Elias Nehme

I am a PhD Candidate at the Electrical Engineering department of the Technion, jointly supervised by Prof. Tomer Michaeli and Prof. Yoav Shechtman. Prior to that, I did my bachelors in Biomedical Engineering also at the Technion.

My research lie at the intersection of computational imaging, computer vision, and machine learning, and their application to the domain of bioimaging/microscopy. During my PhD, I mostly worked on next generation cameras employing Computational Optics and deep neural networks to recover physical properties of the imaged scene. Specifically, I worked on optimizing reconstruction from challenging measurements, on end-to-end learning of optics and image processing for precise depth estimation, and more recently on visualizing and quantifying reconstruction uncertainty in ill-posed inverse problems in imaging. Throughout my studies, I had the privlige of frequently collaborating with Dr. Daniel Freedman from Verily research center in Haifa.

Email  /  CV  /  Google Scholar  /  Linkedin  /  Github  /  Twitter

It is possible to learn tree-based hierarchical representations of reconstruction uncertainty in imaging inverse problems.

Together with Sartorius we have shown that it is possible to make high-throughput microscopes image in 3D much faster.

It is possible to extend DeepSTORM3D to large FOVs by efficiently modelling the spatially-varying point spread function.

It is possible to visualize reconstruction uncertainty in imaging inverse problems by learning informative low dimensional projections of the posterior distribution.

It is possible to learn the input-adaptive top posterior PCs and visualize reconstruction uncertainty in imaging inverse problems.

Multicolor 3D PSF engineering enables measuring distance between 2 fluorescently tagged DNA loci in yeast cells.

Cell-cycle-dependent chromatin dynamics in live cells can be quantified with high spatiotemporal resolution using end-to-end learned optics and image processing.

Efficient numerical combination of a set of differently illuminated images retrieves high-frequency information from a small number of speckle images.

Extracting and analyzing blinking events of fluorophore clusters with neural networks enables quantification of oligomerization level in membrane receptors.

Diffractive Optical Elements fabrication can be made orders of magntiude cheaper by immersing 3D printed designs in a near-index-matched solution.

Given a budget of signal photons, we learn the optimal wavefront codes for recovering the depth of nanometric point sources under challenging conditions.

Self-explanatory collection of Colab notebooks to simplify access and use of deep learning in microscopy.

Leveraging PSF engineering it is possible to profile the topology of dynamic microsurface in 3D without scanning.

We address the challenge of dense 3D localization from a single 2D image by harnessing neural networks for both acquisition design and image processing.

Vectorial phase retreival facilitated by utilizing Wirtinger flow and vectorial optics.

Temporal convolutional networks can be employed to classify many short single-particle trajectories by diffusion type.

Using a physical simulator we learn a neural network that achieves SotA performance in dense 2D localization microscopy.

It is possible to modify objective lenses to perform compact wavefront modulation without a 4f system extension.

The course focuses on deep generative models for images describing practical strategies for learning high-dimensional probability distributions from samples.

The course focuses on fundamental problems in computer vision describing practical solutions including state-of-the-art deep learning approaches.

The course surveys recent computational imaging techniques co-designing optical sensors and data processing algorithms to advance imaging capabilities.

The course focuses on fundamental approaches in biological signal processing including linear filtering, correlation functions, non-stationary signals, and point processes.

Fundamentals of biosignal and bioimage processing delivered to electronics high school teachers.

Reviewing in multiple scientific communities including journals (e.g. Optica, Nature Scientific Reports), and computer vision/machine learning conferences (e.g. CVPR, NeurIPS).

Website credits to Jon Barron.