Fig. 6: Quantitative validation of alignment to reference and inter-section alignment. | Communications Biology

Fig. 6: Quantitative validation of alignment to reference and inter-section alignment.

From: Brainbuilder: a software pipeline for 3D reconstruction of cortical maps from multi-modal 2D data sets

Fig. 6

A Dice scores for two different validations are shown. Reference: Dice score between aligned sections versus corresponding coronal section from reference volume, Inter-section: Average Dice score between aligned section versus neighbors in posterior and anterior direction along coronal axis. Overall Dice score were high 0.89–0.95 but with some sections that appear to be poorly aligned. B Examples of sections that were either very poorly (left) or well (right) aligned. The grey images indicate the non-linearly aligned 2D segmentations and the hot (orange-red) lines are the edges of the target 2D section from the structural reference volume. The poorly aligned sections show that the source of misalignment is frequently 2D sections that are either very small (Slab 1,2,6), poorly segmented (Slab 4), or incomplete sections (Slab 2,3). The sections that have been well aligned consistently demonstrate a Dice score of 0.97, indicating an upper ceiling for BrainBuilder.

Back to article page