Search

CN-122000029-A - Esophageal high-resolution pressure measurement data intelligent diagnosis and risk assessment system based on machine learning

CN122000029ACN 122000029 ACN122000029 ACN 122000029ACN-122000029-A

Abstract

The invention relates to the technical field of machine learning diagnosis, in particular to an esophagus high-resolution pressure measurement data intelligent diagnosis and risk assessment system based on machine learning. According to the invention, stable identification of swallowing key time nodes is realized through a base line pressure analysis and swallowing period boundary generation mode, standardized expression of a pressure time sequence is pushed to be completed in a unified time scale, adaptability deficiency caused by dependence on fixed rules and manual labeling in existing processing is relieved, identification capacity of different sampling state leading pressure change stages is strengthened through construction of resampling sequences under various time sampling conditions and dynamic time rule comparison, judgment deviation caused by neglecting time sequence consistency characteristics in an analysis process is made up, and a sensor channel is screened by combining with esophageal axial pressure distribution.

Inventors

  • LI LINGMIN

Assignees

  • 中国人民解放军联勤保障部队第九六〇医院

Dates

Publication Date
20260508
Application Date
20260127

Claims (10)

  1. 1. Machine learning-based esophageal high-resolution pressure measurement data intelligent diagnosis and risk assessment system is characterized in that the system comprises: The data analysis module is used for collecting multichannel continuous voltage signals of the esophageal high-resolution pressure measuring catheter, calculating multichannel pressure time sequences, analyzing baseline pressure values and continuous pressure changes, judging swallowing starting and ending time points, generating swallowing period data and transmitting the swallowing period data to the standard mapping module; The standard mapping module intercepts the multichannel pressure time sequence based on the swallowing period data, performs interpolation and time axis remapping, calculates a unified time scale pressure value sequence, generates a standardized pressure sequence and transmits the standardized pressure sequence to the time alignment module; The time alignment module is used for aligning resampling sequences at a plurality of time sampling intervals based on the standardized pressure sequence, comparing time indexes of the same sensor pressure entering a dominant change stage, generating a time consistency feature set and transmitting the time consistency feature set to the function evaluation module; The function evaluation module screens sensor channels with stable swallowing period based on the time consistency feature set, divides the sensor into continuous sections along the axial direction of the esophagus, analyzes the number proportion of the sensor channels with dominant pressure change, calculates the pressure responsibility distribution of the multiple sections, combines the time consistency feature set, and judges the swallowing function state and disease risk level data.
  2. 2. The machine learning based esophageal high-resolution pressure measurement data intelligent diagnosis and risk assessment system of claim 1, wherein the swallowing cycle data comprises a swallowing start time point, a swallowing end time point, a single swallowing duration time, the standardized pressure sequence comprises a uniform time scale index, a multi-channel corresponding pressure value sequence and a multi-channel synchronous sequence number, the time consistent feature set comprises a dominant change start time index, an inter-channel time offset set and a resampling consistency identifier, and the swallowing function status and disease risk level data comprises a function status category identifier, a risk level identifier and a section pressure responsibility distribution.
  3. 3. The machine learning based esophageal high-resolution pressure measurement data intelligent diagnosis and risk assessment system of claim 1, wherein the data analysis module comprises: The signal analysis submodule collects multichannel continuous voltage signals of the esophagus high-resolution pressure measuring catheter, performs time stamp alignment on original voltage samples based on multichannel synchronous time sequences, detects sampling intervals and eliminates missing points, and completes signal rearrangement according to catheter channel numbers to generate multichannel voltage time sequences; The pressure sequence sub-module is used for converting the voltage value point by point into a corresponding channel pressure value according to the channel calibration relation based on the multi-channel voltage time sequence, selecting a swallowing-free section, calculating a multi-channel pressure average value as a baseline pressure value and generating a multi-channel pressure time sequence; And the swallowing period sub-module is used for continuously detecting the multi-channel pressure change rate based on the multi-channel pressure time sequence, marking the rising point and the falling end point of the pressure according to the set pressure change threshold value, judging the starting time point and the ending time point of swallowing, and generating swallowing period data.
  4. 4. The intelligent diagnosis and risk assessment system for esophageal high-resolution pressure measurement data based on machine learning according to claim 3, wherein the pressure change threshold is determined by counting the distribution of baseline pressure values and continuous pressure difference values in a multichannel pressure time sequence, selecting a swallowing-free section pressure sample to calculate a pressure fluctuation amplitude mean value and combining with a preset 3-time standard deviation summation calculation.
  5. 5. The machine learning based esophageal high-resolution pressure measurement data intelligent diagnosis and risk assessment system of claim 1, wherein the standard mapping module comprises: The interval intercepting sub-module is used for executing time interval interception on the multi-channel pressure time sequence based on the swallowing period data, carrying out time judgment on multi-channel pressure sampling points at the starting and ending time points of each swallowing period, reserving pressure values in a period boundary, and eliminating samples in other time periods to generate a period pressure fragment sequence; The time interpolation sub-module is used for detecting inconsistent time interval positions according to multichannel sampling time distribution based on the periodic pressure segment sequence, selecting pressure values of adjacent sampling points to conduct time position calculation, supplementing pressure values corresponding to missing time scales, and rearranging to obtain a continuous time pressure sequence; and the scale mapping sub-module is used for selecting the same time length as a mapping reference according to the continuous time pressure sequence, re-marking the multi-channel pressure value positions according to the time scales with equal intervals, and sequentially arranging the pressure values under the same time scales after mapping to generate a standardized pressure sequence.
  6. 6. The machine learning based esophageal high-resolution pressure data intelligent diagnosis and risk assessment system of claim 1, wherein the time alignment module comprises: the resampling construction submodule is used for constructing resampling sequences under the condition of a plurality of time sampling intervals based on the standardized pressure sequences, collecting standardized pressure values and corresponding time marks of each sensor channel, and equally dividing a time axis to generate a multi-interval resampling sequence set; the sequence alignment sub-module detects a channel pressure arrangement starting time mark under the multi-sampling condition according to the multi-interval resampling sequence set, selects a unified reference time point as an alignment standard, carries out translation adjustment on the time index of the multi-sampling sequence, eliminates starting offset and rearranges the time sequence to obtain an alignment resampling sequence set; and the index difference sub-module is used for monitoring the continuous increase section of the pressure change rate for the multi-sensor channels based on the aligned resampling sequence group, judging the starting time index entering the leading pressure change stage, executing the mutual difference calculation between the channels for the corresponding time index under a plurality of sampling conditions, and generating a time consistency feature set.
  7. 7. The machine learning based esophageal high-resolution pressure measurement data intelligent diagnosis and risk assessment system of claim 1, wherein the functional assessment module comprises: The consistency screening submodule screens the stable sensor channels of the swallowing period based on the time consistency characteristic set, acquires a multi-channel pressure time sequence, analyzes the phase alignment degree and the period offset, reserves the serial numbers of the sensor channels meeting the time consistency threshold value, gathers and sorts the serial numbers, and generates a stable channel index set; The axial dividing sub-module is used for acquiring a corresponding channel space sequence along the axial direction of the esophagus according to the stable channel index set, acquiring position and interval data of adjacent channels, performing continuity judgment on interval difference values, and dividing to form a continuous section number set which is not overlapped with each other to obtain an axial continuous section sequence; And the responsibility evaluation sub-module is used for acquiring the pressure change time of the sensors in the multiple sections based on the axial continuous section sequence, calculating the channel quantity ratio of the sensor for leading the pressure change, analyzing the pressure responsibility distribution of the multiple sections, and combining the state discrimination of the time-consistent feature set to generate the swallowing function state and disease risk grade data.
  8. 8. The intelligent diagnosis and risk assessment system for esophageal high-resolution pressure measurement data based on machine learning according to claim 7, wherein the time coincidence threshold is determined by performing a period alignment process on a multi-sensor channel pressure time sequence in a time coincidence feature set, obtaining multi-channel phase offset and period length offset values in a swallowing period, counting all channel corresponding offset distribution intervals, and calculating a median value of the offset distribution intervals.
  9. 9. The machine learning based esophageal high-resolution pressure data intelligent diagnosis and risk assessment system of claim 1, further comprising: the model correction module is used for inputting a two-way long-short-term memory network based on the swallowing function state and disease risk level data, the standardized pressure sequence and the time consistency feature set, correcting and judging boundary adjustment are carried out on the swallowing function state and disease risk level data through learning of result consistency and judging confidence corresponding to a plurality of swallowing modes, and a swallowing function state and disease risk level correction result is generated; The swallowing function state and disease risk level correction result comprises a corrected function state mark, a corrected risk level mark and a state adjustment mark.
  10. 10. The machine learning based esophageal high-resolution pressure data intelligent diagnosis and risk assessment system of claim 9, wherein the model correction module comprises: The data integration sub-module is used for acquiring data items corresponding to multiple swallowing events based on the swallowing function state and disease risk level data, the standardized pressure sequence and the time consistency feature set, performing time index alignment and length consistency processing on multiple data sources and generating a swallowing multi-source feature fusion vector set; The confidence evaluation sub-module calls output labels corresponding to a plurality of swallowing modes according to the swallowing multisource feature fusion vector set, performs consistency judgment on the state judgment result under the same swallowing event, counts the distribution frequency of the state labels, calculates the concentration metric value and gathers the concentration metric value, and generates a judgment confidence sequence; and the boundary correction sub-module is used for executing boundary offset calculation according to the swallowing function state and disease risk level data distribution interval corresponding to the multi-concentration measurement based on the discrimination confidence coefficient sequence, correcting the original state demarcation limit and mapping to the multi-swallowing event mark, and generating a swallowing function state and disease risk level correction result.

Description

Esophageal high-resolution pressure measurement data intelligent diagnosis and risk assessment system based on machine learning Technical Field The invention relates to the technical field of machine learning diagnosis, in particular to an intelligent diagnosis and risk assessment system for esophageal high-resolution pressure measurement data based on machine learning. Background The technical field of machine learning diagnosis is the technical field of learning modeling on medical related data by using a computer algorithm and being used for auxiliary judgment and analysis, and the technical field is generally developed around the core matters of acquisition digitization, feature expression, sample labeling, model training, inference and the like of medical data, covers the numerical representation of physiological signals, image data and clinical indexes, forms a calculation model for medical diagnosis analysis through a statistical learning method and a supervised or unsupervised learning process, and is an important direction of combining medical detection data with an information processing technology. The traditional intelligent diagnosis and risk assessment system for esophageal high-resolution pressure measurement data based on machine learning is a system for analyzing and assessing individual pressure measurement results according to preset risk classification standards on the basis of a multichannel pressure value sequence acquired in the process of esophageal high-resolution pressure measurement inspection by preprocessing pressure data of each measuring point acquired by a pressure measurement catheter by adopting a manually established pressure section division rule, a pressure peak extraction step and a time sequence alignment mode, constructing a training data set according to a manually marked disease type or abnormal state sample, judging esophageal dynamics states through a set classification or regression learning flow. In the prior art, by carrying out traditional pressure section division, pressure peak value extraction and time sequence alignment treatment on esophageal high-resolution pressure measurement data, although a certain reference can be provided for esophageal dynamics state discrimination, the esophageal dynamics state discrimination is dependent on manually set rules, so that the esophageal dynamics state discrimination cannot flexibly cope with measurement conditions of different individuals and changes in practical application. The construction of manually marked disease types or abnormal state samples also easily brings sample deviation, and influences the accuracy of the model. Furthermore, existing systems fail to adequately account for the timing consistency between the sequence of pressure changes during swallowing and the effectiveness of each sensor channel, thereby limiting accurate assessment of individual swallowing function status and accurate stratification of disease risk. Disclosure of Invention In order to solve the technical problems in the prior art, the embodiment of the invention provides an intelligent diagnosis and risk assessment system for esophageal high-resolution pressure measurement data based on machine learning. The technical scheme is as follows: In one aspect, an intelligent diagnosis and risk assessment system for esophageal high-resolution manometry data based on machine learning is provided, the system comprising: The data analysis module is used for collecting multichannel continuous voltage signals of the esophageal high-resolution pressure measuring catheter, calculating multichannel pressure time sequences, analyzing baseline pressure values and continuous pressure changes, judging swallowing starting and ending time points, generating swallowing period data and transmitting the swallowing period data to the standard mapping module; The standard mapping module intercepts the multichannel pressure time sequence based on the swallowing period data, performs interpolation and time axis remapping, calculates a unified time scale pressure value sequence, generates a standardized pressure sequence and transmits the standardized pressure sequence to the time alignment module; The time alignment module is used for aligning resampling sequences at a plurality of time sampling intervals based on the standardized pressure sequence, comparing time indexes of the same sensor pressure entering a dominant change stage, generating a time consistency feature set and transmitting the time consistency feature set to the function evaluation module; The function evaluation module is used for screening sensor channels with stable swallowing period based on the time-consistent feature set, dividing the sensor into continuous sections along the axial direction of the esophagus, analyzing the number proportion of the channels of the sensor with dominant pressure change, calculating the pressure responsibility distribution of the multiple sections, combini