Program

The MLMIR workshop will take place on the 8th of October from 9:00am to 13:00pm UTC. The timeschedule is planned as follows:


  • 9:00am Keynote: Deep Learning in Fast, Low-cost, Low-dose Image Acquisition by Dinggang Shen:

Abstract: This talk will introduce various deep learning methods we developed for fast MR acquisition, low-dose CT reconstruction, and low-cost and low-dose PET acquisition. The implementation of these techniques in scanners for real clinical applications will be demonstrated. Also, comparisons with state-of-the-art acquisition methods will be discussed in this talk.

  • 9:30am Q&A

  • 9:35am Fast forward presentations on MRI reconstruction:

    • Deep Parallel MRI Reconstruction Network Without Coil Sensitivities

    • Neural Network-based Reconstruction in Compressed Sensing MRI Without Fully-sampled Training Data

    • Deep Recurrent Partial Fourier Reconstruction in Diffusion MRI

    • Model-based Learning for Quantitative Susceptibility Mapping

    • Learning Bloch Simulations for MR Fingerprinting by Invertible Neural Networks

    • Weakly-supervised Learning for Single-step Quantitative Susceptibility Mapping

    • Data-Consistency in Latent Space and Online Update Strategy to Guide GAN for Fast MRI Reconstruction

    • AutoSyncoder : An Adversarial AutoEncoder Framework for Multimodal MRI Synthesis

  • 10:00am Panel Q&A

  • 10:45am Coffee break

  • 11:00am Keynote: Deep Learning meets Modeling: Taking the Best out of Both Worlds by Gitta Kutyniok

Abstract: Pure model-based approaches are today often insufficient for solving complex inverse problems in medical imaging. At the same time, we witness the tremendous success of data-based methodologies, in particular, deep neural networks for such problems. However, pure deep learning approaches often neglect known and valuable information from the modeling world and are not interpretable. In this talk, we will develop a conceptual approach by combining the model-based method of sparse regularization by shearlets with the data-driven method of deep learning. Our solvers are guided by a microlocal analysis viewpoint to pay particular attention to the singularity structures of the data. Finally, focussing on the inverse problem of (limited-angle) computed tomography, we will show that our algorithms significantly outperform previous methodologies, including methods entirely based on deep learning.

  • 11:30 Q&A

  • 11:35 Fast forward presentations on general image reconstruction:

    • 3D FLAT: Feasible Learned Acquisition Trajectories for Accelerated MRI

    • A deep prior approach to magnetic particle imaging

    • End-To-End Convolutional Neural Network for 3D Reconstruction of Knee Bones From Bi-Planar X-Ray Images

    • Cellular/Vascular Reconstruction using a Deep CNN for Semantic Image Preprocessing and Explicit Segmentation

    • Extending LOUPE for K-space Under-sampling Pattern Optimization in Multi-coil MRI

    • Improving PET-CT Image Segmentation via Deep Multi-Modality Data Augmentation

    • Stain Style Transfer of Histopathology Images Via Structure-Preserved Generative Learning

  • 11:50 Panel Q&A

  • 12:45 Closing words