arXiv:2606.26991v1 Announce Type: cross Abstract: X-ray computed tomography reconstruction is an ill-posed inverse problem, particularly in low-dose and sparse-angle settings where measurements are noisy and incomplete. While learned reconstruction methods such as the Learned Primal-Dual algorithm achieve strong performance, they typically rely on supervised training with access to ground-truth data, which is often unavailable in practice. In this work, we propose a self-supervised reconstruction method by extending the Noise2Inverse framework to the Learned Primal-Dual algorithm. The resultin
The continuous advancements in AI and the need for more robust, data-efficient reconstruction methods in critical applications like medical imaging drive this innovation.
This development addresses a fundamental limitation in AI for critical applications, enabling high-performance reconstruction without the immense cost and logistical hurdles of acquiring perfect ground-truth data.
The reliance on supervised learning with pristine ground-truth data for complex inverse problems like X-ray CT reconstruction is reduced, potentially democratizing access to performant AI models.
- · Medical imaging companies
- · AI algorithm developers
- · Healthcare providers
- · Low-resource research institutions
- · Companies reliant on large, perfectly curated datasets
- · Traditional supervised learning approaches
Improved and more accessible medical diagnostics, especially in regions with limited resources for data collection.
Accelerated development and deployment of AI in other ill-posed inverse problems beyond medical imaging, such as radar or seismic imaging.
Reduced computational and data acquisition costs for deploying advanced AI, leading to broader adoption and unexpected applications.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG