Skip to main content

Table 5 Performance comparison between the proposed segmentation and other state-of-the-art methods on PH2 dataset

From: DermoNet: densely linked convolutional neural network for efficient skin lesion segmentation

Method

JC (%)

DSC (%)

Adaptive thresholding [29]

72

80

Yen thresholing [30]

45

55

Level set active contours [31]

76

83

Statistical region growing [32]

43

61

Bootstrap segmentation [26]

60

75

Contexual hypergraph [27]

63

77

Multi-scale segmentation[28]

76

86

FCN - (JC) trained with DermoSafe data

75.8

84.7

U-Net - (JC) trained with DermoSafe data

84.8

91.1

DermoNet (JC) trained without DermoSafe data

84.3

90.2

DermoNet (JC) trained with DermoSafe data

85.3

91.5

  1. Here, JC denotes Jaccard coefficient