• Aucun résultat trouvé

1JH@K?JE 8EIK=E=JEIOIJA,ECEJ=0CH=FDE?E?HI?FA +D=FJAH

N/A
N/A
Protected

Academic year: 2021

Partager "1JH@K?JE 8EIK=E=JEIOIJA,ECEJ=0CH=FDE?E?HI?FA +D=FJAH"

Copied!
10
0
0

Texte intégral

(1)

Visualization system: Digital

Holographic Microscope

Contents

2.1 Introduction . . . 31

2.2 Holography as a 3D visualization technique. . . 32

2.3 Microscope set up presentation . . . 34

2.3.1 Setup . . . 34

2.3.2 Experimental parameters . . . 35

2.4 Image processing . . . 36

2.4.1 Automated detection of particles . . . 37

2.4.2 Automated calculus of Z-position . . . 39

2.5 Conclusion . . . 40

2.1 Introduction

Analysis and characterization of ow contents could be achieved through dierent ways. Due to the strong enzymatic selectivity of biological samples, immunology assays are an ecient way to investigate the contents in dierent cells. Follow-up of sh populations in marine environment is usually performed by acoustic ultrasound technology, that enable long-working distance communications. With the recent technological progress of Micro-Electro-Mechanical Systems (MEMS), the devel-opment of sensors has experienced a boom, enabling a wide range of biological, chemical and physical data measurement in ows. However, despite of the high sensitivity of measurement capabilities of these systems, they can not provide an image of the investigated object.

(2)

32 Chapter 2. Visualization system: Digital Holographic Microscope visualization is used as a process tool for quality control of batch sample for the quantication and characterization of protein aggregation, vaccine growth culture, microencapsulation, dry power suspension. Flow visualization is also used in many industrial applications, to analyze food contents in Food and Beverage industries, to determine the quality of a chemical mixture in Chemical industries or to characterize size and shape of powder particles in Plastics industries. All these applications requires a rapid and automatic detection of species, with the capability of recording images for a further post-processing.

2.2 Holography as a 3D visualization technique

Observation of biological samples is a common issue in many applications relative to marine environment and in the eld of water treatments. In particular, the ability to detect the presence of cysts such as Criptosporidium and Giardia Lamblia in freshwater contributes to prevent people from critical diseases even in developed countries. The main challenge in this eld is to analyze a large enough volume of biological sample to make it representative of the selected environment, while characterizing the species of interest whose size is often many order of magnitude smaller. Obtaining detailed information about the observed species implies high magnications; as a consequence, the eld of view is greatly restrained, the rate of analysis is lowered and the characterization of the samples time-consuming. To tackle this issue, an increase of the ow rate is possible by nding an optimal match between the uidic and the visualization systems. Those latter require often dyes or stains to mark the biological objects and make them stand out with respect to the background. Hence, that conducts to the use of uorescence coupled with confocal microscopy for volumetric analysis [110].

One of the issue in digital holography is the accurate detection with large number of particles in the experimental volume. Indeed, the hologram recorded corresponds to the integration of the whole experimental volume along the optical axis. When the particle density becomes high, the amplitude of every refocused particle can be overlapped by amplitude contributions of out of focus particles. This eect can disturb the correct detection and, furthermore, the determination of the 3D locations. This limitation becomes more important in in-line holography as there is a twin image that is created when the digital holographic refocusing is applied. Therefore, with large number of particles, the twin images amplitude is added that makes the detection more dicult with in-line holography than with o-axis holography. This is why the use of an o-axis digital holographic microscope has been preferred in the present project.

The main limitation of usual optical microscopy is the small depth of eld induced by the high numerical apertures of the microscope lenses combined with high mag-nication ratios. As a result, a mechanical scanning along the optical axis has to be performed in order to investigate the complete 3D volume of the sample. Fig. 2.1

(3)

DHM. In the case of classical microscopy, out of focus objects are blurred and will not be further considered. On the other hand, the recorded hologram presents a pattern of concentric rings around out of focus objects and their associated focus plane can be later computed.

Among the dierent particle imaging techniques, some are of great interest for

Fig. 2.1: Comparison of the respective depths of view of classical microscopy and Digital Holographic microscopy

measuring both ow parameters and microparticles characteristics. Particle Image Velocimetry (PIV) is used to measure 2D ow velocity elds. Coupled with Particle Tracking Velocimetry (PTV) technique it is able to provide velocity ow eld of eddy ow. The main limitation of this system is the small depth of imaging, that could be increased by using additional cameras for a 3D volume imaging.

Confocal scanning microscopy is imaging a volume by performing an optical sectioning of a very small depth of focus step, that gives the vertical resolution of the system. The 3D reconstruction process combines then the step-by-step recorded images. The main advantage of this technique is its great resolution capability, however its use is not compatible with high ow through applications since this method requires the simultaneous acquisition of multiple images, which restrict the eld of applications to objects that slowly vary in time. On the other hand, interferometric digital holographic microscopy [111113] is very suitable for high ow through applications since the whole information of the volume is included in a single hologram. The use of a laser source provides a sucient power to work with very short exposure times allowing experimental ow of very high ow rates.

(4)

34 Chapter 2. Visualization system: Digital Holographic Microscope is to enable 3D volume imaging with the use of a single camera.

DHM has been successfully applied to many applications. As a phase-quantitative imaging system this technique is able to estimate a slight change in the refractive index of a transparent medium. Besides, most of the living cells are transparent or semi-transparent objects and, as a result, they are dicult to observe with traditional microscopy. But since the refractive index of their constituents is dier-ent from the one of the medium, DHM is able to visualize them. This technique has been successfully applied to the observation of biological samples [116119], refractometry [115,120], analysis of living cells [121123], such as performing 3D trajectories analysis of the in vitro migration of cancer cells [124] or human sperm [125]. It was also used for metrology application [126] and velocime-try [127,128] such as displaying Poiseuille ow pattern [129] in owcytometry. Because the complex amplitude signal is accurately determined with a digital holography setup (see paragraph 2.3), DHM allows processing to improve the digital holographic reconstruction [130], to study concentration proles inside conned deformable bodies owing in microchannels [131], to perform 3D pattern recognition [132134], to achieve quantitative phase contrast imaging [135139]. Digital holographic microscopy (DHM) is the visualization technique that has been selected to perform observation and analysis of sample during this thesis. As introduced in Chapter 1, this thesis is part of the HoloFlow project where the visualization and analysis of a wide range of particles, from 1µm to a few hundred of micrometers is required.

2.3 Microscope set up presentation

The 3D position of particles in the ow is calculated through numerical reconstruc-tion of holograms recorded from a Digital Holographic Microscope (DHM). DHM is an o-axis Mach-Zender and its working principle is described on Fig. 2.2(b). To improve the image quality, a spatial partial coherent source is used as illumination.

2.3.1 Setup

A coherent source (mono-mode laser diode, λ = 635nm, power 30mW) is modied into a partial spatial coherent source by focusing the beam close to the plane of a rotating ground glass (RGG). The spatial partial coherence is adjusted by changing the position of the focused spot with respect to the RGG plane [108]. It is demon-strated that spatial partial coherent illumination greatly improves the hologram quality by reducing the coherent noise and by removing the multiple interferences due to reections.

The lens L2 (focal length of 50mm, Melles Griot, USA) collimates the beam, which is split by beam splitter BS2. The transmitted part, the object beam, illuminates the channel test section in transmission. The objective L3 (Magnication ×10,

(5)

USA) perform the image of one plane of the sample on the CCD camera sensor. The reference beam, reected by beam splitter BS2 and by mirror M3, is transmitted by microscope lens L4 (Magnication ×10, NA = 0.3, Leica, Germany), by beam splitter BS2, and by lens L5.

The reference beam interferes with the object beam on the CCD sensor. The

cam-(a) Sample test section in DHM set up (b) Optical Setup of the Digital Holo-graphic Microscope, CCD refers to the camera, L to lens, M to mirror, BS to beam splitter, RGG to rotated ground glass

Fig. 2.2: Set Up Presentation

era (CV-M4+CL, JAI, Japan) has a CCD array of 1280 × 1024 pixels. The camera is adjusted with a 200µs exposure time. The reference beam is slanted with respect to the object beam to implement an o-axis conguration that enables to compute the complex amplitude eld on every recorded hologram [140].

2.3.2 Experimental parameters

Before each set of experiment, the recorded focus plane for the experiments has been set manually to the middle of the channel depth. Several magnication have been used, depending on the foreseen application.

In the case of 3D hydrodynamic focusing of particles (see Chapter 3), the total magnication used was ×10. The corresponding eld of view was (720 × 720µm2)

(6)

36 Chapter 2. Visualization system: Digital Holographic Microscope The focusing of particles under acoustic standing waves (see Chapter 4) was investigated using the total magnication of ×5. The corresponding eld of view was (1460 × 1460µm2) and the reconstruction capability was in a range of -960µm

to +960µm from the recorded plane. In this case, the focusing eciency of acoustic standing waves was under investigation, and the spreading of particle was quite large. As a result, the selected magnication should enable an observation of the complete channel cross-section.

The resolution of each hologram is determined by the numerical aperture of the selected objective magnication. When imaging objects of ow in motion, a blurring eect can be observed if the time exposure of the camera is too large. In the other way around a small time exposure will result into a shortage of light caught by the camera sensor and a dark image recording. In the frame of these experiments, the optimal exposure time was found to be 0.5ms that induces a maximum linear velocity up to few mm.s−1for a maximum blur of few micrometers.

Depending of the channel cross-section, the sample ow-rate and the acquisition frame-rate are set accordingly.

These settings are of major importance for a proper detection of characteristic features of biological species, but in the frame of this thesis, synthetic particles were mostly used as they are easier to handle than biological sample.

2.4 Image processing

In image processing, the main focus is on the complex amplitude u(x, y, z) that corresponds to the spatial distribution of the wave of amplitude a and phase φ.

u(x, y, z) = a(x, y, z) e−jφ(x,y,z) (2.1)

Holography relies on the superposition of two waves that interfere on the camera sensor. The intensity resulting from the sum of complex amplitude of the object wave uo(ao, φo) and reference wave ur(ar, φr)is calculated as:

I =|uo+ ur|2

I = a2o+ a22+ 2aoarcos(φo− φr)

I = Io+ Ir+ 2

IoIrcos ∆φ (2.2)

Where Io and I2 are the intensity of the two waves and ∆φ is the phase dierence

between the two waves.

(7)

Holoow project. The extracted intensity is displayed in Fig. 2.3(b). The phase map is a representation of the optical path length that is integrated along the propagation axis of the incident wave. A compensation process could be performed to remove the background phase and consequently the phase discontinuities [131], the compensated phase information is represented on Fig.2.3(c).

The compensated phase enables an easy detection of this organism and the

(a) Recorded hologram of a Pedias-trum sp.

(b) Extracted intensity image

(c) Compensated phase image (d) Pseudo-3D representation with inverted contrast (the z-axis corre-sponds to the optical thickness)

Fig. 2.3: Extraction of the intensity and the phase information of a Pediastrum sp. Cour-tesy of A. El Mallahi

measurement of its corresponding quantitative phase information, as illustrated in Fig. 2.3(d), where a pseudo-3D representation (with inverted contrast) has been built using this phase information.

2.4.1 Automated detection of particles

(8)

38 Chapter 2. Visualization system: Digital Holographic Microscope in-depth screw of a classical microscope. It counteracts the narrow depth of focus of classical imaging techniques. Particles that are recorded out of focus, can be refocused thanks to an appropriate post processing of holograms. This refocusing step is indispensable to extract the 3D position of the recorded objects.

In the frame of this thesis, the aim of holograms post-processing was to compute the 3D coordinates of each particle suspended in the ow in order to quantify the focusing eciency of each method. The accuracy of this method is based on the observation of a sample large enough to enable a statistical treatment of data analysis. However, the software developed so far for holograms post processing oered only a manual detection of objects. The use of similar synthetic opaque particles was stimulated by the necessity of applying one single criterion for particle detection during the reconstruction process. Then thanks to the high reproducibility of this detection and the accuracy of the reconstruction (around 5µm), an automated detection procedure could have been implemented.

After the phase and intensity extraction process, the rst step is to smooth the background of intensity pictures for an enhanced detection of particles. The background presented on the intensity is not fully uniformed that is induced by several sources of optical defects. When investigated the 3D volume of ow through channels, we tend to increase the depth of reconstruction. But in the same time, the volume of cross-over ow increases and the probability to encounter other particles, microbubbles, and fragments increases as well. Interferences due to these small objects results in a changing background. On the other hand, the use of a RGG to reduce the laser coherence could also lead to background inhomogeneities. Indeed, signal phase is integrated over a time shorter than the period of the rotating disk, due to its relatively low angular velocity compared to the acquisition frame rate. As a result, an average image background is calculated based on the total sequence of amplitude frames. Then this resulting background image is subtracted to each intensity image in order to smooth the picture.

(9)

(a) O-axis Hologram (b) Processed amplitude image: Background removed

(c) Minimum calculated from integrated amplitude modulus criterion, X axis represents the Z-plan scanning of the image, Y axis represents the refocusing criterion (arbitrary unit)

Fig. 2.4: From a recorded hologram to (x,y,z) coordinates extraction

focus plane can be calculated by numerically investigating the experimental volume using a criterion.

2.4.2 Automated calculus of Z-position

(10)

40 Chapter 2. Visualization system: Digital Holographic Microscope the literature. In our lab, the usual criteria used are those developed by [140], where the focus plane determination is based on the method of invariance of both energy and integrated amplitude. Thanks to those invariance properties, two focus criteria have been outlined, respectively for pure amplitude and pure phase objects, based on the score of the integrated amplitude modulus. It has been demonstrated that this score is minimized for pure amplitude objects while it is maximized for pure phase ones.

There is no universal criterion for computing the focus plane of any particles. In the frame of this thesis, we have chosen to work mainly with synthetic particles (opaques for a pure amplitude object and transparent for a pure phase object) in order to t one of the two criteria, which is a prerequisite for an automated detection. During the automated post-processing of holograms, the (x,y,z) coordinates as well as the geometry of the detected particle are written in a separate le for further analysis. As a result, a complete set of 3D coordinates of the detected particles is obtained. Removing the resulting background image from each intensity picture prevents from wrong detection of dark pixel correlation from channel defaults. However, light illumination is not perfectly balanced in the eld of view and results in some mistaken detection. Setting a Z-reconstruction range wider than the eective depth of the channel is an easy way to outline the wrong detection and separate them from the good experimental data.

2.5 Conclusion

Among the dierent visualization techniques, digital holography is able to perform a 3D imaging of an experimental volume with the use of a single frame and single camera. The post-processing of each hologram could lead to the extraction of quan-titative phase information, which is critical for the discrimination of many biological species as required by the Holoow project.

Références

Documents relatifs

Einen andereren Satz, der als w 5 in die Abhandlung ~Definition analytischer Funktionen usw." h~tte aufgenommen werden ki~nnen, pflegte W~:E~STRASS unge- f~hr

Dans un premier temps, le voisinage considéré pour chaque pixel est un carré de côté 3 centré sur le pixel. Les neuf valeurs de ce voisinage sont triées. On appelle valeur

[r]

[r]

[r]

[r]

[r]

Quelle est la probabilité d'obtenir au moins trois fois (( face