Systems and method for rapid identification and analysis of cells in forensic samples
AssigneeVirginia Commonwealth University
InventorChristopher EHRHARDT
Life of this patent
Estimating the term…
Abstract
High-throughput methods and systems for using morphological and/or autofluorescence signatures of cells to characterize unknown cell/tissue types within a forensic sample are provided. Machine learning algorithms are used to correlate morphological and/or autofluorescence signatures to characteristics such as cell type.
Claims25
A method of characterizing cells from an unknown contributor or contributors in a forensic sample, comprising
- obtaining a plurality of morphological and/or autofluorescence measurements from a plurality of cells in the forensic sample;
- classifying the plurality of cells into three or more groups using two or more binary classifications, each binary classification comprising
- calculating at least two coordinate values for each cell using respective first and second functions that are weighted combinations of the plurality of morphological and/or autofluorescence measurements, and
- sorting each cell into either a first group or a second group based on ratios of multivariate distances between the calculated coordinate values and multivariate centroids of cell groups in a reference dataset, wherein the three or more groups includes the first group and/or second group;
- and
- outputting information for the three or more groups based on the classifying step.
A method of training a computer for analysis of forensic samples, comprising
- obtaining for a plurality of morphological and/or autofluorescence variables a plurality of measurements from a plurality of cells having one or more known characteristics;
- and
- generating two or more functions which are weighted combinations of the plurality of morphological and/or autofluorescence variables such that the variation between user-defined sample groups is maximized and within group variation is minimized, the two or more functions being usable or used to classify or characterize further cells from an unknown contributor.
A device comprising one or more processors and a computer readable storage medium, the computer readable storage medium having instructions executable by the one or more processors for characterizing cells from an unknown contributor or contributors in a forensic sample, said instructions when executed causing the device to perform:
- obtaining a plurality of morphological and/or autofluorescence measurements from a plurality of cells in the forensic sample;
- with the one or more processors, classifying the plurality of cells into three or more groups using two or more binary classifications, each binary classification comprising
- calculating at least two coordinate values for each cell using respective first and second functions that are weighted combinations of the plurality of morphological and/or autofluorescence measurements, and
- sorting each cell into either a first group or a second group based on ratios of multivariate distances between the calculated coordinate values and multivariate centroids of cell groups in a reference dataset, wherein the three or more groups includes the first group and/or second group;
- and
- outputting information for the three or more groups based on the classifying step.
Dependencies are inferred from the claim text.