Table 1 Uncorrected data results per model.

From: System transferability of Raman-based oesophageal tissue classification using modern machine learning to support multi-centre clinical diagnostics

 

5 × 3 CV

one centre

Centre 1

Centre 2

Centre 3

5 × 3 CV

All centres

PCA-LDA

Centre 1

63.8% +/− 8.0

72.1%

63.9%

78.7%

64.2% +/− 6.9

Centre 2

62.5% +/− 6.4

63.9%

70.5%

63.9%

Centre 3

62.7% +/− 6.2

65.6%

60.6%

68.9%

SVM

Centre 1

66.5% +/− 8.4

75.4%

63.9%

83.6%

63.9% +/− 11.2

Centre 2

67.1% +/− 9.3

63.9%

77.0%

63.9%

Centre 3

66.5% +/− 10.7

67.2%

63.9%

75.4%

CNN

Centre 1

66.5% +/− 13.2

100%

67.2%

60.7%

68.4% +/− 5.6

Centre 2

63.1% +/− 6.4

65.6%

90.2%

63.9%

Centre 3

64.2% +/− 12.6

75.4%

65.6%

91.8%

  1. Top line per centre represents accuracy (%) +/− 1 SD, bottom line represents log loss +/− 1 SD. Italicised entries indicate the training results and so are not indicative of performance.