Skip to main content

Table 10 Optimal alpha values for random phase encoding (SSIM optimization)

From: Optimal compressed sensing reconstructions of fMRI using 2D deterministic and stochastic sampling geometries

Sample rate

Image1

Image2

Image3

Image4

Image5

Image6

Image7

Image8

Image9

Image 10

Image11

Image 12

Median

62.50

3.432*1E-04

5.724*1E-04

1.449*1E-04

2.051*1E-04

−3.137*1E-05

7.519*1E-04

4.396*1E-04

4.749*1E-04

−5.880*1E-06

2.859*1E-04

7.813*1E-06

6.028*1E-04

3.15*1E-04

48.50

−1.633*1E-06

4.126*1E-04

−2.972*1E-05

−6.086*1E-05

−1.590*1E-05

−1.863*1E-04

−1.957*1E-04

4.404*1E-04

−5.236*1E-05

−2.679*1E-04

7.813*1E-06

5.633*1E-04

−2.28*1E-05

40.60

9.952*1E-04

4.536*1E-04

2.723*1E-04

2.109*1E-04

2.979*1E-05

2.140*1E-04

1.763*1E-04

4.544*1E-04

5.251*1E-04

4.368*1E-04

−6.852*1E-05

8.507*1E-04

3.55*1E-04

32.80

5.493*1E-07

−4.938*1E-04

−6.073*1E-06

−1.334*1E-04

−5.382*1E-05

5.497*1E-04

2.771*1E-05

−7.382*1E-05

−4.323*1E-05

−4.177*1E-04

−3.918*1E-05

−5.866*1E-04

−4.85*1E-05

28.10

−3.632*1E-05

2.559*1E-04

−3.033*1E-05

−1.074*1E-05

−3.271*1E-05

4.256*1E-04

7.668*1E-04

1.416*1E-04

−7.233*1E-06

3.029*1E-06

−1.099*1E-05

6.348*1E-06

−2.10*1E-06

26.60

−3.906*1E-06

1.526*1E-07

−3.009*1E-05

−3.311*1E-05

−4.181*1E-05

2.896*1E-05

2.068*1E-03

−4.675*1E-05

−5.447*1E-06

−3.906*1E-06

5.127*1E-06

−7.563*1E-05

−4.68*1E-06

25.00

−1.844*1E-04

−4.150*1E-06

−1.160*1E-04

−4.858*1E-05

8.374*1E-05

1.191*1E-03

6.500*1E-04

1.036*1E-04

9.277*1E-06

2.930*1E-06

−7.629*1E-08

−1.639*1E-05

1.43*1E-06

21.90

9.766*1E-07

7.047*1E-04

1.367*1E-05

1.121*1E-03

3.906*1E-06

2.021*1E-03

6.571*1E-03

9.662*1E-04

1.068*1E-06

9.479*1E-05

1.271*1E-03

7.481*1E-04

7.26*1E-04

20.30

1.318*1E-04

2.374*1E-05

2.148*1E-05

6.586*1E-05

2.138*1E-04

5.205*1E-06

1.572*1E-06

1.978*1E-05

6.895*1E-06

4.661*1E-04

4.562*1E-05

4.753*1E-04

3.47*1E-05

  1. The optimal values are shown for each image pair. The median value for each sampling geometry is also shown. In the leave one out method, the median values were only computed over the training sets. See Table 3 for details.