Near-Exact Recovery for Tomographic Inverse Problems via Deep Learning
This work addresses the problem of achieving high-accuracy reconstructions in tomographic inverse problems for medical imaging and scientific applications, representing a significant advance by providing the first positive evidence for near-exact recovery with deep learning.
The paper tackles the fundamental question of whether deep learning can solve noise-free inverse problems with near-perfect accuracy, demonstrating for a computed tomography (CT) setup that an iterative end-to-end network achieves reconstructions close to numerical precision, comparable to classical compressed sensing strategies, as evidenced by winning the AAPM DL-Sparse-View CT Challenge and state-of-the-art performance on the LoDoPaB CT dataset.
This work is concerned with the following fundamental question in scientific machine learning: Can deep-learning-based methods solve noise-free inverse problems to near-perfect accuracy? Positive evidence is provided for the first time, focusing on a prototypical computed tomography (CT) setup. We demonstrate that an iterative end-to-end network scheme enables reconstructions close to numerical precision, comparable to classical compressed sensing strategies. Our results build on our winning submission to the recent AAPM DL-Sparse-View CT Challenge. Its goal was to identify the state-of-the-art in solving the sparse-view CT inverse problem with data-driven techniques. A specific difficulty of the challenge setup was that the precise forward model remained unknown to the participants. Therefore, a key feature of our approach was to initially estimate the unknown fanbeam geometry in a data-driven calibration step. Apart from an in-depth analysis of our methodology, we also demonstrate its state-of-the-art performance on the open-access real-world dataset LoDoPaB CT.