Table 3.

Overview of the diffusion-model-based architectures for medical image augmentation that have been published to date (to our knowledge, no such studies were released before 2022). The table includes the reference, architecture name, and hybrid status (if applicable), indicating the combination of VAEs, GANs, and DMs used in each study. The table provides a useful summary of the current state of the art in this area and can help guide researchers in selecting appropriate approaches for their specific needs.

Reference	Architecture	Hybrid Status	Dataset	Modality	3D	Eval. Metrics
Classification
[97]	CLDM		UK Biobank	MR	✓	FID, MS-SSIM
[106]	DDPM		ICTS	MR	✓	MS-SSIM
[107]	LDM		CXR8	X-ray		AUC
[108]	MF-DPM		TCGA	Dermoscopy		Recall
[109]	RoentGen	Hybrid (D + V)	MIMIC-CXR	X-ray		Accuracy
[110]	IITM-Diffusion		BraTS2020	MR		-
[111]	DALL-E2		Fitzpatrick	Dermoscopy		Accuracy
[112]	CDDPM		ADNI	MR	✓	MMD, MS-SSIM, FID
[113]	DALL-E2		Private	X-ray		-
[114]	DDPM		OPMR	MR	✓	Acc., Dice
[115]	LDM		MaCheX	X-ray		MSE, PSNR, SSIM
Segmentation
[116]	DDPM		ADNI, MRNet,	MR, CT		Dice
			LIDC-IDRI
[101]	brainSPADE	Hybrid (V + G + D)	SABRE, BraTS2015	MR		Dice, Accuracy
			OASIS, ABIDE			Precision, Recall
[110]	IITM-Diffusion		BraTS2020	MR		-
Cross-modal translation
[117]	SynDiff	Hybrid (D + G)	IXI, BraTS2015	CT → MR		PSNR, SSIM
			MRI-CT-PTGA
[118]	UMM-CSGM		BraTS2019	FLAIR ↔ T1 ↔ T1c ↔ T2		PSNR, SSIM, MAE
[103]	CDDPM		MRI-CT-PTGA	CT ↔ MR		PSNR, SSIM
Other
[119]	DDM		ACDC	MR	✓	PSNR, NMSE, DICE

Note: V = variational autoencoders, G = generative adversarial networks, D = diffusion models.