Optical Projection Tomography (OPT) provides non-invasive 3D mesoscopic imaging in transmission and fluorescence modes, but its use is limited by slow acquisition and computationally expensive reconstruction. We present Tomopari, an open-source, user-friendly OPT software integrating data processing, reconstruction, and 3D visualization within the Napari environment. For the tomographic inverse problem, Tomopari integrates classical and deep learning-based methods, including filtered backprojection (QBI-Radon), compressed sensing (TwIST), deep learning (U-Net), and physics-informed unrolled reconstruction (ToMoDL), with CPU and GPU support. We demonstrate near-real-time 3D reconstructions from up to 20×-accelerated acquisitions.
This study received funds from PT FCT [UID/04326/2025, UID/PRR/04326/2025 and LA/P/0101/2020 (DOI:10.54499/LA/P/0101/2020)], ”la Caixa” Foundation and FCT, I.P. [LCF/PR/HR22/00533]; EU Horizon Europe IMAGINE [GA 101094250], Horizon 2020 MSCA OPTIMAR [GA 867450] and NVIDIA GPU hardware Grant. TMC also received support from Erwin Schrodinger International Institute for Mathematics and Physics (ESI).