75 results about "Logistische regression" patented technology

The invention discloses a pulmonary nodule risk assessment system, and the system carries out the comprehensive assessment of the obtained image analysis data of a patient, the CAC detection data of the patient and the risk factors of the patient through a logistic regression model, a decision tree model or a random forest model, and carries out the assessment of the pulmonary nodule risk of the patient. Whether the pulmonary nodules exist or not can be evaluated, the risk of the existing pulmonary nodules can be predicted, and high accuracy is achieved.

The embodiment of the invention discloses an acute kidney injury patient death rate prediction method and device, a server and a storage medium. The method comprises the steps of generating medical feature data of a specific structure according to clinical medical data of a to-be-tested patient; and selecting a corresponding pre-trained random forest model according to the survival state and survival days of the to-be-tested patient, inputting the medical feature data of the to-be-tested patient into the corresponding pre-trained random forest model, and outputting the death rate of the to-be-tested patient. According to the technical scheme of the embodiment of the invention, the death rate of the acute kidney injury patient is predicted through the random forest model, the structured clinical data is used for training, and the logistic regression algorithm is used to calibrate the model, so that fuzzy clinical definition and data acquisition deviation are effectively avoided, and theaccuracy of predicting the death rate of the acute kidney injury patient is improved.

The invention discloses a ventricular fibrillation recognition method based on a machine learning technique. The ventricular fibrillation recognition method comprises the following steps of treating an original electrocardiosignal so as to extract ventricular fibrillation characteristics; through a marked electrocardio database, training a logistic regression model which is used for judging whether each ECG signal is ventricular fibrillation; and through the trained model, acquiring the probability that each ECG signal is the ventricular fibrillation. According to the ventricular fibrillationrecognition method disclosed by the invention, the most effective combination is selected from various characteristics provided in the prior art, so that the properties of a ventricular fibrillation detection algorithm can be effectively improved. Besides, a heart rate characteristic is applied to ventricular fibrillation detection. In order to solve the problems that QRS detection is difficult and heart rate is possibly quite inaccurate when the ventricular fibrillation occurs, according to the method disclosed by the invention, the quality information of the signal is used, so that the modelcan accurately judge the condition when heart rate information needs to be neglected, and the condition when obvious non-ventricular fibrillation electrocardiogram can be eliminated according to theheart rate.

The invention discloses a colorectal cancer auxiliary diagnosis kit for fecal nucleic acid detection and a use method of the colorectal cancer auxiliary diagnosis kit. The kit comprises a KRAS gene mutation detection reagent, a fusobacterium nucleatum detection reagent, a DNA sulfite conversion reagent and a Septin9, SDC2, HOXA11 and NDRG4 gene methylation detection reagent. The use method comprises the following steps: collecting an excrement sample, and then completing DNA purification, KRAS gene mutation detection, fusobacterium nucleatum abundance detection and Septin9, SDC2, HOXA11 and NDRG4 gene methylation detection of the sample by using reagents in the kit; and calculating the obtained data through a logistic regression model, and performing result evaluation and analysis. The colorectal cancer detection sensitivity reaches 89% or above, the specificity reaches 94% or so, high-sensitivity and high-specificity colorectal cancer detection can be achieved, and doctors can be helped to guide follow-up treatment medication of patients.

The invention discloses a cerebral apoplexy attack risk prediction system. The system comprises a data acquisition unit, a preprocessing unit, a feature screening unit, a model establishing unit and a model prediction unit, wherein the data acquisition unit is used for acquiring health data information of a target crowd and performing desensitization processing; the preprocessing unit is used for performing standardized processing on basic data information of the target crowd and marking a cerebral apoplexy risk category; the feature screening unit is used for screening cerebral apoplexy feature data based on an IV value analysis method to obtain feature data with relatively high model prediction value, and forming a data set; the model establishing unit is used for training data by utilizing an Adaboost enhanced learning method based on logistic regression and establishing a fusion model; and the model prediction unit is used for predicting to-be-tested sample data through the fusion model to obtain the cerebral apoplexy risk category. According to the system, accurate analysis of the cerebral apoplexy risk category is realized based on the health data information of the target crowd, so that the attack risk prediction efficiency can be improved.

The invention discloses a medicine recommendation evidence-based support method and device, equipment and a storage medium. The method comprises the following steps: acquiring a recommended prescription list of a target patient by utilizing a medicine recommendation model; acquiring N similar patients of the target patient, wherein N is a positive integer; acquiring feature information, prescription information and prognosis information of the N similar patients; constructing a logistic regression model according to the feature information, the prescription information and the prognosis information of the N similar patients; sorting the recommended prescriptions in the recommended prescription list of the target patient by using the logistic regression model to obtain a sorting result of the recommended prescriptions. According to the method, the device, the equipment and the storage medium, prescription recommendation can be accurately carried out on the patient.

An embodiment of the invention provides a heart valve abnormality analysis method based on a convolutional neural network. The method comprises the steps of segmenting a collected heart sound, and calculating a time-frequency spectrum of each heart sound segment; inputting the time-frequency spectrum of each heart sound segment into the convolutional neural network, and outputting a first result that the heart sound is normal or abnormal; extracting envelope spectrum characteristics and power spectrum characteristics of the heart sound with the first result that the heart sound is abnormal, inputting a logistic regression hidden semi-Markov model for segmentation, and outputting K states to which all frames in a cardiac cycle of the heart sound belong, wherein K is a natural number; extracting an energy feature of each state of the heart sound; and inputting the energy features into a support vector machine to obtain analysis results of aortic valve stenosis and / or aortic valve regurgitation and mitral valve stenosis and / or mitral valve regurgitation. According to the method, the heart valve abnormality is subjected to two-stage analysis, so that the accuracy of heart valve abnormality judgment can be improved, and the misdiagnosis risk is reduced.

The invention discloses a medical ability evaluation method and device based on an electronic medical record, equipment and a medium. The method comprises the steps of obtaining an electronic medical record initial text issued by a to-be-evaluated doctor; performing text preprocessing on the initial text; carrying out chapter division by utilizing a preset rule template; performing entity feature extraction by using a pre-trained deep learning model to obtain a corresponding entity feature extraction result; performing vectorization encoding on the entity feature extraction result to obtain a feature vector; and inputting the feature vector into a preset logistic regression model to obtain a diagnosis and treatment ability score of the doctor. According to the invention, the feature entities in the electronic medical record are extracted through the model, then the treatment capacity of doctors is evaluated through the logistic regression model, the labor cost can be reduced, the mobility performance is good, and diagnosis and treatment guidance can be provided for medical staff. The invention also relates to a medical ability evaluation method based on the electronic medical record realized by the blockchain in a blockchain network.

The invention provides an esophageal squamous cell carcinoma risk prediction method based on clinical phenotype and logistic regression analysis. The method is used for realizing prognosis survival risk assessment of esophageal squamous cell carcinoma patients. The method comprises the following steps: firstly, screening out characteristic indexes according to clinical detection data of the esophageal squamous cell carcinoma patients, and constructing a decision tree classifier according to the characteristic indexes; secondly, dividing the esophageal squamous cell carcinoma patients into early-stage esophageal squamous cell carcinoma patients and middle-late-stage esophageal squamous cell carcinoma patients by utilizing the decision tree classifier; then, obtaining blood index informationof the esophageal squamous cell carcinoma patient one week before the operation, and screening out blood indexes with high correlation with the survival risk of the esophageal squamous cell carcinomapatient, and constructiung a logistic regression model; inputting the classified blood indexes of the esophageal squamous cell carcinoma patient into the logistic regression model to obtain a prognosis survival risk probability value of the esophageal squamous cell carcinoma patient; and judging the prognosis survival risk. According to the method, the postoperative survival state of the esophageal squamous cell carcinoma patient can be accurately judged, the risk prediction performance is improved, and the risk prediction cost is reduced.

The invention discloses a cerebrovascular disease neurological function damage degree prediction model based on kernel principal component analysis and polynomial characteristics. The method comprisesthe following steps: firstly, acquiring processed medical structured data from a big data platform, preprocessing all the data, and then adopting the following feature extraction method by firstly, constructing features, then extracting feature data with significant differences from all the feature data, then mapping the data to a high dimension by adopting kernel principal component analysis, and then reducing the dimension; and finally, constructing a machine learning model of the stroke condition severity based on logistic regression, a support vector machine and a random forest. Experiments respectively set four control groups, and it is proved that compared with the control groups, the method can effectively improve the prediction performance of each classifier on the damage degree of the neurological function of the cerebrovascular disease, and compared with the control groups, the method greatly shortens the feature extraction time.

[Object] An object is to provide a brain activity analyzing method for realizing a biomarker using brain function imaging for neurological / mental disorders.[Solution] From data of resting-state functional connectivity MRI obtained by measuring groups of healthy subjects and patients, a correlation matrix of degrees of activities among prescribed brain regions is derived. By a sparse canonical correlation analysis (SCCA) of attributes of subjects and the correlation matrix, elements of correlation matrix that connect to canonical variables corresponding only to the diagnosis label are extracted. By sparse logistic regression (SLR) during Leave-One-Out Cross Validation of a first sum-set of elements of correlation matrix obtained as a result of feature extraction by a sparse regularized canonical correlation analysis, a second sum-set of elements of the correlation matrix is extracted. By discrimination analysis using sparse logistic regression on the second sum-set, a discriminator is generated.

The invention discloses an insurance underwriting system, method and device and a computer readable medium, and relates to the technical field of computers. A specific embodiment of the system comprises: a data subsystem used for extracting basic medical data from a medical insurance database; a processing subsystem which is used for screening patients suffering from chronic diseases and malignanttumors in the basic medical data according to the types of the chronic diseases and the types of the malignant tumors; an analysis subsystem which is used for carrying out logistic regression analysis according to the basic medical data of the patient so as to determine the correlation between factors and malignant tumors, wheein the factors comprise the type of chronic diseases and the gender orage of an insured person; and an underwriting subsystem which is used for determining insurance data of the insured person based on the correlation between the factor and the malignant tumor. According to the embodiment, the relationship between the potential factor and the disease can be obtained, so that the underwriting accuracy is improved.

The invention belongs to the technical field of medical diagnosis, and particularly relates to an isolated pulmonary nodule malignant risk prediction system based on a logistic regression model. The invention provides an isolated pulmonary nodule malignant risk prediction system, and the system comprises: a data collection module which is used for acquiring and / or inputting variable data of an isolated pulmonary nodule patient; a data operation module which is used for substituting the variable data into the logistic regression model for calculation to obtain a nodule malignancy probability and / or a judgment conclusion about whether the nodule is malignant; and a data output module which is used for outputting a nodule malignancy probability and / or a judgment conclusion about whether the nodule is malignant or not. The model provided by the invention is suitable for diagnosing whether nodules of SPN patients in Chinese population are malignant or not.

The invention discloses a composition and kit for methylation noninvasive early detection of gynecological tumours. The composition comprises a target gene detection primer used for detecting methylation sites of target genes; and the target genes are PTGER4, HAND2, PAX1, HOXA9, FOXD3, OPCML, HS3ST2 and JAM3. The kit comprises a detection primer and a probe designed for the gene of the composition, a reference gene detection primer and a probe, a PCR reaction liquid and other reagents. By detecting the methylation state of the gene, noninvasive early screening can be simultaneously performed on various gynecological tumours at one time; and the method has the advantages of high detection speed, low cross contamination, final result accuracy ensured by a logistic regression algorithm and the like, and helps to realize the purposes of early discovery, early diagnosis and early treatment of malignant gynecological tumours.

The invention discloses a method for constructing a tumor screening model, which comprises the following steps of: performing classification model construction on a single classification index by adopting a support vector machine to obtain a prediction score of each index on a sample, and integrating the scores by using a logistic regression model to obtain the tumor screening model. The invention also provides a cancer screening, risk prediction and / or diagnosis method of the tumor screening model constructed based on the method, and a related kit, system, device, computer readable storage medium and equipment. Accurate screening of cancers of each stage, including early cancers, is realized.

The invention belongs to the technical field of tumor gene detection, and particularly relates to a gene tag for predicting breast cancer paclitaxel and anthracycline neoadjuvant chemotherapy sensitivity and application thereof. In the invention, based on LASSO logistic regression, a gene tag consisting of 25 genes related to breast cancer neoadjuvant chemotherapy sensitivity is obtained, and a score containing a gene expression quantity is calculated and predicted, so that the sensitivity of a breast cancer patient using paclitaxel and anthracycline neoadjuvant chemotherapy can be accurately predicted, the response of the patient to treatment is predicted, and whether the patient benefits from chemotherapy is discriminated, and thus selection of neoadjuvant chemotherapy regimens for breast cancer is guided, excessive treatment is avoided, and the medical cost is reduced.

The invention discloses an acute myelogenous leukemia drug sensitivity related gene classifier constructed by a machine learning algorithm. The acute myelogenous leukemia drug sensitivity related gene classifier is characterized by comprising sample clustering and gene screening; the sample clustering is to respectively cluster the patient sensitivity of 24 drugs by using a K-means clustering algorithm; the gene screening is to carry out gene screening and verification on methylation and transcriptome data of 24 drugs by utilizing a feature selection model according to the clustering result of a patient. According to the method, the final screening of the target gene is realized by adopting logistic regression, ridge regression, RFECV-SVM and RFECV-RF algorithms. Parameter optimization is carried out on logistic regression and ridge regression by using four-fold layered cross validation, and a threshold value of feature selection is set as an average value of feature weights, that is, features with the feature weights larger than the average value are reserved. The RFECV algorithm uses different learning models SVM and RF for screening.

The invention provides a refractory schizophrenia risk prediction method based on machine learning. The method comprises the following steps: acquiring the pathological feature index data of a non-refractory schizophrenia patient and a refractory schizophrenia patient, wherein the pathological feature index data to comprise an SNP site typing result; training a deep learning model and a logistic regression model according to the pathological feature index data of the non-refractory schizophrenia patient and the refractory schizophrenia patient; and acquiring pathological feature index data ofa to-be-tested person, and inputting the pathological feature index data into a feature evaluation model of the health condition of the to-be-tested person, thereby outputting an evaluation result ofthe risk of the to-be-tested person in suffering from refractory schizophrenia. By means of the refractory schizophrenia risk prediction method based on machine learning, a refractory schizophrenia genetic risk assessment system of Chinese population is constructed, the genetic susceptibility risk of refractory schizophrenia of a subject is assessed, and clinical treatment is assisted to effectively diagnose and treat refractory schizophrenia in time.

The invention discloses an ovarian malignant tumor patient thrombus risk prediction model modeling method and prediction device, and the modeling method comprises the steps: obtaining clinical examination data of a research object; taking whether venous thromboembolism occurs or not as a dependent variable, and establishing a logistic regression model by using the clinical examination data to obtain independent prediction factors of venous thromboembolism; and establishing a column graph according to the independent prediction factors. According to the modeling method and the prediction device, the risk of venous thromboembolism of the patient with the ovarian malignant tumor is predicted through the column diagram constructed by using the clinical examination data, and a quantitative prediction result with higher accuracy can be obtained.

The invention belongs to the technical field of intelligent medical treatment, and particularly relates to a model for predicting the RAS gene state of a patient suffering from advanced colorectal cancer liver metastasis based on radiomics characteristics and application of the model. According to the method, a model for predicting the RAS gene state of the advanced colorectal cancer liver metastasis patient based on the radiomics characteristics is successfully constructed, specifically, an enhanced CT image of the patient before initial treatment is collected, the radiomics characteristics are extracted, 12 radiomics characteristics are obtained through 1000 times of Lasso-Logistic analysis, and a radiomics prediction model is constructed through a multi-factor logistic regression method. According to the method, the enhanced CT image is utilized, the RAS gene state of the colorectal cancer liver metastasis patient can be economically, noninvasively and rapidly predicted, genetic typing is assisted, clinical doctors can be helped to make accurate individualized treatment decisions, and the method has great clinical significance.

The invention belongs to the technical field of operation risk diagnosis and prediction models, and discloses a thoracotomy risk diagnosis and prediction model in thoracoscope lung cancer resection and a construction system. The construction system of the thoracoscope lung cancer resection transfer thoracotomy risk diagnosis and prediction model comprises a data acquisition module, a risk diagnosis model construction module, a data preprocessing module, a data fusion module, a data set training module, a model verification and optimization module, a central control module, a risk diagnosis and prediction module, a data storage module, and an update display module. According to the method, on the basis of tens of thousands of data of the thoracic surgery department of the West China hospital, a risk diagnosis and prediction model for lung cancer thoracoscopic surgery transfer thoracotomy is established through machine learning algorithms such as classical logistic regression, an artificial neural network and a random forest, and after some indexes are input on the basis of an established webpage version nomogram, the size of the chest opening risk in the thoracoscope lung cancer resection can be predicted on line.

The invention discloses a diabetic ICU patient death risk prediction method based on artificial intelligence. The method comprises the following steps: S1, collecting ICU data of a diabetic ICU patient according to a set standard; s2, importing the collected ICU data into a database, and taking out vital sign data of a diabetic ICU patient; s3, deleting data which do not meet the specification in the acquired ICU data and the vital sign data; s4, constructing a prediction model based on a logistic regression algorithm, a prediction model based on a decision tree algorithm, a prediction model based on a random forest algorithm and a model based on hybrid bidirectional LSTM; s5, inputting the ICU data and the vital sign data processed in the step S3 into the model in the step S4 for model training and iterative training, and selecting an optimal model; and S6, inputting the basic information of the diabetic ICU patient and the vital sign data of the patient into the optimal model to predict the death risk. According to the invention, an early warning effect can be achieved, so that the treatment condition of a diabetic ICU patient is improved.

The invention discloses a logistic regression-based pharmacokinetic parameter prediction method for a drug compound in the technical field of drug research and development, and the method comprises the following steps: 1, file input: inputting a molecular three-dimensional structure file of the drug compound; 2, calculation of a descriptor: calculating the descriptor of a molecule through CDK; 3,a training process: dividing text data and label data contained in the data set into a training set and a verification set; training a model by using the training set data, and outputting a parameterprediction model; adopting an Adam optimizer to optimize the model, and optimizing the model through 20-stage training. According to the method, a neural multi-task logistic regression algorithm is utilized, and rules can be automatically designed under the condition of little or no manual intervention. Through the molecular structure of the drug compound, the pharmacokinetic lipophilicity, solubility, plasma protein binding rate, transdermal property and other parameter values of the drug compound are accurately predicted.

A method to quantify and correct for drug-drug interaction, physiologic change, diet, weight, genetic data and compliance on Warfarin dosing. The International Normalized Ratio (INR), a lab value used to follow Warfarin use, will fluctuate in an unpredictable manner due to factors other than the current Warfarin dose. The method mathematically describes these changes and eventually adjusts for these interacting factors through the use of logistic regression (LR) or multiple linear regression analysis. By anticipating changes in INR, Warfarin dosing can be adjusted resulting in patients having their INR be therapeutic range. The technique can be used in any field that requires a specific measured quantity, with variables that change and the need to correct for changes with a mathematical model.

The invention relates to a Chinese herbal medicine recommendation system based on deep learning. The method comprises the following steps: by taking a public 'typhoid theory' data set which is widely accepted in the field of traditional Chinese medicine as a research object, establishing a standardized database through 'Multi-hot' code preprocessing on symptoms and traditional Chinese medicine prescription information; simulating an internal matching rule between a clinical symptom group and Chinese herbal medicines by using a convolutional neural network, designing a clinical traditional Chinese medicine prescription recommendation algorithm by using a Pytorch framework, and comparing with three traditional machine learning models including a support vector machine, a Bayesian classifier and logistic regression; therefore, the effectiveness and application value of the algorithm are explained from two dimensions of quantification and qualification. According to the method, Chinese herbal medicine prescriptions can be accurately recommended, and the method plays an important role in promoting traditional Chinese medicine diagnosis and treatment research based on artificial intelligence.

The invention provides a new coronal pneumonia patient outcome prediction method based on an interpretable machine learning algorithm, wherein the method comprises the steps: extracting COVID-19 patient data from a database, and dividing the patient data into an experimental group and a control group according to the illness state conversion condition of a patient; interpolating the missing value of each index through random forest regression; screening the indexes of the input model, and taking the screened indexes as key risk factors for identifying the deterioration of the patient; inputting the key risk factors of the patient into the XGBoost model and the logistic regression model; selecting an XGBoost model with better prediction expressive force to generate an index combination, and performing prediction by using the XGBoost model and recording the prediction result; defining the early warning range of the key index; when the key risk index of the patient enters the early warning range, giving out an alarm prompt to medical staff. According to the invention, the calculation result of the algorithm and the clinical experience of a doctor are synthesized, and two index combinations composed of 15 first groups of indexes and 5 second groups of indexes are proposed to be used for predicting the condition of the new coronal pneumonia patient.

The invention relates to a telemedicine-aided decision-making method for chronic obstructive pulmonary disease. According to the method, the problem in the prior art that the existing treatment of chronic obstructive pulmonary diseases cannot meet remote medical treatment is mainly solved. The telemedicine-aided decision-making method for the chronic obstructive pulmonary disease comprises the following step: step 1, realizing the input x=[x1, x2, x3, x4] of the telemedicine-aided decision-making method for chronic obstructive pulmonary disease, wherein x1 is the respiratory rate, and the unitof x1 is times / 10 seconds; x2 is pulse oxyhemoglobin saturation, and the unit of x2 is%; x3 is blood pressure, and the unit of x3 is mmHg; x4 is body temperature, and the unit of x4 is DEG C and thelike. According to the telemedicine-aided decision-making method for the chronic obstructive pulmonary disease, a regression model can be trained through a large amount of data through a logistic regression algorithm. The accuracy of data analysis is enhanced, and the realization of automatic medical decision-making of the chronic obstructive pulmonary disease is facilitated.

The invention discloses a correlation detecting system for correlation between a mechanical ventilation driving pressure and a ventilator-related event. The system comprises the components of a data preprocessing module which obtains a mechanical ventilation driving pressure starting value, a final value and a mechanical ventilation driving pressure change value in 48 hours, and filling in an acquired to-be-tested case index; a characteristic selecting module which screens a pathological characteristic related with the ventilator-related event generation from ventilator-related event generation cases as a training set; and a model constructing and detecting module which constructs a correlation detecting module based on the training set by means of a logistic regression algorithm, detectsthe to-be-tested case index based on the correlation detecting model and determines a correlation probability between the mechanical ventilation driving pressure change value and the ventilator-related event generation. Based on the logistic regression algorithm of machine learning, the ventilator-related event (VAE) is correlated with the mechanical ventilation driving pressure change value, thereby monitoring the influence of the mechanical ventilation driving pressure change to the ventilator-related event (VAE).