Table 5 Quantitative assessments (NMSE) of the results in validation set using the benchmark reconstruction algorithms.

From: CMRxRecon: A publicly available k-space dataset and benchmark to advance deep learning for cardiac MRI

Modality

Sequence

Acceleration factor

Reconstruction method

ZF

GRAPPA

SENSE

Cine

Lax

4-fold

0.094 ± 0.033

0.017 ± 0.012

0.029 ± 0.032

8-fold

0.096 ± 0.035

0.072 ± 0.027

0.076 ± 0.028

10-fold

0.098 ± 0.037

0.071 ± 0.026

0.075 ± 0.030

Sax

4-fold

0.081 ± 0.040

0.007 ± 0.006

0.012 ± 0.019

8-fold

0.094 ± 0.045

0.043 ± 0.018

0.049 ± 0.051

10-fold

0.099 ± 0.047

0.054 ± 0.024

0.062 ± 0.076

Mapping

T1 mapping

4-fold

0.113 ± 0.039

0.006 ± 0.004

0.016 ± 0.068

8-fold

0.126 ± 0.043

0.054 ± 0.019

0.075 ± 0.175

10-fold

0.130 ± 0.044

0.087 ± 0.026

0.107 ± 0.212

T2 mapping

4-fold

0.055 ± 0.017

0.004 ± 0.003

0.026 ± 0.108

8-fold

0.062 ± 0.020

0.028 ± 0.009

0.069 ± 0.229

10-fold

0.062 ± 0.020

0.037 ± 0.010

0.085 ± 0.276

  1. NMSE: normalized mean square error; ZF: zero-filling.