Dual-generative synthesis framework: enhancing polyp segmentation in colonoscopy via mask-conditional GANs - Takeaways - MDSpire

Framework for Dual-Generation Synthesis: Improving Polyp Segmentation in Colonoscopy Using Mask-Conditional GANs

Share

1

The study presents a dual generative synthesis framework to enhance polyp segmentation in colonoscopy images.
2

The framework generates realistic masks of small, flat polyps and uses a mask-conditioned GAN for image synthesis.
3

Results showed a Dice score of 0.8786 and an IoU of 0.7835, outperforming baseline U-Net models.
4

The proposed method addresses the lack of diversity in training datasets for small polyps in colonoscopy.
5

The framework separates structure and appearance generation, improving alignment between masks and images.

Frontiers In Oncology

by Mejdl Safran, Sultanul Arifeen Hamim, M. F. Mridha, Dunren Che, Sultan Alfarhood
July 20, 2026

Frontiers In Endocrinology

Frontiers In Oncology

Frontiers In Oncology

1