Matlab 2020a

We processed spectroscopic data in MatLab 2020a (9.8, MathWorks, Inc.) with an automated custom written post-processing pipeline incorporating Gannet^{28 (link)} (ver 3.1) and MR libs (https://github.com/chenkonturek/MRS_MRI_libs) software parts. We reconstructed the non-water-suppressed frequency-aligned measurement series as described previously^{6 (link),27 (link)} and applied Hankel-Lanczos singular value decomposition (HLSVD) residual water suppression and apodization filter (BW = 1 Hz) before quantification by LCModel^{29 (link)} (ver 6.3–1N). Metabolite concentration values are referenced to the voxel’s internal water concentration taking into consideration a full tissue and relaxation correction according to Gasparovic et al^{30 (link),31 (link)}.
The basis set for spectral fitting included the following metabolites: N-acetyl-aspartate and N-acetyl-aspartyl-glutamate (NAA + NAAG = tNAA), GSH, glutamate and glutamine (Glu + Gln = Glx), glycerophosphochline and phosphocholine (GPC + PCh = tCho), Cr, scyllo-Inositol (sI) and myo-Inositol (mI). Metabolites were included based on the MRS consensus recommendations^{32 (link),33 (link)}. All MR spectroscopy acquisitions were performed by a trained spectroscopist (last author with 8 years of experience). Data analysis were performed by using an automated post-processing pipeline.

Represents the correlation between FC and in‐scanner head motion (mFD) since nonneuronal fluctuations can increase the apparent FC between regions by introducing spurious common variance across time series. Here, we calculated Pearson's correlation coefficient (r) using MATLAB's corr function (MATLAB 2020a, The MathWorks, Inc.) between each pair of ROIs FC and the mFD across patients. Then, we compared the proportion of edges where this QC‐FC correlation was statistically significant, as well as the median absolute QC‐FC correlation after applying each denoising pipeline. Higher QC‐FC (either the proportion of significant correlations or the absolute r‐value) represents the inability of a pipeline to mitigate noise in FC.

MATLAB (MATLAB 2020a; The MathWorks, Inc.) was used for statistical analysis. Normality of the data was tested with the Lilliefors test. For the analysis of topographical distribution, measurements from all nine scans were averaged per patient and are presented as a median and interquartile range. The quartile-based coefficient of variation³⁹ (CV) was calculated (equation 5) where Q₂₅ and Q₇₅ denote the 25th and 75th quantiles of the distribution.
$CV=\left(\frac{Q_{75}-Q_{25}}{Q_{75}+Q_{25}}\right)\times 100.$    Correlations of the choroidal parameters with AEL were performed using Spearman rank correlation.⁴⁰ (link) The coefficient of repeatability (CR) from nonnormal data was derived for each ETDRS region as explained elsewhere.^{41 ,42} (link) In short, the median of each of the three measurements per day and its difference to the overall mean across the 3 days were calculated. Subsequently, the resulting deltas were evaluated in a cumulative distribution function. The 95% limits of the cumulative distribution function were determined and multiplied by $\sqrt{3/2}$ as a correction factor for centered data. The CR was considered half the length of the 95% interval. It is noteworthy that only intersession but not intrasession repeatability was considered in the analysis, to reflect most studies that investigate choroidal metrics in an intersession rather than intrasession (consecutive) manner.

Data analysis was carried out with the help of MATLAB (MATLAB 2020a, The MathWorks, Inc., Natick, MA, USA). Normal distribution was ensured with the Lilliefors test [25 (link)]. A t-test was performed in addition to Bland–Altman analysis [26 (link)] and the calculation of the intraclass correlation coefficient (ICC(2,1)) [27 (link)] in order to investigate the statistical differences and agreement between the OCT- and PRX-based retinal curvatures. Bivariate correlation analysis was performed with Pearson correlation [28 ]. Statistical results were interpreted as significant in the case that p < 0.05.