Table 1. Different types of SAM utilized for object segmentation (Kirillov et al., 2023 ▸).

Type	Name of SAM	Description
SAM 1	ViT-H (ViT-Huge)	Most accurate SAM 1 model. Ideal for highly precise segmentation. Requires high computational power and runs slower.
	ViT-L (ViT-Large)	Balances processing speed and accuracy. Ideal for standard 2D X-ray images and moderate CT data.
	ViT-B (ViT-Base)	Fastest SAM 1 model. Requires low computational resources. May have slightly lower accuracy on complex or low-contrast features.
SAM 2	Hiera-L (Hiera-Large)	Most advanced SAM 2 model. Best for highly challenging segmentation tasks. Ideal for massive CT data requiring maximum accuracy.
	Hiera-S (Hiera-Small)	Lightweight model. More accurate than the Tiny version. Maintains high speed for dynamic video data.
	Hiera-T (Hiera-Tiny)	Most compact SAM 2 model. Highly optimized for speed. Best for tracking fast objects in massive in situ X-ray videos.