Deep Learning Image Reconstruction for CT: Technical Principles and Clinical Prospects
Problem
Standard filtered back projection CT reconstruction suffers from noise and artifacts at lower doses, while model-based iterative reconstruction is slow and yields unnatural image appearance.
Why it matters
Lower-dose CT with high-quality images would reduce patient radiation exposure while keeping diagnostic value.
What the AI did
Deep learning reconstruction uses artificial intelligence to reconstruct high-quality images from lower-dose CT faster than model-based iterative reconstruction.
Results
Qualitative claim only: DLR reconstructs high-quality images from lower-dose CT faster than MBIR; no numbers stated.
Impact
Peer-reviewed Radiology review summarising DLR techniques and their clinical prospects over the past 5 years.
Record
| Venue | Radiology |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Lennart R. Koetzier, Domenico Mastrodicasa, Timothy P. Szczykutowicz, Niels R. van der Werf, Adam Wang, Veit Sandfort, Aart J. van der Molen, Dominik Fleischmann et al. |
| Citations | 426 (OpenAlex, at import time) |
| References | 100 |
| Model or method | Deep learning reconstruction (DLR) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1148/radiol.221257 |
| Open access | green |
| Keywords | deep learning, CT reconstruction, low-dose CT, medical imaging, iterative reconstruction |
Importance rating
significant — thousands to millions affected
Lower-dose CT with high-quality images would cut patient radiation exposure while preserving diagnostic value across widespread CT imaging; no scale quantified.