Figure 3
From: Deep-learning-based extraction of circle of Willis topology with anatomical priors
Results of our path extraction experiments. Top: Qualitative results showing maximum intensity projections of the 3D TOF-MRA around the CoW, with 2D projections of the centerlines and automatically detected bifurcation points for two patients. Manually annotated centerlines, with measured diameters in mm (left). Centerlines from Dijkstra’s algorithm using \(\mathcal {C}_{\text {img}}\), \(\mathcal {C}_{\text {entr}}\) and \(\mathcal {C}_{\text {orient}}\) (right). Bottom: Quantitative results showing Fréchet distances between manual centerlines and lines acquired using \(\mathcal {C}_{\text {img}}\), \(\mathcal {C}_{\text {entr}}\), \(\mathcal {C}_{\text {orient}}\) and \(\sum \mathcal {C}_i\) for all arteries (left) and categorized by artery diameter (right).
