Flat panel detector

Pre-tests were performed with a standard laboratory X-ray source (RX Solutions, 3SR Lab, Grenoble, France) equipped with a conical polychromatic and divergent beam (Hamamatsu L12161-07 source), allowing absorption imaging mode. Corresponding imaging parameters are reported in Tables 1 and 2 (italic lines)^{30 (link),31 (link)}. The scans were obtained with a number n_p of X-ray 2D radiographs onto a 1914 × 1580 pixel² Varian flat panel detector, leading to a voxel size V_vox (varying from 12³ to 45³ μm³). Samples were exposed to a 360° rotation with respect to the X-ray source (step angle 360°/n_p varying from 0.07° to 0.25°), with an exposure time of 125 ms (respectively 400 ms) per radiograph for experiments on the whole laryngeal structures (respectively on dissected vocal tissues). To restrain the noise, an average of 6 radiographs per 2D image was used.

The specimen was mounted onto the sample holder and secured using glue, and scanned at the Ghent University Centre for X-ray Tomography (UGCT), Woodlab-UGent, using a scanner developed at UGCT. The scanner consisted of two X-ray tubes and two X-ray detectors, specifically designed to obtain very high-resolution scans as well as scans of larger objects. Scans were carried out using a microfocus X-ray tube in combination with a Varian flat-panel detector with an exposure time of 1500 ms, a rotation angle of 0.25° resulting in an average scan time of 45–60 min and an approximate voxel pitch of 2.5 μm. Details of the scanner are outlined in Masschaele et al. [63 (link)] and Van den Bulcke et al. [64 (link), 65 ]. Due to large specimen size, two stacked scans were performed. The dataset was reconstructed using the Octopus software package with beam hardening correction. The two reconstructed volumes were then loaded in VGStudio MAX and stitched into a single stack of cross-sections. All resulting image and video analysis was performed using visualisation software myVGL.