Search

CN-121983314-A - Dynamic early warning method and system for heart injury

CN121983314ACN 121983314 ACN121983314 ACN 121983314ACN-121983314-A

Abstract

The invention discloses a heart damage dynamic early warning method and system, which comprises the following steps of collecting continuous real-time heart monitoring data, segmenting and aligning according to time window to form a data segment sequence, preprocessing data segments, extracting heart damage monitoring indexes to form a real-time monitoring index sequence, constructing an individual statistical base line based on the time window data meeting stability criteria, calculating central parameters and discrete parameters, generating control limits, carrying out online updating on the statistical base line and the control limits in the operation process, judging the real-time monitoring indexes by using a Shewhart control chart, outputting a heart state evolution stage identifier, reconstructing operation parameters and reset rules of CUSUM variable point detection by taking the stage identifier as a condition, recursively calculating an accumulated statistic sequence, carrying out variable point judgment based on the accumulated statistic sequence, and generating and outputting a dynamic early warning result by combining the stage identifier. The invention belongs to the technical field of medical health data processing and intelligent monitoring and early warning.

Inventors

  • XIANG AILING
  • SHEN XIAOYAN
  • QU QINGMEI

Assignees

  • 青岛西凯生物技术有限公司

Dates

Publication Date
20260505
Application Date
20260128

Claims (9)

  1. 1. A dynamic early warning method for heart injury, comprising the steps of: Collecting continuous real-time heart monitoring data flow, cutting and time-aligning according to a preset time window, and generating a heart monitoring data fragment sequence; Preprocessing the cardiac monitoring data segment sequence, extracting cardiac injury monitoring indexes, and forming a real-time monitoring index sequence corresponding to a time window; selecting time window data meeting stability criteria from the real-time monitoring index sequence to construct an individual statistical base line, calculating a central parameter and a discrete parameter according to the individual statistical base line, and generating a control limit of a Shewhart control chart by utilizing the central parameter and the discrete parameter; In the continuous updating process of the real-time monitoring index sequence, the newly added real-time monitoring index is utilized to update the individual statistical base line on line, and the central parameter, the discrete parameter and the corresponding control limit are synchronously updated; Inputting the real-time monitoring index sequence and the updated control limit into a Shewhart control chart judging flow, and outputting a heart state evolution stage identifier; Reconstructing the operation parameters of CUSUM variable point detection by taking the heart state evolution stage mark as a condition, and performing cumulative statistic recurrence calculation on the real-time monitoring index sequence by using the reconstructed operation parameters to obtain a cumulative statistic sequence; and carrying out variable point judgment based on the accumulated statistic sequence, and generating a dynamic early warning result by combining the heart state evolution stage identification.
  2. 2. The method for dynamically early warning cardiac injury according to claim 1, wherein the generation of the cardiac injury dynamic early warning method is characterized in that the cardiac injury dynamic early warning method comprises the steps of continuously outputting a cardiac injury dynamic early warning data segment sequence in real time by sampling records continuously output in time sequence, establishing a unified time axis, sequentially dividing continuous time windows on the unified time axis according to preset time window lengths, classifying the sampling records with time stamps falling into the same time window into corresponding cardiac injury dynamic early warning data segments, resampling multi-type sampling records with inconsistent sampling frequencies in each cardiac injury dynamic early warning data segment according to a unified time alignment rule to obtain time-aligned cardiac injury dynamic early warning data segments, and arranging the time-aligned cardiac injury dynamic early warning data segments in time window sequence.
  3. 3. The method of claim 1, wherein the preprocessing comprises noise suppression, outlier correction and missing data compensation.
  4. 4. The method for dynamic early warning of cardiac injury according to claim 1, wherein the calculating of the central parameter and the discrete parameter and the generating of the control limit of the shawhart control chart comprise: Acquiring heart injury monitoring indexes corresponding to a plurality of time windows which are continuously arranged from a real-time monitoring index sequence to form a baseline candidate time window set; Performing stability criterion judgment on the heart injury monitoring indexes of each time window in the baseline candidate time window set, wherein the stability criterion comprises the variation amplitude limit of the same heart injury monitoring index between adjacent time windows and the fluctuation range limit of the heart injury monitoring index in the corresponding time window, and removing the time windows which do not meet the stability criterion to obtain a stable time window set; Collecting heart injury monitoring indexes of each time window in the stable time window set according to the monitoring index dimension to obtain an individual statistics baseline data set; Based on the individual statistics baseline data set, respectively calculating a central parameter and a discrete parameter for each cardiac injury monitoring index, wherein the central parameter is the mean value of the cardiac injury monitoring index in the individual statistics baseline data set, and the discrete parameter is the standard deviation of the cardiac injury monitoring index in the individual statistics baseline data set; Generating a Shewhart control chart control limit based on a central parameter and a discrete parameter corresponding to each cardiac injury monitoring index, wherein the control limit comprises an upper control limit and a lower control limit, the upper control limit is calculated by combining the central parameter and the discrete parameter according to a preset upper control coefficient, and the lower control limit is calculated by combining the central parameter and the discrete parameter according to a preset lower control coefficient; And establishing a mapping relation between the upper control limit and the lower control limit corresponding to each monitoring index and the heart damage monitoring index to form a control limit set.
  5. 5. The method of claim 4, wherein the online updating of the individual statistics baseline and the simultaneous updating of the central parameter, the discrete parameter and the corresponding control limits comprises: acquiring a heart damage monitoring index corresponding to a new arrival time window in the real-time monitoring index sequence as newly-increased monitoring index data; Performing baseline updating admission judgment on the newly-added monitoring index data, wherein the baseline updating admission judgment comprises judging whether the newly-added monitoring index data is between an upper control limit and a lower control limit of a corresponding heart damage monitoring index according to a control limit set, and adding the newly-added monitoring index data into an individual statistics baseline data set when an admission condition is met; after the newly added monitoring index data are added into the individual statistics baseline data set, a fixed capacity window is set for the individual statistics baseline data set, and the historical monitoring index data exceeding the fixed capacity window are removed in a first-in first-out mode to form an updated individual statistics baseline data set; Based on the updated individual statistical baseline data set, recalculating the central parameter and the discrete parameter according to the monitoring index dimension; And regenerating an updated upper control limit and an updated lower control limit according to a preset upper control coefficient and a preset lower control coefficient based on the recalculated central parameter and the discrete parameter, and writing the updated upper control limit and the updated lower control limit into a control limit set.
  6. 6. The method of claim 5, wherein the outputting of the cardiac injury dynamic early warning device comprises: acquiring heart damage monitoring indexes corresponding to a new arrival time window in a real-time monitoring index sequence, and reading an upper control limit and a lower control limit corresponding to each heart damage monitoring index from a control limit set; performing point location determination on each cardiac injury monitoring index based on the upper control limit and the lower control limit, wherein the point location determination comprises the steps of comparing the cardiac injury monitoring index of the current time window with the corresponding upper control limit and lower control limit to generate a point location determination result of the current time window; setting the number of preset continuous time windows, establishing a control diagram judging queue, and writing point position judging results of the current time window into the control diagram judging queue according to the time window sequence; Extracting point position judging result sequences of each heart injury monitoring index within the number of preset continuous time windows from a control chart judging queue, and executing continuity statistics on the point position judging result sequences, wherein the continuity statistics comprise the continuous occurrence times of a statistical limit, the continuous occurrence times of upper border crossing and the continuous occurrence times of lower border crossing; Under the condition that the point position judging result sequences are all within the limit, executing single-side approach judgment on each cardiac injury monitoring index, wherein the single-side approach judgment comprises judging whether the corresponding cardiac injury monitoring index continuously meets the upward control limit direction offset or the downward control limit direction offset within the preset continuous time window number, and marking the monitoring index continuously meeting the single-side offset and reaching the preset approach times as a single-side approach state; Performing phase rule matching based on continuity statistics and a single-side approaching state, wherein the phase rule matching comprises outputting a normal fluctuation phase identifier when all cardiac injury monitoring indexes in a control diagram judging queue are within limits and the single-side approaching state does not exist, outputting an edge offset phase identifier when upper and lower out-of-range conditions do not exist in the control diagram judging queue and at least one monitoring index exists in the single-side approaching state, and outputting a remarkable abnormal phase identifier when at least one monitoring index exists in the control diagram judging queue and is in the upper out-of-range or the lower out-of-range condition; and classifying the output normal fluctuation phase identification, the edge deviation phase identification and the remarkable abnormal phase identification into a heart state evolution phase identification.
  7. 7. The method for dynamically pre-warning cardiac injury according to claim 6, wherein the detecting of the change point of the conditional reconstruction CUSUM comprises: Acquiring a heart damage monitoring index corresponding to a new arrival time window in a real-time monitoring index sequence, and acquiring a heart state evolution stage identifier of the corresponding time window, and a central parameter and a discrete parameter corresponding to the heart damage monitoring index; performing standardized processing on heart injury monitoring indexes of a current time window based on the central parameter and the discrete parameter to obtain a standardized offset, and taking the standardized offset as a CUSUM cumulative input; selecting an operation mode according to the heart state evolution stage mark, and generating an operation parameter set corresponding to the selected operation mode, wherein the operation parameter set comprises an accumulation direction, a reference value, an accumulation gain, a reset rule and a judgment threshold value; When the heart state evolution stage mark is a normal fluctuation stage mark, setting an accumulation direction as bidirectional accumulation, setting a reference value and a judgment threshold value as a first reference value and a first judgment threshold value, and executing reset on accumulation statistics when a standardized offset symbol is overturned or the standardized offset falls back to a preset fall-back range according to a first reset rule; when the heart state evolution stage mark is an edge offset stage mark, setting an accumulation direction as single-side accumulation, wherein the single-side direction is determined by a single-side approaching state in a control diagram judging queue, switching a reference value and a judging threshold value into a second reference value and a second judging threshold value, and executing partial attenuation on accumulation statistics when the single-side approaching state is released or the single-side direction is switched according to a second resetting rule; When the heart state evolution stage mark is a significant abnormal stage mark, setting an accumulation direction as single-side accumulation consistent with the out-of-range direction, switching a reference value and a judgment threshold value into a third reference value and a third judgment threshold value, and resetting accumulation statistics when the out-of-range direction disappears or the stage mark is switched according to a third resetting rule; And performing cumulative statistic recurrence calculation on the standardized offset based on the selected operation parameter set, wherein the cumulative statistic recurrence satisfies the following conditions: ; Wherein, the For a normalized offset for the current time window, As a reference value for the value of the reference, In order to accumulate the gain, For the cumulative statistic of the current time window, For the time window index to be used, Is the first The cumulative statistics of the individual time windows, For a truncating operator, defining a lower bound of the cumulative statistic as 0; and writing the accumulated statistics corresponding to each time window into the accumulated statistics sequence according to the time window sequence.
  8. 8. The method for dynamic early warning of cardiac injury according to claim 7, wherein the generating of the dynamic early warning result comprises: Acquiring accumulated statistics corresponding to a new arrival time window in an accumulated statistics sequence, and acquiring a heart state evolution stage identifier corresponding to the corresponding time window and a judgment threshold corresponding to the corresponding stage identifier; comparing the accumulated statistic with the decision threshold, when the accumulated statistic satisfies Determining the corresponding time window as a variable point time window and recording a variable point time window index, wherein For the cumulative statistic of the current time window, A decision threshold corresponding to the current heart state evolution stage identification; And reading the corresponding heart state evolution stage identification based on the variable point time window index, and carrying out combined mapping on the variable point time window index and the heart state evolution stage identification to generate a dynamic early warning result comprising an early warning time window index and an early warning grade.
  9. 9. The heart injury dynamic early warning system is characterized by comprising the following modules: The data segment generation module is used for collecting continuous real-time heart monitoring data streams, segmenting and time-aligning the continuous real-time heart monitoring data streams according to a preset time window, and generating a heart monitoring data segment sequence; the index sequence generating module is used for preprocessing the cardiac monitoring data fragment sequence and extracting cardiac injury monitoring indexes to form a real-time monitoring index sequence; the control limit set generation module is used for selecting time window data meeting a stability criterion from the real-time monitoring index sequence to construct an individual statistical base line, calculating a central parameter and a discrete parameter based on the individual statistical base line, and generating a Shewhart control chart control limit by the central parameter and the discrete parameter to form a control limit set; The limit synchronous updating module is used for carrying out online updating on the individual statistical base line by utilizing the heart injury monitoring index of the new arrival time window in the real-time monitoring index sequence and synchronously updating the central parameter, the discrete parameter and the control limit set; The stage identifier output module is used for executing a Shewhart control chart judging flow based on the real-time monitoring index sequence and the control limit set and outputting a heart state evolution stage identifier; The accumulated statistic sequence generating module is used for reconstructing CUSUM operation parameters by taking the heart state evolution stage mark as a condition, and performing accumulated statistic recurrence calculation on the real-time monitoring index sequence based on the reconstructed operation parameters to generate an accumulated statistic sequence; and the dynamic early warning result generation module is used for carrying out variable point judgment based on the accumulated statistic sequence and generating a dynamic early warning result by combining with the heart state evolution stage identification.

Description

Dynamic early warning method and system for heart injury Technical Field The invention relates to the technical field of medical health data processing and intelligent monitoring and early warning, in particular to a dynamic early warning method and system for heart injury. Background With the development of wearable equipment, remote medical treatment and off-hospital continuous monitoring technology, heart health dynamic monitoring based on multi-source physiological signals such as electrocardio, heart rate, blood oxygen saturation, pulse wave and the like has become an important application scene of chronic disease management and acute event risk identification, and currently, continuous real-time heart monitoring data is subjected to feature extraction in heart injury risk assessment, and an early warning result is output by a statistical detection method combined with threshold judgment or fixed rules. The existing heart injury early warning technology still has obvious limitations in real-time performance, dynamic performance and individual adaptation aspects. On the one hand, the traditional threshold value out-of-range or single-point abnormality alarming mode depends on a fixed threshold value and static state judging logic, so that the evolution process that the heart state shifts from normal fluctuation to edge to obvious abnormality is difficult to be described, and the situation that response to progressive injury signals is delayed or false alarm is generated to short-time physiological fluctuation is easy to occur. On the other hand, the existing variable point detection method mostly adopts fixed reference values, accumulation directions and reset rules to calculate accumulation statistics, and lacks an operation parameter reconstruction mechanism combined with a heart state evolution stage, so that the variable point judgment process is difficult to adaptively adjust along with the distribution change of real-time monitoring data, and real-time and dynamic monitoring and early warning of heart damage risks are difficult to realize. Therefore, how to provide a method and a system for dynamic early warning of heart injury is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention aims at providing a heart damage dynamic early warning method and system, wherein the method combines a Shewhart control chart with CUSUM accumulated variable point detection, performs window alignment, pretreatment and monitoring index extraction on continuous real-time heart monitoring data, generates and updates a control limit on line based on an individual statistical baseline, outputs a heart state evolution stage identifier, reconstructs accumulated statistic recurrence parameters based on the stage identifier to realize variable point judgment and dynamic early warning output, and has the advantages of strong instantaneity, sensitivity to progressive risks, low false alarm rate and adaptation to individual differences. The method and the system for dynamically early warning the heart injury according to the embodiment of the invention comprise the following steps: Collecting continuous real-time heart monitoring data flow, cutting and time-aligning according to a preset time window, and generating a heart monitoring data fragment sequence; Preprocessing the cardiac monitoring data segment sequence, extracting cardiac injury monitoring indexes, and forming a real-time monitoring index sequence corresponding to a time window; selecting time window data meeting stability criteria from the real-time monitoring index sequence to construct an individual statistical base line, calculating a central parameter and a discrete parameter according to the individual statistical base line, and generating a control limit of a Shewhart control chart by utilizing the central parameter and the discrete parameter; In the continuous updating process of the real-time monitoring index sequence, the newly added real-time monitoring index is utilized to update the individual statistical base line on line, and the central parameter, the discrete parameter and the corresponding control limit are synchronously updated; Inputting the real-time monitoring index sequence and the updated control limit into a Shewhart control chart judging flow, and outputting a heart state evolution stage identifier; Reconstructing the operation parameters of CUSUM variable point detection by taking the heart state evolution stage mark as a condition, and performing cumulative statistic recurrence calculation on the real-time monitoring index sequence by using the reconstructed operation parameters to obtain a cumulative statistic sequence; and carrying out variable point judgment based on the accumulated statistic sequence, and generating a dynamic early warning result by combining the heart state evolution stage identification. Optionally, the generation of the cardiac monitoring data segment sequence compris