MRIComp4Flow: Compression of 3D Brain MRI for Training Multi-Modal Generative Models

2026-08-10

AI Summary

This study addresses the high storage and I/O overhead of large-scale multimodal brain MRI data, which hinders training 3D generative models on conventional hardware. It presents the first empirical validation that lossy compression—specifically JPEG2000 and near-lossless JPEG-LS—can preserve 3D generation quality, challenging the conventional assumption that generative models require access to the full, uncompressed data distribution. The proposed approach integrates a wavelet-based flow-matching generative model conditioned on BraTS multimodal sequences. Remarkably, even at a 20:1 compression ratio, the generated outputs exhibit no statistically significant degradation compared to those trained on original data, achieving comparable PSNR (27.3 vs. 27.0 dB) and SSIM (0.95 vs. 0.96), thereby demonstrating that high compression ratios do not compromise synthesis fidelity.

Abstract

Large-scale multi-modal MRI datasets impose substantial storage and I/O costs, limiting the training of 3D generative models on commodity infrastructure. While lossy compression is known to preserve accuracy for discriminative segmentation networks, its effect on generative models, which must learn the full data distribution rather than a decision boundary, is unexplored. We study whether standard image codecs can effectively compress semantically rich brain tumor MRI while preserving the fidelity required to train and deploy a 3D MRI generative model. Each 3D volume is compressed with JPEG2000 or a near-lossless JPEG-LS pipeline. Next, a Wavelet Flow Matching model, conditioned on BraTS image sequences (T1n, T1c, T2, T2f), is trained on compressed data, and the resulting models are evaluated on the validation set. At a 20:1 compression ratio, synthesis quality is statistically equivalent to a model trained on uncompressed data within a pre-specified margin ($Δ$PSNR $<1$,dB, $Δ$SSIM $<0.02$; paired TOST $p=[[p]]$): mean PSNR is 27.3,dB vs. 27.0,dB and mean SSIM is 0.95 vs. 0.96 across modalities. Our results indicate that JPEG2000 compression is a practical step toward scalable 3D MRI generative modeling without degrading synthesis quality. The codebase is available at https://github.com/lisafis/MRIComp4Flow .

Problem

• MRI compression
• generative models
• 3D brain MRI
• multi-modal imaging
• data fidelity

Innovation

• MRI compression
• generative modeling
• JPEG2000
• Wavelet Flow Matching
• multi-modal MRI