CPC A61M 16/026 (2017.08) [A61B 5/02055 (2013.01); A61B 5/14542 (2013.01); A61B 5/7203 (2013.01); A61B 5/7267 (2013.01); A61B 5/7275 (2013.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01); G16H 10/60 (2018.01); G16H 20/40 (2018.01); G16H 40/40 (2018.01); G16H 40/67 (2018.01); G16H 50/20 (2018.01); G16H 50/30 (2018.01); A61B 5/021 (2013.01); A61B 5/024 (2013.01); A61B 5/0816 (2013.01); A61B 5/091 (2013.01)] | 8 Claims |
1. A method for building and using a ventilator-weaning timing prediction system, the ventilator-weaning timing prediction system predicting a try-weaning regimen and a complete weaning regimen for a patient currently connected to a ventilator at a medical facility, the ventilator being connected to a physiological monitor, and the physiological monitor being connected to a medical database server of the medical facility, the medical database server of the medical facility including patient record data of the patient currently connected to the ventilator at the medical facility and of patients previously connected to ventilators at the medical facility, the method comprising:
providing a server host system including a big data database subsystem and a prediction model building subsystem;
connecting the big data database subsystem with the medical database server, and the big data database subsystem capturing weaning-related medical data from the patient record data of patients previously connected to ventilators at the medical facility stored in the medical database server, the big data database subsystem analyzing the captured weaning-related medical data for determining if the captured weaning-related medical data is of a standard data configuration, and upon determining that the captured weaning-related medical data is not of a standard data configuration, the big data database subsystem transforming the captured weaning-related medical data into a standard data configuration;
initiating the big data database subsystem to divide the captured weaning-related medical data in a standard data configuration into a training data set for try-weaning and a testing data set for try-weaning, and further employing the big data database subsystem to extract feature variables for try-weaning from the training data set for try-weaning, each extracted feature variable for try-weaning being a medical data parameter impacting weaning of a corresponding patient previously connected to a ventilator partially therefrom, wherein selection of the extracted feature variables for try-weaning being responsive to at least input provided by medical practitioners at the medical facility, and wherein the extracted feature variables for try-weaning include at least age, disease severity, ventilator parameters, and physiological sign parameters of a corresponding patient previously connected to a ventilator;
initiating the prediction model building subsystem to connect with the big data database subsystem and acquire the extracted feature variables for try-weaning, the training data set for try-weaning, and the testing data set for try-weaning, the prediction model building subsystem operating on the extracted feature variables for try-weaning and the training data set for try-weaning using a plurality of preselected algorithms to thereby generate respective multiple prediction weaning models for try-weaning, each prediction weaning model for try-weaning providing a probability of success of partially weaning at predetermined spans of time;
subsequent to generating the multiple prediction weaning models for try-weaning, the prediction model building subsystem validating the multiple generated prediction weaning models for try-weaning using the testing data set for try-weaning, and responsive to and in accordance with validation results of the multiple prediction weaning models for try-weaning, the prediction model building subsystem selecting at least one of the multiple prediction weaning models for try-weaning to be the first designated weaning prediction model, the first designated weaning prediction model providing success probabilities for partially weaning at the predetermined spans of time; and
initiating the big data database subsystem to divide the captured weaning-related medical data in a standard data configuration into a training data set for complete weaning and a testing data set for complete weaning, and further employing the big data database subsystem to extract feature variables for complete weaning from the training data set for complete weaning, each extracted feature variable for complete weaning being a medical data parameter impacting weaning of a corresponding patient previously connected to a ventilator fully therefrom, wherein selection of the extracted feature variables for complete weaning being responsive to at least input provided by medical practitioners at the medical facility, and wherein the extracted feature variables for complete weaning include at least age, disease severity, ventilator parameters, and physiological sign parameters of a corresponding patient previously connected to a ventilator;
initiating the prediction model building subsystem to connect with the big data database subsystem and acquire the extracted feature variables for complete weaning, the training data set for complete weaning, and the testing data set for complete weaning, the prediction model building subsystem operating on the extracted feature variables for complete weaning and the training data set for complete weaning using the plurality of preselected algorithms to thereby generate respective multiple prediction weaning models for complete weaning, each prediction weaning model for complete weaning providing a probability of success of fully weaning at predetermined spans of time;
subsequent to generating the multiple prediction weaning models for complete weaning, the prediction model building subsystem validating the multiple generated prediction weaning models for complete weaning using the testing data set for complete weaning, and responsive to and in accordance with validation results of the multiple prediction weaning models for complete weaning, the prediction model building subsystem selecting at least one of the multiple prediction weaning models for complete weaning to be the second designated weaning prediction model, the second designated weaning prediction model providing success probabilities for fully weaning at the predetermined spans of time; and
actuating the server host system to provide a medical information system service interface program subsystem, a feature value capturing service program subsystem, and a weaning prediction service program subsystem, connecting the medical information system service interface program subsystem with the feature value capturing service program subsystem, the weaning prediction service program subsystem, and a medical information system server of the medical facility that includes at least an input terminal and an output visual display terminal, the medical information system server being connected to the medical database server, the medical information system service interface program subsystem providing an input and an output interface between the server host system and the medical information system server;
the medical information system service interface program subsystem providing a command prompt via the input interface for the medical practitioner at the medical facility via the input terminal of the medical information system server to selectively initiate the try-weaning regimen for the patient currently connected to the ventilator at the medical facility, and the feature value capturing service program subsystem responsively connecting with the medical database server to capture the patient record data of the patient currently connected to the ventilator, and further, the feature value capturing service program subsystem extracting medical feature values relating to try-weaning from the captured patient record data of the patient currently connected to the ventilator, the extracted medical feature values relating to try-weaning of the patient currently connected to the ventilator corresponding to the feature variables for try-weaning extracted from the training data set for try-weaning, and the weaning prediction service program subsystem operating on the designated first weaning prediction model according to the extracted medical feature values relating to try-weaning of the patient currently connected to the ventilator for generating the try-weaning regimen for the patient currently connected to the ventilator, and the medical information system service interface program subsystem displaying the try-weaning regimen for the patient currently connected to the ventilator to the medical practitioner at the medical facility on the output visual display terminal of the medical information system server via the output interface, the try-weaning regimen for the patient currently connected to the ventilator including success probabilities for partially weaning the patient from the ventilator at the predetermined spans of time;
wherein the ventilator to which the patient currently is connected to is selectively operable by the medical practitioner for partially weaning the patient from the ventilator in correspondence with the try-weaning regimen displayed on the output visual display terminal of the medical information system server;
wherein when the ventilator to which the patient currently is connected to is operated on by the medical practitioner to partially wean the patient therefrom, the ventilator is selectively switched from a control mode to a support mode, and the patient being successfully try-weaned with respect the ventilator to which the patient currently is connected to only if the ventilator remains in the support mode for at least forty eight (48) hours, and if the ventilator to which the patient currently is connected to is selectively switched by the medical practitioner from the support mode back to the control mode responsive to patient record data of the patient being indicative of deteriorating health of the patient, the patient has been unsuccessfully try-weaned with respect the ventilator to which the patient currently is connected to; and
subsequent to the patient currently connected to the ventilator being successfully try-weaned from the ventilator, the medical information system service interface program subsystem providing a command prompt via the input interface for the medical practitioner at the medical facility via the input terminal of the medical information system server to selectively initiate the complete weaning regimen for the patient currently connected to the ventilator at the medical facility, and the feature value capturing service program subsystem responsively connecting with the medical database server to capture the patient record data of the patient currently connected to the ventilator, and further, the feature value capturing service program subsystem extracting medical feature values relating to complete weaning from the captured patient record data of the patient currently connected to the ventilator, the extracted medical feature values relating to complete weaning of the patient currently connected to the ventilator corresponding to the feature variables for complete weaning extracted from the training data set for complete weaning, and the weaning prediction service program subsystem operating on the designated second weaning prediction model according to the extracted medical feature values relating to complete weaning of the patient currently connected to the ventilator for generating the complete weaning regimen for the patient currently connected to the ventilator, and the medical information system service interface program subsystem displaying the complete weaning regimen for the patient currently connected to the ventilator to the medical practitioner at the medical facility on the output visual display terminal of the medical information system server via the output interface, the complete weaning regimen for the patient currently connected to the ventilator including success probabilities for fully weaning the patient from the ventilator at the predetermined spans of time;
wherein the ventilator to which the patient currently is connected to is selectively operable by the medical practitioner for fully weaning the patient from the ventilator in correspondence with the complete weaning regimen displayed on the output visual display terminal of the medical information system server; and
wherein with respect to the extracted medical feature values relating to try-weaning and the extracted medical feature values relating to complete weaning of the patient currently connected to the ventilator, corresponding patient record data of the patient currently connected to the ventilator is continuously captured over a predetermined period of time by the physiological monitor, and responsively, the corresponding patient record data of the patient currently connected to the ventilator is continuously updated in the medical database server, and further responsively, the feature value capturing service program subsystem continuously updates the captured patient record data of the patient currently connected to the ventilator, and the extracted medical feature values relating to try-weaning and the extracted medical feature values relating to complete weaning of the patient currently connected to the ventilator are thereby, respectively, updated on a real-time basis for operation on by the first designated weaning prediction model and the second designated weaning prediction model, and the try-weaning regimen and the complete weaning regimen for the patient currently connected to the ventilator being generated are, respectively, updated on a real-time basis, in accordance with the extracted medical feature values relating to try-weaning, and the extracted medical feature values relating to complete weaning, of the patient currently connected to the ventilator.
|