CONFLUX: A Latent Diusion Model for 3D Chest-CT Synthesis with RL Post-Training

Hugging Face Daily Papers Papers

Summary

CONFLUX is a 3D latent diffusion model for chest CT synthesis that achieves high-fidelity volumetric generation with controllable clinical attributes, enhanced by a reinforcement learning post-training stage to improve conditioning reliability. The model and a synthetic dataset of ~200k chest CT volumes are released.

Controllable generative models of 3D medical images can synthesize volumes with specified clinical attributes, but this demands samples that are simultaneously high-fidelity, natively 3D, and faithful to the requested conditioning. We present CONFLUX, a latent diffusion model for chest computed tomography (CT): a 3D variational autoencoder compresses each volume, and a rectified-flow transformer generates in the latent space. Generation is conditioned on structured radiological metadata (18 abnormality findings, sex, age, and reconstruction kernel) through adaptive layer normalization. The model leads strong volumetric baselines on tri-planar Frechet distance (FID 32.3 vs. 74.6 for MAISI) while exposing direct control over clinical attributes. To strengthen that control we add an online reinforcement-learning post-training stage (group-relative policy optimization) that rewards how reliably a classifier recovers the requested findings from each generated volume. Judged by a separate, independent classifier, post-training removes 47% of the shortfall relative to real-scan reliability. We release the model and a ~200k synthetic chest-CT dataset with conditioning metadata spanning a wide variety of clinical findings.
Original Article
View Cached Full Text

Cached at: 07/07/26, 10:43 AM

Paper page - CONFLUX: A Latent Diusion Model for 3D Chest-CT Synthesis with RL Post-Training

Source: https://huggingface.co/papers/2607.02998

Abstract

A 3D latent diffusion model for chest CT generation that achieves high-fidelity results while enabling control over clinical attributes through metadata conditioning and reinforcement learning refinement.

Controllable generative models of 3D medical images can synthesize volumes with specifiedclinical attributes, but this demands samples that are simultaneously high-fidelity, natively 3D, and faithful to the requested conditioning. We present CONFLUX, alatent diffusion modelforchest computed tomography(CT): a3D variational autoencodercompresses each volume, and arectified-flow transformergenerates in the latent space. Generation is conditioned on structuredradiological metadata(18 abnormality findings, sex, age, and reconstruction kernel) throughadaptive layer normalization. The model leads strong volumetric baselines ontri-planar Frechet distance(FID32.3 vs. 74.6 for MAISI) while exposing direct control overclinical attributes. To strengthen that control we add an onlinereinforcement-learning post-trainingstage (group-relative policy optimization) that rewards how reliably a classifier recovers the requested findings from each generated volume. Judged by a separate, independent classifier, post-training removes 47% of the shortfall relative to real-scan reliability. We release the model and a ~200k synthetic chest-CT dataset with conditioning metadata spanning a wide variety of clinical findings.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2607\.02998

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper1

#### gevaertlab/conflux Unconditional Image Generation• Updatedabout 4 hours ago • 9

Datasets citing this paper1

#### gevaertlab/conflux-chest-ct Viewer• Updatedabout 4 hours ago • 200k • 41 • 1

Spaces citing this paper1

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Surflo: Consistent 3D Surface Flow Model with Global State

Hugging Face Daily Papers

Surflo is a feed-forward 3D reconstruction model that compresses unposed RGB views into latent tokens and decodes consistent 3D surface points via flow matching, enabling variable-resolution output and outperforming existing methods in speed.

Revisiting LLM Adaptation for 3D CT Report Generation: A Study of Scaling and Diagnostic Priors

arXiv cs.CL

This paper investigates parameter-efficient strategies for adapting large language models to 3D CT report generation, introducing RAD3D-Prefix, a lightweight diagnostic-prior conditioning framework that keeps the LLM frozen and requires minimal trainable parameters. It shows that freezing larger LLMs (~1B+) and training only lightweight projection layers provides a superior trade-off between performance, generalization, and computational efficiency.