Artificial intelligence-based anesthesia auxiliary early warning method and system

Abstract

The application discloses an anesthesia auxiliary early warning method and system based on artificial intelligence, and relates to the technical field of medical information, and comprises the following steps: in the anesthesia auxiliary early warning process, collecting the continuous vital sign time series before compression transmission and the continuous vital sign time series after compression transmission, aligning the rhythm according to a unified time scale, generating a time comparison sequence, and recording a time offset track. The application restores the real decline rhythm of vital signs, eliminates the trend distortion caused by compression expression, realizes the early release of critical recognition signals, and improves the trend judgment accuracy and response timeliness of anesthesia auxiliary early warning by constructing the time comparison sequence and the time offset track under the unified time scale, time unfolding and rate reconstruction of the same numerical section, and recovering the real decline rhythm of vital signs.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, specifically to an anesthesia-assisted early warning method and system based on artificial intelligence. Background Technology

[0002] Artificial intelligence-based anesthesia-assisted early warning refers to an intelligent auxiliary technology that utilizes artificial intelligence to continuously analyze and comprehensively assess various physiological data collected in real time during the patient's anesthesia process. This identifies potential abnormal trends or risk signs and alerts the anesthesiologist before the risks develop into significant clinical events. Specifically, this technology typically establishes data correlations and risk characteristic expressions based on multidimensional data such as electrocardiogram signals, blood pressure changes, blood oxygen saturation, respiratory rate, anesthesia depth index, and medication records. Through learning from historical case data, it develops risk identification capabilities and dynamically compares and judges the current patient's condition in actual surgical scenarios, thereby providing early warnings of potential adverse events such as hypotension, hypoxemia, circulatory depression, respiratory depression, and delayed awakening. Its core purpose is not to replace anesthesiologist decision-making, but rather to serve as a clinical decision support tool, providing earlier, more continuous, and more objective risk alerts in complex surgical environments and under high-intensity monitoring tasks, thereby improving perioperative safety and management accuracy.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, during the compression of vital sign data before transmission to the analysis end in anesthesia-assisted early warning systems, if the data compression strategy involves buffer overwriting or short-term repeated output, it can easily create segments of continuously repeating values ​​within a very short period. These repeated segments may appear stable over time, and models often identify them as temporarily stable indicators when assessing trends, thus weakening the evaluation of the true rate of decline. When a patient's vital signs are in a rapidly declining phase, these pseudo-stable segments can easily mask the true, rapid change, causing the system to delay issuing critical warning signals and consequently affecting the timing of clinical intervention.

[0005] The information disclosed in the background section is only for enhancing the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an anesthesia-assisted early warning method and system based on artificial intelligence to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an anesthesia-assisted early warning method based on artificial intelligence, comprising the following steps:

[0008] During the anesthesia-assisted early warning process, the continuous vital signs time series before and after compressed transmission are collected, aligned rhythmically according to a unified time scale, and a time control sequence is generated, and the time offset trajectory is recorded.

[0009] Using the time offset trajectory in the time reference sequence, the continuous identical value segments in the time reference sequence are processed by time expansion. According to the sampling interval of the continuous vital sign time series before compression and transmission, the continuous identical value segments are redistributed to the time axis to generate a change sequence, and the change frequency is marked in the change sequence.

[0010] Based on the frequency of change in the change sequence, the rate of continuous identical value segments and rapidly decreasing segments is compared to determine the location of rate abrupt change, form a rate change trajectory, and mark the location of risk change in the rate change trajectory;

[0011] Based on the location of risk changes in the rate change trajectory, locate the trend compression segment in the change sequence, determine the trend distortion range, and generate a list of trend distortion ranges.

[0012] Time stretching is performed on the time segments corresponding to the trend distortion range list. The true rate change expression is added to the time window where the consecutive segments with the same value are located, and the adjacent time lengths are redistributed to restore the true descent rhythm and release the critical identification signal in advance.

[0013] Preferably, the steps for collecting continuous vital signs time series before compression and after compression and generating a time control sequence are as follows:

[0014] During the anesthesia-assisted early warning process, the continuous vital signs time series before and after compressed transmission are collected synchronously and assigned a first time marker and a second time marker, respectively.

[0015] Using a unified time scale as a time reference, the first time mark and the second time mark are mapped to the same time axis coordinate system;

[0016] The baseline time position of the continuous vital signs time series before compression and transmission and the actual time position of the continuous vital signs time series after compression and transmission are organized hour by hour around a unified time scale, and the time interval difference is recorded to form a time comparison sequence.

[0017] The time interval differences are arranged in a uniform time scale order to form a time offset trajectory, and the time offset trajectory is associated with the time reference sequence and recorded.

[0018] Preferably, the steps for generating a change sequence using the time offset trajectory in the time reference sequence are as follows:

[0019] Using time-referenced sequences as the data structure, continuous segments with the same values ​​are identified within a unified time scale, and time offset trajectories at corresponding time scale positions are associated to form a time-compressed representation set.

[0020] Based on the time-compressed expression set, the theoretical time span is determined by the sampling interval of the continuous vital sign time series before compression and transmission. The continuous segments with the same values ​​are then processed by time expansion and redistributed to the time axis.

[0021] The numerical sequence formed by time expansion is concatenated with the uncompressed segment according to a uniform time scale to generate a change sequence;

[0022] Record the numerical changes on a scale around the change sequence, and mark the frequency of change at the corresponding time scale position.

[0023] Preferably, during the time expansion process, the numerical order of consecutive identical numerical segments is kept consistent with the time progression order, and the time redistribution interval of consecutive identical numerical segments is determined by the time scale range corresponding to the time offset trajectory. The change frequency marker is bound and recorded at the time position corresponding to the change sequence according to a unified time scale.

[0024] Preferably, the steps for determining the rate change trajectory based on the frequency of change in the change sequence are as follows:

[0025] Using the change sequence as the time analysis structure, the values ​​and change frequencies are organized step by step under a unified time scale order. The change per unit time is formed based on the value difference and time interval between adjacent time scales, and the continuous segments of the same value are identified.

[0026] Based on the frequency of change, segments with consecutive identical values ​​and segments with rapid decline are segmented and classified to form corresponding rate distribution sequences under a unified time scale order;

[0027] The rate distribution sequence is interpolated along a uniform time scale to form a rate change trajectory and determine the location of rate abrupt changes.

[0028] Extended analysis is performed around the locations of rate abrupt changes in the rate change trajectory, and risk change locations are marked in the rate change trajectory.

[0029] Preferably, around the risk change location in the rate change trajectory, the time scale corresponding to the risk change location is associated and mapped with the continuous identical numerical segment in the change sequence. The time range of the rate change location is defined based on the distribution of the risk change location on the unified time scale, and the risk change location is used as the time reference for identifying the trend distortion range.

[0030] Preferably, the steps for generating a list of trend distortion ranges based on the location of risk changes in the rate change trajectory are as follows:

[0031] Using the rate change trajectory as the main time analysis line, the risk change location is extracted within a unified time scale and mapped to the change sequence for localization;

[0032] By comparing and organizing the time distribution of consecutive identical value segments and rapidly declining segments around the risk change location in the change sequence, the segment where the trend is compressed is identified.

[0033] Record the start and end time scales around the compressed segment of the trend and arrange them sequentially in conjunction with the location of risk changes;

[0034] A list of trend distortion ranges is formed by summarizing the compressed segments of the trend according to a unified time scale.

[0035] Preferably, the trend distortion range list records the risk change location and frequency statistics of each trend compression segment, and establishes the correspondence between the trend compression segment and the risk change location according to a unified time scale order.

[0036] Preferably, the time stretching process performed on the time segment corresponding to the list of trend distortion ranges is as follows:

[0037] According to the time sequence of the trend distortion range list, the corresponding time segment is extracted from the change sequence and the continuous identical value segment and the risk change position are extracted;

[0038] The theoretical time span is determined by converting the total length of the time scale with the sampling interval of the continuous vital signs time series before compression and transmission, based on the time segment corresponding to the range of trend distortion.

[0039] Time extension is performed around the time window containing consecutive segments of the same value, and the actual rate of change is added.

[0040] By redistributing and adjusting the time scale corresponding to the location of risk changes around the expanded time scale, critical situation identification signals can be released in advance.

[0041] An AI-based anesthesia-assisted early warning system includes a time mapping module, a change reconstruction module, a rate determination module, a trend distortion localization module, and a rhythm recovery module.

[0042] The time mapping module, during the anesthesia-assisted early warning process, collects continuous vital signs time series before and after compressed transmission, aligns them rhythmically according to a unified time scale, generates a time comparison sequence, and records the time offset trajectory.

[0043] The change reconstruction module uses the time offset trajectory in the time reference sequence to perform time expansion processing on the continuous identical value segments in the time reference sequence. According to the sampling interval of the continuous vital sign time series before compression and transmission, the continuous identical value segments are redistributed to the time axis to generate a change sequence, and the change frequency is marked in the change sequence.

[0044] The rate determination module compares the rates of consecutive segments with the same values ​​and rapidly decreasing segments based on the frequency of change in the change sequence, determines the location of rate abrupt changes, forms a rate change trajectory, and marks the location of risk changes in the rate change trajectory.

[0045] The trend distortion localization module locates the compressed segment of the trend in the change sequence based on the location of risk changes in the rate change trajectory, determines the range of trend distortion, and generates a list of trend distortion ranges.

[0046] The rhythm recovery module performs time stretching processing on the time segments corresponding to the trend distortion range list, inserts the true rate change expression into the time window where the consecutive segments with the same value are located, and reallocates the adjacent time lengths to restore the true downward rhythm and release the critical identification signal in advance.

[0047] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0048] This invention constructs a unified time scale mapping relationship between continuous vital sign time series before and after compressed transmission, forming a time reference sequence and recording the time offset trajectory. Based on this, time expansion processing is performed on continuous segments with the same values, so that the time clustering state caused by compressed expression is restored to a continuous time distribution structure that conforms to the original sampling rhythm. This eliminates the masking effect of repeated numerical segments on trend expression from the time dimension, enabling the change sequence to truly reflect the continuous evolution of vital signs, thereby improving the ability to capture the rate of decline in trend judgment and enhancing the response accuracy of anesthesia-assisted early warning in dynamic changing scenarios.

[0049] This invention conducts rate comparison analysis around the frequency of changes to form a rate change trajectory and mark the location of risk changes. It further locates the range of trend distortion and performs time stretching processing. It fills in the true rate change expression within the time window of consecutive segments with the same value, and at the same time, it reallocates the adjacent time lengths so that the restored change sequence presents a true downward rhythm under a unified time scale framework. This moves the trigger position of the critical identification signal on the time axis forward, thereby improving the timeliness of anesthesia-assisted early warning in identifying rapidly declining states and providing a more sufficient time window for clinical intervention. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0051] Figure 1 This is a flowchart of the anesthesia-assisted early warning method based on artificial intelligence according to the present invention.

[0052] Figure 2 This is a schematic diagram of the modules of the anesthesia auxiliary early warning system based on artificial intelligence of the present invention. Detailed Implementation

[0053] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0054] This invention provides, for example Figure 1 The artificial intelligence-based anesthesia-assisted early warning method shown includes the following steps:

[0055] During the anesthesia-assisted early warning process, the continuous vital signs time series before and after compressed transmission are collected, aligned rhythmically according to a unified time scale, and a time control sequence is generated, and the time offset trajectory is recorded.

[0056] In practical applications, to ensure temporal consistency between the vital signs time series before and after compression and transmission, this step constructs a time reference sequence based on a unified scale of time, and then forms a time offset trajectory on this basis to establish a temporal reference framework for subsequent trend recovery processing. The specific implementation steps are as follows:

[0057] During the anesthesia-assisted early warning process, the continuous vital signs time series before and after compressed transmission are collected synchronously. Each sampled data in the continuous vital signs time series before compressed transmission is assigned a first time marker, and each output data in the continuous vital signs time series after compressed transmission is assigned a second time marker. The first time marker is generated continuously and progressively according to the original sampling rhythm, and the second time marker is generated continuously and progressively according to the compressed output rhythm. After the collection is completed, a unified time scale is used as a global time reference, and the first and second time markers are mapped to the same time axis coordinate system, so that the continuous vital signs time series before and after compressed transmission are comparable within the same time coordinate range, thus providing a basic time reference for subsequent rhythm alignment processing.

[0058] After completing the unified time scale mapping, the distribution of continuous vital signs time series before and after compressed transmission on the unified time axis is organized moment by moment. Using the unified time scale as a reference, the time distribution positions of the data points corresponding to the first time mark and the data points corresponding to the second time mark are compared one by one in chronological order. The sampling position of the continuous vital signs time series before compressed transmission on the unified time scale is used as the reference position. The time interval difference between the actual distribution position of the continuous vital signs time series after compressed transmission on the unified time scale and the reference position is calculated. This time interval difference is recorded in chronological order at the corresponding time scale position to form a time comparison sequence. The time comparison sequence simultaneously contains the unified time scale, the reference time position of the continuous vital signs time series before compressed transmission, the actual time position of the continuous vital signs time series after compressed transmission, and the time interval difference between the two. Through this moment-by-moment correspondence organization method, the time comparison sequence can fully reflect the distribution differences between the two rhythms.

[0059] After obtaining the time-control sequence, the time interval differences in the time-control sequence are continuously arranged according to a unified time scale order. The time interval differences on the continuous time scale are connected sequentially to form a time offset change chain. The time interval difference corresponding to each time scale is used as a trajectory point in the time offset trajectory, so that the time offset trajectory presents a continuous distribution state within a unified time scale. In this process, the time progression order of the continuous vital signs time series before compression and transmission is kept unchanged, and the output order of the continuous vital signs time series after compression and transmission is kept unchanged. This allows the time offset trajectory to truly reflect the change process of the compressed output rhythm relative to the original sampling rhythm over the entire time range, thereby constructing a time offset trajectory data structure covering all acquisition periods.

[0060] After the time offset trajectory is formed, the time reference sequence is associated with and stored with the time offset trajectory, so that each uniform time scale position in the time reference sequence corresponds to a unique time offset trajectory point. The time offset trajectory is then extended and recorded at the tail according to the chronological order of the uniform time scale, so that the time offset trajectory forms a continuously extended trajectory expression area at the tail of the time reference sequence. Through this tail extension recording method, the time offset trend of the continuous vital signs time series after compression and transmission can be continuously reflected relative to the continuous vital signs time series before compression and transmission. This allows the time reference sequence to not only reflect the time interval difference at a single moment, but also to express the dynamic evolution of rhythm offset over the entire time range. This provides a continuous, complete, and uniform time reference for subsequent time unfolding processing of continuous identical numerical segments around the time offset trajectory.

[0061] Using the time offset trajectory in the time reference sequence, the continuous identical value segments in the time reference sequence are processed by time expansion. According to the sampling interval of the continuous vital sign time series before compression and transmission, the continuous identical value segments are redistributed to the time axis to generate a change sequence, and the change frequency is marked in the change sequence.

[0062] After constructing the time-referenced sequence and forming the time-off trajectory, in order to eliminate the compression effect of consecutive identical numerical segments in the time dimension, this step performs time expansion processing on the time-referenced sequence around the time-off trajectory. This redistributes consecutive identical numerical segments under the compressed output rhythm to the time axis according to the sampling interval of the continuous vital sign time series before compression transmission, forming a change sequence with real time expression significance. The frequency of change is marked in the change sequence. The specific implementation steps are as follows:

[0063] Using a time-referenced sequence as the basic data structure, the numerical distribution in the time-referenced sequence is scanned sequentially within a unified time scale. Segments where values ​​remain consistent across multiple consecutive time scale positions are identified as consecutive identical value segments. Simultaneously, the time offset trajectory information at the corresponding time scale position is retrieved. The entire time scale range covered by the consecutive identical value segments is correlated with the offset change of the time offset trajectory within the same time range, forming a time-compressed representation set corresponding to the consecutive identical value segments. This ensures that each consecutive identical value segment corresponds to a clear time offset change range, thus providing a specific time reference range for subsequent time expansion processing.

[0064] After obtaining the time-compressed expression set corresponding to consecutive identical numerical segments, the sampling interval of the continuous vital signs time series before compression and transmission is used as the benchmark scale for time redistribution. The coverage length of consecutive identical numerical segments on a unified time scale is calculated in correspondence with the time offset change in the time-compressed expression set to determine the theoretical time span corresponding to the consecutive identical numerical segments under the original sampling rhythm. Then, according to the sampling interval of the continuous vital signs time series before compression and transmission, the consecutive identical numerical segments are expanded one by one within the theoretical time span. This allows the consecutive identical numerical segments, which were originally concentrated in a finite time scale under the compressed output rhythm, to be rearranged into continuous time positions according to the original sampling rhythm. During the expansion process, the numerical order is kept consistent with the time progression order, so that the distribution of consecutive identical numerical segments on the time axis corresponds to the sampling rhythm of the continuous vital signs time series before compression and transmission.

[0065] After redistributing the time of consecutive identical numerical segments, the expanded time scale positions are integrated with the original unified time scale to construct a new time arrangement structure. The numerical sequence after time expansion is spliced ​​with the segments that have not undergone compression expression of consecutive identical values ​​in chronological order to form a variation sequence. This variation sequence presents a complete and continuous expression state within a unified time scale framework. Each time scale position in the variation sequence corresponds to a single numerical expression, and the sampling interval of the continuous vital signs time series before compression transmission is maintained as the time distribution benchmark. This allows the variation sequence to restore the continuous expression characteristics under the original sampling rhythm in the time dimension.

[0066] After the change sequence is formed, the numerical changes in the change sequence are recorded step by step in chronological order. A change frequency marker is added at each time point where a value changes, and the change frequency marker is bound to the corresponding time point. This makes the change sequence contain not only numerical and time information, but also information on the number of times the numerical change occurs. After the continuous identical numerical segments are expanded over time, the change rhythm that was not reflected during the original compressed expression is clearly expressed through the change frequency marker. This allows the change sequence to simultaneously reflect the trajectory of numerical evolution and the distribution of change density, providing a dual reference basis of temporal continuity and change frequency for subsequent rate comparison and rate mutation identification based on the change sequence.

[0067] Based on the frequency of change in the change sequence, the rate of continuous identical value segments and rapidly decreasing segments is compared to determine the location of rate abrupt change, form a rate change trajectory, and mark the location of risk change in the rate change trajectory;

[0068] After constructing the change sequence and generating change frequency markers, to reveal the true evolution characteristics of consecutive segments with the same numerical values ​​after time expansion and the dynamic differences between these segments and rapidly declining segments, this step focuses on rate comparison processing based on the change frequency in the change sequence. Within a unified time scale framework, the location of rate abrupt changes is determined, forming a continuously expressed rate change trajectory. Risk change locations are then marked within this trajectory. The specific implementation steps are as follows:

[0069] Using the change sequence as the basis for time analysis, the numerical changes in the change sequence are organized step by step under a unified time scale order. The numerical value corresponding to each time scale position in the change sequence is extracted synchronously with the change frequency marker. The numerical change per unit time is calculated based on the numerical difference between adjacent time scales and the corresponding time interval. This transforms the change sequence into a rate expression structure containing time information, numerical information, and unit time change information within a unified time scale framework. At the same time, continuous identical numerical segments formed by time expansion in the change sequence are individually identified. The continuous time scale range of continuous identical numerical segments after time expansion and their corresponding unit time change are recorded together, so that the rate expression characteristics of continuous identical numerical segments and the rate expression characteristics of rapidly decreasing segments can be compared on the same time scale.

[0070] After organizing the unit time change, the segments with consecutive identical values ​​and rapidly decreasing segments are categorized based on the frequency of change. Segments with continuously recorded frequency changes are grouped into rapidly decreasing segments, while segments with unchanged frequency changes over consecutive time periods are grouped into segments with consecutive identical values. These two segments are then compared and arranged along the dimension of unit time change, forming a rate distribution sequence for both segments with consecutive identical values ​​and a rate distribution sequence for rapidly decreasing segments under a unified time scale order. This ensures that the unit time change of the two segments corresponds chronologically. This corresponding arrangement allows for a direct comparison between adjacent time scales between the rate expressions of segments with consecutive identical values ​​after time expansion and the rate expressions of rapidly decreasing segments.

[0071] After completing the corresponding arrangement of the rate distribution sequence, the unit time change of consecutive segments with the same value and the unit time change of rapidly decreasing segments are calculated step by step along a unified time scale. The degree of change of unit time change between adjacent time scales is recorded as the rate change. The distribution of the rate change on the time axis is continuously connected to form the rate change trajectory. In the rate change trajectory, when the unit time change changes abruptly between adjacent time scales, the corresponding time scale position is determined as the rate abrupt change position. The rate abrupt change position is recorded in the rate change trajectory in the form of time coordinates, so that the rate change trajectory not only reflects the overall change trend, but also expresses the turning point of the rate structure in the time dimension.

[0072] After obtaining the rate change trajectory and rate abrupt change locations, an extended analysis is performed on the continuous time scale range before and after each rate abrupt change location, based on the distribution of rate abrupt change locations in the rate change trajectory. The time scale corresponding to the rate abrupt change location is correlated and mapped with the numerical changes in the change sequence. Rate abrupt change locations that may trigger physiological risk evolution are marked as risk change locations in the rate change trajectory, so that a one-to-one correspondence is established between risk change locations and a unified time scale. A complete risk change location distribution map is formed in the rate change trajectory, thereby establishing a rate difference expression structure between continuous identical numerical segments and rapidly declining segments in the time dimension. This allows the potential trend distortion phenomenon in the change sequence to be dynamically presented through the rate change trajectory, providing a clear time reference basis for subsequent time structure repair processing around the trend distortion range.

[0073] Based on the location of risk changes in the rate change trajectory, locate the trend compression segment in the change sequence, determine the trend distortion range, and generate a list of trend distortion ranges.

[0074] After obtaining the rate change trajectory and marking the risk change locations, this step, to further reveal the temporal distortion caused by the compressed representation to the original trend structure, focuses on the risk change locations in the rate change trajectory, precisely locates the compressed segment of the trend in the change sequence, determines the trend distortion range, and generates a structured list of trend distortion ranges. This provides a clear temporal boundary basis for subsequent time reconstruction processing. The specific implementation steps are as follows:

[0075] Using the rate change trajectory as the main line of time analysis, all risk change locations are extracted within a unified time scale framework. The time scale corresponding to each risk change location is mapped one-to-one with the same time scale location in the change sequence. The numerical state and frequency of change corresponding to the risk change location are marked in the change sequence. Extending a certain time scale range forward and backward around the risk change location, the numerical change rhythm of the time segment adjacent to the risk change location is continuously organized, so that the time regions before and after the risk change location in the change sequence form a complete time continuous expression structure. In this way, the risk change location in the rate change trajectory is transformed into a positioning anchor point in the change sequence, providing a clear time starting reference for the identification of the compressed trend segment.

[0076] After locating the risk change positions, the temporal distribution of consecutive identical value segments and rapidly declining segments in the adjacent time intervals of the change sequence is compared and organized around each risk change position. The coverage of consecutive identical value segments before and after the risk change position on a unified time scale is fully recorded, and the distribution range of rapidly declining segments on a unified time scale is continuously marked. By sequentially arranging the alternation relationship between consecutive identical value segments and rapidly declining segments in the time dimension, time segments with time expression compression near the risk change position are identified. That is, when there is a temporal misalignment between the coverage of consecutive identical value segments on the time axis and the rate change characteristics of the rapidly declining segment, the corresponding time segment is determined as the trend compression segment, so that the trend compression segment forms a clear time boundary expression in the change sequence.

[0077] After obtaining the time boundaries of the trend compression segments, the start and end time scales of each trend compression segment within a unified time scale range are numbered and recorded. Combined with the distribution order of risk change locations in the rate change trajectory, the trend compression segments are numbered and arranged in chronological order. At the same time, the frequency of numerical changes within each trend compression segment is statistically analyzed, so that each trend compression segment not only contains information on the start and end times, but also information on the density of numerical changes within the segment. Through this numbering and organization method, the trend compression segments form independent segments with clear time spans and change characteristics in the change sequence, thereby concretizing the range of time distortion into a set of recordable time segments.

[0078] After numbering and organizing the characteristics of the compressed trend segments, all compressed trend segments are summarized in a unified time scale order to form a trend distortion range list. The trend distortion range list records the start time scale, end time scale, corresponding risk change location, and frequency statistics of change within each compressed trend segment. This allows the trend distortion range list to comprehensively reflect the time range distribution of the trend expression affected by compression in the change sequence. Through this list-based expression method, the trend distortion range is extracted from the continuous time expression structure to form a structured time list, providing a clear time reference for subsequent time stretching processing around the trend distortion range. This allows the time reconstruction processing to be carried out within a defined time segment, thereby ensuring that the trend recovery process has clear time boundaries.

[0079] Time stretching is performed on the time segments corresponding to the trend distortion range list. The true rate change expression is added to the time window where the consecutive segments with the same value are located, and the adjacent time lengths are redistributed to restore the true descent rhythm and release the critical identification signal in advance.

[0080] After creating a list of trend distortion ranges and defining the start and end time scales for each range, this step performs time stretching around the time segments corresponding to the trend distortion range list to restore the time distortion caused by the compressed expression to the true descent rhythm. The true rate change expression is then inserted into the time window containing consecutive segments of the same value, and adjacent time lengths are redistributed. This restores the true descent rhythm of the change sequence within a unified time scale framework, while simultaneously enabling the early release of critical identification signals. The specific implementation steps are as follows:

[0081] Following the time sequence recorded in the trend distortion range list, the start and end time scales corresponding to each trend distortion range are retrieved one by one. The numerical expression segment that completely corresponds to the time segment is extracted from the change sequence, and the positions of consecutive identical numerical segments and the corresponding risk change positions within the time segment are extracted simultaneously. The total length of the time scale within the trend distortion range is converted to the original sampling rhythm of the continuous vital signs time series before compression and transmission, and the theoretical time span that the trend distortion range should present under the original sampling rhythm is determined. In this way, the trend distortion range is extracted separately from the change sequence, so that the subsequent time stretching processing can be carried out within a clear time boundary and maintain consistency with the original sampling rhythm.

[0082] After determining the theoretical time span of the trend distortion range, time expansion processing is performed on the time windows containing consecutive identical numerical segments within the trend distortion range. The difference between the number of time scales occupied by consecutive identical numerical segments under the compressed output rhythm and the theoretical time span is calculated. The difference is then evenly distributed within the time windows containing consecutive identical numerical segments according to the sampling interval of the continuous vital sign time series before compression transmission. This allows the numerical expressions originally concentrated within a finite time scale to be redistributed into a continuous time expression structure under a unified time scale framework, while maintaining the relative order of risk change positions on the time axis. This ensures that the time scale distribution after time stretching maintains a correspondence with the risk change positions. Through this time expansion method, the time windows containing consecutive identical numerical segments can obtain a time expression space consistent with the original sampling rhythm.

[0083] After completing the time expansion of the time window containing consecutive identical numerical segments, the unit time change information recorded in the rate change trajectory is mapped to the expanded time scale position around the expanded time scale distribution. The true rate change expression is then added to the time window containing consecutive identical numerical segments, so that the values ​​present a continuously decreasing or increasing rate change structure between the expanded time scales. This makes the change rhythm that was originally missing due to the compressed expression replenished after time stretching. During the process of adding the true rate change expression, the order of the change frequency marker and the risk change position on the time axis is kept consistent, so that the change sequence after time stretching contains both the complete time scale distribution and the true rate change expression structure, thus forming a continuously evolving numerical trajectory under a unified time scale framework.

[0084] After the true rate change expression of the time window containing consecutive identical numerical segments is added, the time length of the expanded time scale within the trend distortion range and the adjacent time scale outside the trend distortion range are redistributed. The number of expanded time scales is coordinated with the time scale outside the trend distortion range as a whole, so that the entire change sequence remains in a continuous and progressive state within a unified time scale framework. At the same time, the time scale corresponding to the risk change position is recalculated based on the rate change trajectory after the true rate change expression is added, so that the trigger position of the critical identification signal on the time axis is moved forward. Thus, the critical identification signal is released in advance while maintaining the consistency of the numerical evolution logic, and the warning time of potential risks in the anesthesia-assisted early warning process is consistent with the actual decline rhythm.

[0085] This invention constructs a unified time scale mapping relationship between continuous vital sign time series before and after compressed transmission, forming a time reference sequence and recording the time offset trajectory. Based on this, time expansion processing is performed on continuous segments with the same values, so that the time clustering state caused by compressed expression is restored to a continuous time distribution structure that conforms to the original sampling rhythm. This eliminates the masking effect of repeated numerical segments on trend expression from the time dimension, enabling the change sequence to truly reflect the continuous evolution of vital signs, thereby improving the ability to capture the rate of decline in trend judgment and enhancing the response accuracy of anesthesia-assisted early warning in dynamic changing scenarios.

[0086] This invention conducts rate comparison analysis around the frequency of changes to form a rate change trajectory and mark the location of risk changes. It further locates the range of trend distortion and performs time stretching processing. It fills in the true rate change expression within the time window of consecutive segments with the same value, and at the same time, it reallocates the adjacent time lengths so that the restored change sequence presents a true downward rhythm under a unified time scale framework. This moves the trigger position of the critical identification signal on the time axis forward, thereby improving the timeliness of anesthesia-assisted early warning in identifying rapidly declining states and providing a more sufficient time window for clinical intervention.

[0087] This invention provides, for example Figure 2 The AI-based anesthesia auxiliary early warning system shown includes a time mapping module, a change reconstruction module, a rate determination module, a trend distortion localization module, and a rhythm recovery module.

[0088] The time mapping module, during the anesthesia-assisted early warning process, collects continuous vital signs time series before and after compressed transmission, aligns them rhythmically according to a unified time scale, generates a time comparison sequence, and records the time offset trajectory.

[0089] The change reconstruction module uses the time offset trajectory in the time reference sequence to perform time expansion processing on the continuous identical value segments in the time reference sequence. According to the sampling interval of the continuous vital sign time series before compression and transmission, the continuous identical value segments are redistributed to the time axis to generate a change sequence, and the change frequency is marked in the change sequence.

[0090] The rate determination module compares the rates of consecutive segments with the same values ​​and rapidly decreasing segments based on the frequency of change in the change sequence, determines the location of rate abrupt changes, forms a rate change trajectory, and marks the location of risk changes in the rate change trajectory.

[0091] The trend distortion localization module locates the compressed segment of the trend in the change sequence based on the location of risk changes in the rate change trajectory, determines the range of trend distortion, and generates a list of trend distortion ranges.

[0092] The rhythm recovery module performs time stretching processing on the time segments corresponding to the trend distortion range list, inserts the true rate change expression into the time window where the consecutive segments with the same value are located, and reallocates the adjacent time lengths to restore the true downward rhythm and release the critical identification signal in advance.

[0093] The artificial intelligence-based anesthesia assistance early warning method provided in this embodiment of the invention is implemented through the aforementioned artificial intelligence-based anesthesia assistance early warning system. For details of the specific methods and processes of the artificial intelligence-based anesthesia assistance early warning system, please refer to the embodiments of the aforementioned artificial intelligence-based anesthesia assistance early warning method, which will not be repeated here.

[0094] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An anesthesia-assisted early warning method based on artificial intelligence, characterized in that, Includes the following steps: During the anesthesia-assisted early warning process, the continuous vital signs time series before and after compressed transmission are collected, aligned rhythmically according to a unified time scale, and a time control sequence is generated, and the time offset trajectory is recorded. Using the time offset trajectory in the time reference sequence, the continuous identical value segments in the time reference sequence are processed by time expansion. According to the sampling interval of the continuous vital sign time series before compression and transmission, the continuous identical value segments are redistributed to the time axis to generate a change sequence, and the change frequency is marked in the change sequence. Based on the frequency of change in the change sequence, the rate of continuous identical value segments and rapidly decreasing segments is compared to determine the location of rate abrupt change, form a rate change trajectory, and mark the location of risk change in the rate change trajectory; Based on the location of risk changes in the rate change trajectory, locate the trend compression segment in the change sequence, determine the trend distortion range, and generate a list of trend distortion ranges. Time stretching is performed on the time segments corresponding to the trend distortion range list. The true rate change expression is added to the time window where the consecutive segments with the same value are located, and the adjacent time lengths are redistributed to restore the true descent rhythm and release the critical identification signal in advance.

2. The anesthesia-assisted early warning method based on artificial intelligence according to claim 1, characterized in that, The steps for collecting continuous vital signs time series before and after compression and transmission, and generating time control sequences are as follows: During the anesthesia-assisted early warning process, the continuous vital signs time series before and after compressed transmission are collected synchronously and assigned a first time marker and a second time marker, respectively. Using a unified time scale as a time reference, the first time mark and the second time mark are mapped to the same time axis coordinate system; The baseline time position of the continuous vital signs time series before compression and transmission and the actual time position of the continuous vital signs time series after compression and transmission are organized hour by hour around a unified time scale, and the time interval difference is recorded to form a time comparison sequence. The time interval differences are arranged in a uniform time scale order to form a time offset trajectory, and the time offset trajectory is associated with the time reference sequence and recorded.

3. The anesthesia-assisted early warning method based on artificial intelligence according to claim 2, characterized in that, The steps to generate a change sequence using the time offset trajectory in a time-referenced sequence are as follows: Using time-referenced sequences as the data structure, continuous segments with the same values ​​are identified within a unified time scale, and time offset trajectories at corresponding time scale positions are associated to form a time-compressed representation set. Based on the time-compressed expression set, the theoretical time span is determined by the sampling interval of the continuous vital sign time series before compression and transmission. The continuous segments with the same values ​​are then processed by time expansion and redistributed to the time axis. The numerical sequence formed by time expansion is concatenated with the uncompressed segment according to a uniform time scale to generate a change sequence; Record the numerical changes on a scale around the change sequence, and mark the frequency of change at the corresponding time scale position.

4. The anesthesia-assisted early warning method based on artificial intelligence according to claim 3, characterized in that, During the time expansion process, the numerical order of consecutive identical numerical segments is kept consistent with the time progression order. The time redistribution interval of consecutive identical numerical segments is determined by the time scale range corresponding to the time offset trajectory. The change frequency marker is bound and recorded at the corresponding time position of the change sequence according to the unified time scale.

5. The anesthesia-assisted early warning method based on artificial intelligence according to claim 3, characterized in that, The steps to determine the rate change trajectory based on the frequency of change in a changing sequence are as follows: Using the change sequence as the time analysis structure, the values ​​and change frequencies are organized step by step under a unified time scale order. The change per unit time is formed based on the value difference and time interval between adjacent time scales, and the continuous segments of the same value are identified. Based on the frequency of change, segments with consecutive identical values ​​and segments with rapid decline are segmented and classified to form corresponding rate distribution sequences under a unified time scale order; The rate distribution sequence is interpolated along a uniform time scale to form a rate change trajectory and determine the location of rate abrupt changes. Extended analysis is performed around the locations of rate abrupt changes in the rate change trajectory, and risk change locations are marked in the rate change trajectory.

6. The anesthesia-assisted early warning method based on artificial intelligence according to claim 5, characterized in that, Around the risk change location in the rate change trajectory, the time scale corresponding to the risk change location is associated and mapped with the continuous identical numerical segments in the change sequence. The time range of the rate change location is defined based on the distribution of the risk change location on the unified time scale, and the risk change location is used as the time reference for identifying the trend distortion range.

7. The anesthesia-assisted early warning method based on artificial intelligence according to claim 5, characterized in that, The steps to generate a list of trend distortion ranges based on the location of risk changes in the rate change trajectory are as follows: Using the rate change trajectory as the main time analysis line, the risk change location is extracted within a unified time scale and mapped to the change sequence for localization; By comparing and organizing the time distribution of consecutive identical value segments and rapidly declining segments around the risk change location in the change sequence, the segment where the trend is compressed is identified. Record the start and end time scales around the compressed segment of the trend and arrange them sequentially in conjunction with the location of risk changes; A list of trend distortion ranges is formed by summarizing the compressed segments of the trend according to a unified time scale.

8. The anesthesia-assisted early warning method based on artificial intelligence according to claim 7, characterized in that, The trend distortion range list records the location of risk changes and the frequency of changes within each compressed trend segment, and establishes the correspondence between the compressed trend segments and the locations of risk changes according to a unified time scale.

9. The anesthesia-assisted early warning method based on artificial intelligence according to claim 7, characterized in that, The following steps are taken to perform time stretching processing on the time segments corresponding to the list of trend distortion ranges: According to the time sequence of the trend distortion range list, the corresponding time segment is extracted from the change sequence and the continuous identical value segment and the risk change position are extracted; The theoretical time span is determined by converting the total length of the time scale with the sampling interval of the continuous vital signs time series before compression and transmission, based on the time segment corresponding to the range of trend distortion. Time extension is performed around the time window containing consecutive segments of the same value, and the actual rate of change is added. By redistributing and adjusting the time scale corresponding to the location of risk changes around the expanded time scale, critical situation identification signals can be released in advance.

10. An anesthesia-assisted early warning system based on artificial intelligence, used to implement the anesthesia-assisted early warning method based on artificial intelligence as described in any one of claims 1-9, characterized in that, It includes a time mapping module, a change reconstruction module, a rate determination module, a trend distortion localization module, and a rhythm recovery module. The time mapping module, during the anesthesia-assisted early warning process, collects continuous vital signs time series before and after compressed transmission, aligns them rhythmically according to a unified time scale, generates a time comparison sequence, and records the time offset trajectory. The change reconstruction module uses the time offset trajectory in the time reference sequence to perform time expansion processing on the continuous identical value segments in the time reference sequence. According to the sampling interval of the continuous vital sign time series before compression and transmission, the continuous identical value segments are redistributed to the time axis to generate a change sequence, and the change frequency is marked in the change sequence. The rate determination module compares the rates of consecutive segments with the same values ​​and rapidly decreasing segments based on the frequency of change in the change sequence, determines the location of rate abrupt changes, forms a rate change trajectory, and marks the location of risk changes in the rate change trajectory. The trend distortion localization module locates the compressed segment of the trend in the change sequence based on the location of risk changes in the rate change trajectory, determines the range of trend distortion, and generates a list of trend distortion ranges. The rhythm recovery module performs time stretching processing on the time segments corresponding to the trend distortion range list, inserts the true rate change expression into the time window where the consecutive segments with the same value are located, and reallocates the adjacent time lengths to restore the true downward rhythm and release the critical identification signal in advance.

Images