# Medical Image Synthesis

Medical image synthesis is often treated as filling in a missing modality, yet a realistic image need not help the downstream task. Across five public datasets, a training framework ties synthesis to task learning: the translator is constrained during training by image-adaptive frequency consistency on structure and detail, and the downstream network receives teacher features only when a teacher using synthetic images beats the student by a task-specific margin on labelled samples. Translation fidelity improved for CycleGAN, CUT and SynDiff unpaired translators, with gains in brain-tumour segmentation and lesion centre-point detection, and the training-time components are removed at inference.

Medical image synthesis is often treated as filling in a missing modality, yet a realistic image need not help the downstream task. Across five public datasets, a training framework ties synthesis to task learning: the translator is constrained during training by image-adaptive frequency consistency on structure and detail, and the downstream network receives teacher features only when a teacher using synthetic images beats the student by a task-specific margin on labelled samples. Translation fidelity improved for CycleGAN, CUT and SynDiff unpaired translators, with gains in brain-tumour segmentation and lesion centre-point detection, and the training-time components are removed at inference.

The work proposes FermatSyn, which injects anatomical priors via LoRA+ fine-tuning of a frozen SAM2 vision encoder, preserves high-frequency lesion detail with HRDM and CIN, and builds an approximately isotropic receptive field through continuity-constrained Fermat spiral scanning inside a bidirectional Mamba; on SynthRAD2023 and merged BraTS (including BraTS-MEN and BraTS-MET) it surpasses compared methods on PSNR, SSIM, FID and 3D structural consistency, and segmentation models trained on its synthesized images show no significant difference from real-image training (p>0.05).

The work proposes FermatSyn, which injects anatomical priors via LoRA+ fine-tuning of a frozen SAM2 vision encoder, preserves high-frequency lesion detail with HRDM and CIN, and builds an approximately isotropic receptive field through continuity-constrained Fermat spiral scanning inside a bidirectional Mamba; on SynthRAD2023 and merged BraTS (including BraTS-MEN and BraTS-MET) it surpasses compared methods on PSNR, SSIM, FID and 3D structural consistency, and segmentation models trained on its synthesized images show no significant difference from real-image training (p>0.05).

The work proposes a retrieval-augmented text-to-CT generation method: given a radiology report, a pretrained 3D vision-language encoder retrieves the semantically nearest case from a reference corpus, and that case's anatomical annotation is used as a structural proxy injected into a text-conditioned latent diffusion model through a ControlNet branch; on the CT-RATE dataset (27,514 training and 1,818 test volumes), retrieval-augmented generation improves image fidelity (FID 3D 0.004), clinical consistency (CT-Net AUC 0.787), and spatial controllability (Dice 0.772, HD95 3.072) over text-only baselines, while additionally enabling explicit spatial controllability that such approaches inherently lack.

The work proposes a retrieval-augmented text-to-CT generation method: given a radiology report, a pretrained 3D vision-language encoder retrieves the semantically nearest case from a reference corpus, and that case's anatomical annotation is used as a structural proxy injected into a text-conditioned latent diffusion model through a ControlNet branch; on the CT-RATE dataset (27,514 training and 1,818 test volumes), retrieval-augmented generation improves image fidelity (FID 3D 0.004), clinical consistency (CT-Net AUC 0.787), and spatial controllability (Dice 0.772, HD95 3.072) over text-only baselines, while additionally enabling explicit spatial controllability that such approaches inherently lack.