Validation of a Deep Learning U-Net Algorithm for Multistructure Segmentation of Infrarenal Abdominal Aortic Aneurysms including Lumen, Thrombus, and Calcifications - ScienceDirect

Figure 1. Evolution of pseudo dice metrics and loss during the fine tuning phase. x-axis: number of epochs (i.e., model iterations over the complete training dataset); y-axis: “loss” side: loss values measuring the error or divergence between model predictions and ground truth; “pseudo dice” side: pseudo dice metric indicating model performance in terms of segmentation quality. Metrics: training loss (dash dot line in purple), validation loss (dotted line with circle marker in orange), and pseudo dices for the flow lumen (solid line in red), the ILT (dotted line in brown), and calcifications (dashed line in blue). Figure 2. Representative images of the segmentation of the arterial lumen in red, intraluminal thrombus in brown, and parietal calcifications in light yellow. A Computed tomography angiography multiplanar sections from a patient with an infrarenal abdominal aortic aneurysm annotated with the fully automated segmentation. B, C Frontal and left sagittal views of the 3D aorto-il

Figure 1. Evolution of pseudo dice metrics and loss during the fine tuning phase. x-axis: number of epochs (i.e., model iterations over the complete training dataset); y-axis: “loss” side: loss values measuring the error or divergence between model predictions and ground truth; “pseudo dice” side: pseudo dice metric indicating model performance in terms of segmentation quality. Metrics: training loss (dash dot line in purple), validation loss (dotted line with circle marker in orange), and pseudo dices for the flow lumen (solid line in red), the ILT (dotted line in brown), and calcifications (