Robustness of chosen tuning parameter values w.r.t. cross-sectional geometry.

<p>Reference vessel cross-sectional shapes estimated with <i>n</i><sub>oct</sub> = 30, λ = 20% and <i>σ</i> = 10 μm. Each vessel cross-sectional shape re-estimated after altering either <i>n</i><sub>oct</sub>, λ or <i>σ</i>, while fixing the other two parameters. Relative deviations dev<sub>cross</sub> of the re-estimated vessel cross-sectional shapes from the reference shapes computed with <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1011924#pcbi.1011924.e032" target="_blank">Eq (16)</a> (main text); results for pooled vessel cross-sectional shapes from wild-type and Endoglin-deficient embryos at 48 hpf and 72 hpf. When <i>n</i><sub>oct</sub> = 60, the number of cross-sections was 242 65. In all other computations, the number of cross-sections was 25111. <sup>†</sup>: constraint from <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1011924#pcbi.1011924.e034" target="_blank">Eq (18)</a> (main text) omitted. *: vessel surface not smoothed. When halving or doubling either <i>n</i><sub>oct</sub>, λ or <i>σ</i>, the majority of estimated vessel cross-sections had a relative deviation of less than 5%. Notably, the largest median differences were found when using the mean shape only (λ = 0%), followed by the setting where smoothing was omitted. The largest maximal differences were obtained when not constraining the deviation of the local cross-sectional shape from the mean shape.</p> <p>(PDF)</p>