Table 3 Results of the main architecture with one, two and three branches and modalities.

From: Investigation of an efficient multi-modal convolutional neural network for multiple sclerosis lesion detection

Modalities

Ground Truth Rater 1

DSC

Jaccard

PPV

TPR

LFPR

LTPR

VD

FLAIR, T1w, T2w

0.78 [0.76, 0.79]

0.64 [0.62, 0.66]

0.81 [0.79, 0.82]

0.76 [0.74, 0.79]

0.27 [0.25, 0.30]

0.74 [0.71, 0.78]

0.13 [0.10, 0.16]

FLAIR, T2w

0.78 [0.77, 0.80]

0.65 [0.63, 0.66]

0.81 [0.79, 0.82]

0.77 [0.75, 0.79]

0.27 [0.24, 0.29]

0.74 [0.70, 0.78]

0.13 [0.10, 0.15]

FLAIR, T1w

0.75 [0.73, 0.77]

0.61 [0.59, 0.64]

0.86 [0.84, 0.88]

0.69 [0.66, 0.72]

0.23 [0.20, 0.25]

0.72 [0.68, 0.75]

0.19 [0.15, 0.23]

T1w, T2w

0.65 [0.64, 0.67]

0.49 [0.47, 0.51]

0.71 [0.69, 0.73]

0.62 [0.60, 0.65]

0.39 [0.35, 0.42]

0.65 [0.61, 0.68]

0.18 [0.15, 0.21]

FLAIR

0.75 [0.74, 0.77]

0.61 [0.59, 0.63]

0.81 [0.79, 0.83]

0.72 [0.70, 0.74]

0.34 [0.30, 0.37]

0.70 [0.66, 0.74]

0.16 [0.12, 0.19]

T2w

0.65 [0.63, 0.67]

0.49 [0.47, 0.50]

0.69 [0.66, 0.72]

0.64 [0.61, 0.66]

0.45 [0.41, 0.49]

0.63 [0.60, 0.67]

0.22 [0.20, 0.25]

T1w

0.48 [0.45, 0.50]

0.32 [0.30, 0.35]

0.78 [0.76, 0.80]

0.36 [0.33, 0.38]

0.30 [0.27, 0.32]

0.52 [0.49, 0.56]

0.54 [0.51, 0.57]

  1. This table contains all metrics that were achieved by the two- and one-branch architectures compared with the initial configuration with three branches. The best and second-best results are written in bold and italic, respectively. For all metrics, the 95%-confidence intervals are given in the square brackets.
Back to article page