De-GAN - Dynamic Parameter Tuned GAN for 3D Medical Image Segmentation: A Step Towards Generalisation

Name: De-GAN - Dynamic Parameter Tuned GAN for 3D Medical Image Segmentation: A Step Towards Generalisation
Uploaded: 2026-10-02T12:01:14.877972+00:00
Duration: 371 s
Description: DE-GAN synthesizes slice-adaptive FLAIR images to improve 3D brain tumor segmentation by addressing low contrast and domain shift. It uses a conditional GAN combining dynamic convolution, MixStyle, and coordinate encoding guided by a class-conditional target. This enhancement is best used as an augmentation (D5) to complement original MR modalities for better Dice scores.

DE-GAN synthesizes slice-adaptive FLAIR images to improve 3D brain tumor segmentation by addressing low contrast and domain shift. It uses a conditional GAN combining dynamic convolution, MixStyle, and coordinate encoding guided by a class-conditional target. This enhancement is best used as an augmentation (D5) to complement original MR modalities for better Dice scores.

Open the full episode