US 11,056,236 B2
Methods for using artificial neural network analysis on flow cytometry data for cancer diagnosis
Amit Kumar, San Jose, CA (US); John Roop, Ben Lomond, CA (US); and Anthony J. Campisi, Setauket, NY (US)
Assigned to ANIXA DIAGNOSTICS CORPORATION, San Jose, CA (US)
Filed by Anixa Diagnostics Corporation, San Jose, CA (US)
Filed on Jan. 26, 2018, as Appl. No. 15/881,558.
Application 15/881,558 is a continuation of application No. 15/445,913, filed on Feb. 28, 2017, granted, now 9,934,364.
Prior Publication US 2018/0247715 A1, Aug. 30, 2018
This patent is subject to a terminal disclaimer.
Int. Cl. G16H 50/20 (2018.01); G06N 3/04 (2006.01); G01N 33/574 (2006.01); G16H 15/00 (2018.01); G16B 25/00 (2019.01); G16H 10/40 (2018.01); G06N 3/08 (2006.01)
CPC G16H 50/20 (2018.01) [G01N 33/57492 (2013.01); G06N 3/0454 (2013.01); G16B 25/00 (2019.02); G16H 10/40 (2018.01); G16H 15/00 (2018.01); G06N 3/084 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A method for generating a classification of a sample, comprising:
(a) receiving a sample at a first site;
(b) obtaining flow cytometry data of a plurality of event features for a plurality of events of interest of the sample with a flow cytometer instrument, wherein the flow cytometry data of the plurality of event features comprises four or more flow cytometer measurement channels;
(c) transmitting the flow cytometry data to a second site; and
(d) receiving the classification of the sample from the second site, wherein the classification is determined by performing, by a computer, analysis of the flow cytometry data using an artificial neural network, the analysis comprising
using the four or more flow cytometer measurement channels to define a feature coordinate space, the feature coordinate space comprising four or more axes, each axis corresponding to a different channel of the four or more flow cytometer measurement channels,
using the flow cytometry data for the plurality of event features for the plurality of events of interest to define locations for the plurality of events of interest in the feature coordinate space to form a distribution in the feature coordinate space indicative of an event population of interest,
applying the artificial neural network to the distribution in the feature coordinate space indicative of the event population of interest, and
identifying characteristic features in the sample indicative of the classification, wherein the classification comprises assigning at least one category to the sample, thereby determining the classification,
wherein the distribution in the feature coordinate space indicative of the event population of interest is formed by:
(a) dividing an axis of the feature coordinate space into a plurality of segments, thereby dividing the feature coordinate space into a plurality of hypervoxels; and
(b) for a hypervoxel of the plurality of hypervoxels, determining a count of a number of events of interest comprising an event feature value that locates the event of interest in the hypervoxel.