Table 3 F1-score of RF for classification of AC and SCC reported using the best feature selection and machine learning model ± standard error.

From: Deep representation learning of tissue metabolome and computed tomography annotates NSCLC classification and prognosis

 

TCIA ext val

RMH (n = 320)

GSTT (n = 128)

ICHT (n = 101)

Features

F1-score

P-value

F1-score

P-value

F1-score

P-value

F1-score

P-value

Radiomics

0.63 ± 0.02

0.15

0.58 ± 0.04

0.32

0.59 ± 0.03

0.19

0.57 ± 0.02

0.20

CT_emb

0.60 ± 0.04

0.18

0.54 ± 0.05

0.33

0.53 ± 0.06

0.17

0.56 ± 0.05

0.30

TMR-CT

0.84 ± 0.03

0.09

0.78 ± 0.02

0.21

0.77 ± 0.03

0.16

0.79 ± 0.03

0.23

  1. TCIA, RMH, GSTT and ICHT are The Cancer Imaging Archive, Royal Marsden Hospital (UK), Guy’s and St Thomas’ Hospital (UK) and Imperial College Healthcare Trust (UK), respectively.