Spatially and Compositionally Controlled Multimodal Glioblastoma MRI Synthesis with Latent Diffusion
Organizations: Department of Mathematical Sciences, Stellenbosch University · Department of Computational Medicine, Stanford University
Abstract
Glioblastoma exhibits substantial heterogeneity in tumour morphology and composition across MRI modalities, often making available clinical datasets highly unbalanced in terms of patient pathologies. We propose a two-stage latent diffusion model (LDM) framework for the generation of explicitly controllable 2D glioblastoma MRI scans. First, a segmentation LDM generates tumour maps conditioned on location, spatial extent, axial position, and subregion composition. Second, a modality-conditioned LDM with ControlNet generates T1, T1Gd, T2, and FLAIR images adhering to the requested segmentation. On a held-out test set, generated segmentations achieved a mean centroid error of 0.842 pixels and a subregion proportion error of 0.795 percentage points. ControlNet improved Fréchet inception distance (FID) across all four modalities, reducing pooled FID from 42.59 to 36.12. An expert reader study yielded mean quality ratings of 6.60 for real images, 6.83 for ControlNet-generated images, and 6.77 for fully synthetic images.
Figures & tables
| Guid. | Centroid (px) | Radius (px) | Outside (%) | Prop. MAE (pp) |
| Validation | ||||
| 1.0 | 1.232 | 1.120 | 2.047 | 1.239 |
| 1.5 | 0.585 | 0.841 | 1.739 | 0.966 |
| 2.0 | 0.633 | 0.683 | 1.574 | 0.820 |
| 2.5 | 0.587 | 0.758 | 1.596 | 0.770 |
| Held-out test ( ) | ||||
| Evaluation | Backbone | ControlNet | FID |
|---|---|---|---|
| T1 6918 | 54.53 | 46.52 | |
| T1Gd 6918 | 57.19 | 46.99 | |
| T2 6918 | 41.89 | 31.14 | |
| FLAIR 6918 | 56.18 | 50.41 | |
| Pooled 27672 | 42.59 | 36.12 |
| Modality | Real | ControlNet | Fully Synthetic |
|---|---|---|---|
| T1 | 6.33 | 6.33 | 5.58 |
| T1Gd | 7.08 | 7.50 | 7.17 |
| T2 | 6.25 | 6.33 | 7.33 |
| FLAIR | 6.75 | 7.17 | 7.00 |
| Overall | 6.60 | 6.83 | 6.77 |
Medical Image Synthesis
Medical image synthesis creates, translates, reconstructs, and edits clinical images to fill missing data and support analysis.
See all 331 papers