Speeding up reconstruction of 3D tomograms in holographic flow cytometry via deep learning

Abstract

Tomographic flow cytometry by Digital Holography is an emerging imaging modality capable of collecting multiple views of moving and rotating cells with the aim of recovering their refractive index distribution...

Fourier Imager Network (FIN): A deep neural network for hologram reconstruction with superior external generalization

“…Due to the missing phase information, various computational approaches have been developed to digitally reconstruct holograms 13 – 22 . Recent work has also utilized deep neural networks 23 – 45 to reconstruct the complex sample field from a hologram in a single forward inference step, achieving an image reconstruction quality comparable to iterative hologram reconstruction algorithms that are based on physical wave propagation. Some of the earlier results have also reported simultaneous performance of phase retrieval and autofocusing in a single network architecture, demonstrating holographic imaging over an extended depth-of-field 25 , 34 , 42 .…”

Developing a Reliable Holographic Flow Cyto-Tomography Apparatus by Optimizing the Experimental Layout and Computational Processing

“…In the case of the static illumination beam, the sample rotation approach is commonly assumed. Recently, a completely new step forward has been demonstrated for obtaining phase-contrast tomography images in a flow-cytometry modality by having cells flowing along microfluidic channels [ 22 , 23 , 24 , 25 , 26 ]. The flowing cells can experience self-rotation, thanks to the shear flow or rolling on the channel-side wall [ 24 ], allowing the 3D phase-contrast tomography of every flowing cell within the field of view (FoV).…”

“…This concept simplifies tomographic microscopes to a remarkable degree, as the cell is probed along multiple directions as it passes through the FoV. The optical arrangement assumed in [ 26 ] is based on a Mach–Zehnder interferometer. The main challenge is faces concerns the implementation of a robust strategy for the rotation angle recovery [ 27 ] in order to apply the propagation algorithms for 3D tomographic imaging, which is the slice-by-slice distribution of the RI inside the volume of each flowing cell.…”

“…The main challenge is faces concerns the implementation of a robust strategy for the rotation angle recovery [ 27 ] in order to apply the propagation algorithms for 3D tomographic imaging, which is the slice-by-slice distribution of the RI inside the volume of each flowing cell. Even if the target has been successfully achieved in recent years, in-flow phase-contrast tomography requires a bulky setup, intended for use in optical laboratories with highly skilled personnel; the main hardware equipment is based on an oil-immersion 40x objective with a high numerical aperture (i.e., NA = 1.3) and a 5120 × 5120 pixel camera for large FoV inspection [ 26 ], thus guaranteeing both high-resolution and high-throughput imaging. Moreover, the main disadvantage of Mach–Zehnder based optical systems concerns the independent optical paths of the mutually coherent signal and reference waves that are subsequently mixed to create the interference record.…”

“…Here, we show a comprehensive approach for optimizing various critical issues of a DHT system in combination with a microfluidic cytometer. In order to achieve a more compact and easy-to-use device, the key issues are a long-working-distance objective, and a camera with a relatively small sensor area and a lower recording frame rate than that of the previous bulk configurations [ 26 ]. For the first time, we designed and tested an experimental layout based on a quasi-common-path lateral-shearing architecture, which optimizes the optical configuration for DHT in flow mode.…”