Tree-based data exploration and data-driven protocol
A method for providing a treatment recommendation to a physician for treating a patient is disclosed. The method comprises determining, from a processor in communication with a patient data repository, a first treatment recommendation based on a combination of selected patient demographics from the patient data repository applicable to the patient, and operational parameters of a plurality of ventricular assist devices (VADs) suitable for treating the patient, the first treatment recommendation having a first survival rate and comprising the use of a first VAD. The method then obtains a first signal from using the first VAD on the patient. The method then determines a second treatment recommendation based on the first signal and the first treatment recommendation, the second treatment recommendation having a second survival rate. The method then provides the second treatment recommendation to the physician if the second survival rate is higher than the first survival rate.
1. A method comprising:
obtaining training data from a data repository, wherein the training data comprises patient data and operational parameters of a plurality of ventricular assist devices (VADs), and wherein the data repository comprises data from treatment of acute myocardial infarction (AMI) patients, high-risk percutaneous coronary interventions (PCI) patients, or patients in cardiogenic shock;
training a prediction model with the training data, wherein the trained prediction model is configured to use a tree-based machine learning algorithm to predict a plurality of survival rates, and wherein each one of the predicted survival rates is associated with using a different one of the VADs to treat a patient; and
adjusting a hyper-parameter of the tree-based machine learning algorithm to improve both a cross-validation score and a training score.
2. The method of claim 1 , further comprising:
using the trained prediction model to select one of the VADs for the patient.
3. The method of claim 2 , wherein the selected VAD is associated with a highest one of the predicted survival rates.
4. The method of claim 1 , wherein the hyper-parameter is adjusted to maximize both the cross-validation score and the training score.
5. The method of claim 1 , wherein the hyper-parameter is tree depth.
6. The method of claim 5 , wherein the tree depth is greater than three.
7. The method of claim 5 , wherein the tree depth is less than twelve.
8. The method of claim 5 , wherein the tree depth is six.
9. The method of claim 1 , wherein the tree-based machine learning algorithm is a bagging and random forest algorithm or a classification decision tree algorithm.
10. The method of claim 1 , wherein the patient data comprises patient demographics and treatment data, wherein the patient demographics comprise gender, age, or region, and wherein the treatment data comprises duration of support, indication for use, insertion site, or ejection fraction.
11. The method of claim 1 , wherein each one of the VADs is configured to be percutaneously inserted into a heart and to run in parallel with the heart to supplement cardiac output.
12. A system comprising:
one or more processors configured to use a trained prediction model to select one of a plurality of ventricular assist devices (VADs) for a patient,
wherein the prediction model is trained with training data from a data repository, wherein the training data comprises patient data and operational parameters of the VADs, and wherein the data repository comprises data from treatment of acute myocardial infarction (AMI) patients, high-risk percutaneous coronary interventions (PCI) patients, or patients in cardiogenic shock, and
wherein the trained prediction model is configured to use a tree-based machine learning algorithm to predict a plurality of survival rates, wherein each one of the predicted survival rates is associated with using a different one of the VADs to treat the patient, and wherein a hyper-parameter of the tree-based machine learning algorithm is selected to improve both a cross-validation score and a training score.
13. The system of claim 12 , wherein the selected VAD is associated with a highest one of the predicted survival rates.
14. The system of claim 12 , wherein the hyper-parameter is selected to maximize both the cross-validation score and the training score.
15. The system of claim 14 , wherein the hyper-parameter is tree depth.
16. The system of claim 15 , wherein the tree depth is between three and twelve.
17. The system of claim 12 , wherein the tree-based machine learning algorithm is a bagging and random forest algorithm or a classification decision tree algorithm.
18. The system of claim 12 , wherein the patient data comprises patient demographics and treatment data, wherein the patient demographics comprise gender, age, or region, and wherein the treatment data comprises duration of support, indication for use, insertion site, or ejection fraction.
19. The system of claim 12 , wherein each one of the VADs is configured to be percutaneously inserted into a heart and to run in parallel with the heart to supplement cardiac output.
20. A non-transitory computer readable storage medium having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to:
use a trained prediction model to select one of a plurality of ventricular assist devices (VADs) for a patient,
wherein the prediction model is trained with training data from a data repository, wherein the training data comprises patient data and operational parameters of the VADs, and wherein the data repository comprises data from treatment of acute myocardial infarction (AMI) patients, high-risk percutaneous coronary interventions (PCI) patients, or patients in cardiogenic shock, and
wherein the trained prediction model is configured to use a tree-based machine learning algorithm to predict a plurality of survival rates, wherein each one of the predicted survival rates is associated with using a different one of the VADs to treat the patient, and wherein a hyper-parameter of the tree-based machine learning algorithm is selected to improve both a cross-validation score and a training score.