Preselection of robust radiomic features does not improve outcome modelling in non-small cell lung cancer based on clinical routine FDG-PET imaging

EJNMMI Research

Table 1 Overview of image acquisition characteristics

Robustness study	Delineation (3 methods)	Attenuation correction		Motion
Robustness study	Delineation (3 methods)	CT based	MR based	Average	Gating
Number of patients	9	9		10
Scanner manufacturer	GE Healthcare, Waukesha	GE Healthcare, Waukesha		GE Healthcare, Waukesha	GE Healthcare, Waukesha
Scanner model	Discovery 690	Discovery 690	SIGNA PET/MR	SIGNA PET/MR	SIGNA PET/MR
Reconstruction method	VPFXS	VPFXS	VPFXS	VPFXS	VPFXS
Attenuation correction	MR based: LAVA-flex pulse sequence	CT based	MR based: LAVA-Flex pulse sequence	MR based: LAVA-Flex pulse sequence
Time delay between FDG injection and PET scan start (min)	71.5–92.5	71.5–92.5	40.3–117.6	57.1–75.9	36.8–82.0
Injected activity (MBq)	181.2–252.3	181.2–252.3		136.2–259.3
Acquisition type	3D	3D		3D	4D (phase-gated)
Time per bed position (min)	2	2	2	2	2
Resolution (mm)	2.73 × 2.73 × 3.27	2.73 × 2.73 × 3.27	2.34 × 2.34 × 2.78	2.34 × 2.34 × 2.78	2.34 × 2.34 × 2.78
Delineation	Gradient-based threshold-based manual	Gradient based	Gradient based	Gradient based	Gradient based
Intraclass correlation coefficient (ICC)	ICC(1,1)	ICC(3,1)		ICC(3,2)

PET image dataset acquisition characteristics including reconstruction and delineation protocols for the three datasets are listed