NV-Generate-CT
3D CT Synthesis
NV-Generate-CT (MAISI) is a 3D latent diffusion model that generates synthetic CT volumes, optionally with paired segmentation masks. Volumes reach 512×512×768 voxels at 0.5–5.0 mm spacing, can include up to 127 annotated anatomical classes such as organs and tumors, and support controllable size for 10 classes. It is designed for training-data augmentation where privacy or rare conditions limit real data.
Agent skills
Run it with an agent
Install the skill, then just ask your agent in plain language.
medagent.nv_generate_ct_rflow
npx skills add NVIDIA-Medtech/medical-AI-skills --skill nv-generate-ct-rflowTry saying
“Generate a synthetic abdomen CT and paired segmentation mask from my config at /data/abdomen_request.json using nv-generate-ct-rflow.”
“Before launching rflow-ct inference, check whether this request is valid and estimate cost.”
“Synthesize a chest CT volume with paired masks from configs/ct_request.json and summarize the output dimensions.”
“Do a dry-run preflight of my rflow-ct request and tell me the expected GPU memory and runtime before I commit.”
Finetune it with an agent
Install the skill, then just ask your agent in plain language.
medagent.nv_generate_vae_finetune
npx skills add NVIDIA-Medtech/medical-AI-skills --skill nv-generate-vae-finetuneTry saying
“Fine-tune the NV-Generate VAE using CT and MRI images listed in /data/ctmr/vae_datalist.json.”
“Check whether my VAE finetune datalist is runnable before launching the GPU job.”
“Kick off a smoke-scale MAISI VAE finetune from data/ctmr/vae_datalist.json and report where checkpoints are written.”
“Validate my CT/MRI VAE datalist and stage the VAE configs before I commit to a full GPU run.”
Engineering & research use only — not a medical device, not for diagnosis or clinical decision-making. Skills follow the open Agent Skills spec.