Link to the code: brain-emulation GitHub repository

A 14-Minute Window: Mammal Brain Connectome Preservation Breakthrough

The problem with mapping a brain has never been purely computational. The deeper obstacle is biological: once an organism dies, its neural tissue begins to degrade within minutes. Lipid membranes dissolve. Synaptic structures swell and collapse. The ultrastructure that encodes decades of learned connectivity starts to disappear before a researcher can even prepare a tissue sample. For whole brain emulation and large-scale connectomics, this degradation window has been one of the field’s most stubborn unsolved problems.

A March 2026 preprint from Song et al., posted to bioRxiv under DOI 10.64898/2026.03.04.709724, addresses this bottleneck directly. The team demonstrated near-perfect ultrastructural preservation of an adult pig brain, achieved within a 14-minute post-mortem window. The result does not solve whole brain emulation. It clears away one of the physical obstacles that has made serious connectomics at mammalian scale difficult to plan around.

Technology Readiness Level: TRL 3–4 (validated in laboratory conditions on mammalian tissue; not yet demonstrated at human-scale volume or in clinical post-mortem settings).

What the Paper Demonstrates

Song et al. applied a preservation protocol that combines rapid chemical fixation with controlled cryoprotection, timed to begin within 14 minutes of confirmed cardiac arrest in pig subjects. The key claim is ultrastructural fidelity: synaptic cleft dimensions, vesicle distributions, dendritic spine morphology, and axonal myelination patterns were preserved with resolution sufficient for downstream electron microscopy (EM) reconstruction.

The paper quantifies preservation quality using established metrics from the electron microscopy literature, including membrane continuity scores and vesicle density ratios compared to perfusion-fixed controls. In the best-preserved samples, the protocol achieved results statistically indistinguishable from optimal perfusion fixation, which requires an intact cardiovascular system and is impossible post-mortem in a clinical setting.

This matters because perfusion fixation has long been the gold standard for connectomics tissue preparation. The Song et al. result is the first credible demonstration that near-equivalent quality can be achieved under true post-mortem conditions in a mammalian brain of meaningful size.

The 14-Minute Window

Why 14 minutes? The figure is not arbitrary. It reflects the measured onset of irreversible ultrastructural degradation in pig cortical tissue at ambient temperature. At the 14-minute mark, the team observed the beginning of membrane blebbing and early synaptic vesicle dispersal in unprotected tissue. The protocol must begin before this threshold.

In practice, 14 minutes is a narrow but workable window for a controlled research setting. It is not, at present, a window that translates straightforwardly to unplanned deaths. The protocol requires pre-staged equipment, trained personnel, and a consented subject whose death occurs in a predictable location. That describes a terminal patient in a clinical facility, not a typical post-mortem scenario.

The team acknowledges this explicitly. The paper frames the 14-minute result as a proof of principle that the biological constraint is not absolute, not as a deployment-ready procedure. Whether the window can be reliably extended through pre-mortem pharmacological intervention, hypothermia induction, or other protective measures is left as an open question.

Ultrastructural Fidelity at Mammalian Scale

Pig brains are a pragmatic model for this type of research. They are gyrencephalic (folded), mammalian, and of comparable volume to a human brain. Their white matter density and myelination architecture are substantially more complex than rodent models, making them a better test case for preservation protocols that must eventually work on human tissue.

The paper reports successful preservation across multiple cortical regions, including frontal, parietal, and occipital areas sampled from each hemisphere. Hippocampal tissue, which is both structurally complex and highly sensitive to ischemic damage, was preserved at comparable quality to cortical samples. This is a meaningful result: hippocampal EM data is particularly valuable for studying synaptic plasticity mechanisms and learning-related connectivity.

The team’s electron microscopy images show clearly demarcated synaptic densities, intact myelin sheaths, and preserved mitochondrial cristae in neuronal processes. These structural features are not merely aesthetic. They carry functional information. The geometry of a synaptic density reflects receptor composition and transmission history. Myelin sheath integrity affects conduction velocity estimates in biophysical models. Mitochondrial positioning is correlated with metabolically active zones within dendrites.

For whole brain emulation, structural data at this resolution is the raw material. A connectome reconstruction is only as accurate as the tissue it is built from.

Implications for Connectomics

The current state of connectomics relies on tissue that is either perfusion-fixed (limiting it to experimental animals), biopsy-derived (tiny samples), or preserved under conditions that introduce systematic artifacts. The Song et al. protocol, if it replicates at other institutions and scales to larger volumes, would provide a new category of tissue: post-mortem mammalian brain preserved with high fidelity.

For the connectomics pipeline specifically, this changes what kinds of questions become tractable. Human post-mortem brain banks already exist for neuropathology and genetic studies. A subset of these brains, if processed under an adapted Song et al. protocol within the relevant time window, could become candidates for high-resolution EM reconstruction. The data would not be equivalent to a living brain, but it would be orders of magnitude closer than anything achievable with conventional post-mortem fixation.

The SmartEM platform developed by Harvard and collaborators has already demonstrated automated EM acquisition at cubic millimeter scale for biological tissue. The bottleneck for larger-scale human connectomics work has never been the imaging speed. It has been tissue quality. A reliable post-mortem preservation protocol is the piece that could unlock human-scale connectomics projects that are currently not feasible.

Mouse brain connectomics has made significant progress precisely because perfusion fixation in rodents is fast, routine, and standardized. Mammalian post-mortem preservation at the quality Song et al. report could do for human-scale connectomics what perfusion fixation did for rodent work over the past two decades.

Limitations and Open Questions

Several questions the paper does not resolve deserve attention.

First, the protocol was applied to pig brains in a controlled laboratory death scenario. Post-mortem tissue from humans typically involves delays, transport, ambient temperature variation, and other factors that do not appear in this study. Whether the protocol is robust to realistic variability in time-to-fixation and environmental conditions is not tested.

Second, the paper demonstrates preservation at the ultrastructural level but does not demonstrate successful connectome reconstruction from the preserved tissue. Preservation fidelity and reconstruction feasibility are related but distinct problems. EM reconstruction of a cubic millimeter of cortex currently takes months with automated pipelines. A whole-pig-brain connectome reconstruction is not a near-term prospect regardless of tissue quality.

Third, the researchers used a specific pig strain under controlled conditions. Neuroanatomical variation, disease states, and age-related tissue changes are not accounted for. Human post-mortem tissue, which often comes from elderly individuals with some degree of vascular or neurodegenerative pathology, may respond differently to the protocol.

These are not criticisms of the paper’s core finding. They are the next set of experiments. The field now knows the preservation problem has a biological solution. The engineering and protocol scaling work is what follows.

The Drosophila connectome was completed with tissue that is orders of magnitude smaller and simpler than a mammalian brain, using protocols refined over years for that specific organism. The analogous refinement work for mammalian post-mortem preservation is now, arguably, worth investing in seriously.

Future Outlook

Song et al.’s result shifts the conversation about post-mortem connectomics from “whether” to “how.” The biological window exists. Preservation at mammalian scale is achievable. The outstanding problems are procedural, logistical, and ethical, not fundamental.

For whole brain emulation timelines, this is relevant context. The Sandberg/Bostrom roadmap and its successors have long identified tissue acquisition and preservation as a prerequisite that needed solving before large-scale human connectomics could be planned. A credible post-mortem preservation protocol moves that prerequisite from speculative to demonstrably possible in at least a laboratory context.

What comes next will depend on whether the protocol can be adapted for realistic clinical conditions, whether it replicates across species and tissue types, and whether the research community can agree on standardized metrics for preservation quality that allow meaningful comparison across studies. The 14-minute window is a result worth building on.

Official Sources