3D Reconstruction of Nanoparticle Distribution in Tumor Spheroids With Volume Electron Microscopy

Spatially resolved characterization of nanomaterial (NM) distribution within cellular ultrastructure is essential for understanding NM fate and activity in biological systems. Volume electron microscopy (vEM) is uniquely positioned to address this challenge, yet fully documented quantitative pipelines that simultaneously segment NMs and cellular structures remain scarce. Here, an end-to-end analytical pipeline is presented based on the example of serial block-face scanning electron microscopy (SBF-SEM) data of tumor spheroids containing nanoparticles (NPs). A hybrid segmentation strategy is adopted, comprising a fine-tuned Cellpose-SAM model for cells and nuclei, and an empirical Bayes approach for AuNPs. The fine-tuned model outperforms both the pre-trained baseline and the benchmark experiments in Amira, and shows good generalization to 2D EM datasets of varying sample types, suggesting potential as a general-purpose segmentation model for electron microscopy. Full 3D reconstruction of NP distributions reveals preferential clustering in the perinuclear region, with a median nucleus-to-NP distance of 2.57 $\mu m$ and NM uptake spanning several orders of magnitude across cells. Furthermore, morphological analysis of segmented cells and nuclei using 3D shape descriptors and local curvature metrics provides quantitative access to features inaccessible from single sections. Together, these results establish a reproducible, open framework for the joint quantitative analysis of NM distribution and cellular morphology in vEM data.