Neighborhood thresholding in mixed model density gating

摘要

The present invention provides automatic gating methods that are useful to gate populations of interest in multidimensional data, wherein the populations of interest are only a subset of the populations identifiable in the data. The populations are modeled as a finite mixture of multivariate probability distributions, preferably normal or t distributions. The distribution parameters that provide a best fit of the model distribution to the data are estimated using an Expectation Maximization (EM) algorithm that further includes a dynamic neighborhood thresholding that enables gating of a subset of the clusters present in the data.

主权项

1. A method for performing computer aided flow cytometry experiments, the method comprising:
performing a flow cytometric experiment on a sample containing particles with a flow cytometer; obtaining, from the flow cytometer, measurements of a set of N cytometric events for the particles in the sample subjected to the flow cytometric experiment; modeling, via a processor, the data using a plurality of p-dimensional parametric distributions, the plurality of p-dimensional parametric distributions including at least a number, G, p-dimensional parametric distributions; providing, via the processor, an initial estimate of the parameters of said G p-dimensional parametric distributions; iteratively, via the processor, estimating updated parameters for each of the p-dimensional parametric distributions, wherein said estimating comprises:
calculating, for each event, a posteriori probabilities that said event is a member of each of said parametric distributions,identifying a subset of said events that are within a neighborhood of at least one of said distributions based on a comparison of an event location to center points of respective distributions, wherein a number of events included in the identified subset of said events is less than a number of events in the set, andcalculating, for each of the parameters for each of the p-dimensional parametric distributions, an updated estimate based on the calculated a posteriori probabilities for each event in the identified subset of said events, wherein said estimating is iterated at least once, and wherein subsequent calculation of the a posteriori probabilities is based on the updated estimate for the parameters; and after said iterative estimating, identifying, via the processor, a gate indicating a population of the particles from each of the p-dimensional parametric distributions using the updated parameter estimates.