Research map / Image generation / Autoregressive and latent generation
Controllable diffusion-based generation for multi-channel biological data
Haoran Zhang, Mingyuan Zhou, Wesley Tansey
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 channel-wise attention in UNet blocks (latent space) (ablation) | multi-channel diffusion model · IMC protein imputation | Pearson r: 0.541 -> 0.596 | Improved |
| add output-space channel attention (ablation) | multi-channel diffusion model · IMC protein imputation | Pearson r: 0.581 -> 0.596 | Improved |
| inject condition via SE soft channel attention instead of element-wise addition (ablation) | hierarchical feature injection · IMC protein imputation | Pearson r: 0.516 -> 0.535 | Improved |
| add hierarchical conditional feature injection to unconditional base (ablation) | diffusion model · IMC protein imputation | Pearson r: -0.017 -> 0.516 | Improved |
| train separately for the single missing channel instead of random masking over all channels | proposed model · IMC breast and lung | Pearson r: 0.596 -> 0.667 (breast), 0.647 -> 0.703 (lung) | Improved |
| train on union of protein channels (zero-padded) instead of intersection across lung and breast IMC (ablation) | proposed model · IMC cross-dataset generalization | average Pearson r | Improved |
| distill to 1-step inference with SiD | proposed model · gene-to-protein, PBMC | r_p: 0.673 -> 0.672 | No clear effect |
| use proposed diffusion model instead of ControlNet | IMC lung imputation · IMC lung | Pearson r: 0.537 -> 0.647 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ControlNet — ControlNet++: Improving Conditional Controls with Efficient Consistency Feedback (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ChAda-ViT : Channel Adaptive Attention for Joint Representation Learning of Heterogeneous Microscopy Images (CVPR 2024 · similarity 0.86)
- Diffusion Generative Modeling for Spatially Resolved Gene Expression Inference from Histology Images (ICLR 2025 · similarity 0.86)
- GeneFlow: Translation of Single-cell Gene Expression to Histopathological Images via Rectified Flow (NeurIPS 2025 · similarity 0.85)
- Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning (NeurIPS 2025 · similarity 0.84)
- DUSTED: Dual-Attention Enhanced Spatial Transcriptomics Denoiser (AAAI 2025 · similarity 0.84)
- Bridging Protein Sequences and Microscopy Images with Unified Diffusion Models (ICML 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.
