When
Where
TITLE
Seeing Without Imaging: Physics-Informed Inference from Weak Optical and Thermal Signals
ABSTRACT
Optical and infrared measurements carry information about a source’s location, motion, and physical properties, shaped by emission, propagation, and detector response. Recovering these source properties becomes a challenging inverse problem when signals are weak and multiple source parameters are coupled. At low signal-to-noise ratios, for example, a thermal image may offer too little contrast for reliable tracking. The central example in this talk is the passive detection and trajectory estimation of small or distant thermal sources. I present a framework for estimating trajectories without forming a thermal image, using a distributed network of single-pixel thermal detectors. Each detector records only a time series, yet collectively their measurements encode the source’s strength, location, and motion. The corresponding negative-log-likelihood function may exhibit flat directions and competing minima, producing an optimization landscape resembling those encountered in statistical physics, molecular biophysics, and machine learning. To navigate this landscape, I introduce a new estimation scheme called Hierarchical Superparameter Mapping (HSM), which replaces the difficult joint search with a sequence of simpler inference steps.
Within this framework, a radiometric model supports a generalized likelihood-ratio test (GLRT) for detection and estimation of seven scalar unknowns per source: source strength, position, and velocity vector for a quasilinear trajectory over a short observation window. HSM first extracts amplitude- and motion-related quantities from each detector’s time series and then recovers the physical parameters through smaller geometric and linear inverse problems. In representative simulations, HSM reduced estimation latency by more than an order of magnitude without sacrificing estimation accuracy.
The same physics-informed inference strategy may help recover useful parameters from other weak optical or infrared signals. I will discuss how a GLRT-HSM framework could potentially be adapted for skin cancer detection research, using models of tissue absorption, scattering, and thermal transport to estimate lesion-related parameters and support the noninvasive assessment of suspicious lesions.
BIO
Dr. Asem Hassan earned his M.S. in Physics in 2019 and Ph.D. in Physics in 2022 from Northeastern University, where he worked with Prof. Paul C. Whitford. His doctoral research combined statistical physics, structure-based molecular dynamics, and information-theoretic analysis to characterize the ribosome’s coupled motions and identify experimental strategies for detecting and measuring them. With collaborators, he developed a geometric framework for describing ribosomal ratcheting, swiveling, tilting, and rolling across structures from more than 50 species. This work led to the Ribosome Analysis Database, a comprehensive, searchable resource that is regularly updated and now contains more than 2,000 complete ribosome assemblies and is currently the main database used in the ribosome research field in medical, biophysical and structural biology communities.
During his doctoral work, Dr. Hassan also contributed to the development of SMOG 2 and OpenSMOG, tools that extend structure-based molecular simulations to large molecular assemblies. He performed the first simulations of large-scale conformational motion in a human mitoribosome. Following his Ph.D., he was a Postdoctoral Fellow at the University of Texas at Austin from 2023 to 2024, where he applied methods from statistical physics to molecular-biophysical systems, with a particular focus on RNA condensates.
In 2024, Dr. Hassan joined the University of Arizona’s Wyant College of Optical Sciences as a Research Scientist. His current research applies statistical physics, information theory, and statistical inference to optical sensing and computational photonics. He is developing methods for detecting moving drones and estimating their trajectories using distributed arrays of single-pixel thermal detectors under low-signal-to-noise-ratio conditions. He is also exploring how optical-beam design and statistical inference can improve information recovery in space-to-Earth optical communication after diffraction and atmospheric distortion.