Machine-learning approach to holographic particle characterization

Patent metadata

Field	Value
Publication number	US-2017241891-A1
Application number	US-201515518739-A
Country	US
Kind code	A1
Filing date	Oct 12, 2015
Priority date	Oct 13, 2014
Publication date	Aug 24, 2017
Grant date	—

Abstract

Holograms of colloidal dispersions encode comprehensive information about individual particles' three-dimensional positions, sizes and optical properties. Extracting that information typically is computation-ally intensive, and thus slow. Machine-learning techniques based on support vector machines (SVMs) can analyze holographic video microscopy data in real time on low-power computers. The resulting stream of precise particle-resolved tracking and characterization data provides unparalleled insights into the composition and dynamics of colloidal dispersions and enables applications ranging from basic research to process control and quality assurance.

First claim

What is claimed:

1 . A method identifying a particle of interest comprising:
inputting a collimated laser beam; splitting the beam into a diffracted beam and an undiffracted beam; interacting the diffracted beam with the particle of interest to generate a scattered beam; propagating the scattered beam to the focal plane of a microscope; interfering the propagated scattered beam with an undiffracted portion of the original beam recording a hologram characteristic of the scattering beam; obtain an estimated two-dimensional initial position of the particle of interest based upon a center of rotational symmetry in the recorded hologram; obtaining a radial profile of the hologram of the particle of interest using a machine learning algorithm to analyze the radial profile of the hologram of the particle of interest to obtain information about the particle of interest.

2 . The method of claim 1 , wherein the estimated initial position is determined from averaging around a center of rotational symmetry with single-pixel resolution.

3 . The method of claim 1 , wherein obtaining the radial profile of the hologram comprises averaging the recorded hologram of the particle of interest over angles in the focal plane around the estimated initial position to obtain a radial profile of the hologram of the particle of interest.

4 . The method of claim 1 , further comprising applying a machine learning algorithm to analyze the radial profile of the hologram of the particle of interest to obtain information about the particle of interest.

5 . The method of claim 4 where the information is one of the axial position of the particle (z p ), the radius of the particle (a p ,), and the refractive index of the particle (n p ).

6 . The method of claim 5 , wherein each of the radius (a p ,), the refractive index (n p ), and the axial position relative to the focal plane of the microscope (z p ) are separately estimated by a different machine learning algorithm by comparing b(r) with sets of simulated data.

7 . The method of claim 4 where the machine learning algorithm is a support vector machine.

8 . The method of claim 6 where the simulated data is computed for values of the axial position of the particle, the radius of the particle and the refractive index of the particle that span respective desired range for each.

9 . The method of claim 8 , wherein a desired range for axial position of the particle z p is 13.5 μm≦z p ≦75 μm, a desired range for the radius of the particle a p is 0.4 μm≦a p ≦1.75 μm, and a desired range for the refractive index of the particle n p is 1.4≦n p ≦1.8.

10 . The method of claim 9 , wherein the resolution is 1.35 μm in z p , 0.1 μm in a p and 0.1 in n p .

11 . The method of claim 8 where simulated data is provided to achieve a desired degree of convergence of the machine learning algorithm.

12 . A method of identifying a property of a colloidal particle comprising:
inputting a collimated laser beam; splitting the beam into a diffracted beam and an undiffracted beam; interacting the diffracted beam with the colloidal particle to generate a scattered beam; propagating the scattered beam to the focal plane of a microscope; interfering the propagated scattered beam with an undiffracted portion of the original beam recording a hologram characteristic of the scattering beam; determining an estimated initial position from averaging around a center of rotational symmetry with single-pixel resolution; averaging the recorded hologram of the particle of interest over angles in the focal plane around the estimated initial position to obtain a radial profile of the hologram of the particle of interest; using a machine learning algorithm to analyze the radial profile of the hologram of the colloidal particle to obtain information about the colloidal particle selected from the group consisting of axial position of the particle (z p ), radius of the particle (a p ,), and refractive index of the particle (n p ).

13 . The method of claim 12 , wherein each of the radius (a p ,), the refractive index (n p ), and the axial position relative to the focal plane of the microscope (z p ) are separately estimated by a different machine learning algorithm by comparing b(r) with sets of simulated data.

14 . The method of claim 12 where the machine learning algorithm is a support vector machine.

15 . The method of claim 13 where the simulated data is computed for values of the axial position of the particle, the radius of the particle and the refractive index of the particle that span respective desired range for each.

16 . The method of claim 15 , wherein a desired range for axial position of the particle z p is 13.5 μm≦z p ≦75 μm, a desired range for the radius of the particle a p is 0.4 μm≦a p ≦1.75 μm, and a desired range for the refractive index of the particle n p is 1.4≦n p ≦1.8.

17 . The method of claim 16 , wherein the resolution is 1.35 μm in z p , 0.1 μm in a p and 0.1 in n p .

18 . A computer-implemented machine for identifying a particle of interest comprising, comprising:
a processor; and a tangible computer-readable medium operatively connected to the processor and including computer code configured to:
Input a collimated laser beam;
split the beam into a diffracted beam and an undiffracted beam;
interact the diffracted beam with the particle of interest to generate a scattered beam;
propagate the scattered beam to the focal plane of a microscope;
interfere the propagated scattered beam with an undiffracted portion of the original beam
record a hologram characteristic of the scattering beam;
obtain an estimated two-dimensional initial position of the particle of interest based upon a center of rotational symmetry in the recorded hologram;
obtain a radial profile of the hologram of the particle of interest; and
use a machine learning algorithm to analyze the radial profile of the hologram of the particle of interest to obtain information about the particle of interest.

19 . The computer implemented machine of claim 18 , further comprising applying a machine learning algorithm to analyze the radial profile of the hologram of the particle of interest to obtain information about the particle of interest.

20 . The computer implemented machine of claim 19 where the information is one of the axial position of the particle(z p ), the radius of the particle (a p ,), and the refractive index of the particle (n p ).

Assignees

• Univ New York

Inventors

• Grier David G
• Yevick Aaron
• Hannel Mark

Classifications

• G01N2015/1488

  Methods for deciding · CPC title

• G01N15/1484

  microstructural devices · CPC title

• G01N15/147

  the analysis being performed on a sample stream · CPC title

• G01N15/1429 Primary

  Signal processing · CPC title

• G01N2015/1445

  Three-dimensional imaging, imaging in different image planes, e.g. under different angles or at different depths, e.g. by a relative motion of sample and detector, for instance by tomography · CPC title