Diatom frustules exhibit highly organised silica architectures with submicron features, which are widely studied. Nowadays, scanning electron microscopy (SEM) is commonly used for studying their morphology. However, SEM provides a two-dimensional projection that distorts the true size of three-dimensional features. In addition, samples for SEM observations are usually prepared by removing organic parts and drying the remaining parts made of bio-silica (so-called frustule). Here, we combine high-resolution SEM and liquid-cell SEM imaging with confocal microscopy to construct a curvature-aware correction model that accounts for the surface area distortion in SEM imaging caused by the in-plane projection of three-dimensional frustule geometry, demonstrated using Diploneis didyma frustules as a case study. The paper describes how automated image analysis enabled the detection of individual pores, the localisation of their centroids, and the calculation of curvature-corrected surface areas. True diameters of the frustules are provided, and drying-induced shrinkage of pores is obtained. Species length-based normalisation of morphometric data reduced variability and allow to infer systematic dependence of pore sizes on position on frustules. The results establish a reliable framework for quantitative characterisation of diatom frustules and can be applied across disciplines requiring accurate morphometrics of biogenic silica structures. LAY DESCRIPTION: Curvature-corrected SEM morphometrics, paired with hydrated/dehydrated imaging, offer a reliable quantitative framework for diatom frustule pore-size assessment. Drying shrinks pores by ∼10%, so dry-sample measurements should be corrected when targeting living/wet morphology. Curvature correction compensates SEM projection distortion, with uncorrected pores underestimated by ∼34% on average, enabling realistic edge-to-centre comparisons. The protocol is broadly transferable beyond Diploneis didyma with minor adjustments and supports applications in taxonomy, biomonitoring, and bioinspired design.
Zglobicka et al. (Tue,) studied this question.