Deep learning and compressed sensing for reconstruction of 3D knee MRI

This study explores combining compressed sensing (CS) and artificial intelligence (AI), particularly deep learning (DL), to accelerate three-dimensional (3D) magnetic resonance imaging (MRI) of the knee. Twenty healthy volunteers were examined using a 3-T scanner with a fat-saturated 3D proton density sequence at various acceleration levels. Two reconstruction methods were compared: conventional CS and a new DL-based algorithm (CS-AI). Two blinded readers assessed image quality using seven criteria on a 5-point-Likert-scale, including overall impression, artifacts, posterior cruciate ligament, etc. The results showed that CS-AI images achieved significantly better quality than those reconstructed with CS at the same acceleration factor. Moreover, the DL-based algorithm showed that a tenfold acceleration may be possible without significant loss of quality compared to the reference standard. These findings suggest that DL-based algorithms can effectively accelerate 3D-MRI of the knees while maintaining image quality. However, due to the study’s small sample size and use of only healthy volunteers, there is a need for further research with larger, more diverse populations. Key points: Combination of compressed sensing and DL improved image quality and allows for significant acceleration of 3D knee MRI. DL-based algorithm achieved better subjective image quality than conventional compressed sensing. For 3D knee MRI at 3 T, 54% faster image acquisition may be possible. Article: Reconstruction of 3D knee MRI using deep learning and compressed sensing: a validation study on healthy volunteers Authors: Thomas Dratsch, Charlotte Zäske, Florian Siedek, Philip Rauen, Nils Große Hokamp, Kristina Sonnabend, David Maintz, Grischa Bratke & Andra Iuga