Three-dimensional synthetic image generation with diffusion models for organ segmentation model training

At least one tangible computer-readable storage medium comprising instructions that, when executed, cause at least one processor to at least:

• train a first diffusion model using a first set of images without contours, the first diffusion model being an unconditioned diffusion model;
• fine-tune the first diffusion model using a second set of images with contours to form a second diffusion model, and wherein fine-tuning comprises conditioning the first diffusion model with a contour label map that is independent of image intensity and imaging modality;
• generate synthetic image patches with contours using the second diffusion model and at least one contour, wherein the at least one contour is derived from a first imaging modality and the synthetic image patches correspond to a second imaging modality different from the first imaging modality;
• train a segmentation model using the synthetic image patches;
• and
• deploy the segmentation model to inference on a third set of images.

A segmentation apparatus comprising:

• a first diffusion model trained using a first set of images without contours, the first diffusion model being an unconditioned diffusion model;
• a second diffusion model formed from the first diffusion model tuned using a second set of images with contours, the second diffusion model to generate synthetic image patches with contours using at least one contour, wherein the at least one contour is derived from a first imaging modality and the synthetic image patches correspond to a second imaging modality different from the first imaging modality, and wherein the first diffusion model is fine-tuned by conditioning the first diffusion model with a contour label map that is independent of image intensity and imaging modality;
• and
• a segmentation model trained using the synthetic image patches and deployed to inference on a third set of images.