CN111407244A - Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm - Google Patents

Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm Download PDF

Info

Publication number
CN111407244A
CN111407244A CN202010234811.3A CN202010234811A CN111407244A CN 111407244 A CN111407244 A CN 111407244A CN 202010234811 A CN202010234811 A CN 202010234811A CN 111407244 A CN111407244 A CN 111407244A
Authority
CN
China
Prior art keywords
volume
pressure
current
left ventricular
systolic
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010234811.3A
Other languages
Chinese (zh)
Inventor
肖伟平
辛立明
孙钰
米特什·巴迪瓦拉
郑树森
李兰娟
彭艳
罗均
谢少荣
蒲华燕
郑杰
李建辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Transpacific Technology Development Ltd
University of Shanghai for Science and Technology
Original Assignee
Beijing Transpacific Technology Development Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Transpacific Technology Development Ltd filed Critical Beijing Transpacific Technology Development Ltd
Priority to CN202010234811.3A priority Critical patent/CN111407244A/en
Publication of CN111407244A publication Critical patent/CN111407244A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/02028—Determining haemodynamic parameters not otherwise provided for, e.g. cardiac contractility or left ventricular ejection fraction
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7246—Details of waveform analysis using correlation, e.g. template matching or determination of similarity
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Physiology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Signal Processing (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Cardiology (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Fuzzy Systems (AREA)
  • Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)

Abstract

The invention relates to a left ventricle end-systolic elasticity prediction method and a system based on SVR algorithm, which comprises the steps of extracting current key characteristics from a single current pressure-volume ring, wherein the current key characteristics are end-systolic volume, end-diastolic pressure, area of the pressure-volume ring and end-systolic elastic potential energy; determining the volume when the current left ventricle pressure is zero according to the current key characteristics and a trained volume prediction model when the left ventricle pressure is zero; the volume prediction model is obtained by adopting SVR algorithm training according to a plurality of groups of historical current key characteristics and the corresponding volumes when the historical left ventricle pressure is zero; and calculating the current left ventricular end systolic elasticity according to the volume when the current left ventricular pressure is zero and the constructed left ventricular end systolic elasticity prediction model. The method can reduce invasive injury to a great extent on the premise of accurately predicting the left ventricle end-systolic elasticity, and is very suitable for isolated heart perfusion ESHP experiments.

Description

Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm
Technical Field
The invention relates to the fields of biomedical engineering technology and computer science, in particular to a left ventricle end-systolic elasticity prediction method and a system based on an SVR algorithm.
Background
Heart transplantation is the gold standard for treating patients with end-stage heart failure. Ex vivo heart perfusion (ESHP) is a promising technique for assessing the viability of a donor heart prior to heart transplantation. The technique can maintain normal metabolism of the donor heart in a beating state, and provides an opportunity for assessing the viability of organs before transplantation. Left ventricular end-systolic elasticity (E)es) Is an effective measure of left ventricular systolic function and left ventricular end-systolic elasticity EesHas been shown to be an indicator of the force of contraction that is not sensitive to loading. This gives the left ventricle end-systolic elasticity EesBecomes an ideal parameter for assessing donor heart function in ex vivo heart perfusion ESHP.
Traditionally, left ventricular end-systolic elasticity EesMeasured by altering left ventricular preload or afterload conditions, such as by inferior vena cava occlusion, and obtaining multiple pressure-volume rings (PV loop) simultaneously to establish an end-systolic pressure-volume relationship (ESPVR). However, for ex vivo cardiac perfusion of ESHP, the following problems exist with conventional measurement methods: first, the heart is loweredBlood infusion can cause damage to the donor heart, such as tissue hypoxia, ischemic injury, and cardiac arrhythmia. Secondly, the conventional measurement method cannot realize real-time measurement, so that the heart problem cannot be found and treated in time. Finally, it is difficult to generate multiple stably measured pressure-volume loops PV loop in an ex vivo heart perfusion ESHP environment, because ex vivo hearts typically recover at a relatively low volume loading compared to in vivo hearts, while lower left ventricular volumes typically produce poorer pressure-volume loops PV loop, especially during occlusion. This makes it difficult for conventional methods to extract left ventricular end systolic elasticity E ex vivo through these pressure-volume loops PV loopes. There are several in vivo estimates of left ventricular end-systolic elasticity EesThe method of (1). Some researchers estimate left ventricular end-systolic elasticity E by using the geometry of the left ventricular time-varying elasticity curve or the left ventricular isovolumetric peak pressure (Pmax)es. However, these left ventricular end-systolic elasticity EesThe estimates all rely on having clear, distinguishable isovolumetric systolic and ejection phases on the in vivo measured pressure-volume loop PV loop, which is difficult to obtain in an ex vivo heart perfusion ESHP environment, as shown in fig. 1.
Disclosure of Invention
The invention aims to provide a left ventricle end-systolic elasticity prediction method and a left ventricle end-systolic elasticity prediction system based on an SVR algorithm, which greatly reduce invasive injury on the premise of accurately predicting left ventricle end-systolic elasticity and are very suitable for isolated heart perfusion ESHP experiments.
In order to achieve the purpose, the invention provides the following scheme:
a left ventricle end-systolic elasticity prediction method based on an SVR algorithm comprises the following steps:
obtaining a single current pressure-volume loop measured at steady state;
extracting current key features from all hemodynamic features of the current pressure-volume ring based on a feature recursive elimination algorithm of a filter and a support vector machine; the current key features include end systolic volume, end diastolic pressure, area of the pressure-volume ring, and end systolic elastic potential energy;
determining the volume when the current left ventricular pressure is zero according to the current key characteristics and a trained volume prediction model when the left ventricular pressure is zero; the trained volume prediction model with the left ventricular pressure being zero is obtained by adopting SVR algorithm training according to multiple groups of historical current key features and the volume with the historical left ventricular pressure being zero corresponding to each group of historical current key features; the trained volume prediction model with the left ventricular pressure being zero represents a nonlinear mapping relation between the key features and the volume with the left ventricular pressure being zero;
calculating the current left ventricle end-systolic elasticity according to the volume when the current left ventricle pressure is zero and the constructed left ventricle end-systolic elasticity prediction model; the left ventricular end systole elasticity prediction model is constructed according to the volume when the left ventricular pressure is zero and the pressure-volume coordinates of the end systole on a single pressure-volume ring.
Optionally, before executing a feature recursive elimination algorithm based on a filter and a support vector machine to extract a current key feature from all hemodynamic features of the current pressure-volume loop, the method further includes:
all hemodynamic characteristics on the current pressure-volume loop were extracted using IOX software.
Optionally, the feature recursive elimination algorithm based on the filter and the support vector machine extracts a current key feature from all hemodynamic features of the current pressure-volume loop, and specifically includes:
extracting features with stability greater than a first set threshold from all the hemodynamic features according to an intra-group consistency index, and storing the features in a first set;
according to the correlation coefficient, removing the features with redundancy larger than a second set threshold from the first set, and storing the remaining features in a second set;
and extracting the current key features from the second set based on a filter and a feature recursive clearing algorithm of a support vector machine.
Optionally, the calculating the current left ventricular end systole elasticity according to the volume when the current left ventricular pressure is zero and the constructed left ventricular end systole elasticity prediction model specifically includes:
according to the formula
Figure BDA0002430622750000031
Calculating the current left ventricular end-systolic elasticity;
wherein E isesRepresenting the current left ventricular end-systolic elasticity; v0Represents the volume when the current left ventricular pressure is zero; the last systolic pressure volume coordinate on the pressure-volume ring is (V)es，Pes)，VesDenotes the end-systolic volume, PesIndicating end-systolic pressure.
An SVR algorithm-based left ventricular end systolic elasticity prediction system, comprising:
the current pressure-volume ring acquisition module is used for acquiring a single current pressure-volume ring measured in a steady state;
a current key feature extraction module, configured to extract a current key feature from all hemodynamic features of the current pressure-volume ring based on a feature recursive elimination algorithm of a filter and a support vector machine; the current key features include end systolic volume, end diastolic pressure, area of the pressure-volume ring, and end systolic elastic potential energy;
the volume prediction module is used for determining the volume when the current left ventricular pressure is zero according to the current key characteristics and the trained volume prediction model when the current left ventricular pressure is zero; the trained volume prediction model with the left ventricular pressure being zero is obtained by adopting SVR algorithm training according to multiple groups of historical current key features and the volume with the historical left ventricular pressure being zero corresponding to each group of historical current key features; the trained volume prediction model with the left ventricular pressure being zero represents a nonlinear mapping relation between the key features and the volume with the left ventricular pressure being zero;
the current left ventricle end-systolic elasticity calculation module is used for calculating the current left ventricle end-systolic elasticity according to the volume when the current left ventricle pressure is zero and the constructed left ventricle end-systolic elasticity prediction model; the left ventricular end systole elasticity prediction model is constructed according to the volume when the left ventricular pressure is zero and the pressure-volume coordinates of the end systole on a single pressure-volume ring.
Optionally, the method further includes:
a hemodynamic feature extraction module to extract all hemodynamic features on the current pressure-volume loop using IOX software.
Optionally, the current key feature extraction module specifically includes:
a first set determination unit, configured to extract, according to the intra-group consistency index, features with stability greater than a first set threshold from all the hemodynamic features, and store the features in a first set;
a second set determining unit, configured to remove, according to the correlation coefficient, a feature with redundancy greater than a second set threshold from the first set, and store the remaining features in a second set;
and the current key feature extraction unit is used for extracting current key features from the second set based on a filter and a feature recursive elimination algorithm of a support vector machine.
Optionally, the current left ventricular end-systolic elasticity calculation module specifically includes:
a current left ventricular end-systolic elasticity calculation unit for calculating the current left ventricular end-systolic elasticity according to the formula
Figure BDA0002430622750000041
Calculating the current left ventricular end-systolic elasticity;
wherein E isesRepresenting the current left ventricular end-systolic elasticity; v0Represents the volume when the current left ventricular pressure is zero; the last systolic pressure volume coordinate on the pressure-volume ring is (V)es，Pes)，VesDenotes the end-systolic volume, PesIndicating end-systolic pressure.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a left ventricle end-systolic elasticity prediction method and a system based on SVR algorithm, which mainly comprises the steps of establishing a nonlinear mapping relation between hemodynamic characteristics and the volume when the left ventricle pressure is zero through the SVR algorithm, and calculating the left ventricle end-systolic elasticity by predicting the volume when the left ventricle pressure is zero; in the practical application process, only a single pressure-volume loop PV loop needs to be measured through the steps, so that the left ventricular end systole elasticity can be accurately predicted, invasive injury is reduced to a great extent, and the method is very suitable for isolated heart perfusion ESHP experiments.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a graph comparing an in vivo measured pressure-volume loop PV loop with an ex vivo measured pressure-volume loop PV loop;
FIG. 2 is a flowchart of a method for predicting end-systolic elasticity of left ventricle based on SVR algorithm in accordance with embodiment 1 of the present invention;
FIG. 3 is an exemplary flowchart of an embodiment 2 of the present invention of a method for left ventricular end systolic elasticity prediction based on SVR algorithm;
FIG. 4 is a schematic diagram of the calculation of the end-systolic elasticity of the left ventricle in embodiment 2 of the present invention;
FIG. 5 is a block diagram of the SVR algorithm-based left ventricular end systolic elasticity prediction system according to embodiment 3 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a left ventricle end-systolic elasticity prediction method and a left ventricle end-systolic elasticity prediction system based on an SVR algorithm, which greatly reduce invasive injury on the premise of accurately predicting left ventricle end-systolic elasticity and are very suitable for isolated heart perfusion ESHP experiments.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
SVR is an english abbreviation of Support Vector Regression (Support Vector Regression), and is an important application branch of Support Vector Machines (SVMs).
Example 1
As shown in fig. 2, the present embodiment provides a left ventricular end systole elasticity prediction method based on SVR algorithm, which includes:
step 101: a single current pressure-volume loop measured at steady state is acquired.
Step 102: extracting current key features from all hemodynamic features of the current pressure-volume ring based on a feature recursive elimination algorithm of a filter and a support vector machine; the current key features include end systolic volume, end diastolic pressure, area of the pressure-volume ring, and end systolic elastic potential energy.
Wherein the end systolic volume (V)es): volume after left ventricular systole is over; end diastolic volume (V)ed): the volume of the left ventricle just beginning to contract; end diastolic pressure (P)ed): pressure at the beginning of contraction of the left ventricle; area of pressure-volume loop (PVA): total mechanical energy generated when the ventricles contract; end-systolic elastic Potential (PE): the end-systole stores the elastic potential energy of the ventricular wall.
The characteristic recursion elimination algorithm based on the filter and the support vector machine integrates three indexes of characteristic stability, characteristic redundancy and prediction precision.
Step 103: determining the volume when the current left ventricular pressure is zero according to the current key characteristics and a trained volume prediction model when the left ventricular pressure is zero; the trained volume prediction model with the left ventricle pressure being zero is obtained by adopting an SVR algorithm in a machine learning algorithm for training according to a plurality of groups of historical current key features and the volume with the historical left ventricle pressure being zero corresponding to each group of historical current key features; the trained volume prediction model with zero left ventricular pressure represents a nonlinear mapping relationship between the key features and the volume with zero left ventricular pressure.
Step 104: calculating the current left ventricle end-systolic elasticity according to the volume when the current left ventricle pressure is zero and the constructed left ventricle end-systolic elasticity prediction model; the left ventricular end systole elasticity prediction model is constructed according to the volume when the left ventricular pressure is zero and the pressure-volume coordinates of the end systole on a single pressure-volume ring. The expression of the left ventricle end-systolic elasticity prediction model is
Figure BDA0002430622750000061
Wherein E isesRepresenting the current left ventricular end-systolic elasticity; v0Represents the volume when the current left ventricular pressure is zero; the last systolic pressure volume coordinate on the pressure-volume ring is (V)es，Pes)，VesDenotes the end-systolic volume, PesIndicating end-systolic pressure.
Before step 102 is executed, the method further includes:
all hemodynamic characteristics on the current pressure-volume loop were extracted using IOX software.
Step 102 specifically includes:
features having a stability greater than a first set threshold are extracted from all of the hemodynamic features according to an intra-group consistency indicator and stored in a first set.
And according to the correlation coefficient, removing the features with redundancy greater than a second set threshold from the first set, and storing the remaining features in a second set.
And extracting the current key features from the second set based on a filter and a feature recursive clearing algorithm of a support vector machine.
Step 104 specifically includes:
according to the formula
Figure BDA0002430622750000071
The current left ventricular end-systolic elasticity is calculated.
Example 2
As shown in fig. 3, the present embodiment also provides a left ventricular end systole elasticity prediction method based on SVR algorithm, which is divided into three stages from training to actually using:
the first stage is to train the model, which requires much historical data to be obtained for training the model, and the raw data directly acquired by the sensor is a multi-beat pressure-volume loop PV loop (the measured data portion shown in fig. 3) obtained by a conventional occlusion operation. From which can be extracted: 1. single steady state PV loop data (one of the multi-beat pressure-volume loop PV loop is selected) that will be further extracted as hemodynamic parameters; 2. gold standard V0、Ees. 34 of these hemodynamic parameters and gold standard V0A single sample (1x35) is constructed. All the acquired samples were used as a model data set (Nx35), and key features (5 out of 34 swirl kinetic parameters were selected) were selected using a filter and support vector machine based feature recursive elimination algorithm. The data set (Nx6) containing only key features was divided into a training set (75%) and a test set (25%). The SVR model is then trained in a training set.
The second stage is testing the model, and verifying the reliability of the model in the testing set.
The third stage is to apply the model by first measuring a single PV loop at steady state (where no blocking is required to obtain multiple PV loops to obtain V0And EesThese are obtained by trained models), from which key features are extracted (where there is no need to extract complete hemodynamic parameters, norThe feature recursive elimination algorithm based on the filter and the support vector machine needs to be reused, because the key features in the feature recursive elimination algorithm are analyzed based on a large amount of data in the first stage, and the key features are directly extracted). Taking the extracted key features as input, and outputting V by the model0According to V, and thenes、PesCalculation of Ees。
Comprises the following steps.
The method comprises the following steps: and acquiring and processing the measurement data.
First, pressure data and volume data of the heart are measured by the PV catheter, and a single pressure-volume loop PVloop is obtained.
Secondly, a plurality of pressure-volume loops PV loop of each heart can be obtained through the occlusion operation, and the end-systolic pressure-volume coordinate (V) of each pressure-volume loop PV loop is determinedes(end-systolic volume), Pes(last systolic pressure)) is subjected to least squares regression to obtain a pressure-volume relationship ESPVR line (as shown in the measured data portion of FIG. 3), and the slope of the pressure-volume relationship ESPVR line is the last systolic elasticity EesEnd-of-contraction elasticity EesThe intercept with the volume axis being referred to as volume V when the left ventricle pressure is zero0Here the end-systolic elasticity EesVolume V at zero pressure in the left ventricle0The gold standard value obtained was measured as a conventional method.
Then, a single pressure-volume loop PV loop is extracted from the plurality of pressure-volume loops PV loop as single-beat data measured at steady state (theoretically, any one pressure-volume loop PV loop can be extracted, and the end-systolic elasticity E corresponding to each PV loop isesAll the same, the single-shot data used in this embodiment is a pressure-volume loop PVloop) without operating by occlusion, and hemodynamic characteristics on the single-shot data are extracted using IOX software.
The single beat dataset obtained (containing 34 hemodynamic characteristics, end-systolic elasticity E)esAnd volume V at which left ventricular pressure is zero0) The training set (75%) and the test set (25%) were divided and normalized separately.To analyze the stability of the hemodynamic characteristics, each time the data (raw data acquired by the sensor) will be measured three times in duplicate (the first measurement is taken as data to train the test model, and the other two measurements are used to analyze the reliability of the measurements). Of these, 34 hemodynamic parameters were extracted (using IOX software) from a single pressure-volume loop PV loop as a single sample. The division of the data is performed here based on the numerous history data obtained.
Step two: and (5) feature extraction.
A proposed Filter-based feature-based recursive elimination algorithm (Filter-based SVM-RFE) is used to select a volume V for predicting when the left ventricular pressure is zero from all hemodynamic features extracted from the single beat data0Key hemodynamic characteristics. The process is as follows:
first, the high stability (the ICC score is greater than 0.75) is screened from all hemodynamic characteristics by the ICC as follows:
A＝{c|I(c)≥k1,c∈x}
wherein A is a feature subset screened by an intra-group consistency index ICC, c is a hemodynamic feature, x is a feature subset to be selected in which the hemodynamic feature is stored, I (c) is a stability score of the feature c, and k is1Taken to be 0.75.
Then, the features with high redundancy are removed from the feature subset through the correlation index (the correlation of the two features is higher than 0.92, and one of the two features is considered as a redundant feature), and the formula is as follows:
B＝{ci|R(ci,cj)≤k2,ci∈A,cj∈A,ci≠cj}
where B is the final candidate feature subset, ciAnd cjIs a candidate feature belonging to the feature subset A, R (c)i,cj) Representing the correlation coefficient, k, between two features2Taken to be 0.92.
Finally, a representative feature is selected through a recursive elimination algorithm, the process traverses all possible feature combinations in the candidate feature subset B by using a sequential backward selection (sequential backward selection) method, and evaluation indexes are as follows:
Figure BDA0002430622750000091
wherein, VcalculatedVolume V when left ventricle pressure is zero0Calculated value of (V)meaasuredIs the gold standard of measurement. Volume V at which left ventricular pressure is zero at a final feature subset dimension of 50The prediction accuracy of (2) is highest, and the five characteristics are respectively end-systolic volume (V)es) End diastolic volume (V)ed) End diastolic pressure (P)ed) Area of pressure-volume loop (PVA) and end-systolic elastic Potential Energy (PE).
Step three: construction of left ventricular end-systolic elasticity E based on SVR algorithm in machine learning algorithmesAnd (4) predicting the model. The model is divided into two parts, firstly, five key features selected in the second step and the volume V when the left ventricle pressure is zero are established through an SVR algorithm in a machine learning algorithm0The volume V at which the left ventricular pressure is zero, then predicted from the single beat pressure volume data0And measured end systolic volume VesEnd systolic pressure PesTo calculate the left ventricular end-systolic elasticity Ees(the calculation principle is shown in FIG. 4).
Figure BDA0002430622750000101
EesRepresenting left ventricular end-systolic elasticity; v0Represents the volume at which left ventricular pressure is zero; the last systolic pressure volume coordinate on the pressure-volume ring is (V)es，Pes)，VesDenotes the end-systolic volume, PesDenotes end-systolic pressure, V0Are predicted from five key hemodynamic characteristics extracted from a single pressure-volume loop PV loop.
Step four: step three, middle fiveVolume V at which the key hemodynamic characteristic and left ventricular pressure are zero0And training and verifying a nonlinear mapping relation model. The input sample consists of five key hemodynamic characteristics (V)es、Ved、PedPVA, PE) and the target value (volume V at which the left ventricular pressure is zero0) And (4) composition, randomly dividing the training set into 75% of a training set and 25% of a testing set. And obtaining the optimal parameters of the SVR model by using a five-fold cross validation algorithm in the training set. Testing the trained model in a test set, and obtaining the volume V when the pressure of the left ventricle is zero0And left ventricular end-systolic elasticity EesCompared with the volume V of the left ventricle when the pressure of the left ventricle is zero measured by the traditional method0Left ventricular end-systolic elasticity EesThe gold standard value of (2) is compared and verified.
Step five: and acquiring a single current pressure-volume ring measured in a steady state, extracting features, inputting the extracted 5 key features into the trained model, and predicting the current left ventricular end-systolic elasticity.
Example 3
As shown in fig. 5, the present embodiment provides a left ventricular end systole elasticity prediction system based on SVR algorithm, which includes:
a current pressure-volume loop acquisition module 201 for acquiring a single current pressure-volume loop measured at steady state.
A current key feature extraction module 202, configured to extract a current key feature from all hemodynamic features of the current pressure-volume ring based on a feature recursive elimination algorithm of a filter and a support vector machine; the current key features include end systolic volume, end diastolic pressure, area of the pressure-volume ring, and end systolic elastic potential energy.
The volume prediction module 203 when the current left ventricular pressure is zero is used for determining the volume when the current left ventricular pressure is zero according to the current key characteristics and the trained volume prediction model when the left ventricular pressure is zero; the trained volume prediction model with the left ventricular pressure being zero is obtained by adopting SVR algorithm training according to multiple groups of historical current key features and the volume with the historical left ventricular pressure being zero corresponding to each group of historical current key features; the trained volume prediction model with zero left ventricular pressure represents a nonlinear mapping relationship between the key features and the volume with zero left ventricular pressure.
A current left ventricular end-systolic elasticity calculation module 204, configured to calculate a current left ventricular end-systolic elasticity according to a volume when the current left ventricular pressure is zero and a constructed left ventricular end-systolic elasticity prediction model; the left ventricular end systole elasticity prediction model is constructed according to the volume when the left ventricular pressure is zero and the pressure-volume coordinates of the end systole on a single pressure-volume ring.
The SVR algorithm-based left ventricular end systolic elasticity prediction system provided by this embodiment further includes: a hemodynamic feature extraction module to extract all hemodynamic features on the current pressure-volume loop using IOX software.
The current key feature extraction module 202 specifically includes:
and a first set determination unit, configured to extract, according to the intra-group consistency index, features with stability greater than a first set threshold from all the hemodynamic features, and store the features in the first set.
And the second set determining unit is used for removing the features with redundancy larger than a second set threshold from the first set according to the correlation coefficient, and storing the rest features in a second set.
And the current key feature extraction unit is used for extracting current key features from the second set based on a filter and a feature recursive elimination algorithm of a support vector machine.
The current left ventricular end systole elasticity calculation module 204 specifically includes:
a current left ventricular end-systolic elasticity calculation unit for calculating the current left ventricular end-systolic elasticity according to the formula
Figure BDA0002430622750000121
Calculating the current left ventricular contractionThe final elastic.
Wherein E isesRepresenting the current left ventricular end-systolic elasticity; v0Represents the volume when the current left ventricular pressure is zero; the last systolic pressure volume coordinate on the pressure-volume ring is (V)es，Pes)，VesDenotes the end-systolic volume, PesIndicating end-systolic pressure.
Compared with the existing algorithm, the invention has the following positive effects: by means of the invention, the end-systolic elasticity E can be accurately delineatedesA mathematical model of the calculation and can be accurately predicted from the hemodynamic characteristics extracted only on a single pressure-volume loop measured at steady state, i.e. the end-systolic elasticity EesThe calculation of (a) no longer requires multiple occlusion operations on the vena cava, but can be accurately predicted from the hemodynamic characteristics extracted on the pressure-volume loop measured at a single steady state. The invention can be applied to an isolated heart perfusion system, and can accurately and reliably predict the end-systolic elasticity EesAnd invasive damage to the heart can be reduced to a great extent. The invention can also be associated with early transplantation results (cardiac index) and has clinical application value.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (8)

1. A left ventricular end systole elasticity prediction method based on an SVR algorithm is characterized by comprising the following steps:
obtaining a single current pressure-volume loop measured at steady state;
extracting current key features from all hemodynamic features of the current pressure-volume ring based on a feature recursive elimination algorithm of a filter and a support vector machine; the current key features include end systolic volume, end diastolic pressure, area of the pressure-volume ring, and end systolic elastic potential energy;
determining the volume when the current left ventricular pressure is zero according to the current key characteristics and a trained volume prediction model when the left ventricular pressure is zero; the trained volume prediction model with the left ventricular pressure being zero is obtained by adopting SVR algorithm training according to multiple groups of historical current key features and the volume with the historical left ventricular pressure being zero corresponding to each group of historical current key features; the trained volume prediction model with the left ventricular pressure being zero represents a nonlinear mapping relation between the key features and the volume with the left ventricular pressure being zero;
calculating the current left ventricle end-systolic elasticity according to the volume when the current left ventricle pressure is zero and the constructed left ventricle end-systolic elasticity prediction model; the left ventricular end systole elasticity prediction model is constructed according to the volume when the left ventricular pressure is zero and the pressure-volume coordinates of the end systole on a single pressure-volume ring.
2. The SVR algorithm-based left ventricular end systolic elasticity prediction method of claim 1, wherein before performing a filter and support vector machine-based feature recursive elimination algorithm to extract current key features from all hemodynamic features of said current pressure-volume ring, further comprising:
all hemodynamic characteristics on the current pressure-volume loop were extracted using IOX software.
3. The SVR algorithm-based left ventricular end systolic elasticity prediction method of claim 1, wherein said filter and support vector machine-based feature recursive elimination algorithm extracts current key features from all hemodynamic features of said current pressure-volume loop, specifically comprising:
extracting features with stability greater than a first set threshold from all the hemodynamic features according to an intra-group consistency index, and storing the features in a first set;
according to the correlation coefficient, removing the features with redundancy larger than a second set threshold from the first set, and storing the remaining features in a second set;
and extracting the current key features from the second set based on a filter and a feature recursive clearing algorithm of a support vector machine.
4. The SVR algorithm-based left ventricular end systolic elasticity prediction method of claim 1, wherein said calculating a current left ventricular end systolic elasticity from a volume when said current left ventricular pressure is zero and a constructed left ventricular end systolic elasticity prediction model, specifically comprises:
according to the formula
Figure FDA0002430622740000021
Calculating the current left ventricular end-systolic elasticity;
wherein E isesRepresenting the current left ventricular end-systolic elasticity; v0Represents the volume when the current left ventricular pressure is zero; the last systolic pressure volume coordinate on the pressure-volume ring is (V)es，Pes)，VesDenotes the end-systolic volume, PesIndicating end-systolic pressure.
5. A left ventricular end systolic elasticity prediction system based on SVR algorithm, comprising:
the current pressure-volume ring acquisition module is used for acquiring a single current pressure-volume ring measured in a steady state;
a current key feature extraction module, configured to extract a current key feature from all hemodynamic features of the current pressure-volume ring based on a feature recursive elimination algorithm of a filter and a support vector machine; the current key features include end systolic volume, end diastolic pressure, area of the pressure-volume ring, and end systolic elastic potential energy;
the volume prediction module is used for determining the volume when the current left ventricular pressure is zero according to the current key characteristics and the trained volume prediction model when the current left ventricular pressure is zero; the trained volume prediction model with the left ventricular pressure being zero is obtained by adopting SVR algorithm training according to multiple groups of historical current key features and the volume with the historical left ventricular pressure being zero corresponding to each group of historical current key features; the trained volume prediction model with the left ventricular pressure being zero represents a nonlinear mapping relation between the key features and the volume with the left ventricular pressure being zero;
the current left ventricle end-systolic elasticity calculation module is used for calculating the current left ventricle end-systolic elasticity according to the volume when the current left ventricle pressure is zero and the constructed left ventricle end-systolic elasticity prediction model; the left ventricular end systole elasticity prediction model is constructed according to the volume when the left ventricular pressure is zero and the pressure-volume coordinates of the end systole on a single pressure-volume ring.
6. The SVR algorithm-based left ventricular end systolic elasticity prediction system of claim 5, further comprising:
a hemodynamic feature extraction module to extract all hemodynamic features on the current pressure-volume loop using IOX software.
7. The SVR algorithm-based left ventricular end systolic elasticity prediction system of claim 5, wherein said current key feature extraction module specifically comprises:
a first set determination unit, configured to extract, according to the intra-group consistency index, features with stability greater than a first set threshold from all the hemodynamic features, and store the features in a first set;
a second set determining unit, configured to remove, according to the correlation coefficient, a feature with redundancy greater than a second set threshold from the first set, and store the remaining features in a second set;
and the current key feature extraction unit is used for extracting current key features from the second set based on a filter and a feature recursive elimination algorithm of a support vector machine.
8. The SVR algorithm-based end-systolic elasticity prediction system of claim 5, wherein said current end-systolic elasticity calculation module specifically comprises:
a current left ventricular end-systolic elasticity calculation unit for calculating the current left ventricular end-systolic elasticity according to the formula
Figure FDA0002430622740000031
Calculating the current left ventricular end-systolic elasticity;
wherein E isesRepresenting the current left ventricular end-systolic elasticity; v0Represents the volume when the current left ventricular pressure is zero; the last systolic pressure volume coordinate on the pressure-volume ring is (V)es，Pes)，VesDenotes the end-systolic volume, PesIndicating end-systolic pressure.
CN202010234811.3A 2020-03-30 2020-03-30 Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm Pending CN111407244A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010234811.3A CN111407244A (en) 2020-03-30 2020-03-30 Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010234811.3A CN111407244A (en) 2020-03-30 2020-03-30 Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm

Publications (1)

Publication Number Publication Date
CN111407244A true CN111407244A (en) 2020-07-14

Family

ID=71486343

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010234811.3A Pending CN111407244A (en) 2020-03-30 2020-03-30 Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm

Country Status (1)

Country Link
CN (1) CN111407244A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114740728A (en) * 2022-04-25 2022-07-12 哈尔滨工业大学 Isolated heart perfusion blood pressure self-adaptive control method based on semi-parametric model

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008144490A1 (en) * 2007-05-16 2008-11-27 Massachusetts Instutute Of Technology Systems and methods for model-based estimation of cardiac ejection fraction, cardiac contractility, and ventricular end-diastolic volume
CN104887224A (en) * 2015-05-29 2015-09-09 北京航空航天大学 Epileptic feature extraction and automatic identification method based on electroencephalogram signal
CN105705151A (en) * 2013-07-17 2016-06-22 低温药理Kf有限公司 A method for treating haemorrhage, shock and brain injury
CN105976348A (en) * 2015-01-06 2016-09-28 西门子公司 Personalized whole-body circulation in medical imaging
CN107205667A (en) * 2015-01-19 2017-09-26 谷歌公司 The Noninvasive of the other functional states and trend of health of heart and Human Physiology system is determined

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008144490A1 (en) * 2007-05-16 2008-11-27 Massachusetts Instutute Of Technology Systems and methods for model-based estimation of cardiac ejection fraction, cardiac contractility, and ventricular end-diastolic volume
CN105705151A (en) * 2013-07-17 2016-06-22 低温药理Kf有限公司 A method for treating haemorrhage, shock and brain injury
CN105976348A (en) * 2015-01-06 2016-09-28 西门子公司 Personalized whole-body circulation in medical imaging
CN107205667A (en) * 2015-01-19 2017-09-26 谷歌公司 The Noninvasive of the other functional states and trend of health of heart and Human Physiology system is determined
CN104887224A (en) * 2015-05-29 2015-09-09 北京航空航天大学 Epileptic feature extraction and automatic identification method based on electroencephalogram signal

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WEIPING XIAO 等: "Single-beat Measurement of Left Ventricular Contractility in Normothermic Ex Situ Perfused Porcine Hearts", 《IEEE TRANSACTIONS ON BIO-MEDICAL ENGINEERING》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114740728A (en) * 2022-04-25 2022-07-12 哈尔滨工业大学 Isolated heart perfusion blood pressure self-adaptive control method based on semi-parametric model
CN114740728B (en) * 2022-04-25 2023-03-03 哈尔滨工业大学 Isolated heart perfusion blood pressure self-adaptive control method based on semi-parametric model

Similar Documents

Publication Publication Date Title
CN109003270B (en) Image processing method, electronic device and storage medium
CN109864736A (en) Processing method, device, terminal device and the medium of electrocardiosignal
KR20200006447A (en) Method and program for stroke volume acquisition using arterial pressure waveform
CN105550659A (en) Real-time electrocardiogramclassification method based on random projection
CN103961089A (en) Sinus heart rate turbulence tendency detecting method based on segmented straight line fitting
CN109077720B (en) Signal processing method, device, equipment and storage medium
CN101835421B (en) Feature value candidate generating device and feature value candidate generating method
Zhang et al. A novel encoding Lempel–Ziv complexity algorithm for quantifying the irregularity of physiological time series
CN115581465A (en) Coronary heart disease risk assessment method and device, and sudden cardiac death risk assessment method and system
CN111407244A (en) Left ventricle end-systolic elasticity prediction method and system based on SVR algorithm
Dózsa et al. ECG signal compression using adaptive Hermite functions
CN111067512A (en) Ventricular fibrillation detection device, ventricular fibrillation detection model training method and equipment
Denzinger et al. Deep learning algorithms for coronary artery plaque characterisation from CCTA scans
Weng et al. Myocardial infarction classification by morphological feature extraction from big 12-lead ECG data
Bhattacharya et al. Synthetic data generation through statistical explosion: Improving classification accuracy of coronary artery disease using PPG
KR102149748B1 (en) Method and apparatus for obtaining heart and lung sounds
WO2020220403A1 (en) Fatigue state detection method, apparatus and device, and storage medium
Qu et al. ECG Heartbeat Classification Detection Based on WaveNet-LSTM
Mansor et al. Local wall motion classification of stress echocardiography using a hidden Markov model approach
CN109846473A (en) A method of detecting 10 seconds electrocardiogram noise jamming degree of single lead
CN109394197B (en) Heart rate variability measuring method, device and equipment based on time-frequency analysis
CN110313990B (en) Method for establishing prediction bridge vascular permeability model based on wall shear stress image characteristics in heart bypass surgery
CN109461139B (en) Liver cancer quantitative analysis method based on dynamic MRI information fusion
CN110706803B (en) Method and device for determining myocardial fibrosis, readable medium and electronic equipment
CN110176298B (en) Kernel principal component spectrum Hash method for diabetic fundus image classification

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200714