Research map / Vision-language and multimodal / Cross-modal learning
Unified Cross-Modal Medical Image Synthesis With Hierarchical Mixture of Product-of-Experts
Reuben Dorent, Nazim Haouchine, Alexandra J. Golby, Sarah F. Frisken, Tina Kapur, William M. Wells
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add hierarchical latent variables (L=2 instead of 1) (ablation) | MMHVAE · ReMIND T2 synthesis | PSNR: 21.6 -> 22.6 | Improved |
| add hierarchical latent variables (L=2 instead of 1) (ablation) | MMHVAE · ReMIND FLAIR synthesis | PSNR: 19.2 -> 21.1 | Improved |
| increase latent variables to L=5 (ablation) | MMHVAE · ReMIND T2 and FLAIR synthesis | PSNR: 27.8 dB and 26.8 dB | Improved |
| increase latent variables to L=7 (ablation) | MMHVAE · ReMIND synthesis (iUS, T2, ceT1, FLAIR) | PSNR: 27.0 dB for iUS, 28.6 dB for T2, 29.8 dB for ceT1, 27.4 dB for FLAIR | Improved |
| replace concatenation with zero imputation by probabilistic fusion (ablation) | MMHVAE · ReMIND iUS synthesis | PSNR: 26.1 -> 27.0 | Improved |
| replace concatenation with zero imputation by probabilistic fusion (ablation) | MMHVAE · ReMIND FLAIR synthesis | LPIPS: 8.7 -> 6.7 | Improved |
| replace feature averaging by probabilistic fusion (ablation) | MMHVAE · ReMIND T2 synthesis | PSNR: 27.0 -> 28.6 | Improved |
| replace feature averaging by probabilistic fusion (ablation) | MMHVAE · ReMIND T2 synthesis | SSIM: 89.0 -> 90.7 | Improved |
| add GAN loss (lambda_GAN=0.025) (ablation) | MMHVAE · ReMIND iUS synthesis | LPIPS: 20.6 -> 14.6 | Improved |
| add GAN loss (lambda_GAN=0.025) (ablation) | MMHVAE · ReMIND T2 synthesis | PSNR: 30.1 -> 28.8 | Hurt |
| add GAN loss (lambda_GAN=0.025) (ablation) | MMHVAE · ReMIND iUS synthesis | SSIM: 81.4 -> 79.7 | Hurt |
| MMHVAE instead of ResViT | MMHVAE vs ResViT · Computational cost | MACs / parameters: 13G vs. 274G MACs; 14M vs. 293M parameters | Improved |
Closest papers by meaning
- Simultaneous Super-Resolution and Cross-Modality Synthesis of 3D Medical Images Using Weakly-Supervised Joint Convolutional Sparse Coding (CVPR 2017 · similarity 0.87)
- Multi-modal Vision Pre-training for Medical Image Analysis (CVPR 2025 · similarity 0.87)
- Posterior Samplings are Missing Modalities Generators for Medical Image Translation (ECCV 2026 · similarity 0.87)
- Unity by Diversity: Improved Representation Learning for Multimodal VAEs (NeurIPS 2024 · similarity 0.85)
- Flex-MoE: Modeling Arbitrary Modality Combination via the Flexible Mixture-of-Experts (NeurIPS 2024 · similarity 0.84)
- SSC-VAE: Structured Sparse Coding Based Variational Autoencoder for Detail Preserved Image Reconstruction (AAAI 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
