A Biophysically-Conditioned Generative Framework for 3D Brain Tumor MRI Synthesis

2025-10-10

AI Summary

This work addresses the complementary needs of brain tumor MRI synthesis and healthy tissue restoration. We propose the first 3D generative framework conditioned on voxel-wise continuous tumor concentration fields. Methodologically, we innovatively incorporate a biophysically informed tumor concentration field as a continuous conditioning signal into a latent diffusion model, jointly guided by tissue segmentation maps—enabling unified tumor image synthesis and healthy tissue restoration (via zero-concentration conditioning). This dual-conditioning design substantially improves anatomical consistency and spatial coherence. Quantitative evaluation shows that our method achieves a PSNR of 18.5 for healthy tissue restoration and 17.4 for tumor image completion. The source code is publicly released, ensuring strong reproducibility and demonstrating promising potential for clinical research applications.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Magnetic resonance imaging (MRI) inpainting supports numerous clinical and research applications. We introduce the first generative model that conditions on voxel-level, continuous tumor concentrations to synthesize high-fidelity brain tumor MRIs. For the BraTS 2025 Inpainting Challenge, we adapt this architecture to the complementary task of healthy tissue restoration by setting the tumor concentrations to zero. Our latent diffusion model conditioned on both tissue segmentations and the tumor concentrations generates 3D spatially coherent and anatomically consistent images for both tumor synthesis and healthy tissue inpainting. For healthy inpainting, we achieve a PSNR of 18.5, and for tumor inpainting, we achieve 17.4. Our code is available at: https://github.com/valentin-biller/ldm.git

Problem

• Generates high-fidelity 3D brain tumor MRIs using tumor concentrations
• Restores healthy brain tissue in MRI by setting tumor concentrations to zero
• Produces spatially coherent brain images for synthesis and inpainting tasks

Innovation

• Generative model conditioned on tumor concentrations
• Latent diffusion using tissue segmentation inputs
• Synthesizes 3D spatially coherent brain MRIs