AUTHOR=Winter Patrick , Berhane Haben , Moore Jackson E. , Aristova Maria , Reichl Teresa , Wollenberg Julian , Richter Adam , Jarvis Kelly B. , Patel Abhinav , Caprio Fan Z. , Abdalla Ramez N. , Ansari Sameer A. , Markl Michael , Schnell Susanne 

TITLE=Automated intracranial vessel segmentation of 4D flow MRI data in patients with atherosclerotic stenosis using a convolutional neural network

JOURNAL=Frontiers in Radiology

VOLUME=Volume 4 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/radiology/articles/10.3389/fradi.2024.1385424

DOI=10.3389/fradi.2024.1385424

ISSN=2673-8740

ABSTRACT=4D flow MRI enables quantitative assessment of hemodynamics in patients with intracranial atherosclerotic disease (ICAD). However, quantitative assessments are still challenging due to the time-consuming vessel segmentation. To improve the reproducibility and to accelerate data analysis, we developed a fully automated segmentation for stenosed intracranial vessels using deep learning. 154 dual-VENC 4D flow MRI scans (68 ICAD patients with stenosis, 86 healthy controls) were selected. Manual segmentations were used as ground truth for training. For automated segmentation, deep learning was performed. 20 randomly selected cases (10 controls, 10 patients) were separated and used for testing. Cross-sectional areas and flow parameters were determined in the Circle of Willis (CoW) and the sinuses. Furthermore, the flow conservation error was calculated. Dice scores (DS), Hausdorff distance (HD), average symmetrical surface distance (ASSD), Bland-Altman analyses, and interclass correlations were computed using the manual segmentations from two independent observers as reference. Finally, three stenosis cases were analyzed by comparing the 4D flow segmentations with black blood vessel wall imaging (VWI). Training of the network took approximately 10 hours and the average automated segmentation time was 2.2±1.0sec. No differences in segmentation performance relative to two independent observers were observed. For the controls, mean DS was 0.85±0.03 for the CoW and 0.86±0.06 for the sinuses. Mean HD was 7.2±1.5mm (CoW) and 6.6±3.7mm (sinuses). Mean ASSD was 0.15±0.04mm (CoW) and 0.22±0.17mm (sinuses). For the patients, the mean DS was 0.85±0.04 (CoW) and 0.82±0.07 (sinuses), the HD was 8.4±3.1mm (CoW) and 5.7±1.9mm (sinuses) and the mean ASSD was 0.22±0.10mm (CoW) and 0.22±0.11mm (sinuses). Small bias and limits of agreement were observed in for the flow parameters. The assessment of the cross-sectional lumen areas in stenosed vessels revealed very good agreement (ICC: 0.93) with VWI but an overestimation (bias±LOA: 28.1±13.9%). Deep learning was successfully applied for fully automated segmentation of stenosed intracranial vasculatures using 4D flow MRI. The analysis of segmentation and flow metrics demonstrated very good agreement between the CNN and manual segmentation and good performance in stenosed vessels. To further improve the performance and generalization, more ICAD segmentations will be considered in the future.