Self-supervised missing-wedge correction enables accurate cellular morphology and volume reconstruction in soft X-ray tomography.

Soft X-ray tomography (SXT) bridges the resolution gap between fluorescence and transmission electron microscopy for 3D cellular imaging. However, the missing-wedge artifact from incomplete tilt-series acquisition causes systematic structural elongation, thereby overestimating organelle volumes and compromising quantitative analysis. To address this, we introduce a self-supervised correction model that learns sine-wave trajectories from SXT sinograms to recover unrecorded missing-angle regions, effectively reducing tomographic distortion. We demonstrate significant quantitative improvement, restoring the spherical morphology of 500-nm beads and lipid droplets. When applied to Plasmodium falciparum hemozoin crystals, a key biomarker in antimalarial drug efficacy studies, our model successfully mitigated volume overestimation, achieving up to a 35% reduction in distorted volume. This precision is important for correctly interpreting the mode of action of antimalarial drugs targeting hemozoin crystals.