Back to results
US 8,788,291 B2Granted
System and method for estimation of missing data in a multivariate longitudinal setup
AssigneeRobert Bosch GmbH
InventorsSoundararajan Srinivasan, Priya Kohli
Life of this patent
Estimating the term…
Abstract
A method for estimating values of missing data in partial sets of medical data includes generating a Gaussian mixture with a time-varying mean and time and lag varying covariances. The method generates the estimate for a missing datum with the Gaussian distribution having a selected mean and covariance corresponding to the time of the missing datum. An estimate of the missing datum is generated with reference to the mean of the Gaussian distribution conditioned on other medical data that are observed at the time of the missing datum.
Claims17
A method for estimating a value of a missing datum in a multivariate time series of data comprising:
- generating a plurality of medical data sets for a patient with a telehealth device, each medical data set being generated at one time period in a plurality of times;
- storing the plurality of medical data sets for the patient in a memory in association with the time at which each medical data set is generated with a controller communicatively coupled to the memory;
- retrieving a plurality of variables in the plurality of medical data sets corresponding to the plurality of times from the memory with the controller;
- detecting a missing datum and a time period of the missing datum corresponding to one variable in the plurality of variables for one of the medical data sets and one time in the plurality of times in the retrieved data with the controller;
- generating data corresponding to a function of a mean of each of the plurality of variables with respect to time with the controller;
- identifying, with the controller, a plurality of random variable vectors with reference to the plurality of times, each random variable vector including a plurality of random variables that correspond to each of the plurality of variables;
- generating data corresponding to a covariance matrix of the plurality of random variable vectors with the controller;
- generating, with the controller, a conditional Gaussian mixture model corresponding to the one time of the missing datum with reference to the polynomial of the mean of each variable at the one time, the covariance matrix, and the data corresponding to at least one variable in the plurality of variables other than the one variable;
- identifying an estimated value for the missing datum with reference to the conditional Gaussian mixture model with the controller;
- and
- storing the estimated value for the missing datum in the memory with the controller to enable processing of the plurality of medical data sets for the patient including the data associated with the plurality of variables corresponding to the plurality of times, the estimated value being stored with reference to the one variable and the one time of the missing datum.
8 more elements
A healthcare analysis system comprising:
- a memory configured to store recorded values of a plurality of variables corresponding to medical data of a patient generated at a plurality of times;
- an output device configured to display the recorded values for a healthcare provider;
- and
- a controller operatively connected to the memory and the output, the controller being configured to:
- detect a missing datum corresponding to one variable in the plurality of variables and one time in the plurality of times in the recorded values in the memory;
- generate data corresponding to a polynomial of a mean of each of the plurality of variables with respect to time;
- identify a plurality of random variable vectors with reference to the plurality of times, each random variable vector including a plurality of random variables that correspond to each of the plurality of variables;
- generate data corresponding to a covariance matrix of the plurality of random variable vectors;
- generate a conditional Gaussian mixture model corresponding to the one time of the missing datum with reference to the polynomial of the mean of each variable at the one time, the covariance matrix, and the data corresponding to at least one variable in the plurality of variables other than the one variable;
- identify an estimated value for the missing datum with reference to the conditional Gaussian mixture model;
- store the estimated value for the missing datum in the memory with the controller to enable processing of the data associated with the plurality of variables corresponding to the plurality of times, the estimated value being stored with reference to the one variable and the one time of the missing datum;
- and
- generate a display of the recorded values of the plurality of variables and the estimated value for the missing datum with the output device.
10 more elements
Dependencies are inferred from the claim text.