Deep learning techniques are always bound with big data and large, sophisticated models. In this paper, we show that this is not necessarily true for the task of end-to-end phase retrieval in off-axis interferometric quantitative phase imaging. For this task, we first introduce a new loss function, called bucket error rate (BER), for addressing the problem of imbalanced data distribution by balancing loss-bias of target and background area adaptively. With BER, we demonstrate that a U-Net model can learn the underneath logic for converting a raw interferogram to a phase map from only one training sample. At last, we present a novel mixed-context network (MCN) which can simultaneously aggregate local- and global-contextual information. Experimental results show that compared to U-Net, the proposed MCN is more accurate, more compact, and can be trained faster.
Real-time quantitative phase imaging is beneficial for observation and analysis of living cells. Despite off-axis interferometry-based quantitative phase microscopy (off-axis QPM) offers single-shot image acquisition, it usually requires a calibration image captured at a blank field of view to correct the aberration and a multi-step processing algorithm to reconstruct a phase map. Therefore, it is challenging to achieve real-time phase imaging. To simplify experimental operations and expedite image processing, we propose a lightweight U-Net based deep neural network for calibration-free and fast phase retrieval in off-axis QPM. Output phase maps of the lightweight U-Net achieve high fidelity with an average Structural SIMilarity (SSIM) index value of 90.2%. Via running this lightweight U-Net model on a laptop connected with a portable QPM system, we demonstrate an ease-of-use and compact QPM method that can be used for real-time imaging of living cells.
Rapid assessment of the viability of E. coli and other bacteria pathogens is important for timely monitoring of water quality. Therefore, we propose a label-free method for assessing the viability of E. coli cells in a fast way by using quantitative phase microscopy (QPM) and machine learning. According to the viability levels, E. coli cell populations were divided into two classes that were treated with 0.9% and 25% sodium chloride (NaCl) suspended in phosphate-buffered saline (PBS) solution, respectively. Their high contrast phase images are acquired by a high sensitivity diffraction phase microscope. To determine the viability class of individual E. coli cells, a residual neural network (ResNet) is developed to extract the rich information contained in the phase images. An average testing accuracy as high as 95.5% has been achieved in predicting the two viability classes.
Observation of living plant cells under conventional brightfield microscopy suffers from low imaging contrast. Therefore, fluorescence labeling or fluorescence tag is typically required, but it may not reflect the bona fide state of cellular events and prevent long-term observation due to photobleaching of the fluorescence signal. Therefore, we propose to use quantitative phase imaging (QPI) for label-free imaging of plant cell structures. Using QPI, we have observed vacuoles and nucleus in tobacco BY-2 cells, Arabidopsis cell suspension culture PSB-D, and pollen tubes.
KEYWORDS: Microscopes, Embedded systems, Imaging systems, Image transmission, Image processing, Control systems, Cameras, Power supplies, Mobile devices, Local area networks
Wireless communication can break the limitations of space and enable data transferring between disconnected equipment. Automatic and remote-controlled experimental equipment is required for factories or when it is inconvenient to work in the lab, e.g., during the pandemic. We implemented and demonstrated a wireless and automated quantitative phase microscope (QPM) system which can be used for observation of samples remotely or in a confined space. Microscopic manipulations, such as sample placement and scanning, can be operated through a robotic arm and motorized stages controlled by an embedded computing board equipped with a Wi-Fi connection.
We propose and demonstrate a new phase retrieval method based on a deep neural network (DNN) structure. By inputting only one sample interferogram, measured from an off-axis holography based quantitative phase microscope (QPM), the DNN can output an accurate quantitative phase image of the sample without using a calibration interferogram, therefore significantly simplifying the measurements procedure. Importantly, our method can eliminate the need of performing phase unwrapping, therefore making it easy to achieve real-time phase retrieval in different program platforms. We used different types of cells as test samples to characterize the performance of our method, and we found that the accuracy of our DNNbased phase retrieval method is similar compared with the standard Fourier transform based phase method, while the background phase noise is reduced. Considering the experimental procedures and image processing steps are significantly simplified, we envision this new phase retrieval method will make QPM more easily accessible in bioimaging and material metrology applications in the future.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.