Matlab r2010a

Each surface was displayed on a 24-in. calibrated LED monitor (Viewsonic V3D245) with a linearized gamma. The resolution was set to 1920 × 1080. All stimuli were displayed using Matlab R2010a (The MathWorks Inc., Massachusetts, US) and Psychtoolbox-3 (Brainard, 1997 (link)) that ran on a MAC Pro Quadro-Core Intel Xeon with OSX 10.5.8.

A control experiment was performed after the end of the study, with the aim of estimating the sensitivity of pupil size measurements to manipulations of stimulus strength. The original set-up was inaccessible at the time of testing, and we replicated the conditions of the main experiment as closely as possible in another set-up, using the same eye-tracker (EyeLink 1000 system, SR Research), similar mirror stereoscope and a computer that ensured equal performance. Specifically, stimuli were generated with the PsychoPhysics Toolbox routines (Brainard, 1997 (link)) and MATLAB (MATLAB r2010a, The MathWorks Inc.) housed in a Mac Pro 4.1, and displayed on a 52.5-cm-wide LCD screen with maximum screen luminance of 108 cd/m². Instead of using the maximum and minimum screen output, we reduced luminance levels by about a factor of 10 to allow for modulations of stimulus contrast. We set the background luminance to 15 cd/m² gray and tested six conditions: a no cue and white cued condition in which stimuli where 28 and 2 cd/m² for the white and black disk, respectively, and four conditions where the Michelson contrast of the white disk stimulus was increased by 25%, 50%, 100% and 150% (luminance values: 33, 40, 63, and 108 cd/m²). Each condition was tested in four trials, and all data were collected over a single session.

The coordinates of the particles in the coarse‐grained models were created with Matlab R2010a (Mathworks Inc., Natick, MA, USA) based on the full atomistic geometry described previously.32 Visualization of the simulation is performed with visual molecular dynamics (VMD).41 Stress‐strain curves were computed from the box size and the virial‐stress of the particles. The diameter of the fibril was considered constant and equal to 20 nm. The stress‐strain response was divided into five regions to analyze the mechanical response of the fibril (Fig. 2A). The toughness of the fibril is defined as the area under the stress‐strain curve until ultimate strength (gray area in Fig. 2A).
All mineral densities are expressed as weight percentages of the total model.

Data were processed using Statistical Parametric Mapping software (SPM8) (Wellcome Department of Imaging Neuroscience, London), running in MATLAB R2010a (Mathworks, Sherborn, MA). In SPM8, an image from the control group with no significant atrophy was used as reference to realign images. The images were then segmented into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) using the unified segmentation procedure in SPM8 [38 (link)]. The Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) algorithm [39 (link)] was then used to spatially normalize the segmented images. These fully normalized images were resliced with trilinear interpolation to a final voxel size of 1.5×1.5×1.5 mm³. An additional “modulation” step was applied, consisting of multiplying each spatially normalized GM image by its relative volume before and after normalization; this ensured that the total amount of GM in each voxel was preserved [40 (link)]. We used the ‘normalize_DARTEL’ matlab script [41 (link),42 (link)] to complete the affine transformation of segmented brain tissues into the MNI space. Finally, the resulting GM images were smoothed using an 8-mm isotropic kernel at full-width half-maximum to ensure a normal distribution of the data, as required by subsequent statistical parametric tests.

The SPM8 (Statistical Parametric Mapping; http://www.fil.ion.ucl.ac.uk/spm/; University College London, London, UK) software package in Matlab R2010a (Mathworks, Natick, MA, USA) was used for voxelwise group comparisons. The FA images after normalization and smoothing were analyzed using SPM8 within the framework of a General Linear Model. ANCOVA with age and sex as covariates was performed to check the FA differences between the groups. The FA threshold of the mean WM was set at 0.2 to exclude “contaminated voxels” with composition of gray matter or cerebrospinal fluid. The statistical threshold was set at an uncorrected p < 0.005, with a cluster of >200 contiguous voxels. The extended cluster size was arbitrary and was used to putatively detect significant differences between groups with a cluster size <200 voxels, which might not represent reliable findings [25 (link)]. Additionally, multiple comparisons of FA were corrected by AlphaSim (FWE) to control for type I errors [22 (link)–24 (link), 26 (link)] with the threshold at p < 0.05 and cluster size >1871.
The FSL atlas tool (https://www.fmrib.ox.ac.uk/fsl/fslwiki/Atlases) was facilitated for determining the most probable fiber tracts and anatomic location of each significant cluster.