MediSinGAN

Rajkumar Vaghashiya, Anagha Zachariah, Ekaterina Nepovinnykh, Phuc Nguyen, Md.Sadek Hossain Asif, Amrit Kumar Jethi, Jonas Adler (Senior Research Scientist, DeepMind)
Last updated on Sep 1, 2020

MediSinGAN

The aim is to use an unconditional generative model, SinGAN, to augment medical image datasets using a single natural image.

Current Progress

  • Implemented SinGAN architecture in JAX for the generation of realistic synthetic medical imaging data using a single training image and achieved a 20% reduction in training time
  • Evaluated the model applicability in MRI cross-modality image-to-image translation, Synthetic brain tumor generation, and Medical image segmentation (Histopathology)
Generative Networks Medical Image Analysis Computer Vision Deep Learning Research
Rajkumar Vaghashiya
Rajkumar Vaghashiya
MS in Computer Science

My research interests include applied machine learning and computer vision.

Related