Physically Constrained Signal Reconstruction for Calcium Imaging with Deep-Prior ODEs

by Dr Aleix Boquet I Pujadas (Senior Data Scientist, SDSC hub at PSI)

Wednesday 12 Nov 2025, 11:30 → 13:00 Europe/Zurich

OHSA/B17

Fluorescent sensors are powerful tools for studying biological signaling, designed so their brightness tracks the concentration of chemical messengers inside cells. However, the binding kinetics of these sensors are often overlooked when interpreting the resulting fluorescence data. In this talk, I’ll show how this can distort our view of cellular signals, and how accounting for it allows us to recover their true spatiotemporal dynamics. Our method fits fluorescence data under the constraints of the underlying chemical reactions, using a deep prior for temporal regularization. In calcium imaging experiments, we can reconstruct consistent calcium waveforms across different sensors and even resolve individual calcium events within a single neuron. This approach not only enhances existing sensors but also highlights how incorporating physical constraints can make computational imaging both more accurate and more insightful.

TEAMS link