Ultrasound Imaging as Encoded Sensing: A Unifying Computational Imaging Framework

Forum Details

Date: June 20, 2025 12-1 PM Eastern time

Speakers: Nick Bottenus, PhD

Ultrasound imaging is uniquely positioned in medical imaging to provide a handheld, real-time look inside the body without ionizing radiation. Images are formed by sending acoustic pulses of high-frequency sound into the body and processing the returned echoes to localize the sources of scattering and reflection. This active sensing is a game of trade-offs in time, space, and image quality played by varying the geometry of acoustic beams sent into the body (e.g., steering beams around like a flashlight versus illuminating the entire structure at once). We introduce a matrix-sampling framework that unifies these approaches and enables new optimizations that break free from our geometric intuitions about sampling. We demonstrate improved image quality and sampling strategies that are possible with this new approach and highlight further opportunities it may unlock.