Fig. 5 | Scientific Reports

Fig. 5

From: Clinical feasibility of deep learning-driven magnetic resonance angiography collateral map in acute anterior circulation ischemic stroke

Fig. 5

Images of an elderly patient at admission. The diffusion-weighted image (DWI) (A) shows an acute infarction in the left middle cerebral artery (MCA) territory (circle). Dynamic contrast-enhanced magnetic resonance angiography (DCE-MRA) image (B) reveals occlusion of the M2 segment of the left MCA (arrow). The conventional MRA collateral map derived from DCE-MRA (C–G) and deep learning (DL)-driven MRA collateral map (H–L) demonstrate that hypoperfused area exceeds 50% of the left MCA territory in the capillary phase (circles in D and I) compared to the collateral perfusion of the right MCA territory. Lesions with hypoperfused area in the capillary phase decreasing by less than 50% in the early venous phase (circles in F and J) are categorized as collateral perfusion score 2. The baseline DWI lesion corresponds with the hypoperfused lesion in the early venous phase of the collateral maps (circles in E and J). However, due to the brightness and smoothing of the images, the hypoperfused area area in the capillary phase of the DL-driven MRA collateral map may not be as distinct as that in the conventional MRA collateral map, thus leading to potential underestimation of regions with less than 50% hypoperfused area. Additionally, small hypoperfused areas remaining in the ischemic territory due to contrast washout in the opposite normal MCA territory may be overlooked, thus potentially resulting in a more favorable evaluation of collateral perfusion status in the DL-driven collateral map.

Back to article page