SphereFlow: Missing Modality Imputation via Geometric Transport on Hypersphere
AuthorsHuihan Wang, Juncheng Wang, Yuxiang Feng, Wenlong Hou, Guoqi Yu, Chao Xu, Yang Liu, Baigui Sun, Shujun Wang
AffiliationsDepartment of Biomedical Engineering and Sports Technology (BEST), The Hong Kong Polytechnic University, Hong Kong, China · Zhejiang University, Hangzhou, China · Nov 2026 · IROOTECH TECHNOLOGY, China · Wolf 1069 b Lab, Sany Group, China · Department of Data Science and Artificial Intelligence, The Hong Kong Polytechnic University, Hong Kong, China
SphereFlow uses geometry-aware flow matching to fill in missing medical imaging modalities from the ones that are available.
Key results
dB for MRI modality synthesis.
Reduction versus the runner-up method.
dB for N-to-N modality synthesis.
Compared with 30.91 for the runner-up.
Average Dice with imputed modalities.
Gap closed relative to the zero-fill baseline and full-modality oracle.
What the paper found
SphereFlow reframes missing-modality imputation as short-range transport on a shared unit hypersphere. A frozen DINOv2 encoder maps each modality into this space; instead of starting flow matching from Gaussian noise, the model starts from the observed modality’s latent and predicts the displacement to the target. A time-weighted reconstruction loss trains modality-specific decoders on the intermediate points along that path, addressing their otherwise-untrained behavior off the sphere. The analysis shows that this data-as-source path has a dimension-independent deviation bound governed by cross-modal similarity, unlike the dimension-growing deviation of Gaussian-source flow. On BraTS, SphereFlow reaches 34.83 dB average PSNR for one-to-one MRI synthesis and reduces rFID by 33.6% versus the runner-up; for multi-source synthesis, it achieves 35.88 dB PSNR. On SynthRAD2023, it lowers CT-to-MRI rFID from 30.91 to 13.65. In downstream brain-tumor segmentation, it achieves a Dice score of 0.7473 and closes 79.8% of the gap to the full-modality oracle, indicating that the imputed images preserve features useful for the task.
Original abstract
Missing modality imputation aims to synthesize unobserved modalities from observed ones. Existing methods treat each modality as living on its own latent manifold, forcing the generator to learn a long, unstructured jump between them. We present SphereFlow, which takes a fundamentally different view: by projecting all modalities into a shared hyperspherical latent space via a frozen self-supervised encoder, cross-modal imputation reduces to short-range geometric transport between nearby points on the sphere. Realizing this idea with flow matching, however, exposes two geometric conflicts: the standard Gaussian noise source is catastrophically mismatched with the unit-norm data manifold, and the linear interpolation path departs the sphere surface into a region where the decoder has no training signal. We resolve both with simple, geometry-aware modifications: replacing the noise source with the observed-modality latent so that the velocity field learns only the cross-modal displacement, and training the decoder with a time-weighted loss that extends its domain to the near-surface region traversed by the flow. We prove that this data-as-source formulation reduces the worst-case off-manifold deviation from a quantity that grows with the latent dimension to a small, dimension-independent bound governed by inter-modality similarity. Experiments on BraTS multi-modal MRI imputation and CT-MRI translation demonstrate state-of-the-art generation fidelity across all settings, while downstream brain tumor segmentation with our imputed data closes roughly eighty percent of the gap to the full-modality oracle. Code is available at https://github.com/WTHH031230/SphereFlow.
Read the original paperMore in Multimodal AI
Browse all 79 papers →ALIVE: Making Streaming Video Models Interactive
Jinzhao Li, Yuan Xu, Jinxuan Wu, Yukun Wang, Yijia Lei, Hanyu Ji, Yinuo Chen, Zining Huang, Chi Zhang, Hang Ye, Yiwu Zhong, Honglei Yan, Panwang Pan, Yizhou Wang, Yin Li, Miao Liu
ALIVE aims to make video-streaming AI an interactive partner that can monitor events, coordinate tasks, and ask for feedback.
Let Samples Sing in Chorus: Mitigating Spurious Correlations via Cross-view Consensus
Weiwei Li, Junzhuo Liu, Tong Chu, Hengfu Yu, Yinjie Lei, Wen Li
CHORD combines clues from multiple representation spaces to help models rely less on misleading shortcuts and perform better on underrepresented groups.
LightTAR: Light and Fast Target-Aware Representations for Multimodal Tabular Learning
Eilam Shapira, Alan Arazi, Shoham Grunblat, Roi Reichart
LightTAR adds a small set of target-aware features to standard embeddings, helping tabular models make better predictions with little extra runtime.