Fig. 3: Benchmarking of four 3D hematoma segmentation models. | npj Digital Medicine

Fig. 3: Benchmarking of four 3D hematoma segmentation models.

From: An end-to-end deep learning pipeline for hematoma expansion prediction in spontaneous intracerebral hemorrhage based on non-contrast computed tomography

Fig. 3: Benchmarking of four 3D hematoma segmentation models.The alternative text for this image may have been generated using AI.

A Training curves showing pseudo-Dice (dashed lines) and exponential moving average (EMA) pseudo-Dice (solid lines) over 500 epochs for U-Mamba, nnU-Net, nnFormer, and UNETR++. B Training and validation loss curves over epochs for the four architectures. C Comparison of Dice coefficient and Intersection over Union (IoU) values across models on the preliminary benchmarking dataset. D Representative segmentation outputs from each model compared with ground truth annotations.

Back to article page