Clinico-biological-radiomics (CBR) based machine learning for improving the diagnostic accuracy of FDG-PET false-positive lymph nodes in lung cancer

Table 2 Statistically significant clinico-biological-image of lung cancer patients

Characteristics	Training set (n = 182)		p	Test set (n = 78)		p
Characteristics	LN− (n = 78)	LN + (n = 104)	p	LN− (n = 31)	LN + (n = 47)	p
Age (mean ± SD, years)	65.10 ± 7.19^†	60.46 ± 8.88^†	< 0.01	64.52 ± 6.53^†	59.30 ± 9.66^†	0.01
Weight (kg)	63.12 ± 10.65^†	66.92 ± 10.17^†	0.02	61.45 ± 11.89^†	66.55 ± 10.69^†	0.05
CA153 (U/mL)	12.01 (7.85, 15.77)^‡	13.82 (10.38, 19.81)^‡	0.01	13.08 (12.02, 15.81)^‡	16.94 (11.62, 17.46)^‡	0.09
CEA status			0.01			0.25
Negative	59 (75.64)	59 (56.73)		18 (58.06)	21 (44.68)
Positive	19 (24.36)	45 (43.27)		13 (41.94)	26 (55.32)
LN enlarged			< 0.01			< 0.01
Negative	56 (71.79)	30 (28.85)		26 (83.87)	20 (42.55)
Positive	22 (28.21)	74 (71.15)		5 (16.13)	27 (57.45)
LN SUVmax	4.15 ± 1.67^†	7.86 ± 4.11^†	< 0.01	4.34 ± 1.55^†	7.19 ± 3.96^†	< 0.01

Data in parentheses are percentages unless otherwise noted. LN− lymph node negative, LN+ lymph node positive, SD standard deviation, CA carbohydrate antigen, CEA carcinoembryonic antigen, SUV standardized uptake value
^†Values refer to mean ± standard deviation
^‡Values refer to median (interquartile range). P values were the results of univariate analysis and the bold ones indicated statistical significance

ISSN: 2047-783X