X-ray Microscopy at 4 nm: Is This the Missing Link for Brain Scanning?
X-ray Microscopy at 4 nm: Is This the Missing Link for Brain Scanning?
In 2024, researchers from the Paul Scherrer Institute (PSI) and collaborators published a breakthrough in X-ray nanotomography:
➡️ Phys.org: New X-ray world record — Looking inside a microchip with 4 nanometer precision
➡️ Original paper on Nature (free pdf)
Using a method called ptychographic X-ray computed tomography, they achieved non-destructive 3D imaging at 4 nm resolution—a milestone previously thought exclusive to techniques like electron microscopy (which require slicing and destroy the sample). This was initially demonstrated on a commercial microchip, but the implications go much deeper.
What happened after that 2024 breakthrough?
Since that paper, the same team has pushed the method even further:
🔹 They introduced “burst ptychography”, which dramatically increases scanning speed while preserving nanoscale resolution.
🔹 Presented results at PASC 2025, revealing ~14,000 voxels/sec — a ~170× speedup over prior methods.
🔹 PSI Newsletter Source
🔹 PASC 2025 Abstract by Tomas Aidukas
This combo — high resolution and practical throughput — means we're now looking at a serious contender for 3D scanning of biological tissues.
So... could this scan a brain?
Let’s ask the obvious: Is this usable for mapping or uploading a brain?
Here’s how it stacks up, especially for dead brain tissue:
✅ Pros
- Non-destructive: Unlike electron microscopy, the tissue isn't sliced up or coated in metal. You can re‑scan or apply other imaging later.
- Sub-synaptic resolution: At 4 nm, you can see synapses, axon terminals, dendritic spines, mitochondria, and vesicles.
- True 3D: Isotropic resolution avoids the axial (z-axis) blurring of ssTEM.
- Sample preparation is easier: Chemically fixed or cryo-preserved samples are sufficient — no microtomy or ultrathin slicing.
- Good for light elements: X-ray phase contrast is sensitive to soft biological tissue (C, H, N, O).
🚫 Cons
- Can’t scan living brains: Radiation dose and scan time make it lethal for live tissue (for now).
- Limited scan volume: Scanning an entire mouse brain would still take weeks/months. Full human connectomes? Out of reach unless infrastructure scales dramatically.
- Computationally intense: Teravoxel reconstructions require major GPU resources and careful phase processing.
- No molecular tagging: It shows structure, not specific proteins or genes.
How does it compare to existing brain-mapping methods?
| Feature | FIB-SEM | ssTEM | X-ray Ptychography |
|---|---|---|---|
| Resolution | 5–8 nm | 2 nm (xy), 40 nm (z) | 4 nm isotropic |
| Destructive | ✅ Yes | ✅ Yes | ❌ No |
| Throughput | Slow | Extremely slow | Improving fast |
| Reusability | No | No | Yes |
| Tissue types | Resin-embedded | Heavy metal-stained | Fixed or frozen, easier prep |
This puts ptychographic X-ray scanning in a unique sweet spot between speed, resolution, and sample integrity.
So, what could this mean for mind uploading?
If you want to:
- Digitally reconstruct neuronal circuits
- Map cortical microcolumns
- Study synaptic connectivity in 3D
...then this method is almost ideal — for dead, preserved brain tissue.
It won’t scan a live brain, but it could enable non-destructive connectome reconstruction on post-mortem samples, cortical slabs, or even human surgical discards.
With the right segmentation AI and enough compute, we might eventually push this toward reconstructing functionally relevant neural networks in biologically accurate detail.
Sources / Further Reading
- Phys.org news article
- Nature Paper
- PSI Newsletter – Speed improvement via burst mode
- PASC 2025 presentation abstract
- eenewseurope coverage
If we had full-volume X-ray brain imaging at this resolution — and the AI to segment it — what parts of "mind uploading" become viable? And what challenges remain?