Several CT vendors have introduced deep learning algorithms for the reconstruction of CT images. Deep learning reconstruction methods use deep learning neural networks that were trained to turn low quality image data into high quality target data. They represent the next generation of image reconstruction, following iterative reconstruction which was introduced to denoise images and enable image acquisition with reduced doses.
Deep learning reconstruction was reported to improve image quality in abdominal CT imaging and to maintain noise texture, which was a limitation of iterative reconstruction methods1,2. Images shown above show two acquisitions of phantom 53-01 with a CTDIvol of 1.7 mGy. The image on the left was reconstructed using iterative reconstruction, the image on the right using deep learning reconstruction.
References
- Akagi M, et al. Deep learning reconstruction improves image quality of abdominal ultra-high-resolution CT. Eur Radiol 29, 6163-6171 (2019). Link
- Racine D, et al. Task-based characterization of a deep learning image reconstruction and comparison with filtered back-projection and a partial model-based iterative reconstruction in abdominal CT: A phantom study. Phys Med 76, 28-37 (2020). Link
