CN115331819A - Pancreatitis prognosis data processing method and system based on artificial intelligence - Google Patents
Pancreatitis prognosis data processing method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN115331819A CN115331819A CN202210910750.7A CN202210910750A CN115331819A CN 115331819 A CN115331819 A CN 115331819A CN 202210910750 A CN202210910750 A CN 202210910750A CN 115331819 A CN115331819 A CN 115331819A
- Authority
- CN
- China
- Prior art keywords
- human body
- sign data
- body basic
- pancreatitis
- basic sign
- 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
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
Landscapes
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a pancreatitis prognosis data processing method, a system, equipment and a computer readable storage medium based on artificial intelligence, wherein the method comprises the following steps: acquiring human body basic sign data of a sample; inputting the human body basic sign data into a pre-trained prognosis prediction model to obtain a classification result; the prognosis prediction model comprises: the training method of the pancreatitis grading prediction model comprises the following steps: acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise mild pancreatitis, moderate severe pancreatitis and severe pancreatitis; carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction; and constructing a model by using the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
Description
Technical Field
The invention relates to the technical field of medical biology, in particular to an artificial intelligence-based pancreatitis prognosis data processing method and a system thereof.
Background
Acute Pancreatitis (AP) is a common acute abdomen disease and the first digestive system disease in emergency hospitalization worldwide. Pancreatitis can be classified into severe, moderate and severe, and mild according to systemic and local complications. About 20% of patients with acute pancreatitis develop severe pancreatitis, i.e., with persistent organ failure, with mortality rates as high as 20% -40%. Moderate-severe pancreatitis has local complications and has potential intervention and treatment needs. Mild pancreatitis is more self-limited through supportive treatment. Early prognosis of the severity of acute pancreatitis and timely targeted treatment are important to reducing the fatality rate.
At present, no single variable or scoring system for acute pancreatitis can accurately predict the severity of pancreatitis prognosis at the initial stage of disease onset. Machine learning has been widely used in clinical studies including predicting the risk of severe pancreatitis, but most studies only perform dichotomy prediction, i.e., acute pancreatitis patients will not develop severe pancreatitis or non-severe pancreatitis in the future, and research results that can perform accurate three-classification prediction on the severity of pancreatitis prognosis, i.e., differentiate light, moderate and severe pancreatitis, are not retrieved. In addition, most of the existing researches are single-center researches, the number of samples is small, and the accuracy of results is required to be provided for the merchant.
The invention provides an artificial intelligence-based pancreatitis prognosis data processing method and system, which aims to enable doctors to identify high-risk patients as early as possible and make diagnosis and treatment plans to provide help references.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art. According to the method, a large amount of data from multiple centers are adopted, and a prediction model is established for indexes such as the three-classification prognosis of the acute pancreatitis, whether the acute pancreatitis dies, whether the pancreas is infected and necrotic, whether the acute pancreatitis stays in an ICU (intensive care unit), the length of stay in a hospital and the like by various machine learning methods, so that a help reference is provided for doctors to identify high-risk patients as early as possible and make diagnosis and treatment plans, and related life science problems are solved.
The application discloses an artificial intelligence-based pancreatitis prognosis data processing method, which comprises the following steps:
acquiring human body basic sign data of a sample;
inputting the human body basic sign data into a pre-trained prognosis prediction model to obtain a classification result;
the prognostic prediction model includes: the training method of the pancreatitis grading prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of training set samples, wherein the labels comprise mild pancreatitis, moderate and severe pancreatitis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
The training method of the pancreatitis grading prediction model further comprises the following steps:
calculating a score value based on human body basic sign data of the training set samples; the scoring values include one or more of: APACHE-II score value, marshall score value, sofa score value, qsofa score value and bisap score value;
and performing model construction by using the score value and the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
The training method of the pancreatitis grading prediction model further comprises the following steps:
sequencing the human body basic sign data characteristics to obtain human body basic sign data characteristics after sequencing;
performing model construction by using the sorted human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model;
optionally, performing model construction by using the score value and the sorted human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
The prognostic prediction model further includes: a death prediction model;
the training method of the death prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise death and non-death;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed death prediction model;
optionally, the prognostic prediction model further includes: an infectious necrosis prediction model;
the method for training the infectious necrosis prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise infection necrosis and non-infection necrosis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed infectious necrosis prediction model;
optionally, the prognosis prediction model further comprises: a stay in hospital duration prediction model;
the training method of the stay in hospital prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise a first hospitalization duration and a second hospitalization duration; the first length of stay is less than or equal to the classification threshold, and the second length of stay is greater than or equal to the classification threshold;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed hospitalization duration prediction model;
optionally, the prognostic prediction model further includes: an ICU prediction model;
the ICU prediction model training method comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise an entrance ICU and an exit ICU;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed ICU prediction model.
The human body basic sign data comprises one or more of the following: age, sex, etiology, body temperature, blood pressure, heart rate, respiratory rate, white blood cells, hematocrit, platelets, blood potassium, blood sodium, blood calcium, creatinine, urea nitrogen, satisfaction of consciousness score, presence of peritoneal irritation, presence of pleural effusion;
performing model construction on the human body basic sign data by using a machine learning method to obtain a constructed prognosis prediction model;
optionally, the machine learning method includes one or more of the following: linear regression, logistic regression, linear Discriminant Analysis (LDA), classification and regression trees, naive bayes, KNN, learning vector quantization, support Vector Machines (SVMs), random forests, lightGBM;
optionally, the machine learning method includes: weighted fusion of any two algorithms;
optionally, the machine learning method includes: weighted fusion of random forest and LightGBM.
The sorting process includes:
for one of the human body basic sign data features passing through the human body basic sign data features, calculating all combinations and marginal contribution of a single human body basic sign data feature in the combinations according to a combination sequence formed by the human body basic sign data features;
obtaining the contribution degree of a single human body basic sign data characteristic in the human body basic sign data characteristic according to the marginal contribution;
sorting the human body basic sign data features based on the contribution degrees, and outputting N human body basic sign data features with high contribution degrees;
optionally, N is at least 5;
an artificial intelligence based pancreatitis prognostic data processing apparatus, the apparatus comprising: a memory and a processor;
the memory is to store program instructions;
the processor is configured to invoke program instructions that, when executed, perform one of the artificial intelligence based pancreatitis prognostic data processing methods described above.
An artificial intelligence based pancreatitis prognostic data processing system comprising:
the acquisition unit is used for acquiring human body basic sign data of the sample;
the classification unit is used for inputting the human body basic sign data into a pre-trained prognosis prediction model to obtain a classification result;
the prognostic prediction model includes: the training method of the pancreatitis grading prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of training set samples, wherein the labels comprise mild pancreatitis, moderate and severe pancreatitis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out a method of artificial intelligence-based pancreatitis prognostic data processing as described above.
The application has the following beneficial effects:
1. the method establishes a three-classification prognosis prediction model for the prognosis data index of the acute pancreatitis through a plurality of machine learning methods, classifies the prognosis severity of the acute pancreatitis more accurately, namely mild pancreatitis, moderate and severe pancreatitis or severe pancreatitis, and compared with a two-classification prediction result, the three-classification prediction result ensures more accurate prediction and advanced prediction of the prognosis severity of the pancreatitis, is beneficial to early prediction of the severity of the acute pancreatitis and assists doctors in accurate prediction;
2. the method ingeniously fuses basic physical sign data of a human body with scoring values obtained by 5 traditional scoring systems, a three-classification prognosis prediction model is established for prognosis data indexes of acute pancreatitis by using various machine learning methods, the 5 scoring systems have certain research on SAP classification, the traditional scoring systems and clinical data are integrated, modeling is performed by using a machine learning algorithm, and the method has multi-center and foresight properties, and the accuracy and depth of data analysis and processing are greatly improved;
3. the method has the advantages that the main endpoint for prediction is the pancreatitis severity prognosis, the secondary endpoints are death, pancreatic infection necrosis, ICU admission and length of stay, various indexes are integrated, help reference can be provided for doctors to recognize high-risk patients and make diagnosis and treatment plans as soon as possible, and related life science problems are solved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a schematic flow chart of an analysis of a pancreatitis prognosis data processing method based on artificial intelligence provided by an embodiment of the invention;
FIG. 2 is a schematic diagram of an artificial intelligence-based pancreatitis prognostic data processing apparatus according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of an artificial intelligence based pancreatitis prognostic data processing system provided by an embodiment of the present invention;
FIG. 4 is a diagram of classification results of various machine learning algorithms in human body basic sign data according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of an algorithm according to an embodiment of the present invention, which uses a random forest method for classification;
fig. 6 is a schematic diagram of an algorithm two according to the present invention, which uses a LightGBM method for classification;
FIG. 7 is a graph illustrating the difference in AUC for prognosis of acute pancreatitis (SAP, MSAP, MAP) predicted under different machine learning algorithms provided by an embodiment of the present invention;
FIG. 8 is a graph showing the accuracy of prognosis prediction for acute pancreatitis (SAP, MSAP, MAP) with different machine learning algorithms provided by an embodiment of the present invention;
FIG. 9 is a schematic diagram of the comparison between the machine learning algorithm provided by the embodiment of the present invention and the severity AUC of acute pancreatitis predicted by the traditional scoring model;
FIG. 10 is a schematic diagram of a machine learning algorithm in comparison with a calibration graph of a traditional scoring model for predicting the severity of acute pancreatitis;
fig. 11 is a diagram illustrating the machine learning algorithm to predict a secondary endpoint according to an embodiment of the present invention: schematic diagrams of whether death, pancreatic infection necrosis, admission to the ICU, and length of stay;
fig. 12 shows that when the fusion model provided by the embodiment of the present invention predicts the secondary endpoint: schematic representation of AUC, acuracy, sensitivity, specificity, PPV and dNTPV;
fig. 13 is a diagram illustrating variable contribution values when predicting AP severity classification according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, 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.
In some of the flows described in the present specification and claims and in the above figures, a number of operations are included that occur in a particular order, but it should be clearly understood that these operations may be performed out of order or in parallel as they occur herein, with the order of the operations being indicated as 101, 102, etc. merely to distinguish between the various operations, and the order of the operations by themselves does not represent any order of performance. Additionally, the flows may include more or fewer operations, and the operations may be performed sequentially or in parallel. It should be noted that, the descriptions of "first", "second", etc. in this document are used for distinguishing different messages, devices, modules, etc., and do not represent a sequential order, nor limit the types of "first" and "second" to be different.
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.
Fig. 1 is a schematic flow chart of a pancreatitis prognosis data processing method based on artificial intelligence, which is provided by an embodiment of the present invention, and specifically, the method includes the following steps:
101: acquiring human body basic sign data of a sample;
in one embodiment, the human body basic sign data includes one or more of the following: age (years), gender Male (%), etiology Etiology, body Temperature (deg.C), blood pressure SBP (mmHg), heart rate HR (bpm), respiratory rate RR (bpm), white blood cell WBC (x 109/L), hematocrit HCT, platelet PLT (x 109/L), blood potassium K (mmol/L), blood sodium Na (mmol/L), blood calcium Ca (mmol/L), creatinine Cr (umol/L), urea nitrogen Bun (mmol/L), mental disorder, peritoneal irritation Peritoneal irration, pleural effusion.
Optionally, the basic human body sign data further includes one or more of the following: demographic data (age, sex, cause of acute pancreatitis), baseline laboratory tests (white blood cells, hematocrit, platelets, potassium, sodium, calcium, creatinine, urea nitrogen, glu (mmol/L) blood glucose levels), baseline vital signs (body temperature, systolic pressure, heart rate, respiratory rate), baseline physical signs (presence or absence of mental changes, peritoneal irritation, etc.), and prognostic data (incidence of persistent organ dysfunction, incidence of local complications, incidence of infectious pancreatic necrosis, days Hospital say (days of hospitalization), ICU admission rate ICU allowance and mortality Die, etc.).
In one embodiment, the sample comprises: 930 patients with acute pancreatitis from the multicenter, prospective, random-control acute pancreatitis study database from month 5 in 2018 to month 4 in 2022. The study was approved by the ethical committee (ethical number ZS-1413). The inclusion criteria were: acute pancreatitis is diagnosed within 48 hours of onset, the age is 18-75 years, and informed consent is participated in the experiment. Exclusion criteria were: the acute pancreatitis can be caused by tumor or ERCP operation after onset of over 48 hours, renal insufficiency, gestational or lactation period, hypertension with poor control, cardiovascular disease, abnormal mind, and acute pancreatitis.
The research result shows that: the average age of 930 patients with acute pancreatitis was 48.6 ± 6.0; the male patient ratio was 62.7% (586/930); causes of acute pancreatitis include: biliary =383, lipogenic =291, alcoholic =65, other =191; body temperature of 37.1 +/-1.05; the blood pressure is 130.3 +/-68.0.
102: inputting the human body basic sign data into a pre-trained prognosis prediction model to obtain a classification result;
in one embodiment, the prognostic prediction model includes: the training method of the pancreatitis grading prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise mild pancreatitis, moderate severe pancreatitis and severe pancreatitis; symptoms of the mild pancreatitis MAP include: no systemic or local complications, no organ failure; symptoms of the moderate-severe pancreatitis MSAP include: accompanied with systemic or local complications, the organ failure is transient, namely less than or equal to 48h; symptoms of the severe pancreatitis SAP include: persistent organ failure; the data set was randomly divided into a training set and a test set at a ratio of 4 to 1.
Carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
In one embodiment, the training method of the pancreatitis grading prediction model further comprises the following steps:
calculating a score value based on human body basic sign data of the training set samples; the scoring values include one or more of: APACHE-II score value, marshall score value, sofa score value, qsofa score value and bisap score value;
and carrying out model construction by using the score value and the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
In one embodiment, the training method of the pancreatitis grading prediction model further comprises the following steps:
sequencing the human body basic sign data characteristics to obtain human body basic sign data characteristics after sequencing;
performing model construction by using the sequenced human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model;
optionally, performing model construction by using the score value and the sorted human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
In one embodiment, the basic physical sign data of the human body is input into a pre-trained pancreatitis grading prediction model to obtain classification results of mild pancreatitis, moderate severe pancreatitis and severe pancreatitis.
In one embodiment, the prognostic prediction model further includes: a death prediction model;
the training method of the death prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise death and non-death;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed death prediction model;
optionally, the training method of the death prediction model further includes: calculating a score value based on human body basic sign data of the training set samples; the scoring values include one or more of: APACHE-II score value, marshall score value, sofa score value, qsofa score value and bisap score value;
carrying out model construction by using the score value and the human body basic sign data characteristics to obtain a constructed death prediction model;
optionally, the training method of the death prediction model further includes: sequencing the human body basic sign data characteristics to obtain human body basic sign data characteristics subjected to sequencing;
carrying out model construction by using the sorted human body basic sign data characteristics to obtain a constructed death prediction model;
optionally, a model is built by using the score value and the sorted human body basic sign data characteristics to obtain a built death prediction model.
In one embodiment, the human body basic sign data is input into a pre-trained death prediction model to obtain a classification result of death and non-death.
Optionally, the prognostic prediction model further includes: an infectious necrosis prediction model;
the method for training the infectious necrosis prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise infection necrosis and non-infection necrosis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed infectious necrosis prediction model;
optionally, the method for training the infectious necrosis prediction model further includes: calculating a score value based on the human body basic sign data of the training set sample; the scoring values include one or more of: APACHE-II score value, marshall score value, sofa score value, qsofa score value and bisap score value;
performing model construction by using the score value and the human body basic sign data characteristics to obtain a constructed infectious necrosis prediction model;
optionally, the method for training the infectious necrosis prediction model further includes: sequencing the human body basic sign data characteristics to obtain human body basic sign data characteristics subjected to sequencing;
performing model construction by using the sorted human body basic sign data characteristics to obtain a constructed infectious necrosis prediction model;
optionally, performing model construction by using the score value and the sorted human body basic sign data characteristics to obtain a constructed infectious necrosis prediction model.
In one embodiment, the human body basic sign data is input into a pre-trained infectious necrosis prediction model to obtain classification results of infectious necrosis and non-infectious necrosis.
Optionally, the prognosis prediction model further comprises: a stay in hospital duration prediction model;
the training method of the stay duration prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise a first hospitalization duration and a second hospitalization duration; the first length of stay is less than or equal to the classification threshold, and the second length of stay is greater than or equal to the classification threshold; the classification threshold is the first 24h of admission, and the later prognosis of the patient is predicted by using the data of the first 24 h;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed hospitalization duration prediction model;
optionally, the training method of the stay duration prediction model further includes: calculating a score value based on human body basic sign data of the training set samples; the scoring values include one or more of: APACHE-II score value, marshall score value, sofa score value, qsofa score value and bisap score value;
carrying out model construction by using the score value and the human body basic sign data characteristics to obtain a constructed hospitalization duration prediction model;
optionally, the training method of the hospitalization duration prediction model further includes: sequencing the human body basic sign data characteristics to obtain human body basic sign data characteristics subjected to sequencing;
performing model construction by using the sorted human body basic sign data characteristics to obtain a constructed hospitalization duration prediction model;
optionally, performing model construction by using the score value and the sorted human body basic sign data characteristics to obtain a constructed hospitalization duration prediction model.
In one embodiment, the human body basic sign data is input into a pre-trained stay duration prediction model to obtain classification results of the first stay duration and the second stay duration.
Optionally, the prognostic prediction model further includes: an ICU prediction model;
the ICU prediction model training method comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise an entrance ICU and an exit ICU;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed ICU prediction model.
Optionally, the method for training the ICU prediction model further includes: calculating a score value based on human body basic sign data of the training set samples; the score value comprises one or more of the following: APACHE-II score value, marshall score value, sofa score value, qsofa score value and bisap score value;
carrying out model construction by using the score value and the human body basic sign data characteristics to obtain a constructed ICU prediction model;
optionally, the method for training the ICU prediction model further includes: sequencing the human body basic sign data characteristics to obtain human body basic sign data characteristics subjected to sequencing;
performing model construction by using the sorted human body basic sign data characteristics to obtain a constructed ICU prediction model;
optionally, performing model construction by using the score value and the sorted human body basic sign data characteristics to obtain a constructed ICU prediction model.
In one embodiment, the human body basic sign data is input into a pre-trained ICU prediction model to obtain classification results of the ICU which is in the live state and the ICU which is not in the live state.
In one embodiment, a machine learning method is used for model construction of the human body basic sign data to obtain a constructed prognosis prediction model;
optionally, the machine learning method includes one or more of the following: linear regression, logistic regression, linear Discriminant Analysis (LDA), classification and regression trees, naive bayes, KNN, learning vector quantization, support Vector Machines (SVMs), random forests, lightGBM;
optionally, the machine learning method includes: weighted fusion of any two algorithms;
optionally, the machine learning method includes: weighted fusion of random forest and LightGBM.
In one embodiment, the study uniformly employed 5-rule cross validation, targeting average AUC, with Optuna based on bayesian probability for hyper-parametric adjustment of all models. First, the support vector machine is a classical and efficient method that aims to find an optimal hyperplane to classify a data set. Modeling was attempted by adjusting gamma and dependency term using kernel functions including linear, sigmoid, polymodal and radial basis functions. Secondly, logistic regression is a simple, effective and well-interpretable method, data are assumed to obey Bernoulli distribution, sigmoid is introduced on the basis of a linear model, and a gradient descent method is adopted to update parameters, so that data classification is realized. We try to model by selecting different regularization functions, setting different regularization coefficients, residual convergence conditions, and maximum number of iterations. Thirdly, the decision tree is a basic method of various tree models, simulates human decision judgment ideas, and simply and intuitively carries out decision classification according to characteristic attribution. And performing characteristic division by comparing the information entropy and the gini coefficient, and constructing a model by adopting different tree depths, the minimum sample number of leaf nodes and the minimum impurity degree reduction value. And fourthly, the random forest is a bagging ensemble learning method taking a decision tree as a sub-model, training samples are randomly extracted by adopting a bootstrap sample method to train each sub-model, and each sub-model votes to obtain a final classification result. Besides the basic submodel hyperparameters including the maximum tree depth and the like, we also try to adjust ensemble hyperparameters such as the number of submodels and the like. Fifth, lightGBM is a lightweight and efficient (GBDT) based improved model, which uses histogram algorithm to perform feature discretization, uses Leaf-wise to suppress Leaf growth, and enhances the robustness and generalization capability of the algorithm while increasing the algorithm speed. The hyper-parameters of the LightGBM to be adjusted mainly include maximum tree depth, maximum leaf node number, learning rate, regularization parameters, iteration times, minimum splitting gain, maximum histogram bin, and the like. And sixthly, fusing the random forest and the LightGBM with excellent comprehensive performance by adopting a soft voting fusion method with weight to generate an optimal classification model.
The accuracy of the machine learning algorithm is evaluated by combining the area AUC under the Receiver Operating Characteristic (ROC) curve with the accuracy, the sensitivity, the specificity, the positive predictive value and the negative predictive value. The belibration belt verifies the consistency of the model output probability and the true probability. Model interpretability and contribution of various parameters to classification are very important in clinical research, and feature contribution analysis is performed on the optimal model of each endpoint by using a general machine learning interpretation method, namely ShapleAdtive extension (SHAP). For the tree model, we used a variant TreeSHAP of fast and efficient SHAP for analysis.
In one embodiment, a weighted fusion model of random forest and LightGBM is specifically used; for the patient parameter x input to the model, the output classification result y of the model is shown as
w 1 And w 2 Fusion weights of the random forest and the LightGBM are respectively; for random forests, it is composed of N 1 And each base learner DT is trained by different data subsets to perform decision prediction on input, and all predicted values are averaged to obtain a prediction result. For the reaction of N 2 In the LightGBM formed by the base learners CART, the input of each base learner firstly determines the histogram subblock to which the input belongs through the histogram function H, and then generates corresponding predicted values, all the predicted values and the initial value y through the base learners 0 The summation generates a prediction result. Fusing the random forest and the LightGBM through respective weights to obtain a final fusion classification result;
optionally, the initial value y 0 Can be as follows:
wherein
Is the average value of the training set labels;
optionally, the N is 1 The individual basis learner DT is: ID3, C4.5, CART; said N is 2 The individual base learner CART is: a CART decision tree;
optionally, referring to fig. 5, the method is a method for performing calculation by using a random forest algorithm; referring to fig. 6, a method for performing a calculation using the LightGBM algorithm;
in one embodiment, the sorting process comprises:
for one of the human body basic sign data features passing through the human body basic sign data features, calculating all combinations and marginal contribution of a single human body basic sign data feature in the combinations according to a combination sequence formed by the human body basic sign data features;
obtaining the contribution degree of a single human body basic sign data characteristic in the human body basic sign data characteristic according to the marginal contribution;
sorting the human body basic sign data characteristics based on the contribution degrees, and outputting N human body basic sign data characteristics with high contribution degrees;
optionally, N is at least 5;
in the process of constructing the pancreatitis grading prediction model, basic sign data of a human body are sorted according to contribution degrees from high to low, and the result is as follows: multiple effusion, HR, cr, RR, ca, glu, wbc, na, plt, BUN, age, temperature, GCS =15, period authentication, sbp, k, ethiology, hct, gene as shown in FIG. 13. During the pancreatitis grading prediction, the AUC modeled by using the most important 5 characteristics HR, ca, glu, RR and temperature is 0.792, and the AUC modeled by using the most important 10 characteristics HR, ca, glu, RR, temperature, cr, wbc, BUN, k and plt is 0.836.
When whether death happens or not is predicted, the AUC modeled by adopting the most important 5 characteristics of Ca, HR, wbc, BUN and etiology is 0.911, and the AUC modeled by adopting the most important 10 characteristics of Ca, HR, wbc, BUN, etiology, sbp, cr, age, glu and Na is 0.949; when whether the infection necrosis is predicted, the AUC modeled by the most important 5 characteristics of glu, HR, ca, plt and age is 0.776, and the AUC modeled by the most important 10 characteristics of glu, HR, ca, plt, age, wbc, hct, BUN, temperature and RR is 0.825; when the hospital stay number grading is predicted, the AUC modeled by using the most important 5 characteristics of Ca, glu, wbc, plt and HR is 0.790, and the AUC modeled by using the most important 10 characteristics of Ca, glu, wbc, plt, HR, hct, temperature, na, BUN and age is 0.807; when predicting whether to live in the ICU, the AUC modeled by the most important 5 characteristics glu, RR, HR, ca and plt is 0.863, and the AUC modeled by the most important 10 characteristics glu, RR, HR, ca, plt, cr, wbc, temperature, na and BUN is 0.887.
In one embodiment, the score value is calculated using public scoring rules, and when the presence indicator is not included in the public scoring rules, the indicator is directly discarded; or according to the sorting result of sorting the human body basic sign data features based on the contribution degree, scoring the indexes which are not included in the public scoring rules according to the scoring rules of the indexes which are closest to the contribution degree; for example, when the a-index is not incorporated into a publicly available scoring rule, the result from the ranking process: and if the index closest to the contribution degree of the index A is the index B, the scoring rule of the index A is consistent with the scoring rule of the index B.
FIG. 2 isThe embodiment of the invention provides pancreatitis prognosis data processing equipment based on artificial intelligence, which comprises: a memory and a processor;
the memory is to store program instructions;
the processor is configured to invoke program instructions that, when executed, perform one of the artificial intelligence based pancreatitis prognostic data processing methods described above.
FIG. 3 isThe embodiment of the invention provides an artificial intelligence-based pancreatitis prognosis data processing system, which comprises:
an obtaining unit 301, configured to obtain human body basic sign data of a sample;
a classification unit 302, configured to input the human basic sign data into a pre-trained prognosis prediction model to obtain a classification result;
the prognosis prediction model comprises: the training method of the pancreatitis grading prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of training set samples, wherein the labels comprise mild pancreatitis, moderate and severe pancreatitis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out an artificial intelligence based pancreatitis prognosis data processing method as described above.
It can be clearly understood by those skilled in the art that, for convenience and simplicity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: read Only Memory (ROM), random Access Memory (RAM), magnetic or optical disks, and the like.
It will be understood by those skilled in the art that all or part of the steps in the method for implementing the above embodiments may be implemented by hardware that is instructed to implement by a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
While the invention has been described in detail with reference to specific embodiments thereof, it will be apparent to one skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.
Claims (10)
1. An artificial intelligence based pancreatitis prognostic data processing comprising:
acquiring human body basic sign data of a sample;
inputting the human body basic sign data into a pre-trained prognosis prediction model to obtain a classification result;
the prognostic prediction model includes: the training method of the pancreatitis grading prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of training set samples, wherein the labels comprise mild pancreatitis, moderate and severe pancreatitis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
2. The artificial intelligence based pancreatitis prognostic data processing according to claim 1, wherein said training method of the pancreatitis grading predictive model further comprises:
calculating a score value based on human body basic sign data of the training set samples; the score value comprises one or more of the following: APACHE-II score value, marshall score value, sofa score value, qsofa score value and bisap score value;
and carrying out model construction by using the score value and the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
3. The artificial intelligence based pancreatitis prognostic data processing according to claim 2, wherein said training method of the pancreatitis grading predictive model further comprises:
sequencing the human body basic sign data characteristics to obtain human body basic sign data characteristics subjected to sequencing;
performing model construction by using the sequenced human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model;
optionally, performing model construction by using the score value and the sorted human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
4. The artificial intelligence-based pancreatitis prognostic data processing according to claim 1, wherein said prognostic prediction model further comprises: a death prediction model;
the training method of the death prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise death and non-death;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed death prediction model;
optionally, the prognosis prediction model further comprises: an infectious necrosis prediction model;
the method for training the infectious necrosis prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise infection necrosis and non-infection necrosis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed infection necrosis prediction model;
optionally, the prognosis prediction model further comprises: a stay in hospital duration prediction model;
the training method of the stay in hospital prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise a first hospitalization duration and a second hospitalization duration; the first length of stay is less than or equal to the classification threshold, and the second length of stay is greater than or equal to the classification threshold;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
constructing a model by using the human body basic sign data characteristics to obtain a constructed hospitalization duration prediction model;
optionally, the prognostic prediction model further includes: an ICU prediction model;
the ICU prediction model training method comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise an entrance ICU and an exit ICU;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed ICU prediction model.
5. The artificial intelligence-based pancreatitis prognostic data processing according to claim 1, wherein said human body basal signs data include one or several of: age, sex, etiology, body temperature, blood pressure, heart rate, respiratory rate, white blood cells, hematocrit, platelets, blood potassium, blood sodium, blood calcium, creatinine, urea nitrogen, adequacy score, peritoneal irritation, pleural effusion.
6. The artificial intelligence based pancreatitis prognostic data processing according to claim 1, wherein the basic human body sign data is model-constructed by a machine learning method to obtain the constructed prognosis prediction model;
optionally, the machine learning method includes one or more of the following: linear regression, logistic regression, linear Discriminant Analysis (LDA), classification and regression trees, naive bayes, KNN, learning vector quantization, support Vector Machines (SVMs), random forests, lightGBM;
optionally, the machine learning method includes: weighted fusion of any two algorithms;
optionally, the machine learning method includes: weighted fusion of random forest and LightGBM.
7. The artificial intelligence based pancreatitis prognostic data processing according to claim 3, wherein said ranking process comprises:
for one of the human body basic sign data features passing through the human body basic sign data features, calculating all combinations and marginal contribution of a single human body basic sign data feature in the combinations according to a combination sequence formed by the human body basic sign data features;
obtaining the contribution degree of a single human body basic sign data characteristic in the human body basic sign data characteristic according to the marginal contribution;
sorting the human body basic sign data features based on the contribution degrees, and outputting N human body basic sign data features with high contribution degrees;
optionally, N is at least 5.
8. An artificial intelligence-based pancreatitis prognostic data processing apparatus, the apparatus comprising: a memory and a processor;
the memory is to store program instructions;
the processor is configured to invoke program instructions, which when executed, are configured to perform the artificial intelligence based pancreatitis prognostic data processing of any of claims 1-7.
9. An artificial intelligence based pancreatitis prognostic data processing system comprising:
the acquisition unit is used for acquiring human body basic sign data of the sample;
the classification unit is used for inputting the human body basic sign data into a pre-trained prognosis prediction model to obtain a classification result;
the prognosis prediction model comprises: the training method of the pancreatitis grading prediction model comprises the following steps:
acquiring human body basic sign data and classification labels of a training set sample, wherein the labels comprise mild pancreatitis, moderate severe pancreatitis and severe pancreatitis;
carrying out feature extraction on the human body basic sign data of the training set sample to obtain human body basic sign data features after feature extraction;
and constructing a model by using the human body basic sign data characteristics to obtain a constructed pancreatitis grading prediction model.
10. A computer-readable storage medium, having stored thereon a computer program which, when executed by a processor, carries out artificial intelligence-based pancreatitis prognostic data processing as described in any one of claims 1 to 7 above.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210910750.7A CN115331819A (en) | 2022-07-29 | 2022-07-29 | Pancreatitis prognosis data processing method and system based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210910750.7A CN115331819A (en) | 2022-07-29 | 2022-07-29 | Pancreatitis prognosis data processing method and system based on artificial intelligence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115331819A true CN115331819A (en) | 2022-11-11 |
Family
ID=83920578
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210910750.7A Pending CN115331819A (en) | 2022-07-29 | 2022-07-29 | Pancreatitis prognosis data processing method and system based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115331819A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111081377A (en) * | 2020-01-16 | 2020-04-28 | 四川大学 | Necrotic acute pancreatitis patient operation time prediction model |
| US20210327540A1 (en) * | 2018-08-17 | 2021-10-21 | Henry M. Jackson Foundation For The Advancement Of Military Medicine | Use of machine learning models for prediction of clinical outcomes |
| CN113903460A (en) * | 2021-12-10 | 2022-01-07 | 中国医学科学院北京协和医院 | System for predicting severe acute pancreatitis and application thereof |
| CN114023433A (en) * | 2021-09-29 | 2022-02-08 | 四川大学华西医院 | Early prediction system for predicting severity of acute pancreatitis patient |
| CN114141363A (en) * | 2021-12-07 | 2022-03-04 | 川北医学院附属医院 | Severe pancreatitis prediction model construction method based on machine learning method |
-
2022
- 2022-07-29 CN CN202210910750.7A patent/CN115331819A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210327540A1 (en) * | 2018-08-17 | 2021-10-21 | Henry M. Jackson Foundation For The Advancement Of Military Medicine | Use of machine learning models for prediction of clinical outcomes |
| CN111081377A (en) * | 2020-01-16 | 2020-04-28 | 四川大学 | Necrotic acute pancreatitis patient operation time prediction model |
| CN114023433A (en) * | 2021-09-29 | 2022-02-08 | 四川大学华西医院 | Early prediction system for predicting severity of acute pancreatitis patient |
| CN114141363A (en) * | 2021-12-07 | 2022-03-04 | 川北医学院附属医院 | Severe pancreatitis prediction model construction method based on machine learning method |
| CN113903460A (en) * | 2021-12-10 | 2022-01-07 | 中国医学科学院北京协和医院 | System for predicting severe acute pancreatitis and application thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kohli et al. | Application of machine learning in disease prediction | |
| Bandi et al. | Prediction of Brain Stroke Severity Using Machine Learning. | |
| Gil et al. | Predicting seminal quality with artificial intelligence methods | |
| Chen et al. | Heterogeneous postsurgical data analytics for predictive modeling of mortality risks in intensive care units | |
| KR20210068713A (en) | System for predicting disease progression using multiple medical data based on deep learning | |
| Yang et al. | Predicting coronary heart disease using an improved LightGBM model: Performance analysis and comparison | |
| Al-Mualemi et al. | A deep learning-based sepsis estimation scheme | |
| Kamel et al. | Feature selection using grasshopper optimization algorithm in diagnosis of diabetes disease | |
| Pradhan et al. | Diabetes mellitus prediction and classifier comparitive study | |
| Adi et al. | Stroke risk prediction model using machine learning | |
| Kanwal et al. | A multimodal deep learning infused with artificial algae algorithm–An architecture of advanced E-health system for cancer prognosis prediction | |
| Verma et al. | Breast Cancer Survival Rate Prediction In Mammograms Using Machine Learning | |
| Jaganathan et al. | Machine learning for smartphone-based early detection of diabetic disease in Pima Indians diabetes database | |
| Siddiqa et al. | Robust Length of Stay Prediction Model for Indoor Patients. | |
| CN115240854B (en) | Pancreatitis prognosis data processing method and system | |
| CN115331819A (en) | Pancreatitis prognosis data processing method and system based on artificial intelligence | |
| Bang et al. | Naïve bayes ensemble: A new approach to classifying unlabeled multi-class asthma subjects | |
| Ramdhani et al. | Heart failure prediction based on random forest algorithm using genetic algorithm for feature selection | |
| Riyaz et al. | Improving coronary heart disease prediction by outlier elimination | |
| Uddin et al. | An Explainable machine learning framework for early detection of stroke | |
| Tuba et al. | Design of Decision Support System in the Metastatic Colorectal Cancer Data Set and Its Application | |
| EP4227948A1 (en) | Machine-learning based prediction of the survival potential of cells | |
| Jayaraj et al. | Disease forecasting and severity prediction model for COVID-19 using correlated feature extraction and feed-forward artificial neural networks | |
| Thakkar et al. | Diabetes Classification and Diet Recommendation | |
| Mohammed et al. | Comparison between neural networks and binary logistic regression for classification observation (case study: risk factors for cardiovascular disease) |
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 |