Artificial Intelligence in Digital Holographic Imaging

25Conclusion and Future Work

In this book, we introduced applications of deep learning in digital holographic cell imaging and DHM‐based phenotypic analysis methods. Basically, DHM can provide quantitative phase images if the exact distance between the sensor plane and the reconstruction plane is correctly provided. This process requires an iterative diffraction calculation, which is computationally time consuming. We presented a deep‐learning convolutional neural network with a regression layer as the top layer to estimate the best reconstruction distance. Experimental results obtained using microsphere beads and RBCs showed that the proposed method could accurately predict the propagation distance from a filtered hologram. Additionally, our approach could be used at the single‐cell level for cell‐to‐cell depth measurement and cell adherent studies.

Conventional, numerical phase unwrapping techniques can connect wrapped phases to recover the optical path length of a target object. However, these methods are computationally time consuming. We introduced a new deep‐learning model that can automatically reconstruct unwrapped, focused phase images by combining digital holography and a generative adversarial network (GAN) for image‐to‐image translation. Compared with numerical phase unwrapping methods, the proposed GAN model overcomes the difficulty of accurate phase unwrapping due to abrupt phase changes. It can perform phase unwrapping at twice the rate of numerical methods. ...