Skip to main content

Table 2 Performance characteristics of SELDI profiles in the diagnosis of lung cancer

From: Improving Detection Accuracy of Lung Cancer Serum Proteomic Profiling via Two-Stage Training Process

Training population

Training set

Test set

 

Sensitivity (%)

Specificity (%)

Accuracy (%)

Sensitivity (%)

Specificity (%)

Accuracy (%)

Single-stage decision tree

      

(A) Lung cancer + Healthy control & Inflammatory disease

81.7 (58/71)

85.7 (54/63)

83.6 (112/134)

87.2 (41/47)

37.5* (15/40)

64.4 (56/87)

(B) Lung cancer + Health control

88.7 (63/71)

89.1 (41/46)

88.9 (104/117)

83.0 (39/47)

84.6 (22/26)

83.6 (61/73)

(C) Lung cancer + Inflammatory disease

94.4 (67/71)

70.6 (12/17)

89.8 (79/88)

85.1 (40/47)

57.1 (8/14)

78.7 (48/61)

Two-stage decision tree

      
 

80.3 (57/71)

88.8 (56/63)

84.3 (113/134)

70.1 (33/47)

80.0* (32/40)

74.7 (65/87)

  1. *p < 0.05 comparing single-stage vs. two-stage decision tree patterns in specificity