Near-Infrared II Photoacoustic Imaging with Physics-Informed Neural Network Reconstruction

We demonstrate a NIR-II photoacoustic imaging system operating in the 1000–1700 nm window with a physics-informed neural network (PINN) for image reconstruction. The PINN incorporates the photoacoustic wave equation as a regularization constraint, enabling robust image formation from sparse transducer arrays with 16× fewer elements than conventional systems. We achieve 85 µm lateral resolution at 3 cm imaging depth in tissue-mimicking phantoms and resolve subsurface vasculature in vivo in a murine tumor model. The approach reduces acquisition time by an order of magnitude while maintaining image quality comparable to fully-sampled reconstruction.