Table 3 Impact of different bandwidths on breast density classification for test data.

1

1000 0000

\(70.94 \pm 1.126\)

\(67.38 \pm 1.044\)

\(75.52 \pm 1.575\)

\(0.65 \pm 0.026\)

2

0100 0000

\(67.70 \pm 0.914\)

\(64.43 \pm 0.936\)

\(71.90 \pm 1.483\)

\(0.71 \pm 0.032\)

3

0010 0000

\(68.15 \pm 2.072\)

\(63.62 \pm 2.273\)

\(73.96 \pm 2.518\)

\(0.73 \pm 0.027\)

4

0001 0000

\(67.25 \pm 1.172\)

\(62.95 \pm 1.229\)

\(72.76 \pm 1.716\)

\(0.74 \pm 0.026\)

5

0000 1000

\(68.75 \pm 1.444\)

\(63.62 \pm 1.540\)

\(75.34 \pm 1.962\)

\(0.73 \pm 0.048\)

6

0000 0100

\(64.08 \pm 2.165\)

\(58.93 \pm 2.377\)

\(70.69 \pm 2.604\)

\(0.82 \pm 0.048\)

7

0000 0010

\(66.11 \pm 1.594\)

\(61.61 \pm 1.709\)

\(71.90 \pm 2.098\)

\(0.77 \pm 0.028\)

8

0000 0001

\(66.26 \pm 1.256\)

\(61.88 \pm 1.325\)

\(71.90 \pm 1.793\)

\(0.76 \pm 0.032\)

1, 3, 5, 7

1010 1010

\(68.98 \pm 0.817\)

\(66.71 \pm 0.821\)

\(71.90 \pm 1.399\)

\(0.67 \pm 0.026\)

2, 4, 6, 8

0101 0101

\(70.11 \pm 1.687\)

\(67.11 \pm 1.821\)

\(73.96 \pm 2.176\)

\(0.66 \pm 0.026\)

3, 4, 5, 6

0011 1100

\(69.73 \pm 1.997\)

\(67.52 \pm 2.189\)

\(72.59 \pm 2.450\)

\(0.65 \pm 0.048\)

1, 2

1100 0000

\(71.32 \pm 0.843\)

\(68.46 \pm 0.840\)

\(75.00 \pm 1.429\)

\(0.63 \pm 0.025\)

7, 8

0000 0011

\(64.83 \pm 1.816\)

\(60.67 \pm 1.965\)

\(70.17 \pm 2.298\)

\(0.79 \pm 0.026\)

1, 2, 7, 8

1100 0011

\(70.42 \pm 1.169\)

\(67.79 \pm 1.225\)

\(73.79 \pm 1.715\)

\(0.65 \pm 0.027\)

1, 2, 3, 4

1111 0000

\(70.04 \pm 1.452\)

\(66.44 \pm 1.548\)

\(74.65 \pm 1.971\)

\(0.67 \pm 0.026\)

5, 6, 7, 8

0000 1111

\(70.57 \pm 1.882\)

\(72.08 \pm 2.056\)

\(68.62 \pm 2.347\)

\(0.63 \pm 0.047\)

1, 4, 5, 6

1001 1100

\(71.70 \pm 1.009\)

\(71.54 \pm 1.041\)

\(71.90 \pm 1.572\)

\(0.58 \pm 0.026\)

1, 3, 4, 5, 6

1011 1100

\(73.66 \pm 0.855\)

\(74.50 \pm 0.859\)

\(72.59 \pm 1.436\)

\(0.57 \pm 0.029\)

All

1111 1111

\(71.17 \pm 1.212\)

\(72.21 \pm 1.279\)

\(69.83 \pm 1.751\)

\(0.60 \pm 0.026\)