Vgstudio max 2

The thin-sections were produced following standard petrographic methods (Klein and Sander)⁵⁰ and then studied and photographed with a Leica® DM 750 P compound polarizing microscope equipped with a digital Leica® ICC50HD camera. Histological terminology follows Francillon-Vieillot et al.⁵¹ . Some samples were micro-CT-scanned with a v|tome|xs by GE phoenix|x-ray at the Steinmann Institut für Geologie, Mineralogie und Paläontologie (StIPB) in Bonn (Germany). Image visualization was performed using VGStudio MAX 2.0 software (Volume Graphics GmbH) and Adobe Photoshop. Cross-sections of samples were transformed into black (bone) and white (cavities and vascular spaces) images to measure bone compactness with a custom-designed pixel-counting computer program developed by P. Göddertz (StIPB).

The 32-bit XCT datasets were reconstructed with VGSTUDIO MAX (VGStudio MAX 2.0.5, Volume Graphics, Germany) and rigidly registered with a correlative metric in Avizo (Avizo 9.7, ThermoFisher Scientific, US) using the first preloaded image as a reference. Each image was cropped to include only the specimen structures in the field of view (∼6 mm³) and converted to binary images using Otsu's method [62 (link)].
The corrosion rate (CR) was estimated using the binary 3D tomograms and applying equation (1): $CR=\frac{\Delta V}{A\ast \Delta t}$ where A was the initial total surface area of the fibres exposed to corrosion (533 ± 58 mm²) computed from the 3D XCT images before corrosion using the BoneJ [63 (link)] module of Fiji [58 (link)], Δt the corrosion time (i.e. 2, 8 or 14 days) and ΔV the reduction in volume, equal to the difference between the initial pre-corroded volume the remaining volume. Corrosion maps were obtained by subtracting the registered binary image of pre-corroded samples by the corresponding binary image of corroded samples before loading, resulting in the quantification of lost material volume.
Morphometric properties, such as pore size and solid volume fraction, were computed and corresponding 3D maps were produced using the pore network and volume fraction map modules of Avizo.