Structure-guided, physics-aware reconstruction expands scan-limited multiphoton imaging

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Multiphoton microscopy is increasingly pushed toward deep, large-scale and fast functional imaging, where acquisition speed has become a bottleneck. Here we present SPARC, a structure-guided, physics-aware reconstruction framework for scan-limited multiphoton imaging. Rather than treating undersampled recovery as generic super-resolution, SPARC formulates it as a constrained reconstruction problem tailored to point-scanning microscopy by incorporating a sample-matched structural reference from the same field of view and the anisotropic acquisition physics of scan-limited imaging. SPARC jointly integrates denoising and deep upsampling, enabling stable recovery from noisy sparse measurements. In simulated calcium imaging with ground truth, SPARC improved spatial reconstruction fidelity and temporal signal recovery relative to existing methods. In vivo, SPARC improved temporal readout in three-photon and mesoscale two-photon calcium imaging, and enabled a more temporally informative 400-Hz voltage readout on a standard resonant-galvo two-photon microscope. These results suggest SPARC expands the functional operating range of existing multiphoton microscopes under scan-limited conditions.
Funding Information
  • Brain Science and Brain-like Intelligence Technology – National Science and Technology Major Project (2025ZD0216500)
  • National Natural Science Foundation of China (32471142)
  • Science and Technology Commission of Shanghai Municipality (20JC1419500)
  • Shanghai Pilot Program for Basic Research – Fudan University (21TQ1400100)
  • Shanghai Pilot Program for Basic Research – Fudan University (22TQ019)