MOTIVATION: Morphological analysis of dendritic spines is critical to understanding the function and dysfunction of neural circuits. The growing trends of the large-scale electron microscopy (EM) imaging systems and automatic cellular reconstruction provide unprecedented opportunities to investigate the ultrastructure of dendrites. This morphometric analysis of dendritic spines requires accurate compartment segmentation methods as well as meaningful quantification methods. However, most existing methods rely on surface or volumetric information alone, which may not deliver accurate segmentation results. RESULTS: We developed VSOT, a method based on Volume-Surface Optimization, designed for the accurate structural analysis of dendritic reconstruction. VSOT accurately segments dendritic reconstructions into compartments, including spine, spine head, and spine neck, by leveraging advanced optimization techniques that integrate local surface and global volumetric information. Our tests on public datasets of spine segmentation, as well as on a first-of-its-kind dataset of head-neck segmentation that we manually constructed, show that VSOT offers more accurate results than peer methods. When applied to a large EM dataset of different brain layers, VSOT reveals how the structure of dendrites varies across brain areas. Furthermore, we explored the structural relationships between neurons and astrocytes at tripartite synapses. With the newly developed computation methods, neuroscientists can exploit the large-scale volumetric EM data to address various scientific questions and advance the understanding of neural circuits. AVAILABILITY AND IMPLEMENTATION: VSOT is available at https://github.com/yu-lab-vt/VSOT. The data and codes in this study are available at Zenodo (https://doi.org/10.5281/zenodo.15115542).

The success of deep face recognition (FR) systems has raised serious privacy concerns due to their ability to enable unauthorized tracking of users in the digital world. Previous studies proposed introducing imperceptible adversarial noises into face images to deceive those face recognition models, thus achieving the goal of enhancing facial privacy protection. Nevertheless, they heavily rely on user-chosen references to guide the generation of adversarial noises, and cannot simultaneously construct natural and highly transferable adversarial face images in black-box scenarios. In light of this, we present a novel face privacy protection scheme with improved transferability while maintain high visual quality. We propose shaping the entire face space directly instead of exploiting one kind of facial characteristic like makeup information to integrate adversarial noises. To achieve this goal, we first exploit global adversarial latent search to traverse the latent space of the generative model, thereby creating natural adversarial face images with high transferability. We then introduce a key landmark regularization module to preserve the visual identity information. Finally, we investigate the impacts of various kinds of latent spaces and find that F latent space benefits the trade-off between visual naturalness and adversarial transferability. Extensive experiments over two datasets demonstrate that our approach significantly enhances attack transferability while maintaining high visual quality, outperforming state-of-the-art methods by an average 25% improvement in deep FR models and 10% improvement on commercial FR APIs.