Intraoperative brain oxygen saturation image dynamic monitoring and early warning system and method for elderly patients

Abstract

The present application relates to the technical field of medical image monitoring, in particular to an intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system and method for elderly patients, comprising a bilateral cerebral oxygen acquisition module, a cerebral oxygen waveform image construction module, a cerebral oxygen image feature extraction module, a hierarchical early warning module and a multi-parameter linkage display module, through the bilateral cerebral oxygen saturation time series signals output by the bilateral near-infrared cerebral oxygen saturation sensors are synthesized into a double-channel cerebral oxygen waveform image aligned with the same time window start and end time and sharing the amplitude scale, three types of geometric morphological features of bilateral cerebral oxygen slope, bilateral cerebral oxygen asymmetry index and relative baseline drop amplitude are uniformly extracted in the image domain, three types of high-risk mode hierarchical triggering of unilateral cerebral oxygen threshold persistence, relative baseline drop and bilateral asymmetry are driven, and the invasive arterial pressure, cardiac output, anesthesia depth index waveform and hemodynamic event characteristic root cause pairing display are jointly presented.

Description

Technical Field

[0001] This invention relates to the field of medical imaging monitoring technology, specifically to a system and method for dynamic monitoring and early warning of intraoperative cerebral oxygen saturation in elderly patients. Background Technology

[0002] Elderly patients undergoing general anesthesia and cardiopulmonary bypass during cardiac surgery have a significantly higher risk of postoperative cognitive impairment than the general population, with one of the key contributing factors being the unbalanced decrease in intraoperative cerebral oxygen saturation. Continuous, non-invasive monitoring of bilateral prefrontal cortex oxygen saturation using near-infrared spectroscopy during surgery has become a crucial clinical step in reducing postoperative cognitive impairment. Bilateral cerebral oxygen saturation monitoring devices based on near-infrared spectroscopy are now relatively mature. Clinically, a unilateral absolute oxygen saturation value below 50%, a relative baseline decrease exceeding 20%, or a persistent bilateral asymmetry are typically used as thresholds for diagnosing cerebral oxygen imbalance.

[0003] Chinese patent application CN111481174A discloses an anesthesia and consciousness depth monitoring system and method. The system integrates an anesthesia depth sensor and a cerebral oxygen saturation sensor on a flexible substrate. The cerebral oxygen saturation sensor includes a left-sided cerebral oxygen saturation acquisition group for detecting the left side of the forehead and a right-sided cerebral oxygen saturation acquisition group for detecting the right side of the forehead. A central processing unit within the main controller sends synchronous acquisition signals to an EEG signal acquisition module and an infrared transceiver module, performing synchronous data processing on the acquired bilateral EEG data and bilateral cerebral oxygen saturation data. A joint data analysis module weights and fuses the BIS values ​​from both sides with the cerebral oxygen saturation values ​​according to preset weighting coefficients to obtain a neurovascular coupling index. The left and right BIS indices, left and right cerebral oxygen saturation, right cerebral oxygen saturation, and neurovascular coupling index are simultaneously presented on a display output module in numerical or trend curve form. A discrimination module sets thresholds for the EEG, cerebral oxygen saturation, and neurovascular coupling index; when the corresponding threshold is exceeded, an alarm unit issues a warning signal.

[0004] However, this scheme, in terms of the paradigm for monitoring cerebral oxygen imbalance, still belongs to the paradigm of reducing the time-series signal of bilateral cerebral oxygen saturation to two independent one-dimensional numerical streams for separate processing on the left and right sides. The left and right sides of cerebral oxygen saturation are only presented on the display interface as two independent trend curves. The geometric differences between the waveforms on the left and right sides (including the divergence of the left and right slope directions, the synergy of the left and right waveform envelopes relative to the slope, and the evolution rate of the bilateral asymmetry index itself) are smoothed out in one dimension during the data processing stage. However, this waveform geometric information is precisely the early deformation fingerprint of the differences in independent autoregulation of blood flow in the two cerebral hemispheres before the threshold is crossed. When systemic hypotension, decreased cardiac output, or excessive depth of anesthesia causes a decrease in cerebral perfusion pressure, the cerebral hemisphere with weaker autoregulation ability is the first to experience the collapse of the microvascular autoregulation threshold, resulting in bilateral waveform The divergence of the slope and the gradual increase of the asymmetry index, this bilateral asymmetric collapse process has obvious time-progressive characteristics. First, the slope direction diverges, then the asymmetry index increases significantly, and finally the absolute value on one side falls below the threshold. However, the numerical threshold alarm of this scheme only identifies the terminal event of this gradual process. After the warning is triggered, the physician still needs to manually compare multiple independent curves such as arterial pressure, cardiac output, and anesthesia depth index to determine the hemodynamic root cause of the decline. The lag in the warning time point and the time-consuming manual judgment of the root cause are superimposed. Elderly patients have a weaker ability to regulate cerebral blood flow and their tolerance window for cerebral oxygen imbalance is much shorter than that of general patients. The above lag causes the intraoperative cerebral oxygen management of elderly patients to always be in a passive response mode rather than an active warning mode, which is significantly different from the clinical goal of reducing the occurrence of postoperative cognitive dysfunction. Summary of the Invention

[0005] Addressing the core bottlenecks in existing technologies for intraoperative cerebral oxygen saturation monitoring in elderly patients—namely, the loss of bilateral waveform geometric morphology information after dimensionality reduction of bilateral cerebral oxygen time-series signals to one-dimensional numerical stream processing, the lag of warning points behind numerical threshold crossing events, and the lack of automated correlation analysis of hemodynamic root causes of cerebral oxygen decline—this invention provides a dynamic monitoring and early warning system and method for intraoperative cerebral oxygen saturation in elderly patients. This system synthesizes bilateral cerebral oxygen saturation time-series signals output from bilateral near-infrared cerebral oxygen saturation sensors into dual-channel cerebral oxygen waveform images aligned with the same start and end times and sharing an amplitude scale, using a sliding window. It then extracts three types of geometric morphological features in the image domain: bilateral cerebral oxygen slope, bilateral cerebral oxygen asymmetry index, and relative baseline decline amplitude. These features drive the graded triggering of three high-risk modes and are presented in a paired hemodynamic waveform display. Without adding additional invasive measurement methods, this system achieves automated correlation assistance for advancing the warning point of intraoperative cerebral oxygen imbalance and identifying the root causes of decline in elderly patients from the principle level of image domain geometric morphology analysis.

[0006] The technical solution of this invention is: a dynamic monitoring and early warning system for intraoperative cerebral oxygen saturation images in elderly patients, comprising a bilateral cerebral oxygen acquisition module, a cerebral oxygen waveform image construction module, a cerebral oxygen image feature extraction module, a graded early warning module, and a multi-parameter linkage display module. The bilateral cerebral oxygen acquisition module includes bilateral near-infrared cerebral oxygen saturation sensors respectively attached to the left and right frontal lobes of the patient, and the bilateral near-infrared cerebral oxygen saturation sensors output bilateral cerebral oxygen saturation time-series signals. The cerebral oxygen waveform image construction module segments the bilateral cerebral oxygen saturation time-series signals according to a sliding window to construct a dual-channel cerebral oxygen waveform image. The dual-channel cerebral oxygen waveform image is composed of a left cerebral oxygen waveform channel and a right cerebral oxygen waveform channel aligned side-by-side according to the same start and end times of a time window. The cerebral oxygen image feature extraction module extracts cerebral oxygen state sensitive features from the dual-channel cerebral oxygen waveform image. The cerebral oxygen state sensitive features include bilateral cerebral oxygen slope, bilateral cerebral oxygen asymmetry index, and relative baseline decrease magnitude. The graded early warning module triggers and outputs corresponding graded early warning signals based on the brain oxygenation status sensitivity characteristics for three high-risk patterns: unilateral persistent brain oxygen threshold, relative baseline decline, and bilateral asymmetric pattern. The multi-parameter linkage display module aligns the dual-channel brain oxygen waveform image with the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform on the anesthesia workstation's integrated display interface, and highlights the window segment in the dual-channel brain oxygen waveform image that triggers the early warning when the graded early warning signal is triggered.

[0007] This invention also provides a method for dynamic monitoring and early warning of intraoperative cerebral oxygen saturation images in elderly patients, comprising the following steps: synchronously acquiring bilateral cerebral oxygen saturation time-series signals by using bilateral near-infrared cerebral oxygen saturation sensors respectively attached to the left and right frontal lobes of the patient; segmenting the bilateral cerebral oxygen saturation time-series signals according to a sliding window to construct a dual-channel cerebral oxygen waveform image, wherein the dual-channel cerebral oxygen waveform image is composed of a left cerebral oxygen waveform channel and a right cerebral oxygen waveform channel aligned side-by-side according to the same start and end times of a time window; extracting sensitive features of cerebral oxygen status from the dual-channel cerebral oxygen waveform image, wherein the cerebral oxygen status... The state-sensitive features include bilateral cerebral oxygenation slope, bilateral cerebral oxygenation asymmetry index, and relative baseline decrease. Based on the aforementioned cerebral oxygenation state-sensitive features, three high-risk patterns—unilateral cerebral oxygenation threshold persistence, relative baseline decrease, and bilateral asymmetry—are triggered and corresponding graded warning signals are output. The dual-channel cerebral oxygenation waveform image is combined with the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform on the integrated display interface of the anesthesia workstation, aligned with the time axis, and the window segment in the dual-channel cerebral oxygenation waveform image that triggers the warning is highlighted when the graded warning signal is triggered.

[0008] The beneficial effects of this invention are as follows: First, by synthesizing the bilateral brain oxygen saturation time-series signals into a dual-channel brain oxygen waveform image that is aligned with the same start and end times of the time window and shares the same amplitude scale, the geometric slope difference between the left and right waveforms is explicitly preserved as a visual feature in the image domain. Compared with the scheme that only presents the left and right brain oxygen saturation as two independent trend curves on the same screen, this invention preserves the relative geometric morphology information between the two waveforms during the data processing stage. The mechanism is that the microvascular threshold collapse process caused by the difference in the independent blood flow autonomous regulation of the two cerebral hemispheres is first manifested as the divergence of the slope direction of the two waveforms. The parallel alignment of the two-dimensional dual-channel images preserves the image domain geometric fingerprint of this slope divergence, avoiding the smooth discarding of this fingerprint under the one-dimensional numerical flow processing paradigm. Second, the bilateral short-window slope and bilateral long-window slope of brain oxygen are extracted from dual-channel brain oxygen waveform images using a dual-time-scale slope feature construction method. The sign consistency of the two slopes is used to distinguish between acute brain oxygen decline events and sustained slow decline events. The mechanism is that the slope of a single time scale is contaminated by high-frequency physiological fluctuations such as heart rate coupling and respiratory coupling, and cannot separate instantaneous hemodynamic events from slow brain oxygen consumption imbalance. The parallel calculation of dual time scales and the sign consistency judgment naturally achieve the separation of acute and slow events in the frequency domain, which is an algorithm effect that cannot be equivalently obtained under a one-dimensional numerical flow path. Third, the change rate of the bilateral asymmetric index is further extracted from the dual-channel brain oxygen waveform image. The bilateral asymmetric warning can be triggered in advance based on the change rate before the bilateral brain oxygen asymmetric index crosses the bilateral asymmetric threshold. The mechanism is that in the gradual process of bilateral asymmetric collapse, the evolution rate of the asymmetric index itself reflects the collapse initiation earlier than the static value. The single-point threshold judgment loses this dynamic feature. However, this invention obtains the evolution rate by taking the time difference of the bilateral brain oxygen asymmetric index as an independent judgment dimension, which significantly advances the warning time point relative to the single-point threshold judgment. Fourth, the root cause matching submodule of the multi-parameter linkage display module simultaneously extracts hemodynamic event features synchronized with the trigger window from the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform when the graded warning signal is triggered. The mean arterial pressure decrease, cardiac output decrease, and anesthesia depth index change are displayed together with the graded warning signal using root cause association markers. The mechanism is that the clinical intervention strategy for cerebral oxygen imbalance needs to identify the category of hemodynamic root cause of the decline (such as hypotension, low cardiac output, or excessive anesthesia). Simply displaying cerebral oxygen abnormality itself is insufficient to guide precise intervention. Root cause matching makes the multi-parameter event feature synchronous matching algorithm explicit, upgrading from a data-driven black box to a physical causal gray box. Compared with the solution that only presents multiple curves on the same screen and requires physicians to make comprehensive judgments manually, this invention realizes automated assistance in identifying the root cause of cerebral oxygen decline.Fifth, the above four technical features generate a synergistic effect of 1+1>2: the parallel alignment of the dual-channel waveform images provides an image domain carrying structure for the dual-timescale slope features, and the dual-timescale slope features make the bilateral asymmetric exponential change rate emerge naturally as a new feature. The three types of geometric morphological features drive the three types of high-risk mode hierarchical triggering and then serve as the synchronous triggering window of the root cause pairing sub-module to constrain the extraction range of hemodynamic event features. The five modules form a deep coupling closed loop of acquisition → mapping → features → early warning → root cause, and the overall technical effect far exceeds the effect of the independent accumulation of each module. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the overall architecture of the intraoperative brain oxygen saturation image dynamic monitoring and early warning system for elderly patients according to the present invention.

[0010] Figure 2 This is a schematic diagram of the brain oxygenation image feature extraction module in the system of the present invention extracting the slope features of the two time scales and the rate of change of the bilateral asymmetric index.

[0011] Figure 3 This is a flowchart illustrating the workflow of the graded early warning module in the system of this invention for determining three types of high-risk modes.

[0012] Figure 4 This is a schematic diagram of the root cause matching submodule of the multi-parameter linkage display module in the system of the present invention.

[0013] Figure 5 This is a flowchart illustrating the overall steps of the present invention: a method for dynamic monitoring and early warning of intraoperative brain oxygen saturation in elderly patients. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the specific embodiments of this invention will be further described in detail below with reference to the accompanying drawings. These specific embodiments are only for explaining this invention and are not intended to limit the invention.

[0015] See Figure 1As shown, the intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients provided in this embodiment is deployed on the anesthesia workstation side, working in parallel with existing multi-parameter monitors and invasive hemodynamic monitors as an independent image-based cerebral oxygen monitoring and early warning subsystem. The system includes a bilateral cerebral oxygen acquisition module 1, a cerebral oxygen waveform image construction module 2, a cerebral oxygen image feature extraction module 3, a graded early warning module 4, and a multi-parameter linkage display module 5. The cerebral oxygen waveform image construction module 2, the cerebral oxygen image feature extraction module 3, the graded early warning module 4, and the multi-parameter linkage display module 5 are deployed in software within the processing unit of the anesthesia workstation. The bilateral near-infrared cerebral oxygen saturation sensors in the bilateral cerebral oxygen acquisition module 1 are attached to the patient's head via a medical adhesive patch using a reflective optical path structure. The implementation details of each module are described below.

[0016] The bilateral brain oxygen acquisition module 1 includes bilateral near-infrared brain oxygen saturation sensors that are respectively attached to the left and right frontal lobes of the patient. The bilateral near-infrared brain oxygen saturation sensors are respectively attached to the left frontal lobe FP1 and right frontal lobe FP2 positions of the patient. The distance between the light source and the photodetector of each bilateral near-infrared brain oxygen saturation sensor can be configured within the range of 30mm to 40mm. In this embodiment, 40mm is used. For example, 30mm, 33mm, 35mm, 38mm, and 40mm can all achieve effective penetration of near-infrared light in the prefrontal cortex and effective collection of reflected light. The lower limit of the distance of 30mm is based on the minimum effective optical path required for near-infrared light to penetrate the human prefrontal cortex to a depth of about 15mm. The upper limit of the distance of 40mm is based on the commonly used upper limit of existing medical reflective brain oxygenation light paths within the range that the skin contact pressure can withstand. The emission wavelengths are selected in two bands, 730nm and 810nm, to characterize the absorption characteristics of deoxyhemoglobin and oxyhemoglobin, respectively. The local brain oxygen saturation values ​​of the left and right prefrontal lobe regions are calculated according to the modified Lambert-Beer law. The near-infrared brain oxygen saturation sensor and its related optical path structure, analog front-end, and analog-to-digital conversion circuit used in this embodiment are all conventional implementations of non-invasive brain oxygen monitoring in medical applications, and will not be described in detail here. The bilateral near-infrared brain oxygen saturation sensor continuously acquires brain oxygen saturation at a sampling frequency of 2Hz and outputs bilateral brain oxygen saturation time-series signals; the bilateral brain oxygen saturation time-series signals include left and right brain oxygen saturation time-series signals, which are synchronously acquired according to a unified timestamp and uploaded to the anesthesia workstation via serial communication. In this embodiment, the bilateral near-infrared cerebral oxygen saturation sensors are calibrated preoperatively using the average bilateral cerebral oxygen saturation values ​​within a 60-second time window before anesthesia induction as the baseline values ​​for both sides, which are then written into the workstation storage area. During calibration, the patient is required to maintain calm breathing to avoid stress-induced fluctuations in cerebral oxygen levels that could contaminate the baseline values. After calibration, continuous monitoring mode is activated. The bilateral cerebral oxygen acquisition module 1 packages the original cerebral oxygen saturation values ​​and the acquisition timestamps into a timing signal packet, which is uploaded to the cerebral oxygen waveform image construction module 2 every 250ms. If either near-infrared cerebral oxygen saturation sensor experiences poor contact or detachment during surgery, the bilateral cerebral oxygen acquisition module 1 automatically outputs a flag to the cerebral oxygen waveform image construction module 2 to pause image construction on the affected side, preventing erroneous timing signals from being included in subsequent processing.

[0017] The brain oxygen waveform image construction module 2 receives bilateral brain oxygen saturation time-series signals from the bilateral brain oxygen acquisition module 1, and constructs a dual-channel brain oxygen waveform image by segmenting it according to a sliding window. In this embodiment, the width of the sliding window can be configured within the range of 60s to 300s, preferably 300s (i.e., a long sliding window), for example, 60s, 120s, 180s, 240s, and 300s are all acceptable; the lower limit of the sliding window width of 60s is based on the minimum resolvable window of the characteristic time constant (on the order of 30s to 120s) of the intraoperative acute cerebral oxygen depletion event, and the upper limit of the sliding window width of 300s is based on the reasonable regression window of the characteristic time constant (on the order of several minutes) of the chronic cerebral oxygen consumption imbalance; the overlap rate of adjacent sliding windows is 50%; the update cycle of the dual-channel cerebral oxygen waveform image can be configured within the range of 1s to 5s, preferably 1s, for example, 1s, 2s, 3s, 4s, and 5s are all acceptable; the lower limit of the update cycle of 1s is based on the lower limit of the real-time refresh of the anesthesiologist's visual perception, and the upper limit of the update cycle of 5s is based on the upper limit of the balance between power consumption and readability of the anesthesia workstation display refresh. At the start of each update cycle, the brain oxygen waveform image construction module 2 extracts the left and right brain oxygen saturation time-series signals from the most recent 300s time window and constructs a dual-channel brain oxygen waveform image as follows: The dual-channel brain oxygen waveform image is composed of a left brain oxygen waveform channel and a right brain oxygen waveform channel aligned side-by-side with the same start and end times of the time window. The left brain oxygen waveform channel is placed in the upper half of the image, and the right brain oxygen waveform channel is placed in the lower half of the image. They share the same start and end times of the time axis and the same amplitude scale. The lower limit of the amplitude scale is 40%, and the upper limit is 90%. The time axis direction is positive with the horizontal direction of the image as the positive direction, and the left side is the start time of the window, and the right side is the end time of the window. In the dual-channel brain oxygen waveform image constructed in this way, the geometric slope difference between the left and right brain oxygen waveform channels is explicitly preserved as a visual feature in the image domain. The relative slope of the left and right waveform envelopes and the bilateral height difference at any time can be geometrically measured in the image domain. The left and right brain oxygen waveform channels are aligned with the same time window to form the basic data structure of the so-called dual channels in this invention.

[0018] Specifically, the grayscale value of any pixel in the dual-channel brain oxygen waveform image Construct it according to the following formula:

[0019] ,

[0020] in: For dual-channel brain oxygen waveform images in channel Time axis pixel coordinates Amplitude axis pixel coordinates The grayscale value at that location is a scalar, and its range is [value range missing]. The unit is dimensionless and is constructed from this formula, representing the binarized waveform recording result of the dual-channel brain oxygen waveform image at this pixel position; Here, is the channel index, and is a scalar with a value range of . The unit is dimensionless, 0 represents the left brain oxygen waveform channel and 1 represents the right brain oxygen waveform channel, which is determined by the brain oxygen waveform image construction module 2 according to the channel allocation rules. The technical effect is to distinguish the left and right brain oxygen waveform channels by the channel index. Here are the pixel coordinates on the time axis, and are scalar integers with a value range of . The unit is pixels. The number of pixels in the horizontal direction of the image is set to 300, which is obtained by discretization mapping of the time axis of the sliding window. The technical effect is that the time dimension is carried by pixel coordinates. Here, represents the pixel coordinates of the amplitude axis, and is a scalar integer with a value range of . The unit is pixels. The number of pixels in the vertical direction of the image is set to 128 per channel, which is obtained by discretization mapping of the amplitude scale. The technical effect is that the amplitude dimension is carried by the pixel coordinates. For time Time Channel The corresponding brain oxygen saturation value is a scalar, and its range is [value missing]. The unit is %, determined by the corresponding channel of the bilateral cerebral oxygen acquisition module 1 in time. The real-time output provides a basis for the vertical axis of waveform recording. pixel coordinates on the time axis The corresponding actual time is a scalar, and its value range is... The unit is s. The start time of the current sliding window is a linear mapping from the start and end times of the sliding window to the pixel indices on the time axis. Confirmed, among which With pixel-time resolution, the technical effect lies in establishing a two-way mapping between pixel coordinates and actual time; The lower limit of the amplitude scale is 40%, and the unit is %. It is determined based on the clinical interpretation threshold of perioperative cerebral oxygen imbalance. Too high a value will lose the details of low-value waveforms, and too low a value will waste pixel resolution. The technical effect is to set the lower boundary of the waveform's vertical axis. The upper limit of the amplitude scale is , which is a scalar value of 90% and the unit is %. It is based on the physiological upper limit of cerebral oxygen saturation during the perioperative period. The technical effect is to set the upper boundary of the waveform's vertical axis. This is a rounding function that discretizes the continuous mapping result into integer pixel coordinates. The technical effect is to complete the mapping from continuous amplitudes to discrete pixels. (Right side of the equation) The function returns a dimensionless integer, which corresponds to the left side of the equation. The pixel units are consistent, and the construction result is... Value These are dimensionless grayscale values ​​with consistent dimensions.

[0021] To address potential intraoperative interference from electrocoagulation, non-physiological spikes caused by motion artifacts and poor sensor contact, the brain oxygen waveform image construction module 2 further includes an artifact gating submodule. This submodule performs artifact detection based on median absolute deviation on the bilateral brain oxygen saturation time-series signals to obtain artifact mask values. These mask values ​​are superimposed on the dual-channel brain oxygen waveform image to obtain a dual-channel brain oxygen waveform image with the artifact mask values. The regions corresponding to these mask values ​​are not included in the brain oxygen image feature extraction and are not considered in the triggering determination of the graded early warning module 4. In this embodiment, the artifact gating submodule performs the following calculations on the left and right brain oxygen saturation time-series signals in each update cycle:

[0022] ,

[0023] ,

[0024] in For channel The median absolute deviation of the brain oxygen saturation time-series signal within the current sliding window. For channel The median of a time-series signal. For channel In time The artifact mask value at the location (1 indicates an artifact, 0 indicates normal). The multiplier for artifact detection is set to 4.5. This is the median operator. In the dual-channel brain oxygen waveform image, the pixel column with a value equal to 1 is filled with pure white as a mask, and both the brain oxygen image feature extraction module 3 and the hierarchical early warning module 4 skip this pixel column. Artifact gating based on median absolute deviation is a conventional implementation in time-series signal processing in this field, and its parameter selection and algorithm details will not be elaborated further.

[0025] The brain oxygenation image feature extraction module 3 extracts brain oxygenation state sensitive features from the dual-channel brain oxygenation waveform image. These sensitive features include bilateral brain oxygenation slope, bilateral brain oxygenation asymmetry index, and the magnitude of the relative baseline decrease. The bilateral brain oxygenation slope includes bilateral short-window slope and bilateral long-window slope. In this embodiment, the bilateral short-window slope is obtained by regression analysis of the bilateral brain oxygen saturation time-series signal within a 60-second short sliding window, and the bilateral long-window slope is obtained by regression analysis of the bilateral brain oxygen saturation time-series signal within a 300-second long sliding window.

[0026] The rationale for the parallel computation of short and long windows across two time scales is that the intraoperative cerebral oxygen saturation time-series signal simultaneously carries two different types of imbalance processes. One is an acute drop in cerebral oxygen triggered by sudden hemodynamic events (such as a brief drop in blood pressure caused by pericardial traction during thoracotomy), with a characteristic time constant on the order of 30 to 120 seconds. The other is a chronic imbalance in cerebral oxygen consumption caused by excessive anesthesia, intraoperative blood loss, or insufficient cardiopulmonary perfusion, with a characteristic time constant on the order of several minutes to tens of minutes. The slope feature of a single time scale is severely contaminated by amplitude fluctuations coupled with heart rate (typically with a period of about 1 second) and amplitude fluctuations coupled with respiration (typically with a period of about 4 to 6 seconds), making it difficult to stably distinguish between the two types of imbalances. The cerebral oxygen image feature extraction module 3 uses a dual-time-scale slope feature construction method to separate the two types of imbalances into two independent image domain feature channels, which are then superimposed on the dual-channel cerebral oxygen waveform image.

[0027] The bilateral brain oxygen short window slope With the slope of the bilateral brain oxygenation long window Calculate using the following least squares regression formulas respectively:

[0028] ,

[0029] ,

[0030] in: For channel The bilateral brain oxygen short window slope is a scalar, with a value range of [value missing]. The unit is % / min, determined by a short sliding window. The temporal signals of bilateral brain oxygen saturation were obtained through least-squares linear regression. The technical effect lies in characterizing the channels. Rate of change of brain oxygen saturation on a short timescale of 60 s; For channel The bilateral brain oxygenation long window slope is a scalar, with a value range of [value missing]. The unit is % / min, determined by a long sliding window. The temporal signals of bilateral brain oxygen saturation were obtained through least-squares linear regression. The technical effect lies in characterizing the channels. Rate of change in brain oxygen saturation over a 300-second timescale; The definition is the same as the formula mentioned above; It is a short sliding window. The discrete index set of values ​​corresponds to a time window that goes back 60 seconds from the current update time, with the unit being seconds. It is determined by the brain oxygen waveform image construction module 2 in each update cycle. The technical effect is to limit the sample range of short window regression. It is a long sliding window. The discrete index set of values ​​corresponds to a time window that goes back 300 seconds from the current update time, with the unit being seconds. It is determined by the brain oxygen waveform image construction module 2 in each update cycle. The technical effect is to limit the sample range of long window regression. For the first one inside the window The time of each sampling point is a scalar, and its value range is the time interval of the corresponding window. The unit is seconds, and it is determined by the sampling time. The definition is the same as the formula mentioned above; For short sliding window The mean, in seconds; For long sliding window The mean, in seconds; For channel Within the short sliding window The mean, in % For channel Inside the long sliding window The mean, in % This is the summation operator; the summation range is shown in the indices below. The molecular weight on the right side of the equation is in units of s·%, and the denominator is in units of s². Dividing by s yields % / s, which in this embodiment is uniformly presented in % / min form and processed with a unit conversion factor of 60. The left side of the equation... and All are % / min, with consistent dimensions.

[0031] Furthermore, the graded early warning module 4 distinguishes between acute cerebral oxygen decline events and persistent slow decline events based on the sign consistency of the bilateral short-window slope and the bilateral long-window slope. The sign consistency index is defined as:

[0032] ,

[0033] in: For channel The dual-timescale slope sign consistency index is a scalar with a value range of [value range missing]. The unit is dimensionless and is obtained by multiplying the signs of the short window slope and the long window slope. The technical effect is to determine the same direction of rapid and slow imbalance with a single index. For symbolic functions, defined as when , when , when As defined by mathematical rules, the technical effect lies in discretizing continuous slope values ​​into sign and direction. , , The definition is the same as the formula mentioned above. When Equal to 1 and When it is negative, it is determined to be a continuous slow decline event; when When the value equals -1, it is considered a complex event of acute decline superimposed on chronic recovery (or reverse superposition), requiring further interpretation based on the bilateral asymmetric index; when... Equal to 1 and For negative More than 2 times At that time, it was determined to be an acute cerebral oxygen depletion event. Both sides are dimensionless signed values ​​with consistent dimensions.

[0034] Bilateral brain oxygen asymmetry index Calculate using the following formula:

[0035] ,

[0036] in For time Bilateral cerebral oxygen asymmetry index at time t, in percentage. and They are time The left and right brain oxygen saturation values ​​at any given time are defined as described above. This asymmetry index is a standard calculation formula for brain oxygenation monitoring in this field.

[0037] The brain oxygenation image feature extraction module 3 further extracts the bilateral asymmetry index change rate from the dual-channel brain oxygenation waveform image. The bilateral asymmetry index change rate is the increment of the bilateral brain oxygenation asymmetry index per unit time. The rationale for extracting the bilateral asymmetry index change rate is that, in the gradual process of bilateral asymmetric collapse, the evolution rate of the asymmetry index itself reflects the collapse initiation earlier than the static value; single-point static threshold determination loses this dynamic feature. The bilateral asymmetry index change rate... Calculate using the following difference formula:

[0038] ,

[0039] in: For time The two-sided asymmetric exponential rate of change at time t is a scalar, and its value ranges from t to t. The unit is % / min, which is the asymmetry exponent at the current time and Difference of the asymmetry exponent at previous time points divided by The technical effect is that it uses a single-point rate of change index to carry dynamic information about bilateral asymmetric collapse. The definition is the same as the formula mentioned above; For time The bilateral brain oxygen asymmetry index at time t, in percentage, is calculated and cached by the brain oxygen image feature extraction module 3 at the previous difference time. The differential time interval is a scalar value of 30 seconds. A value that is too short will be contaminated by high-frequency physiological fluctuations, while a value that is too long will lose its early warning advantage. The value is determined based on the characteristic time constant of the bilateral asymmetric collapse process. The technical effect lies in setting the temporal resolution of the differential window. The graded early warning module 4 triggers a bilateral asymmetric early warning if the rate of change of the bilateral asymmetric index exceeds a preset rate of change threshold (0.5% / min) before the bilateral asymmetric index has crossed the 10 percentage point bilateral asymmetric threshold. Compared to the bilateral asymmetric judgment with a fixed threshold, this significantly advances the warning point. The numerator on the right side of the equation is in percentage (%), and the denominator is in seconds (s). Dividing by % / s, and after processing with a unit conversion factor of 60, equals the left side. The percentage / min is consistent with the unit of measurement.

[0040] The relative baseline decrease is determined by the brain oxygen image feature extraction module 3 establishing left and right brain oxygen saturation values ​​before extraction. These left and right brain oxygen saturation values ​​are the average left and right brain oxygen saturation values ​​within 60 seconds before anesthesia induction, respectively. The relative baseline decrease is... Calculate using the following formula:

[0041] ,

[0042] in For time Time Channel The relative baseline decrease, in percentage. For channel The brain oxygen linear value, in percentage. The definition is the same as above. The relative baseline decrease is calculated using the standard formula for brain oxygenation monitoring in this field.

[0043] See Figure 2 As shown, in this embodiment, the brain oxygen image feature extraction module 3 superimposes the bilateral brain oxygen short window slope as the first image domain feature channel onto the top layer of the dual-channel brain oxygen waveform image, superimposes the bilateral brain oxygen long window slope as the second image domain feature channel onto the bottom of the first channel, superimposes the bilateral brain oxygen asymmetry index and the bilateral asymmetry index change rate in a color-coded manner onto the central band between the dual-channel waveforms, and annotates the relative baseline drop amplitude on the right end of the waveform using the right-side scale value.

[0044] See Figure 3As shown, the graded early warning module 4 triggers and outputs corresponding graded early warning signals for three high-risk patterns—unilateral persistent brain oxygen threshold, relative baseline decline, and bilateral asymmetric—based on the brain oxygen status sensitivity characteristics. The specific judgment logic is as follows: The judgment condition for unilateral persistent brain oxygen threshold is that the left or right brain oxygen saturation is below 50% for 60 consecutive seconds, triggering a yellow warning; the judgment condition for relative baseline decline is that the left or right relative baseline decline exceeds 20% and lasts for more than 30 seconds, triggering an orange warning; the judgment condition for bilateral asymmetric includes two parallel sub-conditions: one is that the absolute value of the bilateral brain oxygen asymmetry index exceeds 10 percentage points (i.e., the bilateral asymmetric threshold); the other is that, before the bilateral brain oxygen asymmetry index crosses the bilateral asymmetric threshold, the rate of change of the bilateral asymmetric index exceeds a preset rate of change threshold of 0.5% / min. The fulfillment of either sub-condition triggers a bilateral asymmetric warning; the former outputs a red warning, and the latter outputs a red precursor warning. The graded early warning module 4 outputs a graded early warning signal including an early warning category field, an early warning time field, a trigger window start and end time field, and an early warning level field. The early warning level increases in four levels: yellow, orange, red precursor, and red. During the judgment phase, the graded early warning module 4 compares each of the brain oxygenation status sensitive features with a preset threshold table. The preset threshold table is configured once by the anesthesia workstation system administrator during the system deployment phase and can be fine-tuned by the anesthesiologist during the preoperative calibration phase based on individual patient differences. In the preset threshold table, the default values ​​for unilateral absolute brain oxygenation are 50%, the default values ​​for relative baseline decrease are 20%, the default values ​​for bilateral asymmetry are 10 percentage points, the default values ​​for bilateral asymmetry index change rate are 0.5% / min, the default duration for continuous judgment of unilateral brain oxygenation is 60s, and the default duration for continuous judgment of relative baseline decrease is 30s. When the graded early warning module 4 detects that any high-risk mode has been triggered, in addition to outputting a graded early warning signal, it also simultaneously overlays an early warning sound prompt and a color flashing prompt on the integrated display interface of the anesthesia workstation. The tone of the early warning sound is positively correlated with the early warning level, and the frequency of the color flashing is positively correlated with the early warning level, which makes it easier for anesthesiologists to quickly detect the occurrence of the early warning in a multi-tasking intraoperative environment and switch their focus to the dual-channel brain oxygen waveform image.

[0045] The multi-parameter linkage display module 5 displays the dual-channel cerebral oxygen waveform image together with the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform on the anesthesia workstation integrated display interface, aligned with the time axis. When the graded warning signal is triggered, the window segment in the dual-channel cerebral oxygen waveform image that triggers the warning is highlighted. In this embodiment, the anesthesia workstation integrated display interface displays the dual-channel cerebral oxygen waveform image in the upper half in a horizontal parallel arrangement, and the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform are sequentially placed in the lower half, with all waveforms sharing the same time axis. The highlighting is achieved by surrounding the corresponding pixel column in the dual-channel cerebral oxygen waveform image with a red rectangle in the triggered window segment, and marking the warning level and warning category with a colored logo in the upper right corner of the rectangle. The anesthesia workstation integrated display interface further includes a history scroll bar, allowing anesthesiologists to drag the scroll bar at any point during surgery to review the dual-channel cerebral oxygen waveform image and corresponding hemodynamic waveform at any past time interval, facilitating comparison of the waveform evolution process before the triggering of the graded warning signal. The anesthesia workstation's integrated display interface also provides a single-sided waveform hiding switch, allowing anesthesiologists to hide the contralateral waveform channel to avoid visual interference when only unilateral cerebral oxygenation needs to be monitored. The multi-parameter linkage display module 5 continuously updates the anesthesia workstation's integrated display interface at a refresh rate of 10Hz, ensuring smooth dynamic waveform display and timely warning responses.

[0046] The multi-parameter linkage display module 5 further includes a root cause matching submodule. When the graded warning signal is triggered, the root cause matching submodule extracts hemodynamic event features synchronized with the trigger window from the invasive arterial pressure waveform, the cardiac output waveform, and the anesthesia depth index waveform. These hemodynamic event features include the magnitude of the decrease in mean arterial pressure, the magnitude of the decrease in cardiac output, and the magnitude of the change in the anesthesia depth index. The hemodynamic event features are synchronously matched according to the following formula:

[0047] ,

[0048] in: To trigger window The corresponding root cause pairing result is a discrete scalar with a value range of [value range missing]. The unit is dimensionless and is obtained by finding the subscript of the maximum value in this formula. The technical effect is to automatically determine the category of the maximum relevant hemodynamic root cause of cerebral oxygen deficiency. The trigger window for the graded early warning signal is a time interval. The output is provided by the graded early warning module 4, with the unit being seconds. The technical effect is to constrain the range of synchronous extraction of hemodynamic event features. This is the hemodynamic root cause index, with values ​​from... The sets represent three root cause categories: mean arterial pressure, cardiac output, and anesthesia depth index. root cause In the trigger window The amount of change in event characteristics within, for The time represents the decrease in mean arterial pressure (in mmHg). The time represents the decrease in cardiac output (in L / min). The time represents the change in the depth of anesthesia index (dimensionless), determined by the corresponding invasive arterial pressure waveform, cardiac output waveform, or depth of anesthesia index waveform within the window. The value within the window is calculated by subtracting the value at the end of the window from the maximum value within the window. The technical effect is to quantify the event intensity of each candidate root cause within the triggering window. root cause The normalization threshold is a scalar, for Take 20 mmHg at a time, for The flow rate is taken as 1.5 L / min. Take 15 at a time, unit and The correspondence is consistent, and based on the commonly used threshold values ​​for significant events of each root cause category in perioperative clinical judgment, the technical effect is to normalize the characteristics of three types of events with different dimensions into comparable dimensionless numbers. root cause The weighting coefficients are scalars, and the default value is... , , The unit is dimensionless. It is based on the prior values ​​of clinical causal strength of perioperative cerebral perfusion and various hemodynamic indicators. The technical effect is to sort the results according to the prior causal strength when multiple root causes occur simultaneously. This is the absolute value operator; The operator for finding the maximum value corresponding to the independent variable. The root cause pairing submodule displays the hemodynamic event characteristics and the graded early warning signal together on the anesthesia workstation's integrated display interface using root cause association markers. These markers include root cause category icons, event characteristic values, normalized index bar charts, and root cause explanation phrases, facilitating rapid interpretation by anesthesiologists. (Right side of the equation) and Dividing quantities of the same dimension yields a dimensionless quantity; multiplying by the dimensionless weight... It is still dimensionless. Return the discrete labels, along with the left side of the equation. The discrete labels are consistent, and the dimensions are consistent.

[0049] See Figure 4 As shown, the root cause matching submodule outputs a root cause association marker immediately after the graded early warning signal is triggered and displays it synchronously on the integrated display interface of the anesthesia workstation.

[0050] Taking a 75-year-old male undergoing heart valve replacement surgery with cardiopulmonary bypass as an example, after anesthesia induction, the system described in this embodiment was activated. The bilateral cerebral oxygen acquisition module 1 determined the left cerebral oxygen saturation level to be 73% and the right cerebral oxygen saturation level to be 71%. At 18 minutes after the start of intraoperative cardiopulmonary bypass, the bilateral cerebral oxygen acquisition module 1 output the following temporal signal of bilateral cerebral oxygen saturation: the left cerebral oxygen saturation slowly decreased from 73% to 65%, while the right cerebral oxygen saturation rapidly decreased from 71% to 59%. The dual-channel cerebral oxygen waveform image constructed by the cerebral oxygen waveform image construction module 2 during this period showed that the waveform envelopes on the left and right sides were significantly non-parallel. The waveform slope of the left channel was relatively gentle, while the waveform slope of the right channel was significantly inclined and downward, with the height difference between the two increasing rapidly in the central band of the image. The cerebral oxygen image feature extraction module 3 extracted the bilateral cerebral oxygen short window slope. % / min % / min; bilateral brain oxygenation long window slope % / min The bilateral cerebral oxygen asymmetry index increased from 2.8% after anesthesia induction to 9.6%; the bilateral asymmetry index change rate was 0.8% / min (before the bilateral asymmetry index crossed the 10% threshold); the right side's relative baseline decrease reached 16.9%, and the left side's relative baseline decrease was 10.9%. The graded early warning module 4, based on the bilateral asymmetry index change rate exceeding the preset threshold of 0.5% / min, triggers a bilateral asymmetry red precursor warning in advance; simultaneously, the root cause matching submodule extracts hemodynamic event characteristics within this trigger window, obtaining a mean arterial pressure decrease of 22 mmHg, a cardiac output decrease of 0.7 L / min, and an anesthesia depth index change of 6, and calculates the root cause matching result according to the synchronous matching algorithm. The anesthesia workstation's integrated display interface highlights the corresponding window segment of the dual-channel cerebral oxygen waveform image with a red rectangle, and displays the root cause association marker with the phrase "hypotensive cerebral oxygen imbalance" and an arterial pressure icon. Based on this, the anesthesiologist immediately adjusts the continuous norepinephrine infusion rate. Approximately 75 seconds after the mean arterial pressure recovers, bilateral cerebral oxygen saturation returns to near baseline, preventing the bilateral cerebral oxygen asymmetry index from continuing to rise to the bilateral asymmetry red warning threshold. Under the traditional numerical threshold alarm paradigm, the warning time for this patient would be delayed until the bilateral asymmetry index crosses the 10 percentage point threshold, estimated at approximately 90 seconds based on this case data. Furthermore, root cause determination requires manual comparison of three independent curves—arterial pressure, cardiac output, and anesthesia depth index—by the physician, averaging 45 to 60 seconds. The system described in this embodiment advances the warning time by approximately 90 seconds and achieves automatic root cause matching, significantly expanding the clinical intervention window.

[0051] To further illustrate the ability of the root cause matching submodule to interpret different root cause categories, a case of a 72-year-old female undergoing laparoscopic partial hepatectomy is used as an example. After anesthesia induction, the bilateral cerebral oxygenation module 1 determined the left cerebral oxygenation value to be 68% and the right cerebral oxygenation value to be 69%. At 42 minutes into the operation, the anesthesiologist increased the sevoflurane inhalation concentration to achieve deeper abdominal cavity exposure, and the anesthesia depth index decreased from 55 to 32. The cerebral oxygenation image feature extraction module 3 extracted the bilateral cerebral oxygenation short window slope. % / min % / min (approximately synchronous and slow decrease on both sides); bilateral brain oxygenation long window slope % / min The bilateral cerebral oxygen asymmetry index remained stable at 1.4% (due to near-synchronous decreases on both sides, the asymmetry index change was slight); the bilateral asymmetry index change rate was 0.05% / min; the decrease in the left side relative to the baseline was 13.2%, and the decrease in the right side relative to the baseline was 13.0%. The graded early warning module 4, based on the consistent sign and direction of the bilateral short-window and long-window slopes of cerebral oxygen and their relatively large values, determined it to be a continuous slow decrease event, triggering an orange relative baseline decrease precursor warning; the root cause matching submodule simultaneously extracted hemodynamic event characteristics within this trigger window, obtaining the mean arterial pressure decrease of 7 mmHg, the cardiac output decrease of 0.3 L / min, and the anesthesia depth index change of 23, and calculated the root cause matching result according to the synchronous matching algorithm. The anesthesia workstation's integrated display interface shows the root cause association marker with the phrase "excessive anesthesia and cerebral oxygen imbalance" and an anesthesia depth icon. Based on this, the anesthesiologist immediately adjusts the sevoflurane inhalation concentration. Once the anesthesia depth index recovers to 48, bilateral cerebral oxygen saturation returns to near baseline. This example demonstrates the ability of the root cause matching submodule to accurately identify excessive anesthesia as the root cause category even in scenarios with near-synchronous rather than asymmetrical bilateral decreases, based on a multi-parameter event feature synchronous matching algorithm. This proves the system's robust interpretation capability for different categories of hemodynamic root causes.

[0052] To further illustrate the effectiveness of the artifact gating submodule in typical intraoperative interference scenarios, a case of a 65-year-old male undergoing thoracoscopic lobectomy is used as an example. At the 28-minute mark of the operation, the surgeon applied electrocoagulation to the lesion margin for approximately 15 seconds. During electrocoagulation, significant non-physiological spikes appeared in the bilateral brain oxygen saturation time-series signals output by the near-infrared brain oxygen saturation sensor (the left side momentarily jumped to 93%, and the right side momentarily jumped to 91%). The artifact gating submodule, based on median absolute deviation, outputs an artifact mask value (both sides are identified as artifacts) simultaneously with the occurrence of these non-physiological spikes. This artifact mask value fills the corresponding pixel column in the dual-channel brain oxygen waveform image with pure white. The brain oxygen image feature extraction module 3 skips the pixel column corresponding to this artifact mask value when calculating the brain oxygen status sensitive features, and the graded warning module 4 also skips the trigger determination for this pixel column. After electrocoagulation, the bilateral cerebral oxygen saturation time-series signal quickly returned to normal, and the artifact mask value was subsequently removed, allowing the subsequent judgment process to resume normal operation. In this example, if the artifact gating submodule were not set, the non-physiological spike would cause the cerebral oxygen image feature extraction module 3 to miscalculate the bilateral cerebral oxygen short window slope to a large positive value (incorrectly representing acute recovery of cerebral oxygen), and may interfere with the calculation results of the bilateral asymmetric exponential change rate, thereby triggering an error warning; the artifact gating submodule effectively intercepted the above error propagation in the image domain using a masking method.

[0053] The method for dynamic monitoring and early warning of intraoperative cerebral oxygen saturation images in elderly patients provided in this embodiment corresponds one-to-one with the working steps of each module in the above system embodiment, including the following steps.

[0054] Step S1: Bilateral brain oxygen saturation time-series signals are synchronously acquired using bilateral near-infrared brain oxygen saturation sensors, which are respectively attached to the left and right frontal lobes of the patient. The bilateral near-infrared brain oxygen saturation sensors are respectively attached to the FP1 position of the left frontal lobe and the FP2 position of the right frontal lobe. The distance between the light source and the photodetector of each bilateral near-infrared brain oxygen saturation sensor is configured within the range of 30mm to 40mm; in this embodiment, 40mm is used. Two wavelengths are selected: 730nm and 810nm. Local brain oxygen saturation values ​​in the left and right prefrontal lobe regions are continuously acquired at a sampling frequency of 2Hz, and the bilateral brain oxygen saturation time-series signals are output. The specific implementation details of this step are the same as those of the bilateral brain oxygen acquisition module 1 described in the above system embodiment.

[0055] Step S2: The bilateral brain oxygen saturation time-series signals are segmented using a sliding window to construct a dual-channel brain oxygen waveform image. This dual-channel brain oxygen waveform image is constructed by aligning and placing the left and right brain oxygen waveform channels side-by-side according to the same start and end times of a time window. In this step, the width of the sliding window is configured within the range of 60s to 300s; in this embodiment, 300s is used. The overlap rate of adjacent sliding windows is 50%, and the update cycle of the dual-channel brain oxygen waveform image is configured within the range of 1s to 5s; in this embodiment, 1s is used. At the arrival of each update cycle, the left and right brain oxygen saturation time-series signals are extracted from the nearest 300s time window. Following the image construction rules described in the above system embodiment, the dual-channel brain oxygen waveform image is constructed as follows: the left brain oxygen waveform channel is placed in the upper half of the image, and the right brain oxygen waveform channel is placed in the lower half of the image. Both share the same start and end times of the time axis and the same amplitude scale. In this step, artifact detection based on median absolute deviation is further performed to obtain artifact mask values. These artifact mask values ​​are superimposed on the dual-channel brain oxygen waveform image, and the regions corresponding to the artifact mask values ​​are not involved in subsequent steps. The specific implementation details of this step are the same as those of the brain oxygen waveform image construction module 2 and its artifact gating submodule described in the above system embodiment.

[0056] Step S3: Extract brain oxygenation state sensitive features from the dual-channel brain oxygen waveform image. These features include bilateral brain oxygen slope, bilateral brain oxygen asymmetry index, and the magnitude of the relative baseline decrease. The bilateral brain oxygen slope includes bilateral short-window slope and bilateral long-window slope. The bilateral short-window slope is obtained by regressing the bilateral brain oxygen saturation time-series signal within a 60-second short sliding window, and the bilateral long-window slope is obtained by regressing the bilateral brain oxygen saturation time-series signal within a 300-second long sliding window. The bilateral short-window slope and the bilateral long-window slope are superimposed on the dual-channel brain oxygen waveform image as two independent image domain feature channels. This step further extracts the bilateral asymmetry index change rate from the dual-channel brain oxygen waveform image. The bilateral asymmetry index change rate is the increment of the bilateral brain oxygen asymmetry index per unit time. This step further establishes cerebral oxygen saturation values ​​for both sides before extracting the relative baseline decrease. These values ​​represent the average oxygen saturation of the left and right sides of the brain within 60 seconds prior to anesthesia induction, respectively. The relative baseline decrease is the percentage decrease in the current average oxygen saturation of both sides relative to the cerebral oxygen saturation values. The specific implementation details and all calculation formulas for this step are the same as those in the brain oxygen image feature extraction module 3 described in the above system embodiment.

[0057] Step S4: Based on the aforementioned brain oxygenation status sensitivity characteristics, trigger and output corresponding graded warning signals for three high-risk patterns: unilateral persistent brain oxygen threshold, relative baseline decline, and bilateral asymmetric pattern. In this step, the criterion for unilateral persistent brain oxygen threshold is that the left or right brain oxygen saturation is below 50% for 60 consecutive seconds, triggering a yellow warning. The criterion for relative baseline decline is that the left or right side's relative baseline decline exceeds 20% and lasts for more than 30 seconds, triggering an orange warning. The criterion for bilateral asymmetric pattern includes two parallel sub-conditions: first, the absolute value of the bilateral brain oxygen asymmetry index exceeds 10 percentage points; second, the rate of change of the bilateral brain oxygen asymmetry index exceeds a preset rate of change threshold of 0.5% / min before the bilateral asymmetry index crosses the bilateral asymmetry threshold. The former outputs a red warning, and the latter outputs a red precursor warning. This step further distinguishes between acute brain oxygen decline events and persistent slow decline events based on the sign consistency of the bilateral brain oxygen short window slope and the bilateral brain oxygen long window slope. The specific implementation details of this step are the same as those of the graded early warning module 4 described in the above system embodiment.

[0058] Step S5: The dual-channel cerebral oxygen waveform image is aligned with the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform on the integrated display interface of the anesthesia workstation and presented together on the time axis. When the graded warning signal is triggered, the window segment in the dual-channel cerebral oxygen waveform image that triggers the warning is highlighted. This step further performs root cause matching when the graded warning signal is triggered: hemodynamic event features synchronized with the trigger window are extracted from the invasive arterial pressure waveform, the cardiac output waveform, and the anesthesia depth index waveform. The hemodynamic event features include the magnitude of the decrease in mean arterial pressure, the magnitude of the decrease in cardiac output, and the magnitude of the change in the anesthesia depth index. The hemodynamic event features are calculated according to the synchronous matching algorithm of the root cause matching submodule in the above system embodiment to calculate the root cause matching result. The hemodynamic event features and the graded warning signal are displayed together on the integrated display interface of the anesthesia workstation with root cause association markers. The specific implementation details of this step are the same as those of the multi-parameter linkage display module 5 and its root cause matching submodule in the above system embodiment.

[0059] See Figure 5 As shown, the overall steps of the method for dynamic monitoring and early warning of intraoperative cerebral oxygen saturation images in elderly patients described in this embodiment are executed sequentially from S1 to S5, and steps S2 to S5 are iteratively executed once in each update cycle, which is 1 second. The entire method runs continuously during the operation until the anesthesia is lifted. The clinical implementation scenario of this method is the same as the complete implementation scenario described in the system embodiment, and will not be repeated.

[0060] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic monitoring and early warning system for intraoperative cerebral oxygen saturation images in elderly patients, characterized in that, include: The bilateral brain oxygen acquisition module includes bilateral near-infrared brain oxygen saturation sensors that are respectively attached to the left and right frontal lobes of the patient, and the bilateral near-infrared brain oxygen saturation sensors output bilateral brain oxygen saturation time-series signals. The brain oxygen waveform image construction module divides the bilateral brain oxygen saturation time-series signal into segments according to a sliding window to construct a dual-channel brain oxygen waveform image. The dual-channel brain oxygen waveform image is composed of the left brain oxygen waveform channel and the right brain oxygen waveform channel aligned side by side on the time axis according to the same start and end times of the time window. The brain oxygenation image feature extraction module extracts brain oxygenation state sensitive features from the dual-channel brain oxygenation waveform image. The brain oxygenation state sensitive features include bilateral brain oxygenation slope, bilateral brain oxygenation asymmetry index, and relative baseline decrease. The graded early warning module triggers and outputs corresponding graded early warning signals for three types of high-risk patterns: unilateral persistent brain oxygen threshold, relative baseline decline, and bilateral asymmetric brain oxygen status, based on the aforementioned brain oxygen status sensitivity characteristics. The multi-parameter linkage display module aligns the dual-channel brain oxygen waveform image with the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform on the integrated display interface of the anesthesia workstation, and highlights the window segment in the dual-channel brain oxygen waveform image that triggers the warning when the graded warning signal is triggered.

2. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 1, characterized in that, When the brain oxygen waveform image construction module constructs the dual-channel brain oxygen waveform image, the left brain oxygen waveform channel and the right brain oxygen waveform channel share the same amplitude scale and are aligned according to the same start and end times of the time window, so that the geometric slope difference between the left brain oxygen waveform channel and the right brain oxygen waveform channel is explicitly preserved as a visual feature in the image domain in the dual-channel brain oxygen waveform image.

3. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 2, characterized in that, The brain oxygen waveform image construction module further includes an artifact gating submodule. The artifact gating submodule performs artifact detection based on median absolute deviation on the bilateral brain oxygen saturation time-series signal to obtain artifact mask values. The artifact mask values ​​are superimposed on the dual-channel brain oxygen waveform image. The brain oxygen image feature extraction module only performs brain oxygen image feature extraction on the pixel column with artifact mask value of 0 in the dual-channel brain oxygen waveform image with artifact mask value. The graded early warning module only performs the trigger determination based on the brain oxygen state sensitive features corresponding to the pixel column with artifact mask value of 0.

4. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 3, characterized in that, The bilateral brain oxygenation slope includes the bilateral brain oxygenation short window slope and the bilateral brain oxygenation long window slope. The bilateral brain oxygenation short window slope is obtained by regression of the bilateral brain oxygen saturation time-series signal within a 60s short sliding window, and the bilateral brain oxygenation long window slope is obtained by regression of the bilateral brain oxygen saturation time-series signal within a 300s long sliding window. The bilateral brain oxygenation short window slope and the bilateral brain oxygenation long window slope are superimposed on the dual-channel brain oxygenation waveform image as two independent image domain feature channels. The graded early warning module distinguishes between acute brain oxygenation decline events and persistent slow decline events based on the sign consistency of the bilateral brain oxygenation short window slope and the bilateral brain oxygenation long window slope.

5. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 4, characterized in that, The brain oxygen image feature extraction module further extracts the bilateral asymmetry index change rate from the dual-channel brain oxygen waveform image. The bilateral asymmetry index change rate is the increment of the bilateral brain oxygen asymmetry index per unit time. The bilateral asymmetry high-risk mode triggering judgment of the graded early warning module includes a static threshold judgment branch and a change rate precursor judgment branch. The triggering condition of the static threshold judgment branch is that the bilateral brain oxygen asymmetry index is greater than or equal to the bilateral asymmetry threshold, which is 10 percentage points. The triggering condition of the change rate precursor judgment branch is that the bilateral brain oxygen asymmetry index is less than the bilateral asymmetry threshold and the bilateral asymmetry index change rate is greater than a preset change rate threshold. When the triggering condition of the static threshold judgment branch is met, the graded early warning module outputs a bilateral asymmetry warning. When the triggering condition of the change rate precursor judgment branch is met, the graded early warning module outputs a bilateral asymmetry precursor warning.

6. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 5, characterized in that, The multi-parameter linkage display module further includes a root cause matching submodule. When the graded warning signal is triggered, the root cause matching submodule extracts hemodynamic event features synchronized with the trigger window from the invasive arterial pressure waveform, the cardiac output waveform, and the anesthesia depth index waveform. The hemodynamic event features include the mean arterial pressure decrease, the cardiac output decrease, and the anesthesia depth index change. The root cause matching submodule displays the hemodynamic event features and the graded warning signal together on the anesthesia workstation integrated display interface with root cause association markers.

7. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 1, characterized in that, The bilateral near-infrared brain oxygen saturation sensors are respectively attached to the left frontal lobe FP1 position and the right frontal lobe FP2 position of the patient. The distance between the light source and the photodetector of each bilateral near-infrared brain oxygen saturation sensor is 30mm to 40mm.

8. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 1, characterized in that, The sliding window width used in the brain oxygen waveform image construction module is configurable from 60s to 300s, the overlap rate of adjacent sliding windows is 50%, and the update cycle of the dual-channel brain oxygen waveform image is from 1s to 5s.

9. The intraoperative cerebral oxygen saturation image dynamic monitoring and early warning system for elderly patients according to claim 1, characterized in that, Before extracting the relative baseline decrease, the brain oxygen image feature extraction module first establishes the brain oxygen saturation values ​​on both sides. The brain oxygen saturation values ​​on both sides are the average oxygen saturation values ​​of the left and right sides within 60 seconds before anesthesia induction, respectively. The relative baseline decrease is the percentage decrease of the current average oxygen saturation values ​​on both sides relative to the brain oxygen saturation values ​​on both sides.

10. A method for dynamic monitoring and early warning of intraoperative cerebral oxygen saturation images in elderly patients, employing the dynamic monitoring and early warning system for intraoperative cerebral oxygen saturation images in elderly patients as described in any one of claims 1-9, characterized in that, Includes the following steps: S1: Bilateral brain oxygen saturation time-series signals are simultaneously acquired by bilateral near-infrared brain oxygen saturation sensors that are respectively attached to the left and right frontal lobes of the patient. S2: The bilateral brain oxygen saturation time-series signal is segmented by a sliding window to construct a dual-channel brain oxygen waveform image. The dual-channel brain oxygen waveform image is composed of the left brain oxygen waveform channel and the right brain oxygen waveform channel aligned side by side according to the same start and end times of the time window. S3: Extract brain oxygenation state sensitive features from the dual-channel brain oxygen waveform image. The brain oxygenation state sensitive features include bilateral brain oxygen slope, bilateral brain oxygen asymmetry index, and relative baseline decrease. S4: Based on the brain oxygenation state sensitivity characteristics, trigger judgment and output corresponding graded warning signals for the three types of high-risk patterns: unilateral brain oxygenation threshold persistence, relative baseline decline, and bilateral asymmetric. S5: The dual-channel brain oxygen waveform image is combined with the invasive arterial pressure waveform, cardiac output waveform, and anesthesia depth index waveform on the integrated display interface of the anesthesia workstation, and the window segment in the dual-channel brain oxygen waveform image that triggers the warning is highlighted when the graded warning signal is triggered.

Images