: Step S61: A composite early warning index for the affected side of the finger is constructed by combining the local vascular resistance index, deviation, and current skin color grade; Step S62: Obtain the patient's postoperative time, divide the patient's monitoring stage into the edema stage and the healing stage according to the postoperative time, and identify the patient's current monitoring stage through the patient's corresponding timestamp on the medical terminal; Step S63: Obtain a dynamic threshold model library including the threshold monitoring range during the edema stage and the monitoring range during the healing stage, so as to extract the threshold parameter range of the corresponding stage from the dynamic threshold model library based on the current monitoring stage; Step S64: Compare the composite early warning indicator with the threshold parameter range. When the composite early warning indicator is not within the dynamic early warning threshold range, determine the vascular crisis of the replanted finger and output an early warning signal to the medical terminal.

[0073] Specifically, the dynamic threshold model library was acquired by collecting relevant data from numerous replanted finger patients. This data included age, gender, cause of injury, severity of injury, and postoperative recovery status. Particular attention was paid to recording each patient's local vascular resistance index, deviation, and current skin color grade during the edema and healing phases. Statistical methods and machine learning algorithms were then used to conduct in-depth analysis of the data to uncover potential relationships and patterns, training models to predict reasonable threshold ranges under different conditions. After repeated validation and optimization, the threshold ranges corresponding to different postoperative stages were organized and categorized, ultimately constructing a dynamic threshold model library. This provides a basis for subsequently extracting appropriate threshold parameter ranges based on the patient's current monitoring stage.

[0074] The embodiments of the present invention significantly improve the accuracy and clinical applicability of vascular crisis early warning through the construction of composite early warning indicators and dynamic threshold management mechanism, effectively solving the limitations of traditional single-parameter monitoring and the false alarm problem of fixed thresholds.

[0075] This invention proposes an artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers. By mapping skin temperature monitoring data and blood flow velocity signals to the same coordinate system, and using the R-wave peak of the pulse wave signal as a time reference for interpolation, resampling, and filtering, a high spatiotemporal resolution continuous data stream is constructed, solving the spatiotemporal discreteness problem of traditional single-parameter monitoring. Furthermore, by identifying periodic, trend, and sudden fluctuation characteristics based on the continuous skin temperature monitoring data stream, and combining the blood flow velocity signal with the pulse wave propagation time to calculate the local vascular resistance index, quantitative analysis of hemodynamic abnormalities is achieved, breaking through the limitations of traditional methods that only... This approach overcomes the limitations of relying solely on skin temperature thresholds. It constructs an individualized baseline model based on data from the healthy side and calculates the deviation from the affected side, then adaptively adjusts the judgment threshold in conjunction with postoperative stages to address baseline drift caused by individual differences and postoperative physiological changes. Furthermore, it integrates local vascular resistance index, deviation, and skin color grade to generate a composite early warning index, and generates a dynamic early warning threshold range based on the patient's monitoring stage. This achieves accurate early warning through multi-parameter spatiotemporal alignment fusion and clinical context awareness, significantly improving the sensitivity of vascular crisis identification and reducing false alarm rates, thereby reducing the risk of necrosis in replanted fingers and increasing the success rate of finger replantation.

[0076] Accordingly, refer to Figure 2 Embodiments of the present invention also provide an artificial intelligence-based replanted finger vascular crisis monitoring and early warning system, used to implement the artificial intelligence-based replanted finger vascular crisis monitoring and early warning method in any of the above embodiments. The replanted finger vascular crisis monitoring and early warning system includes: The data acquisition module is used to acquire skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger. The data stream generation module is used to map skin temperature monitoring data and blood flow velocity signals to the same coordinate system, and uses the R-wave peak of the pulse wave signal as the time reference point to interpolate, resample, and filter the skin temperature monitoring data to generate a continuous skin temperature monitoring data stream. The local vascular resistance index calculation module can identify periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points based on continuous skin temperature monitoring data streams, extract and form a skin temperature fluctuation feature set, and calculate the local vascular resistance index based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave conduction time. The deviation calculation module can construct an individualized baseline model based on the skin temperature monitoring data and blood flow velocity signal of the healthy side, and dynamically calculate the deviation between the skin temperature value and blood flow velocity of the affected side based on the individualized baseline model. The current skin color level determination module is used to periodically collect skin color images of the affected side of the finger, compare the skin color images with a predefined skin color matching card, and determine the current skin color level of the affected side of the finger based on the comparison results. The warning signal output module can construct a composite warning index for the affected finger by combining the local vascular resistance index, deviation, and current skin color level. It can also generate a dynamic warning threshold range based on the patient's monitoring stage. When the composite warning index is not within the dynamic warning threshold range, it can determine the vascular crisis of the replanted finger and output a warning signal to the medical terminal.

[0077] The artificial intelligence-based replanted finger vascular crisis monitoring and early warning system proposed in this invention belongs to the same general inventive concept as the artificial intelligence-based replanted finger vascular crisis monitoring and early warning method of this invention. It should at least have the same technical effects as the artificial intelligence-based replanted finger vascular crisis monitoring and early warning method of this invention. Therefore, this invention will not elaborate further here.

[0078] While one or more embodiments of the present invention have been described above, those skilled in the art will recognize that the present invention can be implemented in any other form without departing from its spirit and scope. Therefore, the embodiments described above are illustrative and not restrictive, and many modifications and substitutions will be apparent to those skilled in the art without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for monitoring and early warning of vascular crisis in replanted fingers based on artificial intelligence, characterized in that, include: Acquire skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger; Skin temperature monitoring data and blood flow velocity signals are mapped to the same coordinate system, and the skin temperature monitoring data are interpolated, resampled and filtered using the R-wave peak of the pulse wave signal as the time reference point to generate a continuous skin temperature monitoring data stream. Based on continuous skin temperature monitoring data stream, periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points are identified to extract and form a skin temperature fluctuation feature set. Based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave conduction time, the local vascular resistance index is calculated. An individualized baseline model was constructed based on the skin temperature monitoring data and blood flow velocity signal of the healthy side, and the deviation between the skin temperature value and the blood flow velocity signal of the affected side was dynamically calculated based on the individualized baseline model. Skin color images of the affected finger are collected periodically, and the skin color images are compared with a predefined skin colorimetric card. The current skin color grade of the affected finger is determined based on the comparison results. The composite early warning index of the affected finger is composed of local vascular resistance index, deviation degree and current skin color grade. A dynamic early warning threshold range is generated according to the patient's monitoring stage. When the composite early warning index is not within the dynamic early warning threshold range, a vascular crisis of the replanted finger is determined and an early warning signal is output to the medical terminal.

2. The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers according to claim 1, characterized in that, Acquire skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on both the affected and unaffected sides of the replanted finger, specifically including: Skin temperature monitors were used to monitor the affected and healthy sides of the replanted finger to obtain skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger.

3. The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers according to claim 2, characterized in that, The skin temperature monitor includes a monitoring chip and a flexible skin temperature sensor array and a photoplethysmography (PPG) sensor, both of which are electrically connected to the monitoring chip. The flexible skin temperature sensor array is used to monitor and acquire skin temperature data on the affected and healthy sides, while the PPG sensor is used to monitor and acquire blood flow velocity signals and pulse wave conduction time on the affected and healthy sides.

4. The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers according to claim 3, characterized in that, Skin temperature monitoring data and blood flow velocity signals are mapped to the same coordinate system, and the skin temperature monitoring data is interpolated, resampled, and filtered using the R-wave peak of the pulse wave signal as the time reference point to generate a continuous skin temperature monitoring data stream. Specifically, this includes: Record the physical coordinates of each node in the flexible skin temperature sensor array and the light emission and light receiving points of the photoplethysmography (PPG) sensor to generate a set of mapping points; Obtain the calibration point set of the three-dimensional calibration board, and establish the mapping relationship between the physical coordinates of each mapping point in the mapping point set and the pixel coordinates of each calibration point in the calibration point set through the three-dimensional calibration board and the least squares method. Generate a spatial transformation matrix to map skin temperature monitoring data and blood flow velocity signals to the same coordinate system based on the spatial transformation matrix. Acquire electrocardiogram (ECG) signals and extract their corresponding R-wave peaks. Use the Pan-Tompkins algorithm to extract the R-wave peaks corresponding to the blood flow velocity signals and compare them with the R-wave peaks corresponding to the ECG signals by time difference. Use a dynamic time warping algorithm to align the R-wave peaks of the ECG signals and the R-wave peaks of the blood flow velocity signals to form the R-wave peaks of the pulse wave signals and define them as time reference points. At the time reference point, the skin temperature monitoring data is subjected to cubic spline interpolation and filtering at a sampling rate of 100Hz to generate a continuous skin temperature monitoring data stream.

5. The artificial intelligence-based method for monitoring and early warning of vascular crisis in replanted fingers according to claim 4, characterized in that, Obtain the calibration point set of the 3D calibration board, and establish the mapping relationship between the physical coordinates of each mapping point in the mapping point set and the pixel coordinates of each calibration point in the calibration point set using the 3D calibration board and the least squares method. Generate a spatial transformation matrix to map skin temperature monitoring data and blood flow velocity signals to the same coordinate system based on the spatial transformation matrix. Specifically, this includes: Based on the design spacing of the 3D calibration board, the pixel coordinates of each calibration point in the 3D calibration board are extracted by a corner detection algorithm to generate a calibration point set; Simulated affected and healthy finger bodies with the same shape as the affected and healthy finger bodies were constructed respectively. A three-dimensional calibration plate was attached to the surface of the simulated affected and healthy finger bodies. The mapping relationship between the physical coordinates of each mapping point and each calibration point was established by the least squares method, and a spatial transformation matrix was generated. Skin temperature monitoring data and blood flow velocity signals are mapped to the same coordinate system using a spatial transformation matrix.

6. The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers according to claim 5, characterized in that, Based on continuous skin temperature monitoring data stream, periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points are identified to extract and form a skin temperature fluctuation feature set. Then, based on the skin temperature fluctuation feature set, blood flow velocity signal, and pulse wave propagation time, the local vascular resistance index is calculated, specifically including: The power spectral density of the skin temperature monitoring data stream in each frequency band of the continuous skin temperature monitoring data stream is calculated based on the wavelet transform algorithm, and the frequency of the frequency band corresponding to the maximum power spectral density is selected as the dominant frequency, so as to identify the periodic fluctuation segment based on the dominant frequency. Piecewise linear regression was used to fit the continuous skin temperature monitoring data stream to identify the slope of frequency bands with an upward or downward trend, and the frequency bands whose slopes exceeded the clinical experience threshold were marked as trend fluctuation segments. By calculating the difference between each sampling point in the continuous skin temperature monitoring data stream and the median value of the previous 5 sampling points, the current sampling point is determined to be a sudden change point when the difference is greater than the preset change threshold. Extract the fluctuation features of periodic fluctuation segments, trend fluctuation segments, and abrupt fluctuation points to form a skin temperature fluctuation feature set; Based on skin temperature fluctuation feature set, blood flow velocity signal and pulse wave conduction time, the local vascular resistance index is calculated using a support vector regression model.

7. The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers according to claim 6, characterized in that, An individualized baseline model was constructed based on skin temperature monitoring data and blood flow velocity signals from the healthy side. The deviation between the skin temperature value and blood flow velocity signal on the affected side was dynamically calculated based on this model, specifically including: Using skin temperature monitoring data and blood flow velocity signals from the healthy side as input, and combining machine learning algorithms, an individualized baseline model that reflects the correlation between skin temperature monitoring data and blood flow velocity signals is constructed. The skin temperature and blood flow velocity signals of the affected side are input into the individualized baseline model to obtain the theoretical skin temperature and theoretical blood flow velocity signals corresponding to the affected side. The deviation between the affected side's skin temperature and the theoretical skin temperature, as well as the difference between the affected side's blood flow velocity signal and the theoretical blood flow velocity signal, is obtained by calculating the absolute value of these differences.

8. The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers according to claim 7, characterized in that, Skin color images of the affected finger are collected periodically, and the images are compared against a predefined skin colorimetric chart. The current skin color grade of the affected finger is determined based on the comparison results. Specifically, this includes: Using a camera mounted on a skin temperature monitor, images of the skin color on the affected side of the fingers are periodically collected at preset time intervals. The skin color image of the affected side finger is geometrically corrected and grayscale normalized using the ROI localization algorithm in order to correct the skin color image of the affected side finger. The color gamut of the corrected affected finger skin image is matched with a predefined four-order skin color chart. The Euclidean distance between each pixel in the corrected affected finger skin image in LAB space and the center of each order of the four-order skin color chart is calculated, and the order with the smallest Euclidean distance is extracted and defined as the skin color level of the corresponding pixel. The skin color level of each pixel in the corrected skin color image of the affected side is statistically analyzed, and the level with the highest percentage is determined as the current skin color level of the affected side of the finger.

9. The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers according to claim 8, characterized in that, A composite early warning index for the affected finger is constructed using local vascular resistance index, deviation, and current skin color grade. A dynamic early warning threshold range is generated based on the patient's monitoring stage. When the composite early warning index falls outside the dynamic early warning threshold range, a vascular crisis in the replanted finger is identified, and an early warning signal is output to the medical terminal. Specifically, this includes: A composite early warning indicator for the affected side of the finger is formed by combining local vascular resistance index, deviation, and current skin color grade. The postoperative time of the patient is obtained, and the patient monitoring stage is divided into the edema stage and the healing stage according to the postoperative time. The current monitoring stage of the patient is identified by the corresponding timestamp of the patient through the medical terminal. Acquire a dynamic threshold model library that includes the threshold monitoring range during the edema stage and the monitoring range during the healing stage, so as to extract the threshold parameter range of the corresponding stage from the dynamic threshold model library based on the current monitoring stage; The composite early warning index is compared with the threshold parameter range. When the composite early warning index is not within the dynamic early warning threshold range, a vascular crisis in the replanted finger body is determined and an early warning signal is output to the medical terminal.

10. An artificial intelligence-based monitoring and early warning system for vascular crises in replanted fingers, characterized in that, The artificial intelligence-based method for monitoring and early warning of vascular crises in replanted fingers, as described in any one of claims 1-9, comprises: The data acquisition module is used to acquire skin temperature monitoring data, blood flow velocity signals, and pulse wave conduction time on the affected and healthy sides of the replanted finger. The data stream generation module is used to map skin temperature monitoring data and blood flow velocity signals to the same coordinate system, and uses the R-wave peak of the pulse wave signal as the time reference point to interpolate, resample, and filter the skin temperature monitoring data to generate a continuous skin temperature monitoring data stream. The local vascular resistance index calculation module can identify periodic fluctuation segments, trend fluctuation segments, and sudden fluctuation points based on continuous skin temperature monitoring data streams, extract and form a skin temperature fluctuation feature set, and calculate the local vascular resistance index based on the blood flow velocity signal and pulse wave conduction time in the skin temperature fluctuation feature set. The deviation calculation module can construct an individualized baseline model based on the skin temperature monitoring data and blood flow velocity signal of the healthy side, and dynamically calculate the deviation between the skin temperature value and blood flow velocity of the affected side based on the individualized baseline model. The current skin color level determination module is used to periodically collect skin color images of the affected side of the finger, compare the skin color images with a predefined skin color matching card, and determine the current skin color level of the affected side of the finger based on the comparison results. The warning signal output module can construct a composite warning index for the affected finger by combining the local vascular resistance index, deviation, and current skin color level. It can also generate a dynamic warning threshold range based on the patient's monitoring stage. When the composite warning index is not within the dynamic warning threshold range, it can determine the vascular crisis of the replanted finger and output a warning signal to the medical terminal.

Images