Internal fistula safety detection system and method

Abstract

The invention relates to the technical field of internal fistula safety detection, and discloses an internal fistula safety detection system and method.Image data of an internal fistula area are obtained through an image obtaining module, and physiological data of a patient are monitored in real time; the image processing module is used for processing image data, extracting internal fistula features and judging the existence and severity of internal fistula, so that the diagnosis accuracy and efficiency are improved; the recognition and diagnosis module automatically recognizes internal fistula features and predicts changes of internal fistula, valuable reference information is provided for doctors, and making of a more accurate treatment scheme is helped; the interactive display module visually displays the existence, severity and change of the internal fistula, and generates a diagnosis result and a diagnosis suggestion; and the feedback adjustment module dynamically adjusts an internal fistula judgment strategy and a prediction strategy according to an actual diagnosis result, so that the diagnosis and prediction accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of arteriovenous fistula safety detection technology, and in particular to a system and method for detecting arteriovenous fistula safety. Background Technology

[0002] In existing technologies, combining ultrasound imaging can provide real-time images of arteriovenous fistulas (AVFs), reducing patient discomfort and recovery time while improving the accuracy of AVF detection. The presence and severity of AVFs are assessed by detecting specific proteins or metabolites in the blood. Real-time monitoring of the patient's physiological parameters and symptoms helps in timely problem detection. However, most existing technologies do not address how to determine the presence and severity of AVFs by acquiring image data of the fistula area, how to predict changes in the fistula for visualization, and how to dynamically adjust based on actual diagnostic results. Summary of the Invention

[0003] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0004] In view of the problems existing in the current system and method for detecting arteriovenous fistula safety, the present invention is proposed.

[0005] Therefore, the purpose of this invention is to provide a system and method for detecting arteriovenous fistula safety.

[0006] To address the aforementioned technical problems, this invention provides a safe detection system for arteriovenous fistulas, comprising:

[0007] Image acquisition module, image processing module, recognition and diagnosis module, interactive display module, and feedback and adjustment module;

[0008] The image acquisition module is used to acquire image data of the arteriovenous fistula area and monitor the patient's physiological data in real time;

[0009] The image processing module is used to process image data, extract fistula features, and determine the presence and severity of the fistula.

[0010] The identification and diagnosis module is used to automatically identify arteriovenous fistula characteristics and predict changes in the arteriovenous fistula;

[0011] The interactive display module is used to visualize the existence, severity, and changes of arteriovenous fistulas;

[0012] The feedback adjustment module is used to dynamically adjust the fistula judgment and prediction strategies based on the actual diagnostic results.

[0013] As a preferred embodiment of the fistula safety detection system of the present invention, the system emits near-infrared light to the fistula area through near-infrared spectral imaging, receives the light signal returned after being scattered and absorbed by the fistula tissue, obtains image data of the fistula area by analyzing the differences in absorption characteristics of near-infrared light of different wavelengths, obtains blood flow velocity based on the image data of the fistula area, and monitors the patient's physiological data in real time in conjunction with physiological monitoring equipment, and stores the image data and physiological data synchronously.

[0014] As a preferred embodiment of the fistula safety detection system of the present invention, the system obtains the changes in blood oxygen saturation and hemoglobin concentration based on the absorption characteristics of near-infrared light of different wavelengths in the fistula area, adjusts the wavelength range and detection sensitivity of the near-infrared spectral imaging source, wherein the near-infrared spectral imaging is configured with a dual-wavelength or multi-wavelength light source, and calculates the relative concentration changes of oxyhemoglobin and deoxyhemoglobin by detecting the differences in the absorption characteristics of near-infrared light in the fistula tissue, so as to obtain the changes in blood oxygen saturation and hemoglobin concentration.

[0015] As a preferred embodiment of the fistula safety detection system of the present invention, the fistula features are extracted based on the changes in blood oxygen saturation and hemoglobin concentration in the fistula area. The fistula features include the trend of tissue perfusion volume, oxygenation status, blood oxygen saturation variation characteristics, and perfusion response speed. The presence and severity of the fistula are determined based on the fistula features.

[0016] The variation characteristics of blood oxygen saturation are obtained by quantifying the changes in hemoglobin concentration, reflecting the periodic fluctuation pattern of oxyhemoglobin in the fistula area.

[0017] As a preferred embodiment of the fistula safety detection system of the present invention, the variation characteristics of blood oxygen saturation at adjacent time points are compared to obtain the variation difference value.

[0018] Configure a variation threshold, compare the variation difference with the variation threshold to determine the degree of abnormality of the variation characteristics of blood oxygen saturation. If the variation difference is greater than or equal to the variation threshold, it is judged that the variation characteristics are normal.

[0019] If the difference in variation is less than the variation threshold, it is judged as an abnormal variation feature;

[0020] The concentration change rate is obtained by time series analysis of the hemoglobin concentration change value. A rate threshold is set, and the concentration change rate is compared with the rate threshold. If the concentration change rate is greater than or equal to the rate threshold, the perfusion response rate is judged to be normal.

[0021] If the rate of concentration change is less than the rate threshold, the perfusion response rate is considered abnormal.

[0022] As a preferred embodiment of the fistula safety detection system of the present invention, the system detects the total concentration of hemoglobin in the fistula region and configures the perfusion volume threshold range.

[0023] The total concentration of hemoglobin is compared with the perfusion volume threshold range. If the total concentration of hemoglobin is within the perfusion volume threshold range, the tissue perfusion volume is judged to be normal.

[0024] If the total concentration of hemoglobin exceeds the perfusion threshold range, the tissue perfusion is considered abnormal.

[0025] Analyze the directionality of changes in the ratio of oxyhemoglobin to deoxyhemoglobin concentrations;

[0026] Configure trend judgment rules. If the direction of the ratio change conforms to the preset normal fluctuation pattern, the trend of oxygenation status change is judged to be normal.

[0027] If the direction of the ratio change deviates from the preset normal fluctuation pattern, the trend of oxygenation status change is judged to be abnormal.

[0028] As a preferred embodiment of the arteriovenous fistula safety detection system of the present invention, the presence and severity of the arteriovenous fistula are determined based on the characteristics of the fistula using an arteriovenous fistula determination strategy, the arteriovenous fistula determination strategy including:

[0029] If the tissue perfusion is normal and the trend of oxygenation status is normal, then it is determined that the arteriovenous fistula does not exist;

[0030] If the tissue perfusion volume is abnormal or the trend of changes in oxygenation status is abnormal, then an arteriovenous fistula is determined to exist.

[0031] When an arteriovenous fistula exists, its severity is assessed. If the variation characteristics of blood oxygen saturation are abnormal and the perfusion response rate is abnormal, the severity of the fistula is assessed as the extreme value of severity.

[0032] If the variation characteristics of blood oxygen saturation are abnormal or the perfusion response rate is abnormal, the severity of the arteriovenous fistula is judged by the severe baseline value.

[0033] If the variation characteristics of blood oxygen saturation are normal and the perfusion response rate is normal, then the severity of the arteriovenous fistula is judged to be non-severe.

[0034] In a preferred embodiment of the arteriovenous fistula safety detection system of the present invention, the system automatically identifies arteriovenous fistula characteristics and predicts changes in the fistula using a prediction strategy, wherein the prediction strategy includes:

[0035] Physiological features are extracted from physiological data, including changes in heart rate and blood pressure, and weights of physiological features and arteriovenous fistula features on changes in arteriovenous fistula are configured.

[0036] The changes in the arteriovenous fistula (AVF) are predicted by comprehensively considering physiological characteristics and AVF characteristics, and the AVF change value is obtained. The AVF change value is based on the weighted fusion result of changes in blood oxygen saturation, hemoglobin concentration, heart rate variability, and blood pressure fluctuation.

[0037] Configure the baseline and extreme values ​​of arteriovenous fistula (AVF) changes, and compare the AVF changes with the baseline and extreme values ​​to obtain the AVF status.

[0038] If the change value of the arteriovenous fistula is greater than or equal to the extreme value of the change value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is abnormal and the state of the arteriovenous fistula is at the baseline value of the arteriovenous fistula.

[0039] If the change value of the arteriovenous fistula is less than the extreme value of the change value of the arteriovenous fistula but greater than the base value of the change value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is abnormal and the arteriovenous fistula is in a high value state.

[0040] If the change value of the arteriovenous fistula is less than or equal to the base value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is normal and the state of the arteriovenous fistula is at the extreme value of the arteriovenous fistula.

[0041] As a preferred embodiment of the fistula safety detection system of the present invention, the system configures the range of variation threshold, rate threshold, and perfusion volume threshold based on a dynamic threshold configuration strategy, wherein the dynamic threshold configuration strategy includes:

[0042] Obtain the baseline values ​​of total hemoglobin concentration, baseline variation range of blood oxygen saturation, and baseline perfusion response rate of patients under basic physiological conditions in order to set individualized initial thresholds;

[0043] During continuous monitoring, the initial threshold is offset and corrected according to the changing trend of the patient's physiological data. If the heart rate or blood pressure changes exceed the preset fluctuation range, the corresponding rate threshold and perfusion volume threshold range are adjusted.

[0044] If the baseline variation range of blood oxygen saturation drifts, the variation threshold is adjusted synchronously to ensure that the initial threshold matches the patient's current physiological state.

[0045] A method for detecting arteriovenous fistula safety includes: acquiring image data of the fistula area and monitoring the patient's physiological data in real time;

[0046] Image data is processed to extract arteriovenous fistula features and determine the presence and severity of the fistula;

[0047] Automatically identify arteriovenous fistula characteristics and predict changes in the fistula;

[0048] Visualize the presence, severity, and changes of arteriovenous fistulas;

[0049] The strategies for identifying and predicting arteriovenous fistulas are dynamically adjusted based on the actual diagnostic results.

[0050] The beneficial effects of this invention are as follows: The image acquisition module acquires image data of the fistula area and monitors the patient's physiological data in real time. Real-time monitoring and early warning functions help detect fistula problems early, prevent disease progression, and improve the timeliness and effectiveness of treatment. The image processing module processes the image data, extracts fistula features, and determines the presence and severity of the fistula. Automatic analysis of the image data and identification of fistula features improves diagnostic accuracy and efficiency. The identification and diagnosis module automatically identifies fistula features and predicts changes in the fistula. Based on historical and current data, it predicts the trend of fistula changes, providing valuable reference information for doctors. The interactive display module visually displays the presence, severity, and changes of the fistula, and generates diagnostic results and suggestions, enabling a more intuitive understanding of the patient's condition and treatment progress, thus allowing for more accurate decision-making. The feedback and adjustment module dynamically adjusts the fistula judgment and prediction strategies based on actual diagnostic results, continuously learning and optimizing through feedback mechanisms to improve the accuracy of diagnosis and prediction. Attached Figure Description

[0051] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0052] Figure 1 This is a system structure diagram of a fistula safety detection system according to the present invention;

[0053] Figure 2 This is a flowchart of the prediction strategy for an arteriovenous fistula safety detection system according to the present invention.

[0054] Figure 3 This is a flowchart of a method for detecting arteriovenous fistula safety according to the present invention. Detailed Implementation

[0055] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0056] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention can also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0057] Secondly, the term "one embodiment" or "embodiment" as used 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 different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0058] Example 1:

[0059] This embodiment provides a system structure diagram of an arteriovenous fistula safety detection system, as follows: Figure 1 As shown, a fistula safety detection system includes an image acquisition module, an image processing module, an identification and diagnosis module, an interactive display module, and a feedback adjustment module.

[0060] The image acquisition module is used to acquire image data of the arteriovenous fistula area and monitor the patient's physiological data in real time.

[0061] Furthermore, near-infrared light is emitted into the fistula area through near-infrared spectral imaging, and the light signal returned after being scattered and absorbed by the fistula tissue is received. By analyzing the differences in the absorption characteristics of near-infrared light of different wavelengths, image data of the fistula area is obtained. Blood flow velocity is obtained based on the image data of the monitored fistula area, and physiological data of the patient is monitored in real time in conjunction with physiological monitoring equipment. The image data and physiological data are stored synchronously.

[0062] Near-infrared spectroscopy imaging relies on the specific absorption capacity of tissues for near-infrared light of different wavelengths. The emission of stable and wavelength-accurate light signals is a prerequisite for obtaining analyzable data. The difference in absorption peak values ​​between oxyhemoglobin and deoxyhemoglobin in arteriovenous fistula tissue can only be effectively distinguished by excitation with a light source of a specific wavelength band. A multi-wavelength array of light-emitting diodes is used as the light source system, with at least two near-infrared light emission channels of different center wavelengths configured. The wavelength selection is determined based on the characteristic absorption spectrum of hemoglobin components.

[0063] The light source driving circuit adopts a constant current mode to ensure stable light intensity; the emission angle is mechanically adjusted according to the location of the fistula area on the body surface to ensure that the light spot covers the target detection area; an optical coupling medium layer is configured on the contact surface between the light source and the skin to reduce interface reflection loss; multi-wavelength synchronous emission can acquire multi-dimensional information on tissue oxygenation status at one time, avoiding the time asynchrony problem caused by time-sharing acquisition; constant current drive eliminates measurement errors introduced by light intensity fluctuations and improves the data signal-to-noise ratio; the mechanically adjustable angle design adapts to the differences in anatomical position of different patients.

[0064] After near-infrared light penetrates the skin and subcutaneous tissue, some photons are absorbed by hemoglobin, while others are scattered and return to the surface. Receiving these modulated light signals is the physical basis for extracting tissue blood oxygenation information. Changes in blood flow in the fistula area will dynamically change the absorption coefficient, and this change can only be captured by high-sensitivity detection. Photodetector arrays are deployed at different distances from the light source to form a multi-point spatial sampling layout. The detectors use photoelectric conversion elements with response wavelengths covering the near-infrared band, and narrow-band filters are configured at the front end to suppress ambient light interference.

[0065] The receiving optical path employs either fiber optic coupling or direct contact to ensure efficient collection of returned photons. The detector output signal undergoes impedance matching and primary amplification via a preamplifier circuit before being connected to an analog-to-digital converter for digital acquisition. The sampling frequency setting must meet the requirements for capturing hemodynamic changes. Spatial multi-point sampling can acquire tissue information at different penetration depths, enhancing the ability to detect deep blood flow in arteriovenous fistulas. The filter design effectively isolates interference from external light sources, making it suitable for complex lighting environments such as wards. The high sampling frequency ensures that rapid blood flow fluctuations are not missed.

[0066] The difference in molar absorptivity between oxyhemoglobin and deoxyhemoglobin in the near-infrared band is the theoretical basis for calculating blood oxygen saturation. By comparing the degree of light intensity attenuation at different wavelengths, the relative concentration ratio of the two hemoglobin components in the tissue can be deduced to quantify the oxygenation status of the arteriovenous fistula area. For the multi-wavelength light intensity data collected by each detector channel, attenuation models of photons propagating in the tissue are established respectively. The models take into account the change in average optical path caused by scattering effects, and the calculation deviation of the absorption coefficient is corrected by a correction factor.

[0067] A spectral demodulation algorithm is used to separate the absorption contribution at each wavelength, and a linear equation system is constructed to solve for the concentration of hemoglobin components. Constraints are introduced during the algorithm iteration process to ensure the rationality of the physical meaning of the solution. Attenuation model correction eliminates systematic biases introduced by individual tissue differences, improves the comparability of data between different patients, and the real-time solution capability enables the system to have dynamic monitoring function, which can continuously track the evolution of arteriovenous fistula oxygenation status. Constraints prevent algorithm divergence and ensure stable and reliable output results.

[0068] The raw light intensity data needs to be converted into a spatially meaningful image format to visualize the oxygenation distribution pattern of the arteriovenous fistula area. The image representation makes it easier for clinicians to intuitively identify abnormal perfusion areas, and the image data structure facilitates feature extraction by subsequent digital image processing algorithms. The blood oxygen saturation or hemoglobin concentration values ​​calculated by each detector channel are mapped to grayscale or pseudo-color pixel values. An image coordinate system is established according to the spatial arrangement of the detectors, with each pixel position corresponding to the physiological information of a specific detector sampling point. The image generation process uses an interpolation algorithm to fill the gaps between the detectors, forming a continuously distributed two-dimensional matrix.

[0069] Image frames are arranged in chronological order to form an image sequence, with each frame marked with a timestamp that is synchronously marked with physiological monitoring data; image resolution is related to detector density and interpolation algorithm complexity; pseudo-color encoding enhances the visual contrast of blood oxygenation differences, making low perfusion areas readily apparent; time-series storage supports playback analysis, which helps to capture sporadic abnormal perfusion events; and the image matrix structure provides a standardized input format for subsequent image segmentation, edge detection, and other processing.

[0070] Blood flow velocity in arteriovenous fistulas is a core indicator for assessing patency. Although near-infrared spectroscopy cannot directly measure flow velocity, it can indirectly reflect hemodynamic characteristics by analyzing the temporal gradient of hemoglobin concentration changes. Rapidly changing blood oxygenation signals usually correspond to higher blood flow flushing efficiency, while slow changes suggest blood stasis. Regions of interest are selected in the image sequence, and the time-varying curve of the mean blood oxygen saturation in that region is extracted. Differential operations are used to calculate the unit-time variation of blood oxygenation parameters as a substitute indicator for blood perfusion rate. This indicator is then standardized to eliminate individual differences and converted into a relative blood flow velocity parameter.

[0071] The derivation process can select a single-point area or multiple points along the blood vessel to obtain local and global blood flow information respectively; hemodynamic parameters can be obtained synchronously from spectral images without additional equipment, reducing the examination burden on patients; the relative velocity parameter reflects the changing trend of perfusion efficiency, which has practical guiding significance for clinical monitoring; multi-point area analysis can identify flow velocity differences caused by local stenosis and help locate the lesion site.

[0072] The functional status of an arteriovenous fistula (AVF) is affected by systemic physiological parameters. Local blood oxygenation information alone may lead to misjudgment. Changes in heart rate and blood pressure can alter the blood flow perfusion pressure of the AVF. Combined analysis can distinguish between physiological fluctuations and pathological abnormalities. Multi-source data fusion improves the robustness of diagnostic conclusions and avoids false positives or false negatives. It connects to ECG monitoring and non-invasive blood pressure monitoring devices through a standard medical data interface to obtain real-time heart rate and blood pressure values. The data interface follows medical information transmission protocols to ensure cross-brand device compatibility.

[0073] The acquisition frequency is synchronized with near-infrared spectral imaging, and sampling is triggered by a unified clock source; physiological data is stored in a structured format, and each record is timestamped; the fusion of multi-dimensional physiological parameters reduces the limitations of a single modality and improves diagnostic specificity; standard protocols support the avoidance of device typification, making it easy to integrate with existing hospital equipment; the time synchronization mechanism ensures accurate causal relationship analysis and supports delayed studies of pressure changes and blood oxygen response.

[0074] Image and physiological data originate from different acquisition units and require precise alignment in the time dimension to support joint analysis. A unified storage format facilitates data management and subsequent retrieval, while also meeting medical data traceability requirements. Structured storage provides a standardized dataset for machine learning training. A relational database or dedicated file format is established to merge image frames, blood oxygen parameters, blood flow velocity, and physiological monitoring data into a single record based on timestamps. Each record includes a unique identifier, acquisition time, patient identifier, and data quality label. Data integrity checks are performed before storage, and missing or outlier values ​​are marked as invalid. The database index structure supports fast time range queries and patient-level data aggregation. The storage medium employs a redundant backup mechanism to ensure no data loss, and access control complies with medical information security standards.

[0075] Timestamp fusion eliminates manual intervention in data alignment, improving analysis efficiency; quality labeling filters outomas to avoid interfering with diagnostic results; backup and access control ensure the security and compliance of medical data; and standardized formats provide a data foundation for subsequent algorithm iterations and multi-center studies. Synchronously stored data is the direct input source for the image processing and recognition / diagnosis modules, and its structural integrity determines whether subsequent algorithms can correctly parse it. Time-aligned data enables feature fusion, and high-quality datasets serve as the basis for the feedback adjustment module to optimize strategies. Comparative analysis of actual diagnostic results and stored data forms a closed-loop learning mechanism.

[0076] Specifically, the changes in blood oxygen saturation and hemoglobin concentration are obtained based on the absorption characteristics of near-infrared light of different wavelengths in the monitored fistula area. The wavelength range and detection sensitivity of the light source for near-infrared spectral imaging are adjusted. Near-infrared spectral imaging is equipped with dual-wavelength or multi-wavelength light sources. By detecting the differences in the absorption characteristics of near-infrared light in the fistula tissue, the relative concentration changes of oxyhemoglobin and deoxyhemoglobin are separated and calculated to obtain the changes in blood oxygen saturation and hemoglobin concentration.

[0077] Oxyhemoglobin and deoxyhemoglobin have different characteristic absorption spectra in the near-infrared band. A single wavelength cannot distinguish the contributions of the two. By configuring a dual-wavelength or multi-wavelength light source, a solvable linear equation system can be established under the premise of similar tissue penetration depth, so as to realize the separation calculation of the two hemoglobin components. The selection of the light source wavelength range directly affects the anti-interference ability and signal-to-noise ratio of the solution results. The light source system uses two or more laser diodes or narrow-band light-emitting diodes with adjustable center wavelengths, and the wavelengths correspond to the significant difference range of the absorption curves of oxyhemoglobin and deoxyhemoglobin, respectively.

[0078] The light source driving circuit enables time-division or simultaneous emission control. If time-division mode is used, the switching frequency must be much higher than the hemodynamic change frequency to avoid sampling distortion. The emitted light power is dynamically adjusted according to the patient's tissue thickness. The returned light intensity is monitored through closed-loop feedback, and the driving current is automatically adjusted to keep the signal within the detector's linear response range. An optical isolation layer is configured at the contact end between the light source and the skin to prevent ambient light interference and sweat contamination. The multi-wavelength configuration provides sufficient degrees of freedom for the equations, ensuring that the concentrations of oxyhemoglobin and deoxyhemoglobin can be solved independently. The adaptive power adjustment avoids signal saturation or low signal-to-noise ratio, improving the applicability to patients of different body types. The narrow-spectrum light source reduces absorption interference from other chromophores in the tissue, improving the specificity of hemoglobin detection.

[0079] When near-infrared light propagates in tissues, the differential absorption between oxyhemoglobin and deoxyhemoglobin leads to varying degrees of attenuation at different wavelengths. Detecting this attenuation difference is the physical basis for obtaining blood oxygen saturation information. Sensitivity adjustment ensures that effective information can still be captured under weak signals, while avoiding detector saturation caused by strong signals. After receiving the returned light signal, the photodetector independently quantizes the light intensity of each wavelength. Before quantization, signal conditioning is performed using an adjustable gain amplifier, with the gain coefficient dynamically set according to the returned light intensity level of the corresponding wavelength. The analog-to-digital conversion unit adopts a high-resolution architecture to ensure the resolution capability for subtle light intensity changes. Moving average filtering is applied to the signals acquired by each wavelength channel to suppress random noise.

[0080] The system synchronously records the emission time markers of the light source to accurately match the temporal relationship of each wavelength signal; it continuously monitors the signal quality index during the acquisition process, and triggers recalibration if the signal-to-noise ratio is lower than the preset standard; independent quantization and gain adjustment enable parallel high-precision acquisition of multi-wavelength signals, improving the system response speed; high-resolution analog-to-digital conversion preserves the details of minute changes in light intensity, enhancing the detection capability for weak perfusion; moving average suppresses noise without losing the overall trend of hemodynamic changes; and signal quality monitoring ensures the validity of the data entering the subsequent processing stage.

[0081] The acquired multi-wavelength light intensity data is a comprehensive result of the absorption effects of oxyhemoglobin and deoxyhemoglobin. The mixed signals must be separated using a solution model to obtain the individual hemoglobin component concentrations. Blood oxygen saturation is defined as the proportion of oxyhemoglobin to total hemoglobin, and its calculation depends on the accurate separation of the two components. Concentration changes reflect the increase or decrease in hemoglobin content per unit time and are a key parameter for assessing the dynamic response of blood perfusion. A light transmission model is constructed based on a modified Lambert-Beer law, incorporating a differential path length factor to correct for the increase in optical path length caused by tissue scattering. Absorption equations are established for each wavelength, forming a system of linear equations. The system of equations is solved using least squares or singular value decomposition to obtain the relative concentrations of oxyhemoglobin and deoxyhemoglobin. Blood oxygen saturation is calculated as the ratio of oxyhemoglobin concentration to total hemoglobin concentration. Concentration changes are obtained by differential calculations on the concentration results at consecutive time points, with the differential window length adjusted according to the hemodynamic response velocity.

[0082] The calculation process incorporates constraints, requiring blood oxygen saturation to remain within a physiologically reasonable range to prevent divergence in results caused by noise. A scattering correction model enhances the accuracy of concentration calculations, making the results closer to true physiological values. Least squares solutions strengthen resistance to measurement noise, ensuring stable solutions even with errors in a particular channel. Constraints guarantee the physiological significance of the calculation results, preventing invalid data output. A dynamic difference window automatically adapts to the blood flow response characteristics of different patients, improving the individualized accuracy of concentration changes. The calculated hemoglobin concentration is the source of pixel values ​​for generating functional images of the arteriovenous fistula region; its accuracy directly affects the image's ability to reflect the true perfusion state. The accuracy of blood oxygen saturation and concentration changes determines the reliability of assessing the presence and severity of the arteriovenous fistula. Continuous concentration time-series data provides a basis for subsequent trend analysis, supporting early warning of arteriovenous fistula functional deterioration.

[0083] The image processing module is used to process image data, extract fistula features, and determine the presence and severity of the fistula.

[0084] Furthermore, arteriovenous fistula (AVF) characteristics are extracted based on the changes in blood oxygen saturation and hemoglobin concentration in the AVF region. These characteristics include tissue perfusion volume, trends in oxygenation status, variations in blood oxygen saturation, and perfusion response rate. The presence and severity of the AVF are then determined based on these characteristics.

[0085] The variation characteristics of blood oxygen saturation are obtained by quantifying the changes in hemoglobin concentration, reflecting the periodic fluctuation pattern of oxyhemoglobin in the fistula area.

[0086] Tissue perfusion volume is a fundamental indicator for assessing arteriovenous fistula (AVF) function, directly reflecting the total amount of blood flowing through the AVF area per unit time. Insufficient perfusion volume will lead to suboptimal blood flow rate during dialysis, indicating AVF stenosis, thrombosis, or anastomotic abnormalities. Extracting this feature can quickly screen for functional and nonfunctional AVFs. Spatial integration of the total hemoglobin concentration values ​​of all pixels within the near-infrared spectral imaging field of view yields the total hemoglobin concentration in that region. This total concentration is then divided by the equivalent volume of the imaging region, which is estimated using preset detection depth and imaging area, and multiplied by an individualized calibration coefficient to convert it into standardized perfusion volume units.

[0087] Personalized calibration coefficients are obtained through measurements of the symmetrical position of the patient's healthy arm or statistical analysis of historical stable period data, eliminating individual differences in tissue thickness and hematocrit. During calculation, the penumbra region at the edge of the image is automatically removed to avoid overestimation caused by partial volume effects. The output results are stored in continuous numerical form and bound to timestamps. Spatial integration is used to comprehensively evaluate the overall perfusion status, avoiding random errors caused by single-point sampling. Volume normalization and individual calibration improve the comparability of data between different patients, making the threshold setting more universal. The edge removal mechanism ensures the effectiveness of the calculation area, and the results are closer to the true perfusion level.

[0088] The trend of oxygenation status reflects the dynamic process of the imbalance between oxygen supply at the arterial end and return at the venous end of the arterial fistula. A continuous downward trend suggests increased resistance in the arterial inflow tract or obstruction of venous return, which is an early signal of fistula function decline. Identifying the trend direction can distinguish between acute events and chronic progression, and guide the timing of clinical intervention. The average blood oxygen saturation data of the target region in consecutive frames of images along the time axis are extracted to construct a time series curve. The curve is smoothed using the sliding window averaging method or polynomial fitting method to preserve the trend change.

[0089] The algorithm calculates the slope sign of the curve within a unit time window. If the slopes of multiple consecutive windows are in the same direction, it is determined to be an upward or downward trend; if the slope signs alternate, it is determined to be a stable fluctuation. Confidence intervals are introduced to assess the statistical significance of the trend. The existence of a trend is confirmed only when the slope is significantly non-zero. The analysis results are output in the form of trend type labels and confidence scores. The sliding window and fitting algorithm effectively suppress instantaneous interferences such as respiration and exercise, improving the robustness of trend identification. The significance test mechanism reduces the risk of small fluctuations being misjudged as trends, avoiding false positive alarms. Trend analysis captures early compensatory dysfunction, providing prospective early warning information for clinical practice.

[0090] The periodic fluctuations in blood oxygen saturation originate from the pulsatile perfusion driven by cardiac pulsation. The variability directly reflects the intensity of pulsatile blood flow. The attenuation of the variability is a sensitive indicator of anastomotic stenosis or inflow obstruction, appearing earlier than changes in the mean. Quantifying the variability can identify subclinical hemodynamic abnormalities. The heart rate-corresponding frequency band signal is extracted from the blood oxygen saturation time series, and bandpass filtering is used to retain the 0.5-2Hz component. The peak and trough values ​​of each cardiac cycle are identified, and the peak-trough difference is calculated as the variability amplitude. The variability characteristic value is obtained by averaging over multiple consecutive cardiac cycles. To eliminate the influence of the baseline level, the variability amplitude is normalized by dividing by the mean blood oxygen saturation of the cycle to obtain the relative coefficient of variation.

[0091] Fourier transform was introduced for spectral verification to confirm that the dominant frequency component was consistent with the real-time heart rate, eliminating interference from respiratory or motion artifacts. The final output was the relative coefficient of variation and the spectral matching index. Bandpass filtering and spectral verification ensured that the extracted variability signal originated from the heartbeat, improving the feature specificity. Normalization processing made the degree of variation comparable among patients with different baseline blood oxygen levels, supported uniform threshold setting, and multi-cycle averaging reduced randomness. The results showed good repeatability, and this feature was highly specific for arterial stenosis.

[0092] Perfusion response velocity reflects the rapid adjustment ability of the arteriovenous fistula system to changes in systemic blood flow demand. It is a dynamic parameter for assessing vascular resistance and compliance. Response delay indicates decreased vascular bed compliance or the presence of occult stenosis and is an early sign of diminished functional reserve. Calculating this characteristic can identify potential failure risks even with normal static parameters. When physiological monitoring equipment detects a step change in heart rate or blood pressure, it is marked as an intrinsic stimulation event point. The dynamic change curves of blood oxygen saturation or hemoglobin concentration before and after stimulation are extracted, and the time interval from the stimulation initiation point to the parameter response initiation point is calculated as the response delay. The maximum rate of change is calculated as the response velocity.

[0093] To improve stability, the results of multiple natural physiological fluctuation events are averaged. Two sub-parameters, response delay and response speed, are output, along with the stimulus type, including heart rate or blood pressure. External intervention is avoided by triggering natural physiological fluctuations. The monitoring process is non-invasive and can be performed continuously. The two parameters, response delay and response speed, respectively assess compliance and resistance, providing more comprehensive information on functional reserve. Averaging multiple events improves the reliability of the results and reduces interference from accidental factors. This feature can identify a decline in reserve function even when routine indicators are still normal. Perfusion response speed is used as an independent dimension for severity assessment, evaluated in parallel with variability features. The prolongation trend of response delay provides a basis for predicting the time of functional decline in the prediction module. Abnormalities in this parameter indicate the need to increase monitoring frequency or intervene earlier, affecting the formulation of subsequent follow-up strategies.

[0094] Specifically, the variation characteristics of blood oxygen saturation at adjacent time points are compared to obtain the variation difference value;

[0095] Configure a variation threshold, compare the variation difference with the variation threshold to determine the degree of abnormality of the variation characteristics of blood oxygen saturation. If the variation difference is greater than or equal to the variation threshold, it is judged that the variation characteristics are normal.

[0096] If the difference in variation is less than the variation threshold, it is judged as an abnormal variation feature;

[0097] The concentration change rate is obtained by time series analysis of the hemoglobin concentration change value. A rate threshold is set, and the concentration change rate is compared with the rate threshold. If the concentration change rate is greater than or equal to the rate threshold, the perfusion response rate is judged to be normal.

[0098] If the rate of concentration change is less than the rate threshold, the perfusion response rate is considered abnormal.

[0099] The stability of blood oxygen saturation variability is crucial for assessing the pulsatile perfusion quality of arteriovenous fistulas. During continuous monitoring, sudden decreases or increases in the amplitude of variability often precede changes in the mean, serving as early and sensitive indicators of arterial inflow obstruction or abnormal venous return. Calculating the difference in variability between adjacent time points can transform continuous trends into discrete abnormal events, facilitating the automation of threshold triggering mechanisms. The system uses a fixed-time sliding window, such as 5 minutes, to calculate the mean of the blood oxygen saturation variability coefficient within the current window and perform a difference operation with the mean of the previous window to obtain the variability difference. The window sliding step size is set to 1 minute to achieve sequential updates. The difference calculation uses an absolute difference method to eliminate directional influences and focus only on the amplitude of change.

[0100] For each new data point entering the window, the oldest data point is removed to keep the window length constant and reduce the amount of computation. The calculation results are temporarily stored in the cache for the threshold comparison module to retrieve in real time. The sliding window mechanism balances response speed and stability, which can capture fast anomalies and avoid instantaneous interference and misjudgment. The incrementally updated differential method has low computational overhead and supports real-time monitoring and deployment. The absolute differential focuses on the intensity of change, making the anomaly judgment more intuitive and simplifying the threshold setting logic.

[0101] Differences in physiological baselines among different patients make it difficult to apply a uniform threshold. It is necessary to establish individualized or adaptive variation thresholds. The threshold is the decision boundary that distinguishes between normal physiological fluctuations and pathological abnormalities. The rationality of the configuration directly affects the false positive rate and sensitivity. The dynamic comparison mechanism transforms the abstract difference into a clear binary classification, providing a clear action signal for clinical decision-making. The threshold configuration adopts the baseline calibration method. When the patient uses the system for the first time, the mean of the coefficient of variation of 10-20 windows in the resting state is collected, its standard deviation is calculated, and the threshold is set as the baseline mean minus a preset multiple of the standard deviation, such as 3 times, to ensure that normal fluctuations are not misjudged.

[0102] During continuous monitoring, if multiple consecutive windows of data are normal, the baseline is updated using an exponential moving average method to adapt to the slow changes in the patient's physiological state. The comparison logic is implemented in the form of hardware interrupts or software events. When the difference in variance is lower than the threshold, an abnormal event flag is immediately triggered, and the start time stamp and duration of the abnormality are recorded. The system is equipped with a hysteresis mechanism to avoid frequent flips near the threshold boundary. The baseline calibration method achieves individualized thresholds, significantly improving the accuracy at the time of first use and reducing adaptation costs. The exponential moving average dynamically tracks physiological drift and maintains the validity of the threshold during long-term monitoring. The event triggering mechanism has a low response delay and supports real-time early warning. The hysteresis mechanism prevents signal jitter and avoids clinical interference caused by frequent alarms.

[0103] The rate of change in hemoglobin concentration reflects the perfusion efficiency and compliance of the vascular bed, not just the perfusion volume. During fluctuations in systemic blood pressure or heart rate, the arteriovenous fistula area should respond rapidly to maintain stable blood flow. A delayed response indicates increased vascular resistance or decreased compliance. Time-series analysis can extract rate parameters from the dynamic process of concentration changes, quantifying this responsiveness. The system establishes a time-series queue of concentration change values, with a queue length covering at least 30 seconds to include several cardiac cycles. The concentration difference between adjacent time points is calculated using the first-order difference method, and divided by the sampling interval to obtain the instantaneous rate of change. A moving average filter is applied to the instantaneous rate of change, with the window length matched to the heart rate cycle, eliminating pulsatile interference and preserving the overall trend.

[0104] When the physiological monitoring module detects a step change in heart rate or blood pressure, it marks it as the starting trigger point for analysis. The maximum rate of change within 10 seconds after the trigger is extracted as a representative value of the perfusion response rate. The analysis results are output in the form of rate values ​​and trigger event type labels. Moving average filtering effectively separates pulsatile changes from trend responses, so that the velocity parameter represents macroscopic perfusion regulation rather than a single heartbeat. Triggered analysis focuses on the dynamic response after physiological disturbances, improving parameter specificity. The maximum rate value reflects the peak regulatory capacity of the arteriovenous fistula and is more sensitive to early functional reserve decline.

[0105] The threshold for perfusion response rate needs to reflect the functional reserve baseline under different physiological states. The response requirements at rest and under stress are different, and a fixed threshold cannot adapt to the complexity of clinical practice. A reasonable configuration of the threshold can transform the rate parameter into a normal / abnormal classification, and help identify static indicators, such as perfusion volume, and the early risk of failure when dynamic reserve is impaired. The rate threshold adopts a hierarchical configuration strategy, which is divided into resting threshold and stress threshold. The resting threshold is obtained based on the patient's historical stable period data, taking the low percentile of the response rate under healthy conditions, such as the 10th percentile. The stress threshold is obtained through artificial or physiologically triggered blood pressure change tests, and the response rate of the arteriovenous fistula under normal conditions is recorded as a reference.

[0106] During routine monitoring, the resting threshold is activated when physiological data is within the resting range. If heart rate or blood pressure exceeds the stress threshold, the system switches to the stress threshold standard. The comparison logic adopts a continuous multiple judgment principle. A single reading below the threshold does not trigger an alarm immediately; an abnormal response flag is only triggered when there are three consecutive abnormal readings to prevent occasional noise interference. The hierarchical configuration adapts to different physiological scenarios, avoiding misjudgments caused by normal physiological fluctuations. The resting threshold is based on its own historical data, which is highly individualized. The stress threshold provides a functional reserve test benchmark. The continuous multiple judgment principle significantly improves specificity and reduces unnecessary clinical intervention. The perfusion response abnormality flag is an important dimension for judging severity and complements the variation abnormality. After the abnormality flag is triggered, the system automatically increases the monitoring frequency and suggests performing imaging structural examinations. The real-time value of the response speed provides a quantitative parameter for functional reserve in the prediction module, supporting the estimation of the remaining lifespan of the arteriovenous fistula.

[0107] Furthermore, the total hemoglobin concentration in the arteriovenous fistula region was measured to determine the perfusion volume threshold range;

[0108] The total hemoglobin concentration is compared with the perfusion volume threshold range. If the total hemoglobin concentration is within the perfusion volume threshold range, the tissue perfusion volume is considered normal.

[0109] If the total concentration of hemoglobin exceeds the perfusion threshold range, the tissue perfusion is considered abnormal.

[0110] Analyze the directionality of changes in the ratio of oxyhemoglobin to deoxyhemoglobin concentrations;

[0111] Configure trend judgment rules. If the direction of the ratio change conforms to the preset normal fluctuation pattern, the trend of oxygenation status change is judged to be normal.

[0112] If the direction of the ratio change deviates from the preset normal fluctuation pattern, the trend of oxygenation status change is judged to be abnormal.

[0113] Total hemoglobin concentration is a direct proxy for tissue perfusion, and its concentration level is positively correlated with the effective blood volume of the arteriovenous fistula area. Compared with single oxyhemoglobin or deoxyhemoglobin concentrations, total hemoglobin concentration is not affected by local oxygen metabolism and can more stably reflect the hemodynamic baseline. Spatial integration of the detection area can eliminate random fluctuations from single-point sampling and obtain statistically representative perfusion assessment values. The system extracts the calculated oxyhemoglobin and deoxyhemoglobin concentration values ​​from all effective pixels within the near-infrared spectral imaging field of view, sums them pixel by pixel, and generates a total hemoglobin concentration distribution matrix. An effective tissue detection depth threshold is set to remove pixels in the superficial skin and deep background noise layers, retaining only the data located in the vascular bed depth range.

[0114] The total concentration values ​​after screening are spatially averaged according to the number of pixels to obtain a representative value of the total hemoglobin concentration in the region. The standard deviation of this value is added to the output result as an evaluation index of spatial uniformity. The spatial integration and averaging mechanism reduces the influence of local microcirculation heterogeneity and measurement noise, and improves the repeatability and reliability of perfusion assessment. The depth screening mechanism automatically excludes non-target tissue signals to ensure that the calculation results are for the arteriovenous fistula bed. The standard deviation output reflects the spatial uniformity of perfusion, helps to identify focal hypoperfusion areas, and provides additional dimensional information for refined diagnosis.

[0115] Significant differences exist in hematocrit, vessel diameter, and tissue thickness among different patients. Using a uniform absolute threshold would lead to numerous misjudgments. The threshold should reflect the individualized physiological baseline range and distinguish between normal fluctuations and pathological deviations. Range-based thresholds better reflect the inherent variability of physiological parameters and improve the clinical applicability of the judgment. When a patient's monitoring record is established for the first time, hemoglobin total concentration data under stable conditions for several consecutive days are collected, and its mean and standard deviation are calculated. The perfusion volume threshold range is set to the mean plus or minus a preset multiple of the standard deviation, such as 2 times, and saved as the individual baseline. During continuous monitoring, the system updates the baseline parameters every 24 hours using effective data from the past 72 hours to achieve adaptive tracking of slow physiological drift.

[0116] The comparison and judgment adopts a lag logic, triggering the abnormal tissue perfusion indicator only when the total concentration exceeds the perfusion volume threshold three times consecutively and lasts for more than a preset duration, such as 5 minutes, thus avoiding false positives caused by transient interference. After the abnormality is triggered, the duration and deviation are recorded as a quantitative basis for risk assessment. Individual baseline calibration significantly improves threshold adaptability, allowing for reliable judgments without manual adjustments for different patients. The dynamic update mechanism addresses natural fluctuations in hemoglobin levels, maintaining long-term monitoring accuracy. The lag and duration judgment improve specificity and reduce invalid alarms triggered by transient factors such as movement and changes in body position. The quantification of deviation magnitude supports risk stratification and provides a basis for prioritizing the urgency of clinical interventions.

[0117] Simply analyzing isolated changes in oxyhemoglobin or deoxyhemoglobin concentrations cannot reflect the dynamic relationship of oxygen supply and demand balance. The direction of the ratio change can reveal the relative imbalance between arterial oxygen supply and venous return. A continuous decrease in the ratio suggests insufficient oxygen content in arterial inflow or aggravated venous congestion, which is a sensitive indicator of arteriovenous fistula dysfunction. Directional analysis transforms discrete concentration data into trending physiological events, supporting early warning. The oxyhemoglobin and deoxyhemoglobin concentrations of all pixels in each detection cycle are averaged regionally to calculate the ratio of oxyhemoglobin to deoxyhemoglobin concentrations. This ratio is arranged in a time series, and the moving window slope method is used to determine the direction of change. The window length is set to cover at least 10 cardiac cycles to smooth out pulsatile interference.

[0118] The system calculates the mean difference in ratios between adjacent windows. A positive difference indicates an upward trend, while a negative difference indicates a downward trend. If the absolute value of the difference is less than a preset noise threshold, it is marked as stable. The direction judgment result is output as the trend type (upward / downward / stable) and the quantified value of the slope intensity. The system has a built-in circadian rhythm correction mechanism to avoid misjudgments caused by physiological rhythms. Ratio analysis comprehensively reflects the relative relationship between oxygenation and metabolism, improving the specificity for hypoperfusion states. Moving windows smoothly eliminate instantaneous interference from heartbeat and respiration, making direction judgment more robust. Slope intensity quantification includes both directional and rate information in trend changes, supporting more refined decompensation stages. Circadian rhythm correction ensures that direction judgment is not affected by normal physiological fluctuations during long-term monitoring.

[0119] The normal fluctuation pattern of oxygenation status needs to be defined based on individual physiological characteristics and monitoring scenarios. Static rules cannot adapt to different dialysis stages, activity states, and circadian rhythms. Rule configuration needs to integrate prior medical knowledge and individual historical data to achieve adaptive judgment. Abnormal judgment results should be clinically operable to avoid overly sensitive invalid alarms. The normal fluctuation pattern rule base includes resting rules (nighttime during interdialysis), activity rules (daily activity periods), and stress rules (during dialysis or after exercise). When configuring rules, the system first identifies the current patient status pattern. The patient status pattern is determined by accessing wearable device data or physiological parameter thresholds and then calling the corresponding rule set. The resting rule requires the ratio to oscillate within a preset small range, the activity rule allows the ratio to rise slightly, and the stress rule allows the ratio to drop briefly and then recover quickly.

[0120] Each rule defines the allowable amplitude threshold and duration limit for directional changes. During judgment, if the current ratio direction and amplitude exceed the rule range and continue to exceed the rule limit duration, it is judged as an abnormal trend in oxygenation status. Anomaly judgment is supplemented with a confidence score, calculated based on the consistency of multiple consecutive cycles. If the confidence score exceeds a preset level, an alert is triggered. The scenario-specific rule configuration significantly improves the adaptability of judgment and reduces the false positive rate under different physiological states. The modular design of the rule base allows clinicians to add custom rules based on experience. The confidence scoring mechanism ensures the robustness of anomaly judgment and avoids misjudgment caused by occasional data jumps. The recovery time judgment of stress rules can identify insufficient functional reserve of arteriovenous fistula and provide suggestions on the timing of intervention.

[0121] Specifically, the presence and severity of arteriovenous fistulas (AVFs) are determined based on their characteristics using AVF assessment strategies. These strategies include:

[0122] If the tissue perfusion is normal and the trend of oxygenation status is normal, then it is determined that the arteriovenous fistula does not exist;

[0123] If the tissue perfusion volume is abnormal or the trend of changes in oxygenation status is abnormal, then an arteriovenous fistula is determined to exist.

[0124] When an arteriovenous fistula exists, its severity is assessed. If the variation characteristics of blood oxygen saturation are abnormal and the perfusion response rate is abnormal, the severity of the fistula is assessed as the extreme value of severity.

[0125] If the variation characteristics of blood oxygen saturation are abnormal or the perfusion response rate is abnormal, the severity of the arteriovenous fistula is judged by the severe baseline value.

[0126] If the variation characteristics of blood oxygen saturation are normal and the perfusion response rate is normal, then the severity of the arteriovenous fistula is judged to be non-severe.

[0127] The determination of the existence of arteriovenous fistulas (AVFs) needs to avoid the risks of false positives and false negatives from a single indicator. Tissue perfusion volume reflects the adequacy of overall blood flow, while the trend of oxygenation status reflects the dynamic direction of the balance between arterial oxygen supply and venous return. The two complement each other and can cover different types of AVF failure modes. The existence of AVFs can only be ruled out if both conditions are met, thus improving the specificity of the judgment. Any abnormality triggers the existence confirmation to ensure sensitivity. The system establishes independent parallel processing threads to receive normal / abnormal perfusion volume indicators and normal / abnormal oxygenation status trend indicators, respectively. Each indicator is accompanied by a timestamp and duration information. The logic judgment unit adopts a state machine mechanism, with the initial state being pending confirmation. When both inputs are normal and the duration is greater than the preset stabilization period, such as 3 minutes, the state machine transitions to the state of AVF non-existence, outputs a negative result, and locks the judgment until any subsequent abnormal indicator appears.

[0128] If an abnormal perfusion rate or abnormal trend in oxygenation is received at any time, the state machine immediately transitions to the state of arteriovenous fistula presence, outputs a positive result, and activates the severity assessment branch. The state transition process is logged, including the specific type and value of the indicator that triggered the abnormality. Parallel threads ensure independent judgment of dual conditions, avoiding delays caused by data waiting. The state machine locking mechanism prevents frequent result jumps, improving output stability. The duration requirement eliminates instantaneous fluctuation interference, resulting in higher reliability of positive results. Log recording supports post-event traceability, facilitating clinical validation and algorithm optimization.

[0129] In clinical practice, arteriovenous fistula failure can manifest as simple hypoperfusion without oxygenation imbalance, such as in the early stage of complete venous occlusion, or as simple oxygenation imbalance with perfusion still within the normal range, such as mild stenosis at the arterial end. The logic ensures that any abnormality can trigger an existence warning to avoid missing subclinical cases. This design reflects the inclusiveness of pathophysiological diversity. The state machine has an embedded OR logic gate, and the two input flags are ORed to output a comprehensive abnormal signal. The system has a preset priority register. When two channels are abnormal at the same time, the weight is calculated according to the magnitude and duration of the abnormality. The one with a larger magnitude and longer duration is given a higher priority. This priority is passed to the severity assessment module for feature weight configuration.

[0130] Simultaneously, a single-occurrence anomaly suppression counter is set. If an anomaly occurs briefly and then automatically recovers, and the cumulative number of occurrences within 24 hours is less than the preset tolerance value (e.g., 3 times), it is determined to be an occasional interference and will not trigger existence confirmation. The suppression counter is periodically reset to zero. Alternatively, the logic can be fully implemented to maximize clinical sensitivity. The priority weighting mechanism guides the focus on the main issues when assessing the severity, improving diagnostic accuracy. Occasional interference suppression avoids over-treatment, reduces patient anxiety and unnecessary examinations, and the adjustable tolerance value is designed to adapt to the sensitivity requirements of different clinical scenarios.

[0131] The severe extreme value corresponds to severe decompensation of arteriovenous fistula function, which requires the simultaneous disappearance of pulsatility (abnormal variation) and depletion of dynamic reserve (abnormal response). Both conditions and logic ensure rigorous judgment and avoid overdiagnosis of transient abnormalities in a single indicator. This state indicates that the arteriovenous fistula is on the verge of failure and requires immediate manual intervention. High specificity is crucial. The system receives two output flags: the abnormal variation flag and the abnormal perfusion response flag, along with their duration and amplitude information. When both flags are abnormal at the same time and the duration exceeds the severity judgment threshold, such as 10 minutes, the severity state machine transitions from the mild state to the severe extreme value state. Before the transition, a secondary verification process is initiated to retrieve the original data and recalculate the coefficient of variation and response rate to eliminate false flag triggering.

[0132] Once the verification is successful, the severe extreme value state is locked, a high-level alarm is triggered, and the current monitoring configuration is automatically frozen as evidence. After the state is locked, it can only be reset by manual confirmation by a clinician or by a system power failure and restart. The dual verification mechanism eliminates misjudgments caused by software logic errors or transient sensor failures, and the alarm has extremely high reliability. The duration threshold avoids overreaction to transient reversible abnormalities, ensuring that the extreme value state truly represents continuous failure. The state locking and evidence archiving meet the requirements for tracing medical disputes, and the high-level alarm ensures the timeliness of clinical response and reduces the rate of arteriovenous fistula abandonment.

[0133] The severe baseline corresponds to moderate impairment of arteriovenous fistula function, manifested by an abnormality in a single indicator of pulsatility or reserve capacity. This status ensures that such subclinical failures are not overlooked, while avoiding delayed diagnosis due to alarms only being triggered by double abnormalities. This state suggests the need for enhanced monitoring and preventative intervention to prevent progression to the severe extreme value. When only one of the two indicators—abnormal variation and abnormal perfusion response—is abnormal, or when both are abnormal but the duration does not reach the severe extreme value standard, the severity state machine transitions to the severe baseline state. The system records the specific type, amplitude, and duration of the abnormal indicators, generating a structured risk report. The report includes trend charts and baseline comparison data, supporting clinicians' intuitive judgment.

[0134] A severe baseline state triggers a medium-level alarm, and the monitoring frequency is moderately increased. This state allows for automatic reset. When the abnormal flag disappears and the situation remains stable for a period of time, the state machine can automatically downgrade to mild or non-existent. Alternatively, logic can be used to implement early warnings, providing a window of opportunity for preventive intervention. Structured risk reports improve the efficiency of doctor-patient communication, allowing patients to understand their own fistula status. The automatic reset mechanism avoids the paralysis caused by long-term alarm suspension, maintains the dynamic responsiveness of the system, and moderately increases the monitoring frequency to balance the needs of early warning and the consumption of system resources.

[0135] The "non-severe" state is the final confirmation of normal arteriovenous fistula function. It is necessary to ensure that if dynamic indicators recover after a positive presence test, the severity should be downgraded in a timely manner to avoid overtreatment. The dual normality and logical approach provide a downgrade path and maintain the dynamic balance of the assessment system. This state allows for the release of system resources and reduces the psychological burden on patients. When the severity state machine is at the severe baseline or severe extreme value, the system continuously monitors abnormal variation markers and abnormal perfusion response markers. If both markers remain normal during the continuous monitoring period, and the trends of tissue perfusion and oxygenation status also return to normal synchronously, the state machine transitions to the "non-severe" state.

[0136] Before transfer, a smooth transition mechanism is initiated, gradually reducing the monitoring frequency to baseline levels and sending status recovery notifications to clinicians; all transfer processes are logged in detail, including the downgrade trigger time, recovery curves of each indicator, and possible causes; downgrade operations must be confirmed by a physician or executed automatically without manual intervention; the dynamic downgrade mechanism fully covers the entire lifecycle of arteriovenous fistula function, the assessment system is closed-loop, the smooth transition avoids data loss due to sudden frequency drops, ensuring the safety of the downgrade process, recovery notifications provide positive feedback, enhance patient treatment compliance, and downgrade logs support intervention effect evaluation, providing data support for clinical protocol optimization.

[0137] Specifically, the ranges for variation threshold, rate threshold, and perfusion volume threshold are configured based on a dynamic threshold configuration strategy. This dynamic threshold configuration strategy includes:

[0138] Obtain the baseline values ​​of total hemoglobin concentration, baseline variation range of blood oxygen saturation, and baseline perfusion response rate of patients under basic physiological conditions in order to set individualized initial thresholds;

[0139] During continuous monitoring, the initial threshold is offset and corrected according to the changing trend of the patient's physiological data. If the heart rate or blood pressure changes exceed the preset fluctuation range, the corresponding rate threshold and perfusion threshold range are adjusted.

[0140] If the baseline variation range of blood oxygen saturation drifts, the variation threshold is adjusted synchronously to ensure that the initial threshold matches the patient's current physiological state.

[0141] Significant individual differences exist in physiological parameters such as hematocrit, baseline blood pressure, and autonomic nervous system regulation among different patients, leading to varying hemodynamic baseline levels in their arteriovenous fistulas. Using a uniform fixed threshold can cause continuous false alarms or missed alarms for some patients, reducing the reliability of the system. By collecting data from patients during their physiologically stable periods to set initial thresholds, the assessment starting point can be ensured to match the individual's true state, providing a reliable reference benchmark for subsequent dynamic adjustments and forming the basis for personalized and accurate monitoring. The system has a built-in automatic resting state recognition algorithm. This algorithm monitors heart rate variability and blood pressure variability for 30 consecutive minutes, and if these values ​​are all below the preset thresholds (e.g., HRV SDNN < 20ms, blood pressure variability < 5%, and no body movement signal during the period), it is determined to be in the baseline physiological state. In this state, data is continuously collected for 24 hours. For total hemoglobin concentration data, the median and the 5th / 95th percentile are calculated. The 5th percentile is set as the lower limit of the perfusion threshold, and the 95th percentile is set as the upper limit of the perfusion threshold.

[0142] For the coefficient of variation of blood oxygen saturation, its mean and standard deviation are calculated, and the mean minus twice the standard deviation is set as the lower limit of the variation threshold. For the perfusion response rate, its average response time under physiological stimulation is calculated, and 1.5 times the average time is set as the upper limit of the rate threshold. All thresholds are stored in the patient configuration file, marked as individual baseline thresholds, and accompanied by data collection timestamps and quality scores. Threshold settings based on individual data eliminate systematic bias caused by individual differences, achieving high accuracy on the first use. 24-hour diurnal data collection ensures that the thresholds reflect the range of physiological rhythm fluctuations, avoiding misinterpretation of normal fluctuations at night or in the early morning. The percentile and standard deviation methods do not require assumptions about data distribution, adapting to non-normal physiological data from different populations. The quality scoring mechanism ensures the reliability of the thresholds, prompting for re-collection when the score is too low.

[0143] The patient's physiological state is not static. Systemic changes such as dehydration, blood pressure fluctuations, and arrhythmias can directly affect the hemodynamics of the arteriovenous fistula. Static thresholds cannot adapt to such changes. For example, when blood pressure rises acutely, if the perfusion threshold is not adjusted in time, the normal compensatory increase in blood flow may be misjudged as abnormal. Real-time tracking of heart rate and blood pressure trends and dynamic correction of thresholds can maintain the matching between the assessment criteria and the current physiological state, ensuring that the accuracy of the judgment is not affected by the systemic state. The system uses a dual sliding window to monitor physiological data trends. The short-term window, such as 5 minutes, detects acute changes, and the long-term window, such as 1 hour, assesses steady-state drift. When the rate of change of systolic blood pressure is >10 mmHg / 5 min or the change of heart rate is >10 beats / min within the short-term window, it is judged as an acute stress event, triggering immediate threshold correction. The correction algorithm uses a linear offset model: new threshold = initial threshold × (1 + physiological parameter change × correction coefficient). The correction coefficient is obtained through regression analysis of historical data.

[0144] For chronic drift, i.e., long-term mean shift >15%, a step-by-step adjustment is adopted, adjusting the threshold by 5% every 24 hours until a new match is achieved. Boundary protection is set during the adjustment process, and the corrected threshold must not exceed the physiological safety range, such as the perfusion volume upper limit not exceeding 200% of the baseline value, to prevent over-correction leading to missed diagnoses. All correction operations are logged, including the original threshold, correction amount, correction reason, and post-correction effect evaluation. The dual-window mechanism distinguishes between acute and chronic changes, making the correction strategy more targeted. The linear drift model responds quickly, and the step-by-step adjustment avoids oscillations during chronic drift correction, maintaining monitoring stability. The boundary protection mechanism ensures that the correction does not exceed the physiologically reasonable range, and the correction log provides real-world data for subsequent model optimization, supporting continuous iteration of the correction coefficients.

[0145] Oxygen saturation variability is influenced by multiple factors, including the patient's breathing pattern, hematocrit, and limb movement on the side of the arteriovenous fistula. Its baseline will slowly drift over time. If the variability threshold remains fixed, false alarms or missed alarms may persist after the baseline drifts. For example, after correction of anemia, hemoglobin concentration increases, naturally increasing the variability of oxygen saturation. The original upper limit of the threshold may be too low, continuously triggering false positives. Real-time monitoring of the baseline variability drift and synchronous adjustment of the threshold ensure that the judgment criteria evolve in sync with current physiological characteristics. The system accumulates and tests the coefficient of variation of oxygen saturation, continuously calculating the cumulative deviation between the coefficient of variation and the initial baseline. When the cumulative deviation exceeds the preset drift threshold and persists for more than 6 hours, it is determined as a baseline drift event. After drift confirmation, the variability threshold is updated using a proportional adjustment method: updated variability threshold = original threshold × (current median coefficient of variation / initial median coefficient of variation), ensuring that the threshold is adjusted proportionally to the drift amplitude.

[0146] The adjustment process incorporates gradient constraints, limiting single adjustments to no more than 20% to prevent drastic threshold fluctuations caused by drift misjudgments. The adjustment cycle is limited to once every 12 hours to avoid introducing noise through frequent adjustments. The upper and lower thresholds are adjusted synchronously to maintain a constant threshold width. After adjustment, the current median coefficient of variation is relabeled as a new benchmark value for subsequent drift detection. The cumulative sum test is sensitive to slow drifts but robust to instantaneous fluctuations, resulting in high drift detection accuracy. The proportional adjustment method maintains the relative position of the thresholds, ensuring consistent judgments after adjustment. Gradient constraints and cycle limits prevent over-adjustment, resulting in strong system stability. Synchronous adjustment of the upper and lower limits avoids threshold width distortion and maintains a reasonable definition of the normal fluctuation range.

[0147] The identification and diagnostic module is used to automatically identify arteriovenous fistula characteristics and predict changes in the fistula.

[0148] Specifically, such as Figure 2 As shown, the characteristics of arteriovenous fistulas are automatically identified and changes in the fistulas are predicted using a prediction strategy. The prediction strategy includes:

[0149] Physiological features are extracted from physiological data, including changes in heart rate and blood pressure, and the weights of physiological features and arteriovenous fistula features on changes in arteriovenous fistula are configured.

[0150] The changes in the arteriovenous fistula (AVF) are predicted by comprehensively considering physiological characteristics and AVF characteristics. The AVF change value is a weighted fusion result based on changes in blood oxygen saturation, hemoglobin concentration, heart rate variability, and blood pressure fluctuation.

[0151] Configure the baseline and extreme values ​​of arteriovenous fistula (AVF) changes, and compare the AVF changes with the baseline and extreme values ​​to obtain the AVF status.

[0152] If the change value of the arteriovenous fistula is greater than or equal to the extreme value of the change value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is abnormal and the state of the arteriovenous fistula is at the baseline value of the arteriovenous fistula.

[0153] If the change value of the arteriovenous fistula is less than the extreme value of the change value of the arteriovenous fistula but greater than the base value of the change value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is abnormal and the arteriovenous fistula is in a high value state.

[0154] If the change value of the arteriovenous fistula is less than or equal to the base value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is normal and the state of the arteriovenous fistula is at the extreme value of the arteriovenous fistula.

[0155] Fluctuations in heart rate and blood pressure directly affect the arteriovenous fistula (AVF) by altering systemic hemodynamic pressure. However, the degree of this impact varies among individuals. Fixed weights cannot adapt to changes in the patient's cardiovascular function, such as arrhythmias and decreased blood pressure regulation, leading to inaccurate predictions. Dynamically extracting physiological features and assigning weights can quantify the real-time impact of systemic status on the AVF, improving the individual adaptability of the prediction model. ECG and blood pressure waveform data are acquired in real-time from physiological monitoring devices. Time-domain analysis is used to extract the RMSSD (root mean square of the difference between adjacent normal heartbeats), and frequency-domain analysis is used to extract the LF / HF ratio (low-frequency to high-frequency power ratio) to quantify autonomic nervous system regulation. Blood pressure fluctuation features are used to extract the standard deviation and coefficient of variation of systolic and diastolic blood pressure, and step change events are identified.

[0156] The weight configuration employs an online learning mechanism. The initial weights are set based on prior clinical data, such as the highest weight for perfusion response speed. During continuous monitoring, the fluctuation patterns of various physiological characteristics before each change in the arteriovenous fistula status are recorded. By calculating the mutual information between the features and the changes in the arteriovenous fistula values, the weights are dynamically adjusted, giving higher weights to highly correlated features. The weight update cycle is set to once a day to avoid frequent fluctuations that could lead to predictive instability. Multi-dimensional physiological feature extraction comprehensively captures the cardiovascular regulatory state, improving the richness of predictive information. Dynamic online weight learning enables individualized adaptation of the predictive model, avoiding reliance on manual parameter tuning. The mutual information calculation is based on actual physiological and pathological correlations, and the weight adjustment is interpretable, meeting the requirements of evidence-based medicine. The standardized processing of features and weights provides a unified dimensional input for subsequent fusion calculations.

[0157] Relying solely on blood oxygen saturation or hemoglobin concentration can only reflect local oxygenation status and cannot comprehensively assess hemodynamics, oxygen supply and demand balance, and dynamic reserve capacity. Weighted fusion compresses multimodal information into a single change value, facilitating threshold comparison and trend tracking. This value needs to comprehensively reflect the overall deterioration trend of arteriovenous fistula function, rather than instantaneous fluctuations. Therefore, the fusion process must balance sensitivity and stability. A four-input weighted fusion model is established, with inputs including: blood oxygen saturation (normalized to 0-1), hemoglobin concentration change rate (standardized to individual baseline), heart rate variability characteristics (corresponding to RMSSD normalized value), and blood pressure variability characteristics (corresponding to coefficient of variation). Each input is multiplied by its dynamic weight and summed to obtain the original arteriovenous fistula change value. Subsequently, smoothing is performed, using an exponentially weighted moving average to filter the original value. The attenuation factor is adjusted according to the prediction time window, with a smaller value for short-term predictions and a larger value for long-term predictions.

[0158] The exponentially weighted moving average (TWMA) result is used as the final arteriovenous fistula (AVF) change value, while its acceleration is calculated as a trend strength indicator. The fusion calculation is performed every 30 seconds in the real-time operating system to ensure timely prediction. The model output includes an additional confidence score, which is dynamically calculated based on the input signal quality and historical prediction accuracy. Multi-feature fusion achieves information complementarity, and abnormalities in a single indicator will not dominate the prediction result, reducing false positives. The TWMA smooths out high-frequency interference such as heartbeat and respiration, resulting in more stable change values ​​and strong trend tracking. The confidence score provides clinicians with a reference for the reliability of the prediction, supports decision-making priority ranking, and balances response speed and stability between calculation frequency and filtering parameters to meet the needs of continuous monitoring.

[0159] The baseline and dynamic range of arteriovenous fistula (AVF) changes vary significantly among different patients. Static thresholds cause some patients to remain in an abnormal state for a long time or fail to trigger warnings. Dynamically configuring individualized baseline and extreme values ​​allows the prediction system to adapt to individual physiological ranges and achieve accurate risk stratification. Baseline value configuration uses a statistical method based on the stable period of individual historical data. The system collects data from the past 30 days of patients, filters out periods marked as non-severe states, and calculates the 90th percentile of AVF changes during these periods as the baseline. Extreme value configuration uses a risk extrapolation method, selecting changes in previously severe extreme states and calculating their 10th percentile as the extreme value to ensure that the extreme value represents the true severe state. During continuous monitoring, the baseline and extreme values ​​are automatically updated monthly. If a patient has recently undergone intervention, the post-intervention data is statistically analyzed separately to avoid confounding.

[0160] The comparison logic employs a three-zone division: a change value ≤ baseline is considered normal, corresponding to the extreme value state of the arteriovenous fistula; a change value < baseline < extreme value is considered moderately abnormal, corresponding to the high value state of the arteriovenous fistula; and a change value ≥ extreme value is considered severely abnormal, corresponding to the baseline value state of the arteriovenous fistula. Each comparison is accompanied by trend analysis. If the change value continuously rises and approaches the extreme value, an early warning is triggered even if it has not been exceeded. Individualized baseline / extreme value configuration ensures that risk stratification is physiologically reasonable, avoiding over-monitoring or under-monitoring caused by a one-size-fits-all approach. The monthly update mechanism adapts to disease progression or the rehabilitation process after intervention, maintaining the timeliness of the thresholds. The extreme values ​​set by the risk extrapolation method have clear clinical significance and are highly correlated with serious adverse events. The early warning function provides a buffer period, making clinical intervention more proactive.

[0161] The interactive display module is used to visualize the presence, severity, and changes of arteriovenous fistulas, and to generate diagnostic results and suggestions.

[0162] Clinically, it is essential to intuitively grasp the status of multiple parameters such as fistula perfusion, oxygenation, variability, and response to avoid information overload leading to misdiagnosis. Layered display enables rapid identification of key information, while detailed data can be retrieved as needed. The main view adopts a dashboard design, with a central three-color status light (red / yellow / green) indicating the severity, and four surrounding quadrants displaying perfusion volume, oxygenation trend, variability characteristics, response speed values, and trend arrows in real time. The topology map mode is based on a vascular diagram, using color coding to overlay the spatial distribution of perfusion volume and blood oxygen heterogeneity. The time series mode supports coaxial comparison of multi-parameter curves and event annotation. Views can be switched via gestures or shortcut keys, and the system remembers user preferences. The spatial topology map assists in puncture point selection and lesion localization. Time traceability facilitates the analysis of event correlations.

[0163] Medical documentation requires standardized and traceable diagnoses. Free text is prone to missing information, while structured formats facilitate EMR integration and quality control audits. Results use JSON templates with fields including: existence boolean value, severity enumeration, duration, trend prediction objects (including 6h / 24h / 72h risk probabilities), a list of key feature anomalies, and confidence scores. Version records are automatically created when severity is upgraded or downgraded, including the reason for the change and a data snapshot. This ensures diagnostic completeness, avoids underreporting, and supports efficacy evaluation and accountability through version management.

[0164] Different severity levels require differentiated treatment recommendations, including observation, follow-up examination, or emergency intervention. Tiered push notifications ensure priority delivery of urgent information, while individualized recommendations improve medical standardization and implementation efficiency. The recommendation database is labeled with applicable status, priority, and evidence level. During generation, a candidate set is selected based on severity, filtered for abnormal characteristics, and adjusted for urgency based on trend prediction (risk adjustment within 72 hours). Recommendations are tiered into three levels: red (emergency status, alerted via pop-up and SMS), yellow (important status, relevant information pushed), and green (general status, prompting relevant information). Content covers examination items, medication adjustments, and behavioral guidance, each with an evidence-based support button. This approach combines guidelines with individualization, enhancing clinical applicability. Tiered recommendations ensure timely response to urgent situations, and evidence-based presentation strengthens doctor-patient communication and adherence.

[0165] The feedback adjustment module is used to dynamically adjust the fistula assessment and prediction strategies based on the actual diagnostic results.

[0166] The system's predictions require clinical validation. Multiple sources of results, including imaging, physician assessments, and laboratory indicators, can form the basis for error analysis. Physicians can confirm / modify the system's diagnosis and indicate the reasons on the interface. Laboratory data exceeding the standard is automatically labeled. Results are graded according to their reliability and stored in the feedback database, linked to the original diagnostic records. Reliability labels avoid interference from low-quality data, and physician feedback reveals defect patterns.

[0167] Simple comparisons cannot pinpoint defects, but cluster analysis can identify patterns of misjudgment and make targeted corrections. The comparison engine calculates the difference in existence and severity, clustering errors into four categories: false positives, false negatives, underestimation, and overestimation. It analyzes the mean of each category's features to identify common patterns, such as false positives often being due to overly sensitive variation thresholds. This accurately locates the defective link, avoids blind parameter tuning, and provides a clear direction for strategy correction.

[0168] Misjudgments require local correction rather than global reset. Adjusting weights after identifying contradictory rule points can maintain overall stability. For false positives / negatives, identify triggering rules and calculate their true positive capture rate and false trigger rate. If the false trigger rate is >20%, mark it as a contradictory point and adjust the input feature weight of the rule step by step. Local correction avoids performance fluctuations, and weight adjustment is smoother than threshold adjustment.

[0169] False positives due to trend misjudgment require model fine-tuning. Online updates can quickly adapt to error patterns, and a rollback mechanism prevents performance degradation. For data with misjudgments of severity, SGD is used to fine-tune the model weights online with a learning rate of 0.01, updating every 50 data points. After updating, validation set testing is performed. If the false negative rate increases by more than 5%, the system automatically rolls back to the previous version. Weight version libraries are archived, and performance and applicable populations are labeled. The system can quickly adapt to new data, with small-step updates ensuring stability. The rollback mechanism reduces medical risks, and multiple versions support individualized strategies.

[0170] After the strategy is adjusted, the long-term effect needs to be continuously verified. The full-cycle evaluation provides objective evidence. Sensitivity, specificity, and false positive / negative rate are automatically calculated weekly and statistically analyzed by subgroup. The performance trend is displayed through control charts. A decrease in specificity of >5% for 4 consecutive weeks triggers a review. The effect is continuously quantified. Subgroup statistics reveal problems in specific populations. Early warnings prevent performance deterioration. Monitoring data guides R&D improvements. Error point heatmaps locate the next round of adjustment targets.

[0171] Example 2:

[0172] This embodiment provides a flowchart of a method for detecting arteriovenous fistula safety, as follows: Figure 3 As shown, a method for detecting arteriovenous fistula safety includes:

[0173] Acquire image data of the arteriovenous fistula area and monitor the patient's physiological data in real time;

[0174] Image data is processed to extract arteriovenous fistula features and determine the presence and severity of the fistula;

[0175] Automatically identify arteriovenous fistula characteristics and predict changes in the fistula;

[0176] Visualize the presence, severity, and changes of arteriovenous fistulas, and generate diagnostic results and suggestions;

[0177] The strategies for identifying and predicting arteriovenous fistulas are dynamically adjusted based on the actual diagnostic results.

[0178] For details on a method for detecting the safety of arteriovenous fistulas, please refer to a system for detecting the safety of arteriovenous fistulas; these details will not be repeated here.

Claims

1. A detection system for arteriovenous fistula safety, characterized in that: include: Image acquisition module, image processing module, recognition and diagnosis module, interactive display module, and feedback and adjustment module; The image acquisition module is used to acquire image data of the arteriovenous fistula area and monitor the patient's physiological data in real time; The image processing module is used to process image data, extract fistula features, and determine the presence and severity of the fistula. The identification and diagnosis module is used to automatically identify arteriovenous fistula characteristics and predict changes in the arteriovenous fistula; The interactive display module is used to visualize the existence, severity, and changes of arteriovenous fistulas; The feedback adjustment module is used to dynamically adjust the fistula judgment and prediction strategies based on the actual diagnostic results.

2. The arteriovenous fistula safety detection system as described in claim 1, characterized in that: Near-infrared light is emitted into the fistula area through near-infrared spectral imaging, and the light signal returned after being scattered and absorbed by the fistula tissue is received. By analyzing the differences in the absorption characteristics of near-infrared light of different wavelengths, image data of the fistula area is obtained. Blood flow velocity is obtained based on the image data of the fistula area, and physiological data of the patient is monitored in real time in conjunction with physiological monitoring equipment. The image data and physiological data are stored synchronously.

3. The arteriovenous fistula safety detection system as described in claim 2, characterized in that: The changes in blood oxygen saturation and hemoglobin concentration are obtained by monitoring the absorption characteristics of near-infrared light of different wavelengths in the fistula area. The wavelength range and detection sensitivity of the near-infrared spectral imaging source are adjusted. The near-infrared spectral imaging is equipped with a dual-wavelength or multi-wavelength light source. By detecting the differences in the absorption characteristics of near-infrared light in the fistula tissue, the relative concentration changes of oxyhemoglobin and deoxyhemoglobin are separated and calculated to obtain the changes in blood oxygen saturation and hemoglobin concentration.

4. The arteriovenous fistula safety detection system as described in claim 3, characterized in that: Arterial fistula features are extracted based on changes in blood oxygen saturation and hemoglobin concentration in the fistula region. These features include tissue perfusion volume, trends in oxygenation status, variations in blood oxygen saturation, and perfusion response rate. The presence and severity of the fistula are determined based on these features. The variation characteristics of blood oxygen saturation are obtained by quantifying the changes in hemoglobin concentration, reflecting the periodic fluctuation pattern of oxyhemoglobin in the fistula area.

5. The arteriovenous fistula safety detection system as described in claim 4, characterized in that: The variation characteristics of blood oxygen saturation at adjacent time points are compared to obtain the variation difference value; Configure a variation threshold, compare the variation difference with the variation threshold to determine the degree of abnormality of the variation characteristics of blood oxygen saturation. If the variation difference is greater than or equal to the variation threshold, it is judged that the variation characteristics are normal. If the difference in variation is less than the variation threshold, it is judged as an abnormal variation feature; The concentration change rate is obtained by time series analysis of the hemoglobin concentration change value. A rate threshold is set, and the concentration change rate is compared with the rate threshold. If the concentration change rate is greater than or equal to the rate threshold, the perfusion response rate is judged to be normal. If the rate of concentration change is less than the rate threshold, the perfusion response rate is considered abnormal.

6. The arteriovenous fistula safety detection system as described in claim 5, characterized in that: Detect the total hemoglobin concentration in the arteriovenous fistula area and configure the perfusion volume threshold range; The total concentration of hemoglobin is compared with the perfusion volume threshold range. If the total concentration of hemoglobin is within the perfusion volume threshold range, the tissue perfusion volume is judged to be normal. If the total concentration of hemoglobin exceeds the perfusion threshold range, the tissue perfusion is considered abnormal. Analyze the directionality of changes in the ratio of oxyhemoglobin to deoxyhemoglobin concentrations; Configure trend judgment rules. If the direction of the ratio change conforms to the preset normal fluctuation pattern, the trend of oxygenation status change is judged to be normal. If the direction of the ratio change deviates from the preset normal fluctuation pattern, the trend of oxygenation status change is judged to be abnormal.

7. The arteriovenous fistula safety detection system as described in claim 6, characterized in that: The presence and severity of an arteriovenous fistula are determined based on its characteristics using an arteriovenous fistula assessment strategy, which includes: If the tissue perfusion is normal and the trend of oxygenation status is normal, then it is determined that the arteriovenous fistula does not exist; If the tissue perfusion volume is abnormal or the trend of changes in oxygenation status is abnormal, then an arteriovenous fistula is determined to exist. When an arteriovenous fistula exists, its severity is assessed. If the variation characteristics of blood oxygen saturation are abnormal and the perfusion response rate is abnormal, the severity of the fistula is assessed as the extreme value of severity. If the variation characteristics of blood oxygen saturation are abnormal or the perfusion response rate is abnormal, the severity of the arteriovenous fistula is judged by the severe baseline value. If the variation characteristics of blood oxygen saturation are normal and the perfusion response rate is normal, then the severity of the arteriovenous fistula is judged to be non-severe.

8. The arteriovenous fistula safety detection system as described in claim 7, characterized in that: The prediction strategy automatically identifies arteriovenous fistula characteristics and predicts changes in the fistula. The prediction strategy includes: Physiological features are extracted from physiological data, including changes in heart rate and blood pressure, and weights of physiological features and arteriovenous fistula features on changes in arteriovenous fistula are configured. The changes in the arteriovenous fistula are predicted by comprehensively considering physiological characteristics and fistula characteristics, and the changes in the fistula are obtained. The changes in the fistula are a weighted fusion result based on changes in blood oxygen saturation, hemoglobin concentration, heart rate variability, and blood pressure fluctuation. Configure the baseline and extreme values ​​of arteriovenous fistula (AVF) changes, and compare the AVF changes with the baseline and extreme values ​​to obtain the AVF status. If the change value of the arteriovenous fistula is greater than or equal to the extreme value of the change value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is abnormal and the state of the arteriovenous fistula is at the baseline value of the arteriovenous fistula. If the change value of the arteriovenous fistula is less than the extreme value of the change value of the arteriovenous fistula but greater than the base value of the change value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is abnormal and the arteriovenous fistula is in a high value state. If the change value of the arteriovenous fistula is less than or equal to the base value of the arteriovenous fistula, it indicates that the change of the arteriovenous fistula is normal and the state of the arteriovenous fistula is at the extreme value of the arteriovenous fistula.

9. The arteriovenous fistula safety detection system as described in claim 8, characterized in that: The variation threshold, rate threshold, and perfusion volume threshold range are configured based on a dynamic threshold configuration strategy, which includes: Obtain baseline values ​​for total hemoglobin concentration, baseline variation range for blood oxygen saturation, and baseline perfusion response rate under basal physiological conditions in order to set individualized initial thresholds; During continuous monitoring, the initial threshold is offset and corrected according to the changing trend of physiological data. If the heart rate or blood pressure changes exceed the preset fluctuation range, the corresponding rate threshold and perfusion threshold range are adjusted. If the baseline variation range of blood oxygen saturation drifts, the variation threshold is adjusted synchronously to ensure that the initial threshold matches the current physiological state.

10. A method for detecting arteriovenous fistula safety, implemented based on the arteriovenous fistula safety detection system according to any one of claims 1-9, characterized in that: include: Acquire image data of the arteriovenous fistula area and monitor the patient's physiological data in real time; Image data is processed to extract arteriovenous fistula features and determine the presence and severity of the fistula; Automatically identify arteriovenous fistula characteristics and predict changes in the fistula; Visualize the presence, severity, and changes of arteriovenous fistulas; The strategies for identifying and predicting arteriovenous fistulas are dynamically adjusted based on the actual diagnostic results.

Images