Fig. 6
From: Hippocampus segmentation after brain tumor resection via postoperative region synthesis
Visualization of each comparative method for hippocampus segmentation. Baseline: Only using images from \(\mathbb {D}_{T}\) to train. MUNIT [22], DiscoGAN [23], CycleGAN [21]: Using synthetic images from the model and real images from \(\mathbb {D}_{T}\) as training samples. Ours: Using synthetic images from our proposed VAE-GAN framework and real images from \(\mathbb {D}_{T}\) as training samples