Moving Beyond Diffusion: Hierarchy-to-Hierarchy Autoregression for fMRI-to-Image Reconstruction
arXiv:2510.22335v2 Announce Type: replace-cross Abstract: Reconstructing visual stimuli from fMRI signals is a central challenge bridging machine learning and neuroscience. Recent diffusion-based methods typically map fMRI activity to a single neural embedding, using it as static guidance throughout the entire generation process. However, this fixed guidance collapses hierarchical neural information and is misaligned with the stage-dependent demands of image reconstruction. In response, we propose MindHier, a coarse-to-fine fMRI-to-image reconstruction framework built on scale-wise autoregress
This research is emerging as AI's capability in image generation and understanding fMRI data matures, suggesting a critical juncture for brain-computer interfaces.
This development represents a significant step towards decoding complex thought processes and could form the basis for advanced brain-computer interfaces or diagnostic tools.
The ability to reconstruct visual stimuli from fMRI signals with greater fidelity and a hierarchical approach marks a departure from static guidance methods in fMRI-to-image reconstruction.
- · Neuroscience researchers
- · AI development companies
- · Medical diagnostics sector
- · Brain-computer interface developers
- · Developers focused on single-embedding fMRI-to-image reconstruction
- · Traditional diagnostic imaging techniques
Improved understanding of visual processing in the brain.
Development of more sophisticated neuro-prosthetics or communication devices for locked-in patients.
Ethical and privacy debates around decoding and potentially reconstructing thoughts or visual experiences.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI