CN117672460A - Mitral valve repair difficulty assessment method and system based on ultrasonic and machine learning - Google Patents
Mitral valve repair difficulty assessment method and system based on ultrasonic and machine learning Download PDFInfo
- Publication number
- CN117672460A CN117672460A CN202311624840.0A CN202311624840A CN117672460A CN 117672460 A CN117672460 A CN 117672460A CN 202311624840 A CN202311624840 A CN 202311624840A CN 117672460 A CN117672460 A CN 117672460A
- Authority
- CN
- China
- Prior art keywords
- mitral valve
- clinical
- data
- repair
- sample
- 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
Links
- 230000008439 repair process Effects 0.000 title claims abstract description 114
- 210000004115 mitral valve Anatomy 0.000 title claims abstract description 95
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000010801 machine learning Methods 0.000 title claims abstract description 13
- 238000012216 screening Methods 0.000 claims abstract description 28
- 238000013210 evaluation model Methods 0.000 claims abstract description 14
- 238000002604 ultrasonography Methods 0.000 claims abstract description 13
- 238000000556 factor analysis Methods 0.000 claims abstract description 8
- 206010067171 Regurgitation Diseases 0.000 claims description 35
- 230000003902 lesion Effects 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 22
- 230000002861 ventricular Effects 0.000 claims description 22
- 238000012795 verification Methods 0.000 claims description 15
- 230000035488 systolic blood pressure Effects 0.000 claims description 13
- 210000001147 pulmonary artery Anatomy 0.000 claims description 12
- 238000010992 reflux Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 7
- 230000001419 dependent effect Effects 0.000 claims description 6
- 238000005194 fractionation Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000007405 data analysis Methods 0.000 claims description 4
- 238000001543 one-way ANOVA Methods 0.000 claims description 3
- 238000012706 support-vector machine Methods 0.000 abstract description 6
- 238000003745 diagnosis Methods 0.000 abstract description 3
- 239000000523 sample Substances 0.000 description 30
- 238000011156 evaluation Methods 0.000 description 9
- 238000001356 surgical procedure Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 7
- HVYWMOMLDIMFJA-DPAQBDIFSA-N cholesterol Chemical compound C1C=C2C[C@@H](O)CC[C@]2(C)[C@@H]2[C@@H]1[C@@H]1CC[C@H]([C@H](C)CCCC(C)C)[C@@]1(C)CC2 HVYWMOMLDIMFJA-DPAQBDIFSA-N 0.000 description 6
- DDRJAANPRJIHGJ-UHFFFAOYSA-N creatinine Chemical compound CN1CC(=O)NC1=N DDRJAANPRJIHGJ-UHFFFAOYSA-N 0.000 description 6
- 230000003412 degenerative effect Effects 0.000 description 5
- 206010027727 Mitral valve incompetence Diseases 0.000 description 4
- 238000012552 review Methods 0.000 description 4
- 102000007330 LDL Lipoproteins Human genes 0.000 description 3
- 108010007622 LDL Lipoproteins Proteins 0.000 description 3
- 210000004369 blood Anatomy 0.000 description 3
- 239000008280 blood Substances 0.000 description 3
- 235000012000 cholesterol Nutrition 0.000 description 3
- 229940109239 creatinine Drugs 0.000 description 3
- 238000002592 echocardiography Methods 0.000 description 3
- 208000018578 heart valve disease Diseases 0.000 description 3
- 235000003715 nutritional status Nutrition 0.000 description 3
- 230000002980 postoperative effect Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 210000002966 serum Anatomy 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 206010058178 Aortic occlusion Diseases 0.000 description 2
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 description 2
- 241000404883 Pisa Species 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 2
- 239000008103 glucose Substances 0.000 description 2
- 230000007170 pathology Effects 0.000 description 2
- 239000013610 patient sample Substances 0.000 description 2
- 208000011682 Mitral valve disease Diseases 0.000 description 1
- 208000025747 Rheumatic disease Diseases 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000001746 atrial effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 208000016569 congenital mitral valve insufficiency Diseases 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000005786 degenerative changes Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 208000019622 heart disease Diseases 0.000 description 1
- 210000003709 heart valve Anatomy 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 208000005907 mitral valve insufficiency Diseases 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 231100000915 pathological change Toxicity 0.000 description 1
- 230000036285 pathological change Effects 0.000 description 1
- 230000002685 pulmonary effect Effects 0.000 description 1
- 230000000552 rheumatic effect Effects 0.000 description 1
- 208000004124 rheumatic heart disease Diseases 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000009423 ventilation Methods 0.000 description 1
Abstract
The invention provides a mitral valve repair difficulty assessment method and system based on ultrasound and machine learning. Obtaining variable index data according to the collected success and failure samples; determining a plurality of clinical characteristics according to the correlation coefficients between the success samples and the failure samples; screening from successful and failed samples by adopting single-factor analysis of variance to obtain difference characteristics; and constructing a repair difficulty assessment model according to the clinical characteristics, the variable index data and the difference characteristics, and determining the accuracy of the repair difficulty assessment model according to an ROC curve output by the model and the area under the ROC curve. The method has the advantages that the mitral valve repair operation difficulty evaluation model is established by selecting specific continuous variable indexes and classified variable indexes through the linear support vector machine classifier, complex cases can be accurately identified in clinical work, the patient operation difficulty is layered, diagnosis and treatment schemes are formulated in a targeted manner, and the method is suitable for surgeons; and meanwhile, the heart center medical quality can be monitored, and the method provides help for related medical schemes.
Description
Technical Field
The invention relates to the field of biomedicine, in particular to a mitral valve repair difficulty assessment method and system based on ultrasound and machine learning.
Background
Mitral valve pathology is one of the most common heart valve diseases, with degenerative lesions being the leading cause of mitral valve pathology. With the increasing population aging, the number of patients who need heart valve surgery increases year by year, and the proportion of degenerative changes is also increasing. At present, the international guidelines related to heart valve diseases recommend mitral valve repair operation as a first-choice treatment measure of degenerative mitral valve lesions, and meanwhile, partial researches report that the surgical mitral valve repair operation achieves satisfactory long-term follow-up results in rheumatic mitral valve lesion application.
Mitral valve lesions, especially rheumatic heart disease lesions, are complex, the repair operation difficulty is high, the secondary operation proportion is high, the requirements on experience of operators are high, the number of mitral valve repair operators with abundant experience is insufficient, and the above factors limit the wide development of mitral valve repair. Therefore, a heart valve disease operation complexity scoring system is established, the lesion complexity and operation difficulty are accurately evaluated, and the method has great significance in guiding surgeons to select the optimal operation scheme, evaluating heart center medical quality and improving clinical diagnosis and treatment level.
With the increasing incidence of degenerative mitral valve lesions, the number of mitral valve repair operations is increased, the importance of surgeons is increased, and the maturing and application of research methods such as logistic regression are performed, so that a plurality of mitral valve repair operation difficulty scoring systems, such as Mount Sinai score, are established at home and abroad based on large-scale mitral valve repair operation patient databases, but the problems of poor differentiation, accuracy, applicability, subjectivity and the like still exist.
Therefore, how to provide a repair difficulty assessment method with high discrimination, good adaptability and objective assessment angle is called a problem to be solved.
Disclosure of Invention
The invention provides a mitral valve repair difficulty assessment method and a mitral valve repair difficulty assessment system based on ultrasound and machine learning, which are used for solving the problems of poor differentiation and strong subjectivity of a mitral valve repair operation scoring system in the prior art.
In order to achieve the above object, the technical scheme of the present invention provides a mitral valve repair difficulty evaluation method based on ultrasound and machine learning, comprising: obtaining variable index data according to the collected success samples and failure samples; the variable index data includes: age, left and right and anterior-posterior valve systole and left and right end diastole, lesion in region A2, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, 4+ grade regurgitation fractionation, pulmonary artery systolic pressure, leaflet height in region A2 and region P2, and neck width; determining a number of clinical features from a correlation coefficient between a successful sample and the failed sample; screening the successful sample and the failed sample by adopting single-factor analysis of variance to obtain difference characteristics; and constructing a repair difficulty assessment model according to the clinical characteristics, the variable index data and the difference characteristics, and determining the accuracy of the repair difficulty assessment model according to an ROC curve output by the model and the area under the ROC curve.
Preferably, before the variable index data is obtained, the method further comprises:
collecting clinical data and preoperative ultrasonic data of a patient who has completed mitral valve repair; and determining a success sample and a failure sample according to the clinical data and the ultrasonic data, wherein the failure sample is a reflux of the mitral valve which is indicated by the moderate and above by rechecking an echocardiogram before discharge of a mitral valve which fails in a mitral valve repair operation, and the rest samples are the success samples.
Preferably, in the above technical solution, the determining a plurality of clinical features according to the correlation coefficient between the success sample and the failure sample includes using the clinical feature as an independent variable for constructing the repair difficulty assessment model if the correlation coefficient is greater than 0.9.
As a preferred aspect of the above technical solution, it is preferable that the screening for the difference features from the successful sample and the failed sample by using a one-way analysis of variance includes: and taking whether mitral valve repair of the patient with completed mitral valve repair fails or not as a dependent variable, and obtaining the difference characteristic by using single-factor variance analysis.
Preferably, after obtaining the difference feature, the method further includes adopting an L1 feature selection algorithm to the variable index data, and screening clinical feature data still having a non-zero coefficient after verification from the result for constructing the repair difficulty assessment model.
As a preferred aspect of the above technical solution, it is preferable to screen clinical feature data still having a non-zero coefficient after verification from the results for constructing the repair difficulty assessment model, including: calculating a model function of each feature in the variable index data by adopting a minimum loss function strategy; and when the loss function is minimum, obtaining the weight and the bias constant of each variable index data, so that the optimal model function corresponding to each characteristic is obtained, and the repair difficulty evaluation model is obtained.
The invention also provides a mitral valve repair difficulty evaluation system based on ultrasound and machine learning, comprising: the extraction module is used for extracting variable index data from the collected success samples and failure samples; the variable index data includes: age, left and right and anterior-posterior valve systole and left and right end diastole, lesion in region A2, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, 4+ grade regurgitation fractionation, pulmonary artery systolic pressure, leaflet height in region A2 and region P2, and neck width; the feature module is used for determining a plurality of clinical features according to the correlation coefficient between the successful sample and the failed sample, and if the correlation coefficient is greater than 0.9, the clinical features are used as independent variables for constructing the repair difficulty evaluation model; the screening module is used for screening the successful samples and the failed samples by adopting single-factor analysis of variance to obtain difference characteristics; the model building module is used for building a repair difficulty assessment model according to the clinical characteristics obtained by the characteristic module, the variable index data obtained by the extraction module and the difference characteristics obtained by screening by the screening module, and determining the accuracy of the repair difficulty assessment model according to an ROC curve output by the model and the area under the ROC curve.
Preferably, before the collecting module, the method further includes: the device comprises a collection and classification module, a data analysis module and a data analysis module, wherein the collection and classification module is used for collecting clinical data and preoperative ultrasonic data of a patient with completed mitral valve repair, and determining a success sample and a failure sample according to the clinical data and the preoperative ultrasonic data, wherein the failure sample is a mitral valve repair failure, a valve replacement operation is performed, an ultrasonic cardiogram is rechecked before discharge to prompt moderate and above reflux of the mitral valve, and the rest samples are the success samples.
Preferably, the above technical solution is that the screening module is further configured to obtain the difference feature by using a one-way variance analysis with whether the mitral valve repair of the patient having completed mitral valve repair fails as a dependent variable.
As the optimization of the technical scheme, the screening module is preferably further used for screening the clinical characteristic data which still have non-zero coefficients after verification from the results by adopting an L1 characteristic selection algorithm on the variable index data to construct the repair difficulty assessment model.
The technical scheme of the invention provides a mitral valve repair difficulty assessment method and system based on ultrasound and machine learning. Obtaining variable index data according to the collected success samples and failure samples; determining a plurality of clinical features according to correlation coefficients between the success sample and the failure sample; screening a success sample and a failure sample by adopting single-factor analysis of variance to obtain difference characteristics; and constructing a repair difficulty assessment model according to the clinical characteristics, the variable index data and the difference characteristics, and determining the accuracy of the repair difficulty assessment model according to an ROC curve output by the model and the area under the ROC curve.
The invention has the advantages that a mitral valve repair operation difficulty evaluation system is established by selecting specific continuous variable indexes and classified variable indexes through a linear support vector machine classifier, the operation difficulty evaluation model can accurately identify complex cases in clinical work, and the operation difficulty of patients is layered, so that diagnosis and treatment schemes are formulated in a targeted manner, and the model is selected to be suitable for surgeons; and meanwhile, the heart center medical quality can be monitored, and the method provides help for related medical schemes.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief description will be given below of the drawings required for the embodiments or the prior art descriptions, and it is obvious that the drawings in the following description are some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a mitral valve repair operation difficulty evaluation method provided by the invention.
Fig. 2 is a second schematic flow chart of the mitral valve repair operation difficulty evaluation method provided by the invention.
Fig. 3 is a schematic structural diagram of a mitral valve repair operation difficulty evaluation system provided by the invention.
FIG. 4 is a web page calculator of the mitral valve repair surgery difficulty assessment model provided by the invention.
Fig. 5 is a ROC graph of the present invention for internal verification of a mitral valve repair procedure difficulty assessment model.
Fig. 6 is a ROC graph of the present invention, external verification of a mitral valve repair procedure difficulty assessment model.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 is a schematic flow chart provided in an embodiment of the present invention, and as shown in fig. 1, the evaluation method provided in this embodiment includes the following steps:
s1: clinical data and preoperative ultrasound data are collected for a patient undergoing mitral valve repair surgery.
Clinical data includes, but is not limited to: age, sex, height, weight, complications, nutritional status, NYHA grade, past heart history, past surgical history, preoperative critical status, surgical mode, extracorporeal circulation time, aortic occlusion time, cumulative assisted ventilation time, hospitalization time, ICU residence time, complications, drainage volume, and other perioperative data, and test results of pre-and post-operative total cholesterol, low density lipoprotein, blood glucose, serum creatinine, and the like.
S2: and (3) determining a patient sample with successful mitral valve repair and a patient sample with failure according to the clinical data and the preoperative ultrasonic data obtained in the step (S1).
Among them, mitral valve repair failure is defined as: the operation immediately failed mitral valve repair (mitral valve repair failure with valve replacement) and recurrence after mitral valve repair (pre-discharge review of echocardiography suggests moderate and superior mitral regurgitation).
S3: variable index data of a sample of a patient with successful mitral valve repair and a sample of a patient with failure are obtained.
The variable index data comprises age, left and right end systole, anterior and posterior diameter and left and right end diastole, A2 zone lesion, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, regurgitation grading, pulmonary artery systolic pressure, A2 zone and P2 zone leaflet height and constriction neck width.
Further, the variable index data includes, in years, the left and right and anterior and posterior valve systole and the left and right valve diastole diameters in mm, the A2 region lesions in the presence or absence of regurgitation beam width in mm, left atrial volume in ml, left ventricular ejection fraction in 1% per mm, left ventricular end diastole volume in mm, regurgitation grading in 2,3,4 stages, pulmonary artery systolic pressure in mmHg, A2 region and P2 region leaflet height in mm, and the runback neck width in mm.
In the variable index data of the present application, the age, the left and right and anterior-posterior diameters of the end systole and the end diastole, the regurgitation beam width, left Fang Rongji, the left ventricular ejection fraction, the left ventricular end diastole volume, the pulmonary artery systolic pressure, the leaflet heights in the A2 region and the P2 region, and the regurgitation neck width are continuous variable indexes, and the lesions in the A2 region and regurgitation are classified variable indexes.
S4: and (3) establishing a mitral valve repair operation difficulty evaluation model by using the variable index data obtained in the step (S3) and applying a linear support vector machine classifier algorithm.
And constructing a repair difficulty evaluation model after feature selection is performed on the variable index data, specifically, adopting an L1 feature selection algorithm to screen clinical feature data which still has non-zero coefficients after verification from the variable index data, and constructing the repair difficulty evaluation model. And calculating weights and bias constants in the model functions of each feature in the training set by adopting a minimum strategy of the loss function. And determining optimal model functions corresponding to the clinical features according to different weights and bias constants, wherein the optimal model functions form a repair difficulty evaluation model.
Wherein the ultrasound data is preoperative ultrasound data.
The technical solution of the present invention will now be described with reference to a specific embodiment, as shown in fig. 2:
and 101, acquiring data.
The subject of this example was a degenerative mitral valve disease patient undergoing mitral valve repair surgery. The present example enrolls a patient diagnosed with primary degenerative mitral insufficiency in the preoperative echocardiogram, who is planning mitral valve repair.
And 102, collecting information, defining ending indexes, and training model groups.
Specifically, crowd source and information collection: all study subjects were from the Fulvic hospital mitral valve repair patient database, and hospitalized cases meeting the criteria were enrolled to obtain clinical data and preoperative ultrasound data for the patient. The acquired patient data is the patient data for which the operation has been completed.
Clinical data includes: patient age, sex, height, weight, complications, nutritional status, NYHA classification, past heart history, past surgical history, preoperative critical status, surgical mode, extracorporeal circulation time, aortic occlusion time, hospitalization time, ICU residence time, complications, and other perioperative data, and pre-and post-operative total cholesterol, low density lipoprotein, blood glucose, serum creatinine, and other test results.
The preoperative ultrasound data includes: preoperative ejection fraction, each atrioventricular inner diameter, each atrioventricular volume, leaflet height, valvular lesion site, regurgitation beam width (regurgitant width, RW), constriction width (vena contracta width, VCW), regurgitation volume (regurgitation volume, RVol), effective regurgitation orifice area (effective regurgitant orifice area, EROA), proximal isovelocity surface area (proximal isovelocity surface area, PISA), lesion condition, etc., for a total of 231 cases.
Definition of ending index: mitral valve repair success/failure. Specific mitral valve repair failures are defined as: the operation immediately failed mitral valve repair (mitral valve repair failure with valve replacement) and recurrence after mitral valve repair (pre-discharge review of echocardiography suggests moderate and superior mitral regurgitation).
Obtaining a sample of a patient undergoing mitral valve repair surgery for establishing a mitral valve repair surgery difficulty assessment model, comprising: 210 patients in the successful group and 21 patients in the failed group. And 103, building a machine learning model of the inclusion variable.
Screening the index incorporated into the machine learning predictive model:
(1) according to clinical significance of the variables and whether the prior research supports, the following possible prediction independent variables are screened out together for preliminary construction of a model: age, preoperative ejection fraction, each atrioventricular inner diameter, each atrioventricular volume, leaflet height, valve lesion site, regurgitation beam width, constriction neck width, regurgitation volume, effective regurgitation port area, proximal isovelocity surface area, carpentier functional typing, regurgitation grading, mitral valve lesion condition, contemporaneous surgery and the like.
(2) In the training queue, whether mitral valve repair of a patient fails or not is taken as a dependent variable, and feature screening is carried out by using a single-factor analysis of variance and an L1 feature selection algorithm, and the method is specifically as follows:
(1) Regression model, defining the outcome variable (grouping basis): 1: failure of surgery; 0: the operation is successful.
And constructing a mitral valve operation difficulty assessment model by taking the indexes as inclusion variables and applying a linear support vector machine classifier method.
Given a dataset T of known results (training set comprising all the arguments), wherein:
T＝{(x 1 ,y 1 ),(x 2 ,y 2 )…(x m ,y m ) And respectively weighting and summing the dimensions of T, and giving a weight w to each dimension, wherein the bias constant b, b= -threshold is the negative value of the threshold.
The calculation method of the quantized value y comprises the following steps: when the mitral valve repair is successful, y=1, and when the mitral valve repair is failed, y=0.
For data samples x representing arguments in the training set, the corresponding model function is:
f(x)＝sign(w x +b), wherein sign () is a sign function,
sign(x)＝+1，x＞0；sign(x)＝-1，x＜0；
assuming that the training dataset is linearly separable, the goal of model learning is to find a classification hyperplane that can completely separate the positive and negative instances in the training dataset, take a loss function minimization strategy, i.e., define a loss function, and find w and b by minimizing the loss function.
The loss function in this model is the total distance from the misclassification point to the classification hyperplane S. The distance from any point x in the input space to the hyperplane S is:
when the loss function is minimum, the distance between all misclassification points and the segmentation hyperplane is nearest, and the corresponding function model is optimal.
And repeating the steps, screening to obtain clinical feature data with optimal function models, and obtaining a mitral valve operation difficulty assessment model according to each optimal function model corresponding to the clinical features.
(2) The repair operation difficulty is evaluated through a mitral valve operation difficulty evaluation model: the following variables are used as the input of a model, and the failure risk of the mitral valve repair operation is calculated by using a linear support vector machine classifier method so as to evaluate the repair operation difficulty: age, end systole left and right and anterior-posterior and end diastole left and right diameters, A2 region lesions, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, regurgitation fractionation, pulmonary arterial systolic pressure, A2 region leaflet height and P2 region neck width.
Wherein, the lesion in the A2 area is sometimes 1; when the lesion in the A2 area is absent, the value is 0; age, left and right and anterior-posterior valve systole and left and right end diastole, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, regurgitation fraction, pulmonary artery systolic pressure, leaflet height in the A2 and P2 regions, and constriction neck width are practical observations.
In the variable index data, the age is measured in years, the left and right diameters of the left and right of the end systole and the left and right of the end diastole of the annulus are measured in mm, the lesions of the A2 area are measured in mm or not, the reflux beam width is measured in mm, the left room volume is measured in ml, the left room ejection fraction is measured in mm, the left end diastole volume is measured in mm, the reflux classification is measured in 2,3 and 4 grades, the pulmonary artery systolic pressure is measured in mmHg, the leaflet heights of the A2 area and the P2 area are measured in mm, and the neck width is measured in mm.
First, an embodiment is illustrated:
if the patient is 15 years old, the left and right diameters of the end systole, the front and back diameters and the end diastole are respectively 21, 19 and 22mm, the regurgitation beam width is 3mm, the left room volume is 38ml, the left room ejection fraction is 45%, the left room end diastole volume is LVEDV, the regurgitation grade is grade 2, the pulmonary artery systolic pressure is 12mmHg, the leaflet heights of the A2 area and the P2 area are respectively 23 and 12mm, the systole neck width is 5mm, and the mitral valve repair failure probability of the patient is 82.71% calculated by a webpage calculator.
(3) A web page calculator: and generating a webpage calculator according to the constructed model, wherein the webpage calculator of the mitral valve operation difficulty evaluation model is shown in fig. 4.
The values of all variables in the webpage calculator are measured: the pathological changes of the A2 area are sometimes taken as 1; when the lesion in the A2 area is absent, the value is 0; age, left and right and anterior-posterior valve systole and left and right end diastole, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, regurgitation fraction, pulmonary artery systolic pressure, leaflet height in the A2 and P2 regions, and constriction neck width are practical observations.
(4) Model internal verification: the ROC curve and its AUC value were used to evaluate the discrimination of the predictive model in (2) above. The ROC curve is shown in fig. 5. AUC is 0.996, 95% CI (confidence interval): 0.991-1.000.
And 104, carrying out external verification on the constructed model.
(1) Collecting information for verification: all verification objects come from a mitral valve repair patient database of a Furthotopic hospital, hospitalization cases meeting standards are selected, clinical data of patients are obtained, the clinical data comprise patient ages, sexes, heights, weights, complications, nutritional status, NYHA classification, past heart disease history, past operation history, preoperative critical status, operation mode, extracorporeal circulation time, aortic blocking time, hospitalization time, ICU residence time, complications and other perioperative data, and test results of preoperative and postoperative total cholesterol, low density lipoprotein, blood sugar, serum creatinine and the like, and the ultrasonic data comprise preoperative ejection fraction, inner diameters of all chambers, volumes of all chambers, valve heights, valve lesion sites, reflux beam widths (regurgitant width, RW), reflux neck widths (vena contracta width, VCW), reflux volumes (regurgitation volume, RVol), effective reflux mouth areas (effective regurgitant orifice area, EROA), proximal constant speed surface areas (proximal isovelocity surface area, PISA), lesion conditions and the like, and the total 21 cases are obtained.
(2) End index definition: mitral valve repair failure is defined as: the operation immediately failed mitral valve repair (mitral valve repair failure with valve replacement) and recurrence after mitral valve repair (pre-discharge review of echocardiography suggests moderate and superior mitral regurgitation).
(3) Model external verification
Verifying the number of cases: 1 (repair failed) patient 2, 0 (repair successful) patient 19;
the ROC curve is shown in fig. 6.AUC is 0.739, 95% CI (confidence interval): 0.673-0.805, indicating that the mitral valve shaping procedure difficulty assessment model performs well.
The invention also provides a mitral valve repair difficulty evaluation system based on ultrasonic and machine learning, as shown in fig. 3, comprising:
an extraction module 201, configured to extract variable index data from the collected success samples and failure samples; the variable index data includes: age, left and right and anterior-posterior valve systole and left and right end diastole, lesion in region A2, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, 4+ grade regurgitation fractionation, pulmonary artery systolic pressure, leaflet height in region A2 and region P2, and neck width;
the collection and classification module 201 is configured to collect clinical data and pre-operative ultrasonic data of a patient with completed mitral valve repair, and determine a success sample and a failure sample according to the clinical data and the pre-operative ultrasonic data, where the failure sample is a failure of mitral valve repair operation, and performing valve replacement operation and pre-discharge review of an echocardiogram to prompt moderate and above regurgitation of the mitral valve, and the remaining samples are the success samples.
The feature module 202 is configured to determine a plurality of clinical features according to a correlation coefficient between the success sample and the failure sample, and if the correlation coefficient is greater than 0.9, take the clinical features as independent variables for constructing the repair difficulty evaluation model;
a screening module 203, configured to screen the success sample and the failure sample by using a single-factor analysis of variance to obtain a difference feature; and the method is also used for obtaining the difference characteristic by using a one-factor variance analysis by taking whether mitral valve repair of a patient with completed mitral valve repair fails as a dependent variable. And the variable index data is also used for adopting an L1 feature selection algorithm to screen clinical feature data which still has non-zero coefficients after verification from the results for constructing the repair difficulty assessment model.
The model building module 204 is configured to build a repair difficulty assessment model according to the clinical features obtained by the feature module, the variable index data obtained by the extraction module, and the difference features obtained by the screening module, and determine the accuracy of the repair difficulty assessment model according to an ROC curve output by the model and the area under the curve.
According to the invention, a mitral valve repair operation difficulty evaluation system is established by selecting specific continuous variables and classification variables through a linear support vector machine classifier, a complex case can be accurately identified by an operation difficulty evaluation model in clinical work, the operation difficulty of a patient is layered, a diagnosis and treatment scheme is formulated in a targeted manner, and a doctor is selected to be suitable for the operation; meanwhile, the method is beneficial to the health management department to accurately master the overall level of heart operations in China, monitor the medical quality of each heart center and provide assistance for the formulation and implementation of related medical policies.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (10)
1. A mitral valve repair difficulty assessment method based on ultrasound and machine learning, the method comprising:
obtaining variable index data according to the collected success samples and failure samples; the variable index data includes: age, left and right and anterior-posterior valve systole and left and right end diastole, lesion in region A2, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, 4+ grade regurgitation fractionation, pulmonary artery systolic pressure, leaflet height in region A2 and region P2, and neck width; wherein, the classified variable indexes in the variable index data comprise lesions in the A2 region and reflux classification, and the rest is continuous variable indexes;
determining a number of clinical features from a correlation coefficient between the successful sample and the failed sample;
screening the successful sample and the failed sample by adopting single-factor analysis of variance to obtain difference characteristics;
and constructing a repair difficulty assessment model according to the clinical characteristics, the variable index data and the difference characteristics, and determining the accuracy of the repair difficulty assessment model according to an ROC curve output by the model and the area under the ROC curve.
2. The method of claim 1, comprising, prior to said obtaining variable index data:
collecting clinical data and preoperative ultrasonic data of a patient who has completed mitral valve repair;
and determining a success sample and a failure sample according to the clinical data and the preoperative ultrasonic data, wherein the failure sample is a reflux of the mitral valve middle and above by performing valve replacement operation after mitral valve repair operation failure and rechecking an ultrasonic cardiogram before discharge, and the rest samples are the success samples.
3. The method according to claim 1, characterized in that it comprises: the determining of clinical features based on correlation coefficients between the successful sample and the failed sample includes,
if the correlation coefficient is greater than 0.9, the clinical feature is used as an independent variable for constructing the repair difficulty assessment model.
4. The method of claim 1, wherein the screening for difference features from the successful sample and the failed sample using one-way analysis of variance comprises:
and taking whether mitral valve repair of the patient with completed mitral valve repair fails or not as a dependent variable, and obtaining the difference characteristic by using single-factor variance analysis.
5. The method of claim 4, further comprising, after deriving the difference feature, further comprising, employing an L1 feature selection algorithm to perform feature screening on the clinical feature, comprising:
and (3) adopting an L1 feature selection algorithm to the variable index data, and screening clinical feature data which still have non-zero coefficients after verification from the results to construct the repair difficulty assessment model.
6. The method of claim 5, wherein the screening the results for clinical signature data that still has non-zero coefficients after verification for constructing the repair difficulty assessment model comprises:
calculating a model function of each feature in the variable index data by adopting a minimum loss function strategy;
and when the loss function is minimum, obtaining the weight and the bias constant of each variable index data, so that the optimal model function corresponding to each characteristic is obtained, and the repair difficulty evaluation model is obtained.
7. The mitral valve repair difficulty assessment system based on ultrasound and machine learning by adopting any one of the methods 1-6, which is characterized by comprising:
the extraction module is used for extracting variable index data from the collected success samples and failure samples; the variable index data includes: age, left and right and anterior-posterior valve systole and left and right end diastole, lesion in region A2, regurgitation beam width, left Fang Rongji, left ventricular ejection fraction, left ventricular end diastole volume, 4+ grade regurgitation fractionation, pulmonary artery systolic pressure, leaflet height in region A2 and region P2, and neck width;
the feature module is used for determining a plurality of clinical features according to the correlation coefficient between the successful sample and the failed sample, and if the correlation coefficient is greater than 0.9, the clinical features are used as independent variables for constructing the repair difficulty evaluation model;
the screening module is used for screening the successful samples and the failed samples by adopting single-factor analysis of variance to obtain difference characteristics;
the model building module is used for building a repair difficulty assessment model according to the clinical characteristics obtained by the characteristic module, the variable index data obtained by the extraction module and the difference characteristics obtained by screening by the screening module, and determining the accuracy of the repair difficulty assessment model according to an ROC curve output by the model and the area under the ROC curve.
8. The system of claim 7, comprising, prior to the extraction module:
the device comprises a collection and classification module, a data analysis module and a data analysis module, wherein the collection and classification module is used for collecting clinical data and preoperative ultrasonic data of a patient with completed mitral valve repair, and determining a success sample and a failure sample according to the clinical data and the preoperative ultrasonic data, wherein the failure sample is a mitral valve repair failure, a valve replacement operation is performed, an ultrasonic cardiogram is rechecked before discharge to prompt moderate and above reflux of the mitral valve, and the rest samples are the success samples.
9. The system of claim 7, wherein the screening module is further configured to obtain the variance feature using a one-way analysis of variance with whether mitral valve repair fails in a completed mitral valve repair patient as a dependent variable.
10. The system of claim 9, further comprising the screening module further configured to use an L1 feature selection algorithm on the variable index data to screen the results for clinical feature data that still has non-zero coefficients after verification for use in constructing the repair difficulty assessment model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311624840.0A CN117672460A (en) | 2023-11-30 | 2023-11-30 | Mitral valve repair difficulty assessment method and system based on ultrasonic and machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311624840.0A CN117672460A (en) | 2023-11-30 | 2023-11-30 | Mitral valve repair difficulty assessment method and system based on ultrasonic and machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117672460A true CN117672460A (en) | 2024-03-08 |
Family
ID=90065486
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311624840.0A Pending CN117672460A (en) | 2023-11-30 | 2023-11-30 | Mitral valve repair difficulty assessment method and system based on ultrasonic and machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117672460A (en) |
-
2023
- 2023-11-30 CN CN202311624840.0A patent/CN117672460A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gandhi et al. | Automation, machine learning, and artificial intelligence in echocardiography: a brave new world | |
| Thompson et al. | Artificial intelligence-assisted auscultation of heart murmurs: validation by virtual clinical trial | |
| CN109009222A (en) | Intelligent evaluation diagnostic method and system towards heart disease type and severity | |
| US20110191283A1 (en) | Method and System for Medical Decision Support Using Organ Models and Learning Based Discriminative Distance Functions | |
| US11735321B2 (en) | System for the prognostics of the chronic diseases after the medical examination based on the multi-label learning | |
| KR102439082B1 (en) | Method for predicting liver disease of ordinary perseon using ecg analysis data based on deep running | |
| JP7278323B2 (en) | A method for generating a predictive model for predicting a subject's glaucoma risk, a method for determining a subject's glaucoma risk using the predictive model, a device for predicting a subject's glaucoma risk, a computer program, and a computer-readable medium | |
| WO2022126800A1 (en) | Symptom, blood data and medical image fused method for comprehensively recognizing various suspected diseases | |
| CN113243887B (en) | Intelligent diagnosis and treatment instrument for macular degeneration of old people | |
| Levine et al. | The risk of having additional obstructive lesions in neonatal coarctation of the aorta | |
| Jalali et al. | Advanced analytics for outcome prediction in intensive care units | |
| Detrano et al. | Computer probability estimates of angiographic coronary artery disease: Transportability and comparison with cardiologists' estimates | |
| CN117672460A (en) | Mitral valve repair difficulty assessment method and system based on ultrasonic and machine learning | |
| CN116564521A (en) | Chronic disease risk assessment model establishment method, medium and system | |
| Bhalla et al. | A novel method for medical disease diagnosis using artificial neural networks based on backpropagation algorithm | |
| CN113273959B (en) | Portable diabetic retinopathy diagnosis and treatment instrument | |
| CN114864097A (en) | Method and device for establishing aorta dissection patient postoperative death prediction model | |
| RU2742429C1 (en) | Method for rapid assessment of changes in lung tissue with covid-19 without using computer tomography of thorax organs | |
| WO2021012203A1 (en) | Multi-model complementary enhanced machine leaning platform based on danger early warning in perioperative period | |
| TWI795928B (en) | System and method for prediction of intradialytic adverse event and computer readable medium thereof | |
| Duwairi et al. | A deep learning model and a dataset for diagnosing ophthalmology diseases | |
| Kaur et al. | Brain Stroke Prediction System–A Review | |
| Enériz et al. | Exploring a Segmentation-Classification Deep Learning-based Heart Murmurs Detector | |
| CN116884631B (en) | Comprehensive liver failure prediction and treatment reference system based on AI and similar patient analysis | |
| Demchenko et al. | The Use of Machine Learning Methods to the Automated Atherosclerosis Diagnostic and Treatment System Development. |
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 |