Interrogation trajectory optimisation for Fabry-Perot based photoacoustic tomography

TL;DR

This work addresses slow FP-PAT imaging caused by spatial variations in FP cavity thickness that necessitate per-point spectral tuning. It introduces an interrogation-trajectory optimization framework that combines bias-wavelength binning with a fast, modified traveling salesman approach (2opt*) to minimize tunings and path length, achieving substantial speedups. A quantitative relationship between effective scan frequency $f_{eff}$, effective tunings $n_{eff}$, and laser lag $\tau_{eff}$ is derived and validated, showing that modest sensitivity loss can yield multi-fold speed improvements and about a 7× speedup in practice. The proposed method enables near PRR-limited FP-PAT performance with imperfect cavities and lagged lasers, potentially accelerating time-critical biomedical imaging workflows.

Abstract

Fabry-Pérot based photoacoustic tomography (FP-PAT) is a promising all-optical imaging modality for a wide range of preclinical and clinical applications. However, there exist several challenges in routinely applying FP-PAT in time-critical experiments. Among those, the need for spectral tuning of the laser between each scan position can severely limit the effective imaging speed. Here, we present an interrogation trajectory optimization approach which allows to increase the overall speed in a way that is independent of the type of interrogation laser used as well as the FP quality. Our approach provides a way to tackle speed degradation caused by hardware limitations and simplify the use of FP-PAT systems.

Figures (2)

• Figure 1: A Effect of changes in FP cavity thickness on the spectral position of the transfer function and it's consequences for the optimal interrogation wavelength during raster scanning the FP surface. B Dependence of the effective scan frequency ($f_{eff}$) on the laser scanning lag time ($\tau_{eff}$). C Experimentally determined dependence of the effective sensitivity on the spectral distance of the optimal bias wavelength. D Effective sensitivity dependence on spectral bin width for binned interrogation. E Theoretical dependence on of effective scan frequency ($f_{eff}$) on the laser scanning lag time ($\tau_{eff}$) and the effective number of tunings ($n_{eff}$), see Equation \ref{['eq:f_eff']}. F Comparison between the experimentally measured effective scan frequency ($f_{eff}$) and prediction from Equation \ref{['eq:f_eff']} assuming an experimentally determined $\tau_{eff}=69 \ ms$. G Experimentally quantified tradeoff between $f_{eff}$ and the effective sensitivity $S_{eff}$.
• Figure 2: A Scan trajectory across the FP surface after bias wavelength sorting. B Scan trajectory across the FP surface after TSS optimization. C Comparison of trajectory length between the sorted wavelength trajectory and the ones optimized using 2opt and the modified 2opt*, the raster scan length shows a lower bound estimation. D Comparison between the optimization calculation time between the 2opt and 2opt* algorithms. E Comparison between images taken using a normal raster scan with wavelength tuning between each step ($f_{eff}=14\ Hz$) and an image taken using an 2opt* optimized trajectory ($f_{eff}=93\ Hz$) showing no visible reduction in image quality.