Research

Research AREAS:

Our research focuses on the development of novel computational methods and analysis tools for application areas, including biomedical imaging, biophotonics, machine vision, RF, healthcare, and so on. Some research areas we focused on are:

AI, Deep Learning & Machine Learning Super-resolution Microscopy

Signal and  Image Processing Computer Vision

Biomedical Imaging Healthcare

Autonomy Wireless and Optical Communication Systems

DsSMLM: Deep-learning Algorithm for Super-resolution Microscopy

 The 2014 Nobel Prize-winning technology, Super-resolution fluorescence microscopy, enables the imaging of cellular structures beyond the optical diffraction limit resolution (~250 nm), closer to the molecular scale. Among various super-resolution microscopy techniques, single-molecule localization microscopy (SMLM) such as Stochastic Optical Reconstruction Microscopy (STORM), Photo-activated Localization Microscopy (PALM), DNA Points Accumulation for imaging in nanoscale topography (DNA-PAINT), provides a spatial resolution of approximately 20 nm. Using spatial along with spectral information, the recently developed spectroscopic single-molecule localization microscopy (sSMLM) / spectroscopic photon localization microscopy (SPLM) technique achieves even higher spatial resolution (sub-10 nm resolution). This technique also offers the unique advantage of simultaneous multicolor super-resolution imaging, visualizing multiple structures within the cell in a single round of acquisition. 

Dr. Gaire is a trained researcher to use STORM, sSMLM, and other super-resolution microscopy techniques for single-color and multicolor imaging. He worked on all aspects of SMLM imaging: instrumentation, sample preparation, image acquisition, algorithm development (classical and machine learning), and image analysis. His lab develops new computational approaches for super-resolution microscopy. Our recent work is a novel deep learning-based algorithm, DsSMLM for post processing of sSMLM imaging data. 

DsSMLM, a post processing algorithm for sSMLM imaging. (Image from publication [1]

DsSMLM performance on simulated sSMLM Data (a) Spatial image with ground truth coordinated (green plus) and DsSMLM predicted coordinates (orange circle); (b) Representative DsSMLM reconstructed spectral PSF compared with simulated and ground truth spectral PSF of AF647 dye. Scale bars=0.5  μm. (c) Spectral plots of images in (b). Noisy is the spectrum of simulated PSF. The emission spectrum from DsSMLM is smooth and very close to the ground-truth image with the matching peak.  (Image from publication [1])

Related Publication:


Computational approaches to accelerate SMLM techniques. 

SMLM technique provides higher spatial resolution but is inherently slow due to the requirements of imaging an extremely large number of frames (>10,000 frames) of biological samples to generate a high-quality super-resolution image. Therefore, accelerating image acquisition in SMLM has been of perennial interest. We explored fast SMLM imaging (using a reduced number of acquired frames) leveraging computational approaches. We used computational approaches such as deep learning and blind sparse inpainting to accelerate SMLM techniques. Specifically, during his Ph.D. Dr. Gaire worked on developing approaches to accelerate the following imaging techniques:

For more details, see the links below to access the related publications (open-access).

Blind sparse inpainting reconstruction of a 3D Tubulin SMLM image. (a) Low-density; (b) reconstructed; and (c) high-density super-resolution 3D image with color indicating the depth of z. (Image from publication [1])

Deep learning reconstruction of dual-color simultaneously imaged sSMLM image of microtubules and mitochondria. (a) Low-density; (b) reconstructed; and (c) high-density super-resolution two-color image. (Image from publication [2])

Related publications:

Multilevel Orthogonal Coded Modulation:

Orthogonal codes are widely used in CDMA communication. These codes can also be used as error control codes in noisy communication systems. For channel coding using orthogonal codes, these codes are used to encode the information signal. The encoded data are modulated using spectrally efficient modulation techniques. In order to improve the error correction capability and improve the spectral efficiency, the orthogonal codes can be used as multilevel encoding combined with multilevel modulation techniques. The resulting system is spectrally efficient as well as able to correct a large number of errors. The encoder of the MOCM system is shown in the figure.


OptoRadio: a method of wireless communication using orthogonal M-ary PSK (OMPSK)   modulations:

OptoRadio is a laser-based radio communication system using Orthogonal M-ary PSK Modulation. In this scheme, when a block of data needs to be transmitted, the corresponding block of the biorthogonal code is transmitted by means of multi-phase shift keying. At the receiver, two photodiodes are cross-coupled. The effect is that the net output power due to ambient light is close to zero. The laser signal is then transmitted only into one of the receivers. With all other signals being canceled out, the laser signal is an overwhelmingly dominant signal. The general block diagram of OptoRadio is shown in the figure.