It’s the great honor that our team’s work in medical image segmentation for Organ at Risk (OAR) for NPC (鼻咽癌) was publicly recognized in the center stage of crowded NVIDIA Taiwan GTC 2018 and presented by CEO/Co-founder Jensen Huang.
The work is reported in Chinese in the NVIDIA Technical Blog — “強大的 GPU 運算效能 大幅提升微小的危急器官標記精準率”
We are new to medical image segmentation. However, we did find it hugely important to remove the massive data burdens for the oncologists. Thanks to the luxury resources from NVIDIA and hearty supports from our long-term partner Microsoft Research Asia (MSRA), we are able to contribute for the health of many local friends.
It’s very challenging for segmentation for NPC due to very small organs in the area. For improving the challenging segmentation problem, we aimed to leverage the complementary information between CT and MR images and solved the alignment issue for the two modalities by a GAN-based model. Happy to see the significant improvement from the pilot experiment.