Adaptive Image Reconstruction for Optoacoustic Tomography: A Partial FPGA Reconfiguration Approach

A Widespread clinical adoption of optoacoustic tomography (OAT) depends on technological advancements in handheld probes that can offer immediate image feedback during the scanning process. Field programmable gate arrays (FPGA) can facilitate the design of lightweight, dynamically configurable handheld probes at an affordable cost and power envelope. Optimal image rendering from different areas of the body may involve adjustment of OAT image characteristics during a handheld scan, which requires the implementation of reconstruction algorithms that can be adjusted in real time. However, FPGA implementations of OAT reconstruction algorithms have been constrained by relatively low computational speed and limited flexibility for parameter adjustment. This work exploits FPGAs' partial reconfiguration (PR) capabilities to enable real-time OAT reconstruction and facilitate dynamic updates to data processing blocks without interrupting data acquisition. Implementing PR on a Xilinx Zynq Ultrascale+ FPGA using high-level synthesis (HLS) achieved a 2.5-fold increase in reconstruction speed over state-of-the-art methods, reaching 234 FPS while preserving comparable power efficiency. The proposed approach switches between reconstruction parameters, with 433 ms latency, allowing for on-the-fly configuration changes that support flexible imaging applications and facilitate the adoption of OAT in the clinical setting.