Phosphoenix B.V.
Dutch (EU) neurotechnology company developing the Fountain Probe — a high-density deep-brain implant targeting the lateral geniculate nucleus (LGN) of the thalamus to restore functional vision in blind people. The company pivoted from cortical (V1/V4) stimulation, which underpins the scientific foundation, to LGN stimulation as the clinical target — a strategically important distinction from all other cortical prosthesis companies (Cortigent, ReVision Implant, Neuralink, Monash).
Key facts
- Founded 2019 as spin-off from the Netherlands Institute for Neuroscience (NIN, Royal Netherlands Academy of Arts and Sciences), Amsterdam, NL.
- Founders: Prof. Pieter Roelfsema (NIN Director; scientific lead), Prof. Xing Chen (xing-chen; PhD, NIN Senior Researcher 2020–22; now tenure-track Asst. Prof. of Ophthalmology, Univ. of Pittsburgh; co-PI BCI Vision Lab; NIH Director’s New Innovator Award 2024; Top 5 Young American Scientists, Scientific American 2026), and Dr. Bert Monna (CEO).
- Registered address: Meibergdreef 5, 1105 AZ Amsterdam, NL.
- Funding:
- Pre-seed Sep 2022 — TTT Medtech Fund (Innovation Industries), FIRST Fund (BioGeneration Ventures), Innovatiefonds Noord-Holland.1
- Health-Holland PPP Allowance €211k (POSITIONED project, NIN + Phosphoenix).2
- EU EIC Transition Open 2024 “SIGHTED” grant — Phosphoenix BV lead (€1.7M), Sorbonne Université/Roelfsema scientific coordinator (€777k), Fondation Voir et Entendre (€22k). Total EU contribution >€2.5M.3
- Actively fundraising next private round (2026).
- Awards: HBP Innovation Award (Mar 2023) — “The Brain Prosthesis for the Blind”. Falling Walls ‘23 Science Breakthrough.
Why LGN, not V1? — The strategic pivot
The founding science (Chen et al., Science 2020) demonstrated 1,024-channel V1/V4 stimulation enabling letter recognition in macaques. But Phosphoenix chose the LGN as its clinical target, for reasons now well-articulated in the literature:23
| Dimension | V1 (visual cortex) | LGN (thalamus) — Phosphoenix target |
|---|---|---|
| Location | Surface/buried sulcus; spreads 25 cm² per hemisphere | Deep thalamic nucleus, ~120 mm³ total volume |
| Surgical access | Requires large craniotomy or multiple penetrations | Single burr hole (≈ like DBS) |
| Signal abstraction | High — complex feedforward + feedback, context-modulated | Lower — closer to retinal ganglion cell output; concentric ON/OFF RFs |
| Encoding scheme | Complex stimulus-specific patterns required | More straightforward — stimulus = phosphene pattern at known retinotopic location |
| Visual field coverage | Large V1 area → must tile many craniotomies for full field | Entire visual field represented in one compact 120 mm³ structure |
| Patient eligibility | Requires intact V1 (optic nerve and LGN must be working) | Works for any blindness distal to LGN (retinal degeneration, traumatic eye loss, AMD, RP); does not work for glaucoma or optic nerve damage |
| Existing implant scale | 8–20 channels (Orion); 1,024 in monkey research | 1,000+ channels planned with Fountain Probe |
| Human translational data | Multi-year human chronic data (Cortigent Orion) | First human LGN single-unit recordings published Nov 2025 (Self et al., Roelfsema group)4 |
Key rationale (NIH VA grant, Fried PI, 2023–2027): “LGN is more spatially expansive than the retina and thus allows for a larger number of stimulation sites and higher acuity. At the same time, the neural signaling patterns used by LGN neurons are much less abstract than those of the visual cortex, thereby allowing for more straightforward encoding schemes.”
Important caveat: LGN stimulation cannot help patients with glaucoma, optic nerve damage, or other pre-LGN pathway pathology — the LGN itself must receive retinal input. This means LGN is complementary to, not a replacement for, cortical prostheses for the most severely injured patients (e.g. war veterans with bilateral traumatic eye and optic nerve injury — for those, cortical is the only option).
Fountain Probe — device
- Form factor: Ultra-flexible “fountain-shaped” shank that deploys (expands like an umbrella) inside the target structure after insertion through a single burr hole, distributing hundreds to >1,000 micro-electrodes throughout the LGN’s 3D volume.3
- Electrode density: volumetric; <200 mm³ footprint covers ~120 mm³ LGN entirely — orders of magnitude higher density than earlier thalamic or cortical arrays.2
- Channel count: >1,000 (vs. 8–20 in existing approved implants like Orion).
- Signal chain: camera glasses → video-processing algorithm (relevance filtering, contour extraction) → wireless transmission → Fountain Probe stimulation → phosphene patterns perceived by user as outlines of objects.
- Surgical risk: comparable to DBS (single burr hole, established neurosurgical workflow).
- Predecessor probe: FET-Open NeuraViPeR project (EU, concluded Feb 2025) produced the TRL-3 prototype on which SIGHTED advances.
Scientific foundation and key papers
1. Macaque V1/V4 — shape perception (Chen et al., Science 2020)
Chen X, Wang F, Fernández E, Roelfsema PR. Shape perception via a high-channel-count neuroprosthesis in monkey visual cortex. Science 2020;370(6521):1191-1196. DOI: 10.1126/science.abd7435.
1,024-channel Utah-array stimulation of macaque V1+V4. Phosphenes matched receptive fields; simultaneous multi-electrode patterns produced percepts monkeys recognised as shapes, motion, and letters. The founding proof-of-concept. Note: target was V1/V4 (cortex), not LGN — the pivot to LGN for clinical deployment came afterwards.5
2. First human LGN single-unit recordings (Self et al., Nat Commun 2025)
Self MW, Vilela-Filho O, Neuenschwander S, et al. (senior: Roelfsema PR). Effects of eye closure on the spiking activity of human lateral geniculate neurons. Nat Commun 16, 10402 (2025). DOI: 10.1038/s41467-025-65383-x. Open access CC BY 4.0. Published 24 November 2025.
First-ever characterisation of spiking activity in the human LGN — historically only accessible via fMRI or animal models. DBS electrodes placed in the LGN of 2 epileptic patients in Brazil; Roelfsema was alerted serendipitously via colleague Neuenschwander.
Key findings directly relevant to LGN prosthesis design:
- Human LGN neurons respond with high-frequency bursts to strong visual stimulation.
- Circular ON/OFF subfields — same as macaque; retinotopic map well-organised.
- Dorsal (parvocellular) layers: red-green opponency, high spatial frequency.
- Ventral (magnocellular) layers: high temporal frequency, motion-tuned.
- Responses largely monocular.
- Human LGN organization closely resembles macaque → animal stimulation data transfers.
This paper is the direct human translational evidence underpinning Phosphoenix’s LGN target choice.4
3. NEUmap phosphene mapping (Lozano, Chen, Roelfsema et al., Brain Stimulation 2026)
Lozano A, Chen X, La Grouw M, Li B, Wang F, van der Grinter M, Soto-Sánchez C, Morales-Gregorio A, Fernández E, Roelfsema PR. Large-scale mapping of artificial perceptions for neuroprostheses using spontaneous neuronal activity in macaque and human visual cortex. Brain Stimulation 2026;19(1):103019. DOI: 10.1016/j.brs.2025.103019. Open access CC BY 4.0.
A key clinical enabler: NEUmap (NEural Unsupervised electrode mapping) automatically maps hundreds of phosphene locations from <0.5 s of resting-state neural activity — no vision, no task, no stable fixation required. Evaluated on 896-electrode macaque arrays and 96-electrode human (3 blind volunteers) arrays. Maps 300–700 electrodes (macaque) and 73–91 electrodes (human). Removes the major bottleneck preventing clinical deployment of high-channel-count implants — essential for any >100-channel system including Fountain Probe.6
4. NIH VA grant — Functional analysis of an LGN-based visual prosthesis
PI: Shelley Fried. VA Boston Health Care System. Grant 5I01BX005959-03. Active July 2023 – June 2027. Addresses how to effectively stimulate LGN neurons — optimal parameters, cell-type specificity per laminar layer, downstream V1 activation from LGN stimulation, chronic implant stability. Independent US validation of the LGN prosthesis approach.
Clinical roadmap
| Milestone | Date |
|---|---|
| Pre-seed funding | Sep 2022 |
| HBP Innovation Award | Mar 2023 |
| POSITIONED project (mouse preclinical, NIN + Phosphoenix) | 2023–ongoing |
| FET-Open NeuraViPeR concludes (TRL-3 prototype complete) | Feb 2025 |
| EIC SIGHTED grant begins (TRL 3 → TRL 5) | 2025 |
| First human LGN recordings published (Self et al.) | Nov 2025 |
| First-in-Human study (5 blind patients), SIGHTED | Early 2027 |
| Commercial launch target | 2032 |
Related pages
- xing-chen — Phosphoenix co-founder; Chen Lab (Pitt); Science 2020 macaque shape perception
- pieter-roelfsema — NIN Director; Phosphoenix co-founder; SIGHTED Sci. Coord.
- cortical-visual-prosthesis, visual-prosthesis — anatomical context
- cortical-visual-prosthesis-companies — schema (Phosphoenix row)
- revision-implant — competing EU cortical company (targets V1 intracortical, not LGN)
References
- BioGeneration Ventures press release (2022-09-22); phosphoenix.nl.
- Chen X, Wang F, Fernández E, Roelfsema PR. Science 2020;370(6521):1191-1196. DOI: 10.1126/science.abd7435.
- Self MW et al. Nat Commun 16, 10402 (2025). DOI: 10.1038/s41467-025-65383-x. NIN news: https://nin.nl/news/first-ever-recording-of-a-crucial-visual-structure-in-the-human-brain/
- Lozano A, Chen X, Roelfsema PR et al. Brain Stimulation 2026;19(1):103019. DOI: 10.1016/j.brs.2025.103019.
- CORDIS SIGHTED (101212687): https://cordis.europa.eu/project/id/101212687
- NeuroTech-NL POSITIONED: https://neurotech-nl.com/project/positioned/
- Bionic Vision (Fountain Probe profile, Jul 2025): https://www.bionic-vision.org/implants/fountain-probe
- NIH VA grant 5I01BX005959-03 (Fried SI, 2023–2027): https://reporter.nih.gov/project-details/11130978