Nerve Image Segmentation
Nerve segmentation in medical imaging is a complex process to conduct manually due to the different nerve structures. The naked human eye is not strong enough to completely differentiate the nerves during a surgery which is a crucial requirement. Hence, computing solutions are used with expert monitoring to help in this matter. This research work will be useful to eliminate the existing uncertainty levels in the current research results in the literature. They are unable to apply in clinical usage due to lack of accuracy yet. The elimination of even a tiny amount of uncertainty is important in the health science domain.
Deep probabilistic programming concatenates the strengths in deep learning to the context of probabilistic modelling for efficient and flexible computations in practice. Being an evolving area, there exist only a few expressive programming languages for uncertainty management. The research addresses ultrasound nerve segmentation based biomedical image analysis application using the probabilistic programming language Edward with the U-Net model and Generative Adversarial Networks under different optimizers.