Lozano et al. 2026 — NEUmap phosphene mapping (Brain Stimulation)
Lozano A, Chen X, Roelfsema PR et al. “Large-scale mapping of artificial perceptions for neuroprostheses…” Brain Stimulation 19(1), 103019 (2026). DOI: 10.1016/j.brs.2025.103019 Open access CC BY 4.0.
Problem addressed
High-channel-count prostheses require mapping where each electrode’s phosphene falls in the visual field — currently a tedious, patient-demanding process that blocks clinical deployment of many-electrode implants. Blind users may also have poor fixation stability and spontaneous visual phenomena confounding detection.
Method: NEUmap (NEural Unsupervised electrode mapping)
- Exploit the fact that nearby electrodes have correlated spontaneous activity.
- Dimensionality reduction (unsupervised) → map of relative electrode positions from <0.5 s of resting-state neural data (no vision, no task required).
- User manually reports locations of a small “anchor” subset → NEUmap aligns and extrapolates to all electrodes.
Results
- 2 sighted macaques: 896 V1 electrodes each — NEUmap generated high-quality maps for 300–700 electrodes.
- 3 blind human volunteers: 96 V1/V2 electrodes each — maps for 73–91 electrodes.
- Evaluated across LFP bands and multi-unit activity.
Significance for Phosphoenix / LGN prosthesis
Roelfsema’s group (Chen, Lozano, Roelfsema are co-authors; co-senior with Fernández) developed NEUmap as a core clinical enabler. The same tool will be needed for a 1,000+ channel LGN implant where manual phosphene mapping would be clinically infeasible. NEUmap removes one of the key bottlenecks on the path to clinical deployment identified in SIGHTED.