Figure 1

a: Diagram of the soft tissue dissection procedure. b: Diagram of the craniectomy procedure. The craniectomy is done following similar lines as the soft tissue removal, aiming for approximately the blue lines in this diagram.

Figure 2

Diagram of steps related to the dura removal procedure. a: The removal of the falx cerebri with Adson forceps and scissors. b, c: Arrows indicate the direction that the malleable spatula is inserted to free the dural attachments from the interior of the skull, when necessary.

Figure 3

Example images showing a brain part of the way through the dural removal process. a: Lateral view showing the left hemisphere with the dura removed. b: Superior view showing the dura remaining on the right hemisphere. Donor #325, a 35-year-old female donor with a PMI of 35 hours.

Figure 4

Figure 4. Sequential images during the nerve removal and final brain extraction procedure in one example case. a: The cephalon is positioned prone so the frontal lobes fall away from the frontal bone. b: The stumps of the optic nerves and internal carotid arteries are visible after they have been separated from the brain. c: The brain after falling further posteriorly out of the cranium. The brainstem and nearby structures are more visible. d: View into the empty cranial vault after the brain removal, showing the residual attachments of the dura, nerves, and blood vessels that were separated. Donor #297, a 90+-year-old female donor with a PMI of 72 hours.

Brain immersion fixation

Once the brain is fully isolated, it is placed in a 7.5-liter high-density polyethylene (HDPE) container with approximately 4 liters of neutral buffered formalin (either 10 % or 20 %) so that it is fully immersed for further fixation. A suspension apparatus is placed within this container to keep the brain suspended in neutral buffered formalin. This prevents the bottom of the brain from resting against the bottom of the container, which would impede fixative penetration and could distort the shape of the tissue. To create this apparatus, a simple frame was crafted from 3/16" stainless steel rod and bent into a ring approximately 20 cm in diameter with four welded legs standing about 1.5" tall (Figure 5, Table 2). This size allows the apparatus to fit at the bottom of the container while still allowing easy placement and removal. The frame is covered with two layers of cheesecloth cut in an X pattern, with the flaps tied together underneath to create a flexible yet sturdy surface. This arrangement allows the brain to rest securely in place while absorbing fixative solution from all sides, as the cheesecloth does not meaningfully impede penetration. It also allows a magnetic stir bar to be placed beneath the frame, which is used to gently circulate the fixative solution.

We also created a different apparatus design to hold the brain above the bottom of the container during immersion, using more easily sourced items (Figure 6, Table 3).

Histology

We collected a sample from the frontal lobe of a 79-year-old male donor with a PMI of 23 hours that had a craniectomy artifact in this area caused by the oscillating saw during the craniectomy. Brain tissue sampled for light microscopy was placed into a cassette for processing and embedded in paraffin. A paraffin-embedded brain section 6 μm thick was baked, deparaffinized, and stained for Hematoxylin and Eosin (H&E). A digital image of the stained section was captured at 40X as whole slide images (WSIs) using the Aperio GT450 high-resolution scanner (Leica Biosystems).

Test of diffusion speed with in situ fixation

To assess whether in situ immersion fixation affects the diffusion rate of chemicals into brain tissue, we compared two non-perfused donor brains: one immersion fixed in situ (i.e. the brain remaining in the cranial vault after removal of the calvarium and dura on the surface of the brain) and one immersion fixed ex situ (i.e. with the brain fully extracted and suspended in fixative). Both brains were immersed in 10 % neutral buffered formalin containing iohexol 300 mgI/mL (Omnipaque, Med-Plus 0407-1413-63) as a CT-visible tracer to visualize the relative penetration rate. The in situ brain was immersed in approximately 5 L of formalin with iohexol at ~3 mgI/mL, while the ex situ brain was placed at the bottom of a container immersed in approximately 4 L with iohexol at ~3.75 mgI/mL (without the suspension apparatus). Magnetic stirring of the fluid was not used in either case. Serial CT scans were acquired daily for several days using a CT scanner (OmniTom® Elite; Neurologica, Danvers, MA). Images were reviewed using the Osimis Web Viewer. Diffusion was assessed qualitatively by visual inspection of contrast penetration into the brain parenchyma across timepoints.

Data analysis

For each case, we measured the time to perform each step and two types of artifactual tissue disruption. First, craniectomy artifacts were defined as tissue disruption caused by contact with the oscillating saw blade or osteotome during the craniectomy. Second, skull edge artifacts were defined as tissue disruption caused by contact of the brain with the cut edges of the skull during brain removal.

Results

Development of our brain extraction method

We made changes to our brain extraction protocol over time through iterative refinement. We initially attempted to perform a standard circumferential craniectomy without a midline cut, but we found that this does not allow for sufficient access to the dura, especially the falx cerebri and falx cerebelli, and therefore increases the risk of mechanical damage to the brain tissue during removal of the skull. Therefore, we added a sagittal cut through the midline of the skull, which allows for a better ability to remove the dura, while minimizing the risk of craniectomy artifact, because it is made in the natural space created by the hemispheres. In our experience, adding the midline cut may actually reduce the total procedural time, because it decreases the amount of time needed to do the dural removal, potentially more than making up for the amount of time needed to make the midline skull cut itself. However, we do not have the data to quantify this hypothesis.

Another procedural modification we adopted is to use the oscillating saw to create a shallow guiding groove, rather than cutting completely through the bone. We then complete the separation with an osteotome and mallet. We found that this provides better control and that it qualitatively appears to decrease the risk of brain craniectomy artifact.

Finally, we position our circumferential cut approximately 3–4 cm below the occipital protuberance, which is more inferior than in some previously described protocols (23). Notably, we do not perform a separate wedge cut (or "hinge cut") in the occipital region, which we found creates more sharp edges of the skull that could potentially damage the brain or pose an injury risk to the staff member performing the procedure. The more inferior occipital cut that we perform exposes the posterior fossa and cerebellum to view, providing better access to the tentorium cerebelli. If the cut were to instead be made more superiorly, then the cerebellum would remain enclosed within the skull and would only be able to be accessed from an anterior approach after elevating the cerebrum. From our observations, attempting to cut the tentorium cerebelli from this anterior position, with the cerebellum still seated in the posterior fossa, increases the risk of inadvertent dissection artifact in the cerebellum due to limited visualization and constrained instrument angles. By first removing the dura from the posterior aspect with clear visualization, we can then reposition the cephalon prone, allowing the brain to fall away from the frontal bone due to gravity. This also reduces the force required to elevate the frontal lobes and provides a more controlled brain extraction.

Speed of brain extraction

We recorded the time it takes to perform each different step of the procedure (Figure 7a). We found that the craniectomy step required the most time, with a mean of 23.3 minutes (SD = 12.0, range: 7–95 minutes). Soft tissue removal averaged 10.3 minutes (SD = 5.5). Dura removal and brain isolation were the fastest steps, averaging 7.2 minutes (SD = 6.0) and 3.5 minutes (SD = 2.9), respectively. The total procedure time across all steps was on average 44.2 minutes (SD = 19.2, range: 15–105 minutes). Although there was within-technician variability, some technicians tended to perform some steps generally faster than other technicians, likely in part due to differences in prior experience (Figure 7a). We also tested the relationship between the PMI prior to the procedure and the time it took to extract the brain, finding that there was no significant rank correlation between the PMI and the extraction time (ρ = 0.19, p = 0.052; Figure 7b).

Artifactual tissue disruption during extraction

Craniectomy artifacts, primarily caused by the oscillating saw, occurred in 65 of the 105 cases (61.9 %), with a mean of 0.98 craniectomy artifacts per case (SD = 0.99, range: 0–4). We note that craniectomy artifacts also differ in transverse length and depth in the tissue, which we did not quantify here. Among technicians with at least 5 cases, craniectomy artifact frequency did not differ significantly across technicians in our sample (Kruskal-Wallis H = 7.49, p = 0.19). In contrast, the total procedure time did vary significantly by technician (Kruskal-Wallis H = 42.6, p < 0.001). Notably, the fastest technicians in our sample were not the ones with the most craniectomy artifacts. For example, one technician (Tech 4) who had 20 cases in this series, but by far the most prior experience, had both the shortest mean procedure time (29.1 minutes) and slightly below-average craniectomy artifacts (0.90 per case).

Skull edge artifacts were more difficult for us to code reliably. Recorders appeared to interpret these tissue distortions differently, as they are often subtle and harder to visualize than frank craniectomy artifacts. Also, the degree to which they cause histological damage to underlying tissue can be unclear. Nevertheless, we recorded them and we therefore report this data for completeness. We observed a mean of 1.04 skull edge artifacts per case (SD = 0.87, range: 0–4), with 78.1 % of cases having at least one apparent skull edge artifact.

We next examined the histological effects of a craniectomy artifact in one case using H&E staining (Figure 8). At lower magnification, an area of tissue loss and/or displacement corresponding to the site of the craniectomy artifact can be seen in the cortical surface, measuring approximately 4.5 mm in depth. At higher magnification, the tissue at the craniectomy artifact margins appears jagged and disrupted, with displacement of parenchymal fragments. Deeper to the base of the cut, the zone of apparent tissue disruption is present beyond the craniectomy artifact site, possibly as a result of thermal and mechanical damage caused by the oscillating saw to the adjacent parenchyma.

Technician training procedure

We found that the total removal time decreased with technician experience, with the steepest improvement occurring within the first 5–10 cases (Figure 9). After approximately 5–10 cases, removal times generally stabilized around 30–40 minutes. The decrease in the time required to perform an extraction over time was correlated with a subjective feeling of greater anatomical familiarity and confidence in technique among the technicians.

Immersion diffusion speed with in situ fixation

In two cases, we compared the diffusion rates of a contrast agent (iohexol) during in situ versus ex situ immersion fixation (Figure 10). The brain immersed ex situ showed a rim of bright signal at the cortical surface by Day 1 that progressively expanded inward, with contrast penetrating several centimeters by Day 5. In contrast, the brain immersed in situ showed minimal parenchymal uptake throughout the observation period. While iohexol (~821 Da) is larger than formaldehyde (~30 Da), this finding suggests that in situ fixation likely slows fixative diffusion as well, presumably due to a smaller surface area allowing for diffusion when the brain is still inside of the cranial vault. Note that the ex situ brain was resting at the bottom of the container rather than on a suspension apparatus, so some inferior surface area was also occluded, yet diffusion was still markedly faster than in the in situ condition. However, because we only have one observation from each group, this comparison should be regarded as preliminary.

Variations in anatomy

Anatomical variations encountered during brain extraction can affect the procedure. In our experience, some anatomical variations have minimal impact on the extraction procedure, while others present significant challenges. One important variable is the degree to which the dura is attached to the skull. We have found that in younger donors, this is minimal, while it becomes more adherent in older donors, corroborating previous reports (29,30). Another example of anatomical variation is hyperostosis frontalis interna, which is a benign overgrowth of the inner table of the frontal bone (31). This condition was observed in one donor (Figure 11a). This did not substantially impede brain extraction or appear to increase the risk of tissue artifacts. In contrast, prior neurosurgery can pose more significant obstacles. In one case, a donor with a history of brain tumor resection had residual tumor tissue adherent to a surgical mesh embedded in the skull (Figure 11b). This attachment prevented complete removal of the affected brain region using our standard dissection techniques. The full extraction of this tissue would have required en bloc excision of the overlying skull segment.

Discussion

In this study, we describe our approach to brain extraction in the context of whole-body donation. Three procedural modifications distinguish our approach from some other protocols. First, we incorporate a midline sagittal cut through the skull in addition to the standard circumferential cut. While this adds an additional step, we qualitatively found that it can reduce the potential for damage during extraction, and that it can save procedural time by making subsequent dura removal easier. Another recent study also reported benefits of a mid-sagittal cut (9). Second, we use the oscillating saw to create only a guiding groove rather than cutting completely through the bone, prior to completing the separation with an osteotome and mallet. In our experience, this appears to provide better tactile control and reduce the risk of inadvertent craniectomy artifacts. Third, we position the posterior aspect of our circumferential cut approximately 3–4 cm below the occipital protuberance, which is more inferior than in some other protocols. This exposes the posterior fossa and allows dura removal to be performed from the posterior aspect, allowing for better visualization, which we believe can help to reduce dissection artifacts. This posterior approach to cutting the tentorium cerebelli has been previously reported as well (6,7), but instead of creating a separate "occipital wedge" in the skull, we favor simply making the circumferential incision more inferior instead.

Our timing data shows that there is a moderate degree of case-to-case variation in how long the procedure takes, especially in the craniectomy step. This variability is due to many factors, including differences in skull thickness, skull or brain pathology, dural adhesion, and technician experience. Our learning curve analysis suggests that with regular practice, in our sample, technicians can achieve procedural fluency after approximately 5–10 cases, after which removal times tended to take around 30-40 minutes. Qualitatively, it is obvious that there could be a theoretical trade-off between the duration of the procedure and the risk of tissue artifacts in the brain, because there are different steps in the procedure wherein one could take more or less time. However, we did not see this as a major factor in our data, perhaps because our technicians naturally prioritized tissue integrity over speed, or because of our limited sample size. We note that the development of virtual reality training tools has the potential to accelerate the learning curve for brain extraction (32). In the more distant future, it may also be possible to automate aspects of the brain extraction process with robotics, but it is currently expensive and challenging to automate even partial craniectomy of rodent skulls (33), which is a much easier problem.

We also performed preliminary testing of an in situ immersion fixation protocol as a potential method to help stiffen relatively softer brains prior to extraction, reasoning that some degree of surface fixation might help to mitigate mechanical damage during removal. Low stiffness is very commonly seen in brains with longer PMIs, especially those that have not been refrigerated, because refrigeration both slows decomposition and helps stiffen the brain. However, with CT imaging, we found that this approach substantially slows the diffusion of chemicals into the brain parenchyma, thus delaying the fixation of deeper structures. If these preliminary results are corroborated, they suggest that this approach likely only makes sense if the in situ immersion period is kept brief, or if perfusion fixation quality is very high (in which case fixation via immersion is not as necessary). This approach is also only feasible in the context of whole-body donation and whole brain fixation (as opposed to hemi-secting the brain and cryopreserving half of it), which are both major additional practical barriers.

Our study has several limitations. First, our protocol requires whole body donation rather than brain-only donation, which limits its applicability to most existing brain banking operations. The ability to isolate the cephalon enables positioning advantages that may not translate to standard brain extraction performed with the body intact. However, several elements of our approach could likely also be applied in more conventional brain extraction settings. Second, we did not systematically compare our method to alternative protocols in a controlled manner. Instead, our conclusions about which approaches lead to better outcomes are based on our subjective experience, which is less reliable than randomized studies. Related to that, because we iteratively refined our technique over time and did not systematically record how these changes were made over time, we could not quantitatively assess whether a given change led to improvements in outcomes. Third, our assessment of tissue artifacts relied on coarse-grained, qualitative macroscopic observation and very limited histological sampling of one sample with a craniectomy artifact. Future studies could more thoroughly investigate the degree of histological artifacts that may be differentially present based on the method and effectiveness of brain extraction, such as dark neurons (14).

Conclusions

Brain extraction is a fundamental aspect of brain banking. Qualitatively, our experience suggests that certain procedural refinements, such as using a midline sagittal cut, separating the skull with an osteotome, and cutting the tentorium cerebelli from a posterior approach, can improve the effectiveness of brain extraction. Based on our timing data, technicians appear to reach a basic proficiency after approximately 5–10 cases. Future directions include extending our methods to banking additional structures such as the spinal cord, as well as developing better tools for more rapid and artifact-free brain removal. We anticipate that improving brain extraction methods may support downstream research applications ranging from molecular profiling to connectomics, ultimately advancing our understanding of neurological and psychiatric disease.

Author contributions

J.F.C., K.F., and A.T.M. conceptualized the study. M.W., A.B., L.P., S.D., A.P., G.A.T., M.G., E.T., C.D.S., and A.T.M. performed brain extraction and histology experiments. M.W. and A.T.M. performed data analysis. M.W., A.B., and A.T.M. wrote the initial draft of the manuscript. All authors reviewed the manuscript and approved the final manuscript.

Acknowledgements

We would like to acknowledge the Neuropathology Brain Bank & Research CoRE at the Icahn School of Medicine at Mount Sinai for their histology and tissue processing services. The Icahn School of Medicine at Mount Sinai provided access to library resources.

Conflict of interest statement

Mads Wolf, Autumn Beck, Laura Paredes, Sarah Darcy, Alexander Parra, Gabriel Taylor, Macy Garrood, and Andrew McKenzie are employees of Sparks Brain Preservation, a non-profit brain preservation organization.

Funding statement

This work was supported by the Rainwater Charitable Foundation as well as NIH grants P30 AG066514, RF1 AG062348, K01 AG070326, and R01 NS146414. The funders had no role in the design of the study or in the collection or interpretation of the data.

Declaration of generative AI technologies

In the preparation of this manuscript, the authors used Claude (Anthropic) for both programming assistance and to improve the manuscript’s language. All AI tool-assisted content was reviewed and edited by the authors, who take full responsibility for the final publication.

Data availability

Whole slide image data can be accessed in a public repository on Zenodo, available here: https://zenodo.org/communities/brainextraction. Code and data is available at https://github.com/andymckenzie/Brain_extraction.

Copyright: © 2026 The author(s). This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited, a link to the Creative Commons license is provided, and any changes are indicated. The Creative Commons Public Domain Dedication waiver (https://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.