Rage

In order to build a probabilistic atlas of the thalamic nuclei and surrounding tissue, we used our previously presented atlas construction method (Iglesias et al., 2015a (link)). Compared with its conventional counterpart (Van Leemput, 2009 ), which cannot mix manual delineations at different levels of detail, this method relies on a generative model in which joint segmentations of the thalamic nuclei and the surrounding structures are assumed to exist, but are hidden. Instead, we observe a modified version in which sets of labels have been deterministically merged into coarser labels: in the in vivo scans, all the thalamic nuclei are merged into a generic thalamus label, whereas in the ex vivo images, the extrathalamic labels are fused into a single, generic background. Within this framework, Bayesian inference can be used to derive the most likely atlas of the thalamic nuclei and surrounding (whole) structures that generated the observed coarse labels, effectively combining the segmentations made on the in vivo and ex vivo images.
In this study, we combined our reconstructions of the thalamic nuclei with manual delineations at the whole structure level made on 39 in vivo, T1-weighted scans, acquired at 1 $\times$ 1 $\times$ 1.25 mm resolution (sagittal) on a Siemens 1.5 T plaform with an MP-RAGE sequence. Thirty-six structures, including the left and right whole thalamus, were manually delineated using the protocol described by Caviness et al. (1989) (link). We note that this is the dataset that was used to build the atlas in Freesurfer (Fischl et al., 2002 (link); Fischl, 2012 (link)); further details on the acquisition can be found in the original publication. Both the in vivo and ex vivo datasets were augmented with left-right flips to increase their effective size. The atlas is represented as a tetrahedral mesh, which is locally adaptive to the complexity of the anatomy (Van Leemput, 2009 ). Sample slices of the atlas are displayed in Fig. 7.Fig. 7

Probabilistic atlas, with tetrahedral mesh superimposed. The color of each voxel is a linear combination of the colors in Table 2, weighted by the corresponding label probabilities. For the surrounding structures, we used the standard FreeSurfer color map. (a) Sample coronal slice. (b) Axial slice. (c) Sagittal slice.

Fig. 7

Distribution of continuous variables was examined using histograms, box-whisker plots and Kolmogorov–Smirnov tests. Numerical variables were summarized using descriptive measures using counts and percentages, means and standard deviation for normally distributed continuous variables as well as median and interquartile rage (IQR) for not normally distributed data. For within DensironXTRA analyses paired sample t-test was used. For comparisons between and the comparator gas tamponade group, the independent samples t-test was used for normally distributed continuous data and the Wilcoxon signed-rank test as a non-parametric test equivalent. For categorical data, Chi-squared test and Fisher’s exact test were used. Snellen visual acuity was converted to logarithm of the minimum angle of resolution (logMAR) values. The logMAR values for visual acuity of “counting fingers,” “hand motion,” “light perception” and “no light perception” were assigned 2, 2.3, 2.7 and 3, respectively, based on previously published literature^{28 (link)}. The change in visual acuity from baseline to last follow-up was calculated and compared between the DensironXTRA and comparator gas tamponade group using the Wilcoxon signed-rank test.
A multivariable linear regression was used to determine the association of final visual acuity (continuous) with presence of DensironXTRA while adjusting for confounders. Relevant covariates were identified a priori for inclusion in the model based on clinical relevance and existing literature. The following covariates were included: age, baseline visual acuity, history of previous retinal detachment and extent of retinal detachment. Results were reported as slopes or “parameter estimates” and 95% confidence intervals (95% CI).
For the OCT imaging analysis, affected eyes were compared to contralateral eyes without RRD in both the long and short-term using two-sample t-tests. A separate analysis was performed on affected eyes with OCT scans taken during and after DensironXTRA tamponade. The latest OCT on record with DensironXTRA, was used for the “DensironXTRA in situ” group and was compared to the most recent OCT on record after DensironXTRA. These groups were compared using a paired-sample t-test.
All analyses were performed using SAS software (SAS ONDEMAND FOR ACADEMICS, 3.8 (Enterprise Edition)). A p-value of 0.05 was considered for statistical significance.

For all participants, serum was harvested from blood samples obtained while participants were in a fasted state at the time of diagnosis. Serum samples for the analysis of IGF-1, IGF-1R, AGE, RAGE and sRAGE were collected by the immunology laboratory and stored at -80°C prior to analysis. Enzyme-linked immunosorbent assay (ELISA) kits (AGEs: Shanghai jingkang,JLC6718; RAGE: Shanghai jingkang,JLC 6087;Srage: Shanghai jingkang,JLC5525;IGF-1: Shanghai jingkang,JLC 7140;IGF-1R: Shanghai jingkang,JLC7141) stored in a refrigerated environment were removed and left at room temperature for 30 minutes prior to experimental use according to the instructions, with three measurements taken as an average.