Learning-based optical diffraction tomography for label-free 3D cell imaging

Optical diffraction tomography (ODT) is a powerful 3D imaging technique with immense potential in fields like cancer diagnosis and drug treatment. However, traditional ODT systems face limitations like the "missing cone" problem, affecting 3D resolution and cancer classification. To address this, fiber-optic dual-beam technology employs controlled laser beams for stable cell rotation, improving tomographic imaging. This improvement is further enhanced by a novel tomographic workflow that incorporates optical flow and deep learning, replacing manual interventions with automated processes. This novel method is validated by reconstructing 3D images of simulated cell phantoms, HL60 human cancer cells, and artificial cell phantoms. Its adaptability extends to diverse imaging techniques, promising advancements in cell biology, innovative therapeutics, and enhanced early-stage cancer diagnostics.