Search for: All records

Supervised deep-learning models have enabled super-resolution imaging in several microscopic imaging modalities, increasing the spatial lateral bandwidth of the original input images beyond the diffraction limit. Despite their success, their practical application poses several challenges in terms of the amount of training data and its quality, requiring the experimental acquisition of large, paired databases to generate an accurate generalized model whose performance remains invariant to unseen data. Cycle-consistent generative adversarial networks (cycleGANs) are unsupervised models for image-to-image translation tasks that are trained on unpaired datasets. This paper introduces a cycleGAN framework specifically designed to increase the lateral resolution limit in confocal microscopy by training a cycleGAN model using low- and high-resolution unpaired confocal images of human glioblastoma cells. Training and testing performances of the cycleGAN model have been assessed by measuring specific metrics such as background standard deviation, peak-to-noise ratio, and a customized frequency content measure. Our cycleGAN model has been evaluated in terms of image fidelity and resolution improvement using a paired dataset, showing superior performance than other reported methods. This work highlights the efficacy and promise of cycleGAN models in tackling super-resolution microscopic imaging without paired training, paving the path for turning home-built low-resolution microscopic systems into low-cost super-resolution instruments by means of unsupervised deep learning. 

Digital holographic microscopy (DHM) is a cutting-edge interferometric technique to recover the complex wavefield scattered by microscopic samples from digitally recorded intensity patterns. In off-axis DHM, the challenge is digitally generating the reference wavefront replica to compensate for the tilt between the interfering waves. Current methods to estimate the reference wavefront's parameters rely on brute-force grid searches or heuristics algorithms. Whereas brute-forced searches are time-consuming and impractical for video-rate quantitative phase imaging and analysis, applying heuristics methods in holographic videos is limited since the phase background level occasionally changes between frames. A semi-heuristic phase compensation (SHPC) algorithm is proposed to address these challenges to reconstruct phase images with minimum distortion in the full field-of-view (FOV) from holograms recorded by a telecentric off-axis digital holographic microscope. The method is tested with a USAF test target, smearing red blood cells and alive human sperm. The SHPC method provides accurate phase maps as the reference brute-force method but with a 92-fold reduction in processing time. Furthermore, this method was tested for reconstructing experimental holographic videos of dynamic specimens, obtaining stable phase values and minimal differences in the background between frames. This proposed method provides state-of-the-art phase reconstructions with high accuracy and stability in holographic videos, allowing the successful XYZ tracking of single-moving sperm cells. 

Six different methods for phase compensation in Digital Holographic Microscopy are compared using a calibrated test target and aToxocara canislarva sample regarding processing time, measurement accuracy, and usefulness in biological imaging. 

One of the major drawbacks of confocal microscopy is its limited spatial resolution. This work assesses the performance of an unpaired learning-based model to provide confocal images with improved resolution. 

We propose two Match-Zehnder interferometers coupled to create a structured illumination digital holographic microscope with tunable modulation frequency capability, expanding the system's numerical aperture regardless of the microscope objective lens used. 

This outreach undergraduate research project presents a low-cost method to distinguish the quality of different olive oils. The proposed method is based on an indirect measurement of the chlorophyll molecules present when a green laser diode illuminates the oil sample. Oil blends can be classified into five classes (no olive oil, light olive oil, medium olive oil, olive oil, and extra virgin olive oil) by quantifying the ratio of the red channel versus the green channel along the laser illumination path from a color image. After labeling each oil blend, a convolutional neural network has been implemented and trained to automatically classify oil blends from a color image. The trained convolutional neural network has an accuracy of 90% in identifying and categorizing oil blends. This undergraduate research project introduces students to an interdisciplinary application requiring the combination of optical spectroscopy (i.e., multicolor imaging), image processing, and machine learning. In addition, due to the simplicity of the optical apparatus and computational analysis, high school students could implement and validate their own costeffective oil-quality classification device.