MRI is a powerful imaging modality that is especially useful for imaging soft tissue without exposing the subject to dangerous radiation that is common in other modalities. However, acquisition is slow and sensitive to motion, making it difficult to acquire a sufficient amount of data to reconstruct high quality images. In this talk, we will explore the potential of deep learning methods to create high quality images from under-sampled data. Specifically, we will examine the use of data consistency layers that enforce fidelity to the originally acquired data. In addition to the reconstructed images, we will assess pixel-wise confidence intervals generated around each prediction with statistical guarantees for uncertainty quantification.

Place: Hybrid: Math, 402

Zoom: https://arizona.zoom.us/j/83541348598  Password:  BB2022