Dynamic quantitative photoacoustic tomography under sparse sampling via residual-driven Kalman filter assisted reconstruction

Abstract

Mechanically scanned photoacoustic tomography (PAT) provides high detection sensitivity and flexible system configurations, but its application to dynamic quantitative imaging is limited by the trade-off between temporal resolution and spatial sampling density. We propose a coordinated acquisition–estimation–reconstruction (AER) framework for dynamic quantitative PAT under sparse-pattern sampling. Sparse-pattern cyclic (SPC) acquisition collects baseline-referenced differential signals, while residual-driven adaptive Kalman filtering (RDAKF) adapts location-specific state-transition matrices at sampled locations and uses neighborhood-weighted transition information to propagate states at unsampled locations. The recovered spatially complete differential PA waveforms are then used to reconstruct absorption-coefficient perturbations through adjoint-based, PDE-constrained optimization. The framework was evaluated in numerical simulations, dynamic phantom experiments, and ex vivo tissue experiments using a single-element transmission-mode PAT system. Across multiple sparse-pattern configurations in the dynamic phantom experiment, structural similarity index values of 0.81–0.84 and root-mean-square errors of 0.017–0.023 mm⁻¹ were obtained. In the single-location-per-pattern configuration, the mean sparse-pattern acquisition rate was approximately 0.84 Hz. The framework tracked gradual, abrupt, periodic, and motion-associated variations, supporting the feasibility of dynamic quantitative PAT under measurement-limited conditions.