SEG.A 2023 Aorta Segmentation Challenge: Ensemble U-Nets Win on CT Angiography

Ensemble deep learning (3D U-Net) won the SEG.A 2023 aortic vessel tree segmentation challenge on CT angiography, with model ensembling proving key for robust performance. A new public multi-institutional benchmark established.

• The winning solution used an ensemble-based strategy, with 3D U-Net architectures being most effective.
• Performance strongly correlated with algorithmic design, notably the use of customized post-processing and training data characteristics.
• The challenge introduced a large, multi-institutional public dataset for aortic vessel tree analysis.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.