on-chip multiple particle velocity and size measurement
TRANSCRIPT
On-chip multiple particle velocity and size measurement using single-shot two-wavelength differential image analysis
Mitsuru Sentoku 1,*, Shuya Sawa 1, and Kenji Yasuda 1,2
1 Department of Physics, School of Advanced Science and Engineering, Waseda University, Tokyo, 169-8555, Japan; 2 Department of Pure and Applied Physics, Graduate School of Advanced Science and Engineering, Waseda University, Tokyo, 169-8555, Japan
* Corresponding author: [email protected]
Graphical Abstract
2
Title: On-chip multiple particle velocity and size measurement using single-shot two-wavelength differential image analysis
4994 µs590 nm
2500 µs530 nmRestoration
L2
590 nm 530 nm
L1
t1 t2
<latexit sha1_base64="5BVKTeJ8WECKpAhOJYtDVX97a3M=">AAACgXicSyrIySwuMTC4ycjEzMLKxs7BycXNw8vHLyAoFFacX1qUnBqanJ+TXxSRlFicmpOZlxpaklmSkxpRUJSamJuUkxqelO0Mkg8vSy0qzszPCympLEiNzU1Mz8tMy0xOLAEKxQvIqJTZxqQVJSZX+8Qb6vrEG9VWlwAZJUCGCle8gLKBngEYKGAyDKEMZQYoCMgXWM4Qw5DCkM+QzFDKkMuQypDHUAJk5zAkMhQDYTSDIYMBQwFQLJahGihWBGRlguVTGWoZuIB6S4GqUoEqEoGi2UAyHciLhormAfkgM4vBupOBtuQAcRFQpwKDqsFVg5UGnw1OGKw2eGnwB6dZ1WAzQG6pBNJJEL2pBfH8XRLB3wnqygXSJQwZCF143VzCkMZgAXZrJtDtBWARkC+SIfrLqqZ/DrYKUq1WM1hk8Bro/oUGNw0OA32QV/YleWlgatBsBlAEGKIHNyYjzEjP0EzPMNBI2cEJGhUcDNIMSgwawPA2Z3Bg8GAIYAgF2tvEsIJhI8MmJmYmTSYDJiOIUiZGqB5hBhTAZA0A8CSSzQ==</latexit>
v = L1�L2t1�t2
Multi-view module
Wavelength
Bright-field
Binarized
AbstractPrecise and quick measurement of samples' flow velocities is essential for cellsorting timing control and reconstruction of acquired image-analyzed data. We havedeveloped a simple technique for single-shot measurement of flow velocities ofparticles simultaneously in a microfluidic pathway. Microparticles were injectedthrough an imaging flow cytometer and two wavelengths of light with differentirradiation times were irradiated to the particles simultaneously. The mixture of twowavelengths transmitted lights was divided into two wavelengths, and the images ofthe same microparticles for each wavelength were acquired in a single capture. Thespeed was calculated from the difference in the particles' elongation in an acquiredimage which arose when applying two wavelengths of light with differentirradiation times. The distribution of polystyrene beads' velocity was parabolic andhighest at the center of the flow channel, consistent with the expected velocitydistribution of the laminar flow. Exploiting the calculated velocity, we restored theaccurate shapes and cross-sectional areas of particles in the images, demonstratingthe capability of this simple method for improvement of imaging flow cytometryand cell sorter for diagnostic screening of circulating tumor cells.
3
Keywords: Imaging flow cytometer; precise velocity measurement; particle shape reconstruction; multi-view imaging
Introduction
4
Please see previously published paper, Odaka et al.,Micromachines, 2019, for more detail
• CTC forming clusters have been observed inside bloodstream.
CTC cluster. Bar, 10 μm
• CTC clusters are 23~50 times more likely to cause metastasise.g., A.Fabisiewicz, Medical Oncology, 2016
Advancement in image recognition and analysis technologies in the field of imaging cell sorting,
Difficulty overcoming the precise measurement ofmicroparticle flow speed for correct target collection
Machine learning based sorting Droplet microfluidics
e.g., M. Sesan, et al.,Sci. Rep., 2020e.g., N. Nitta, et al., Cell., 2018
Introduction
5
Chip design
Image recognition+Applied voltageSorting by shape (area)
Development of a universal detection method of CTC clusters
System design
Fig. 1. Schematic of imaging cell sorter system.
cf. M., Odaka, et al. Micromachines, 2019 cf. H., Kim, et al. PLOS, 2014
Solutionl Development of a velocity measurement technology requiring
only single-capture of the object.l Estimation of the elongation and restoration via image
processing
Objective
Development of a universal detection method of CTC clusters
Task• Movement of particles while the shutter is open• shape is not consistent → not the accurate size measurement
Essential to know the flow velocity
Causes elongation
after t seconds
Difficulty identifying the particles
Results
7
Image acquisition flow cytometer with simultaneous two-wavelength differential imaging
(ii) LED(530 nm)
(iv) Microchip
(v) Objectivelens (x20)
(vi) Mirror
(ix) Controller
(vii) Multi-viewmodule
(vii) CCD camera
(i) Halogen lamp(590 nm)
(iii) Dichroic mirror
Sheath flow
Sampleflow
Sheath
flow
Cell
System set up
Sheath flow to focus the dispersed particles
Microchip design
Fig. 2. Schematic of imaging acquisition flow cytometer
Proof of concept
Principle
8
(b)(a)
2500 µs4994 µsRestorationStationary Moving
Irradiation time: 2500 µs4994 µs 2500 µs4994 µs
(i) (ii) (iii)
530 nm590 nmWavelength: 530 nm590 nm 530 nm590 nm
0
500
1000
1500
0 500 1000 1500
Mea
sure
d ve
loci
ty [1
0-3
m/s
]
Setting velocity [10-3 m/s]
Photograph(480 px × 640 px)
Mirror A
Dichroicmirror A
Dichroicmirror B
CCDelement
Mirror B
2 3⁄ frame size window
590 !" 530 !"
image
Mirror C
Separates the image into two wavelength
12 µm single polystyrene beads Pre-set velocity vs Measured velocity
Linearly correlated
Multi-view unit
Bright-field
Binarized
Stationary Flowing Restoration
Image of a sample flowing inside microchip
Fig. Schematic of the multi-view unit
Fig. 3. Original and binarized images of polystyrene beads
590 nm
530 nm
Results
Results
9
0123456789
0 10 20 30 40 50
n [%
]
X position [μm]
0
2
4
6
8
10
12
14
0 10 20 30 40 50 60 70
n [%
]
X position [μm]
(a) (c)
N = 91
(b)
(f)
N = 142
0
5
10
15
20
25
30
35
40
0 0.5 1 1.5 2 2.5 3
n [%
]
Velocity [10-3 m/s ]
(e)
Data acquisition position 149.3 μm
73.9 μm
0
10
20
30
40
50
60
70
0 0.5 1 1.5 2
n [ %
]
Velocity [10-3 m/s]
(d)
Data acquisition position 2
0
0.5
1
1.5
2
2.5
0 10 20 30 40 50 60 70
Velo
city
[10-3
m/s
]
X position [μm]
(g)
00.20.40.60.8
11.21.41.61.8
0 10 20 30 40 50
Velo
city
[10-3
μm/μ
s]
X position [μm]
Ø Accurate flow velocity measurement is a prerequisite for determining the precise cell sorting timing to shift a target at the sorting point.
Even after the hydrodynamic focusing, the variation in flow velocity distribution remained present.
Data acquisition position
Flow velocity distribution in microfluidic pathway
Number of particles vs X position Velocity of particles vs X position Number of particles vs velocity
Lower
Upper
Focused Focused
Laminar flow like distribution
Fig. 4. Flow velocity distribution of particles for upper and lower stream
Results
10
0
0.5
1
1.5
2
2.5
0 10 20 30 40 50Ve
loci
ty [1
0-3
m/s
]X position [μm]
(a) (b)
1
23
45
1
2
3
4
5
1
23
4
5
Wavelength :
Irradiation time :
590 nm
4994 μs
530 nm
2500 μs
Simultaneous flow velocity measurement of multiple particles
• Two-wavelength images of five particles were obtained • The elongation was different depending on the flow velocities of
each particles. • The velocity was dependent on the location of microchannel.
Bar, 15 µm
Five particles flowing simultaneously Measurements of the five observed particle velocities
Quadratic approximation of laminar flowFig. 5。5. ow velocity measurement of multiple particles and its X positions
Results
11
Irradiation time : 4994 μs 2500 μsWavelength: 590 nm 530 nm 590 nm 590 nm
4994 μs 2500 μs590 nm 530 nm
Conventional restoration method versus the proposed method
Stationary particle
Elongation caused by flow velocity
Conventional method vs this method
Compressed particle image
Properly restored image
Moving particle
• Conventional method compresses the particle to restore the image• The proposed method lifts the elongated particle up according to the
calculated velocity
Fig. 6. Shape comparison between conventional and our restoration method
Results
12
90
110
130
150
170
190
210
0 5 10 15 20 25 30 35 40 45 50
Area
[μm
2]
X position [μm]
90
110
130
150
170
190
210
0 5 10 15 20 25 30 35 40 45 50
Area
[μm
2]
X position [μm]
90
110
130
150
170
190
210
0 5 10 15 20 25 30 35 40 45 50
Area
[μm2 ]
X position [μm]
Area of the restored particles using the method
• The size of the polystyrene beads is more consistent at various X positions of the microchannel
Conventional simple image reduction methodDistribution of the averaged area of the particles at different X positions
Without using any restoration method
Fig. 7. Graph of area of the particles at X positions
Discussion
13
(a) (b)
0
2000
4000
6000
8000
0 0.5 1 1.5 2 2.5
Max
irra
diat
ion
time
[μs]
Velocity [10-3 m/s]
5.50
6.00
6.50
7.00
0 0.5 1 1.5 2 2.5
Max
den
sity
[107
cells
/ m
L]
Velocity [10-3 m/s]
Limitation of the method for imaging flow cytometry measurement
• Resolution and preciseness of the measurement is reliant on the hardware capability
• The difference in irradiation times of the two wavelengths acquisition must be greater than the curve (left graph) for elongation to appear
• If multiple particles appears within 80x80 pixel image, accurate particle recognition cannot be performed Ø must adjust the density and the velocity of the flowing particles
(less than the linear plot)
Fig. 8. Relation between velocity and max irradiation time (left) and max density (right)
Discussion
14
590 nm 590 nm530 nm 530 nm
(a) (b)
Wavelength:Irradiation time: 4994 μs 4994 μs2500 μs 2500 μs
Application to flowing Hela cells
• The area of the captured cell image after binarization is 293.5 µm2
• The area size of the Hela cell is 170.6 µm2 after restoration.
Bar, 15 µm
Velocity was determined to be 1.41×10!" m/s
Fig. 9. Images of the flowing Hela cell before and after restoration
RestorationOriginal
True size acquired
1) Developed a system for measurement of flow velocities of each samples and reconstruction of size information from single image acquisition
2) Contribute to precise target recognition and target collection timing for diagnostic screening, such as circulating tumor.
In detail, please visit our publication of this issue:Sawa, S.; Sentoku, M.; Yasuda, K. On-Chip Multiple Particle Velocity and Size Measurement Using Single-Shot Two-Wavelength Differential Image Analysis. Micromachines 2020, 11, 1011.
Conclusions
Acknowledgments
16
This research was funded by research and development projects of the
Industrial Science and Technology Program, the New Energy and Industrial
Technology Development Organization (NEDO, P08030), JSPS KAKENHI
Grant Number JP17H02757, JST CREST program, and Waseda University
Grant for Special Research Projects(2016S-093, 2017B-205, 2017K-239,
2018K-265, 2018B-186, 2019C-559), the Leading Graduate Program in
Science and Engineering for Waseda University from MEXT, Japan.
We would like to thank A. Hattori, M. Odaka, and all the members ofYasuda laboratory for their kind support and collaborations.