Reconstruction of undersampled radial free‐breathing 3D abdominal MRI using stacked convolutional auto‐encoders

• DOI: 10.1002/mp.12870
• Publication Date: 2018-03-25T03:26:54Z
• Authors: Jun Lv; Kun Chen; Ming Yang; Jue Zhang; Xiaoying Wang
• Keywords: Artifact (error); Streaking; Autoencoder

Purpose Free‐breathing three‐dimensional (3D) abdominal imaging is a challenging task for MRI , as respiratory motion severely degrades image quality. One of the most promising self‐navigation techniques 3D golden‐angle radial stack‐of‐stars ( SOS ) sequence, which has advantages in terms speed, resolution, and allowing free breathing. However, streaking artifacts are still clearly observed reconstructed images when undersampling applied. This work presents novel reconstruction approach based on stacked convolutional auto‐encoder SCAE network to solve this problem. Methods Thirty healthy volunteers participated our experiment. To build dataset, reference artifact‐affected were using 451 spokes first 20, 40, or 90 corresponding acceleration rates 31.4, 15.7, 6.98, respectively. In training step, we trained by feeding it with patches from images. The outputs testing applied map each input patch patch. Result ‐based 6.98 15.7 show nearly similar quality Additionally, calculation time below 1 s. Moreover, proposed preserves important features, such lesions not presented set. Conclusion preliminary results demonstrate feasibility strategy correcting undersampled free‐breathing negligible time.