CC BY
5DOI 10.24412/cl-37136-2023-1-90-95
A STUDY ON THE FLOW OF CELLS THROUGH MICROFLUIDIC CHANNEL USING
DIGITAL HOLOGRAPHIC MICROSCOPY
ASWATHY VIJAY1, PAWAN KUMAR1, NIJAS MOHAMED1, AND RENU JOHN1
biomedical Engineering department, Indian Institute of Technology Hyderabad, India
bm18reschn018@iith.acin
ABSTRACT
Micro optofluidic platform is explored to image and track flow of objects through a microfluidic channel. Digital holographic microscopy is integrated with computer vision techniques for this technique. The method is noninvasive and label free. It can be used in biological platforms as most of them are relatively transparent. The surface profile of the objects flowing through a microfluidic channel are reconstructed numerically from the hologram recording. The phase images are tracked using blob analysis and optical flow methods to track the motion of the objects. This technique finds a wide range of applications in lab-on-a-chip platforms.
INTRODUCTION
Digital holographic microscopy (DHM) is a class of quantitative phase imaging techniques. It is used for the label free imaging of weakly scattering objects. It allows the quantitative estimation of parameters like refractive index, thickness etc. It has several applications in microfluidics, which is deals with the handling and manipulation of small volumes of samples in a system of micron sized channels and chambers. It can be used in inline, off axis and on axis geometries [1,2,3]. Computer vision (CV) techniques can be used in combination with DHM for tracking and analyzing the movement and behavior of particles flowing through the channels [4,5,6]. DHM is an ideal candidate to be coupled with a microfluidic device owing to the transparent nature of the device thus generating a opt microfluidic device [7,8,9]. Computer vision techniques are applied to the reconstructed wavefront from the hologram for 3-dimensional cell tracking. This method of optical tracking of cells using holography offers advantages such as non-invasive three-dimensional imaging, Quantitative analysis, and study dynamic processes, and quantitative phase information. It allows for the study of cell migration, interactions, and other complex phenomena with high spatiotemporal resolution, providing insights into cellular dynamics and biological processes. This method incorporates the following steps: A) image pre-processing to improve the signal to noise ratio, image quality and visibility. B) cell tracking from the extracted frames of a sequence of images or video and employing tracking algorithms like centroid tracking, particle filtering etc. C) feature extraction of the tracked cells, such cell position, velocity, shape, or optical properties, to characterize their behavior and properties over time. D)Data analysis and interpretation. The tracked cell data can be analyzed to extract meaningful information about cell dynamics, migration patterns, cell-cell interactions, or other parameters of interest. Statistical analysis, machine learning techniques, or other analytical methods can also be applied for interpretation and extraction of relevant insights. The current study deals with the phase reconstruction from holograms generated from an off-axis hologram, quantitative estimation and dynamic tracking of the particles flowing through the fabricated channel. We have used two computer vision-based methods of cell tracking - blob analysis and optical flow method [10,11,12]. The flow of polystyrene microbeads and human embryonic kidney (HEK) cells in a microfluidic channel are imaged and tracked using holography base 3d particle tracking method incorporating computer vison.
THEORY
An inhouse fabricated 'S' shaped microfluidic channel of depth 50^m and width 300^m is used. It is a PDMS based channel. The set up for holographic microscopy consists of a Mach-Zehnder arrangement (FIG 1). The source is a 532 nm laser (green light) (Verdi V2, power 10 mW, Coherent make) from which light spatially filtered and collimated using an arrangement of spatial filter and beam expander assembly. It is split into two paths via beam splitter. One arm act as the object beam path and the other acts as the reference beam path. The object beam after passing through the sample is collected by a microscope objective (63x, 0.9 NA). The light
from both the beams interferes and the pattern thus created is called a hologram which is recorded using a digital camera sensor (CMOS camera 1920 px x 1200 px. 2.3 MP, 164 fps).
Figure 1: set up for digital holographic microscopy. SF - spatial filter, L M - dichroic mirror, MO -microscope objective, CMOS - sensor
lens,
The reconstructed complex wave front (y (£,, n; z)) from a hologram after propagating a distance z using angular spectrum propagation method can be expressed as follows
^ z) = / / A (fx,fy-, z) exp [i (fxx + fyy)] dfxdfy Angular spectrum at plane z is given as A (fx, fy; z): ^ are coordinates of image plane, fx
(1)
(2) (3)
and fy are the
spatial frequencies. Amplitude contrast image (A(x,y)) and phase 0(x,y) is obtained from the array of complex numbers, T(x, y).
A(m,n) = Re[r(m,n)]2 + Im[r(m,n)]2 Hm,n) = arctan {Re[r(mn)]^
0(m,n) could be wrapped as it corresponds to modulo of 2^ which can be corrected using phase unwrapping algorithms.
The sequence of phase reconstruction should be segmented by thresholding, (contrast-limited adaptive histogram equalization, binarization) and connected component analysis (perimeter overlay). Blob detection involves identifying and characterizing regions or objects of interest, often referred to as blobs or regions of interest (ROIs), within an image. It involves the steps of blob filtering, (where relevant blobs are retained based on their size, shape etc.) and blob characterization (where appropriate features or properties of the particles can be extracted and analyzed). The tracking of particles over time can also be done using optical flow algorithms. It is the estimation of apparent motion of the scene points from an image sequence. In Images we can measure the motion of brightness patterns or optical flow. It may or may not correspond to motion flow. It is difficult to measure the optical flow uniquely at each pixel solely based on the brightness variation of a particular pixel. The length of the vector gives an estimate of the speed of motion and the direction is indicated by the arrow in the vector. A method based on constraint equation can be used to constrain the optical flow at a pixel for constraining the problem and solving for the optical fluid at each pixel using a neighborhood of pixel. It is an under-constraint problem This is done using Lucas - Kanade method. It is based on the local derivatives of the image. To solve for the optical flow it uses the premise that optical flow in a very small neighborhood in the scene is same for all points within that neighborhood around that pixel. Consider an image I(x,y), where (x,y) indicate pixel positions. Consider a smaller motion where pixel displacement be (u,v). The new image be H(x,y). For (k,l) e W: Ix(k.l)u + Iy (k.l)u + It(k.l) =0. If we assume W has a size n x n. In matrix form
UU) W)
№1)
Jx( n,n) /x(n,n).
[/t(U)l
H = W)
which can be expressed in the form of A u = B, which can be solved using least squares using pseudo -inverse method when ATA is be invertible and well-conditioned.
METHODOLOGY
The suspension of microbeads and HEK cells in distilled water was kept in a syringe pump and was passed through the microfluidic channel at flow rate of 1300 nL/min. The hologram is recorded sequentially digitally. The frames are extracted and reconstructed numerically. The particle tracking is performed using the computer vision toolbox of MATLAB R2020a. The phase images are segmented, and objected perimeter is overlayed. Cells are detected as individual blobs based on their properties compared to the surrounding background. It focuses on identifying regions or objects with distinct characteristics, such as intensity, size, and centroid location. We have also employed optical flow method as mentioned above using Lucas - Kanade method to track the particles under observation to describe the image motion.
RESULTS
After numerical reconstruction the channel is amplitude and phase image of the channel is reconstructed. The thickness map is also plotted (Fig 3). The depth of the channel is 50 micron, and the obtained reconstruction agrees the actual value. The hologram vedio is recorded for microbeads and HEK cells flowing through the channel and reconstructed to obtain the phase images. Figure 4 shows the phase reconstruction at random time points of flow. The phase reconstruction of objects (microbeads and HEK cells) during the observation period are also shown.
Figure 2: (a) hologram of the microfluidic channel (b) amplitude image (c) phase image (d)
thickness map
Figure 4: (a) Phase reconstruction ofpolystyrene microbeads (b) HEK cells flowing through the channel
Tracking of the cell using blob analysis method is performed by detection of the phase images and extraction of the features. The phase images are segmented by applying a threshold to binarize the image to separate out the individual blobs. and the blob filtering is performed based on the shape and size of the blobs to mark the cell boundaries. It is followed by the characterization of the individual blobs by extracting the features like area, diameter, centroid, orientation extra for the 8 -connected components in the image. This is done using the computer vision toolbox in MATLAB. In the current work centroid is used as the characterization property. The result from the tracking is depicted in Fig 5.
Figure 5: Tracking HEK cells using blob detection method at time points (a) t=24 ms (b) t= 54 ms . Red ircle indicates the detected centroid of the cell and the green outline is the detected perimeter of the cell. Optical flow method has also been employed for the tracking of the cells through the channels. Optical flow of the moving object is calculated from the solution of the optical flow constraint equation using Lucas-Kanade method. The determined flow vectors are overlayed on the corresponding points on the phase image (Fig 6) for tracking.
(a) (b)
Figure 6: Tracking HEK cells using optical flow method at time points (a) t=24 ms (b) t= 54 ms . Blue arrow indicate the speed and direction of motion of dynamic objects in the image Blob detection primarily focuses on the identification of objects or region of interests based on distinct features that separates them from the surroundings. It can precisely localize particles like cells with from these features. It is based on simple processes like thresholding and connected component analysis methods which makes it faster compared to optical flow method. Each cell is considered as a separate entity in this method and the interactions or association of the objects are disregarded. It is not an effective method when cells are overlapping or closely packed together, and to distinguish between cells having similar characteristics. Optical flow method is used to track the apparent motion of the pixels between consecutive frames by capturing the motion of the objects in the image to generate dense motion vectors. The flow of individual cells can be tracked continuously while maintaining the cell identity. The challenge in this method arises in case of occlusions, appearance changes and complexities in motion patterns. It is a computationally intensive process.
CONCLUSION
A label free 3D tracking methodology for micro-optofluidic platforms has been developed using digital holography and compute vision. It is a fast and effective technique for imaging and tracking non-invasively. It is also possible to extract features of the object from the individual frames of the phase image. Two methods of tracking are applied in this study - blob analysis and optical flow method. It was possible to detect and track the cells under study accurately using both the methods.. The method of tracking of cells in a microfluidic platform depends on the cell characteristics, motion of the fluid and cells, computational efficiency, and accuracy requirements. An integrated approach promises an improved performance and robustness. Future directions in this study will be on tracking specific cells from a population of several objects in lab-on-a-chip applications for cell growth and disease screening
REFERENCES
[1] U. Schnars and W. Juptner. Digital Holography: Digital Hologram Recording, Numerical Reconstruction, and Related Techniques. Springer, 2005.
[2] M.K.Kim, Principles and techniques of digital holographic microscopy, SPIE reviews, 7(1), 018005,(2010).
[3] L. Yaroslavsky, Digital holography and digital image processing: principles, methods, algorithms. Springer Science & Business Media, (2003).
[4] G. Danuser, Computer vision in cell biology. Cell, 147(5), 973-978, (2011).
[5] K. Li, E. D.Miller, M. Chen, T.Kanade, L. E. Weiss, & P. G. Campbell, Computer vision tracking of stemness. In 2008 5th IEEE international symposium on biomedical imaging: from nano to macro (pp. 847850). IEEE. (2008, May).
[6] Galan, E. A., Zhao, H., Wang, X., Dai, Q., Huck, W. T., & Ma, S. (2020). Intelligent microfluidics: The convergence of machine learning and microfluidics in materials science and biomedicine. Matter, 3(6), 18931922.
[7] V. P Pandiyan,.and R. John, Optofluidic bioimaging platform for quantitative phase imaging of lab on a chip devices using digital holographic microscopy. Appl. Opt., 55(3), A54-A59 (2016).
[8] D. Mark, S. Haeberle, G. Roth, F.Von Stetten, and Zengerle, R. Microfluidic lab-on-a-chip platforms: requirements, characteristics, and applications. Microfluidics based microsystems: fundamentals and applications, 305-376, (2010).
[9] S. Kumar, S. Kumar, M. A. Ali, P. Anand, V. V. Agrawal, R. John, and B. D. Malhotra, Microfluidic- integrated biosensors: Prospects for point- of- care diagnostics. Biotechnol. J., 8(11), 1267-1279, (2013).
[10] K. T. M. Han, and B. Uyyanonvara, A survey of blob detection algorithms for biomedical images, In 2016 7th International Conference of Information and Communication Technology for Embedded Systems (IC-ICTES) (pp. 57-60). IEEE, (2016, March).
[11] J. L. Barron, D. J. Fleet, S. S. Beauchemin, and T. A. Burkitt, Performance of optical flow techniques, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR),236-242. Champaign, IL: CVPR, (1992).
[12] I.,Kajo, A. S. Malik, and N. Kamel, Motion estimation of crowd flow using optical flow techniques: A review. In 2015 9th International Conference on Signal Processing and Communication Systems (ICSPCS) (pp. 1-9). IEEE, (2015, December).
