A brainstem-cerebellum section was acquired from the brain of an 85-year-old male cadaver with a postmortem interval (PMI) of 174 h (7 days) (Figure 1A). The whole-brain specimen was preserved in formalin for 6 months before imaging and then rehydrated for 1 week in a standard saline solution to maintain structural integrity. Ethical review and approval were not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the participant's next of kin was not required in accordance with the national legislation and the institutional requirements. The brain specimen was unidentifiable except for sex, age, medical history, and cause of death. There was no history of neurological disease or condition. The tissue specimen was acquired from a reputable nonprofit research tissue provider (Science Care, Phoenix, AZ) approved by and agreed to by our institution (St. Joseph's Hospital and Medical Center, Phoenix, AZ).

The lateral portions of the cerebellum were sharply cut, facilitating a secure fit within the custom-designed, 3D-printed container (outer diameter: 6.8 cm) (Figure 1B). This container was tailored to fit the MRI bore using a 12-cm gradient insert. The container featured dual outlet ports connected to a 3-way stopcock (Figure 1C). The container was filled with Fomblin Y-L-VAC 25/6 PFPE Pump Inert Oil (Subspecialty Fluids Co., Castaic, CA) and nonradiosensitive spacers (Figure 1B) to reduce the volume of the inert solution used.

The degasser was designed with a central cylinder featuring external threads on both ends and was sealed with internal rubber O-rings (Figure 1C). This setup, including a concave surface on the plugs, directed air bubbles out of the container, with 1 plug incorporating 2 ports for fluid management during the degassing process, followed by continuous vibration using a standard vortex mixer (VWR Scientific Products, Radnor, PA). The cylinder and plugs were manufactured using a Formlabs Form 2 SLA 3D printer with BioMed Amber Resin. The caps were produced on a Flashforge Creator Pro FDM 3D printer using polylactic acid. Oil-resistant Buna-N Rubber O-rings were used to ensure an airtight seal (Figure 1C).

Diffusion tensor imaging (DTI) lasted 48 h and 48 min, and subsequent anatomical imaging was performed using a 7-Tesla (7T) MRI system (Bruker Biospec 70/30 with 30-cm bore size) equipped with a 70-mm volume coil. Initially, DTI images were obtained to map neural pathways using a high b-value of 3,105 s/mm² to enhance contrast and resolution. The gradient settings with a duration (Δ) of 7 ms and a separation (δ) of 14 ms were selected to capture the microstructural features of the brainstem and cerebellum. The in-plane resolution was set at 0.55 mm with a slice thickness of 0.5 mm to obtain high-resolution imaging conducive to fiber analysis and tractography.

After DTI, high-resolution anatomical images were acquired using a fast low-angle shot sequence (Figures 2, 3), with TR/TE set to 1,500/9 ms for axial and coronal views and 3,500/15 ms for sagittal views. This step, performed after diffusion imaging, provided precise anatomical references and facilitated the accurate overlay of diffusion data onto the structural images.

A deterministic algorithm within DSI Studio (https://dsi-studio.labsolver.org/) was used to identify and track neural fibers (Yeh et al., 2010; Abhinav et al., 2014; Wade et al., 2021; Yeh et al., 2013; Yeh and Tseng, 2013; Yeh et al., 2011). DTI data were used to inform the placement of more than 1 million seeding points, ensuring detailed visualization of the brain's connectivity. Fibers not meeting the specified length requirements were removed by filtration.

Imaging data acquired in multiple contiguous blocks were combined by aligning the 3D volumes from each scan in their correct order. This alignment process effectively mirrored a continuous acquisition process and constructed an accurate 3D model of the brainstem-cerebellum complex.

DIPY (https://dipy.org/) was used to generate diffusion and fractional anisotropy (FA) maps to quantify the directional preference of water diffusion within tissues. FA is calculated from the diffusion tensor, indicating the extent of preferential water diffusion along the principal direction vs. perpendicular axes. These maps were used to validate the anatomical accuracy of our models against established Montreal Neurological Institute (MNI) space standards in terms of spatial orientation and resolution (Aggarwal et al., 2013).

Segmentation was performed on the basis of high-resolution images, references from histological images in Gray's Anatomy (Gray et al., 2005), FA, and diffusion-weighted imaging of the specimen at relevant locations. Clinical and anatomical significance was considered when selecting the segmented structures, and references from Rhoton's Cranial Anatomy and Surgical Approaches (Rhoton, 2003) and Gray's Anatomy (Gray et al., 2005) were consulted for better delineation of anatomical and topological extensions (Table 1) (Gray et al., 2005; Rhoton, 2003).

3D Slicer (https://www.slicer.org/) was used for segmentation and reconstruction. Segmentation was primarily oriented in the transverse plane, which is closer to the appearance of histological images. Orientation in the sagittal and coronal planes, along with the “smooth” module within 3D Slicer, were used where necessary to reduce slice-to-slice inconsistency. The color codes for the segmented anatomical regions are given in Supplementary Table 1.

After death, autolysis degrades tissue quality and can alter certain microstructural characteristics. The PMI between death and chemical fixation appears to be critical for the preservation of tissue quality (Nagy et al., 2015; Tijssen et al., 2022; Sillevis Smitt et al., 1993). Anatomical changes such as myelin loosening can be observed as early as 4 h after death (Krassner et al., 2023; de Wolf et al., 2020), depending on the temperature of the tissue. Radiological signs of autolysis, including a decrease in anisotropy and diffusion, become more pronounced with the lengthening of time between death and fixation, observable within at least 4 h (Thicot et al., 2023). Therefore, rapid fixation is essential for preserving tissue integrity (Fox et al., 1985). Challenges in global laboratory standardization, along with the potentially longer transit times of specimens before postmortem imaging, are crucial factors for obtaining ultrahigh-resolution imaging data from cadavers (Aggarwal et al., 2013; Barrett et al., 2023; Blezer et al., 2007; Oishi et al., 2020; Schilling et al., 2018; Shepherd et al., 2009). We demonstrated that diffusion MRI data and tractography information could be obtained from a cadaver specimen that was formalin-fixed 7 days postmortem and subsequently stored in formalin for 6 months. Our results indicate that proper specimen preparation, MRI techniques, and standardized data collection can yield accurate results even with longer storage times. Furthermore, our use of a higher b-value of 3,000 s/mm² increases contrast in both gray and white matter, reflecting the advancements made with the diffusion imaging modalities. Our findings validate earlier studies (Aggarwal et al., 2013; Adil et al., 2021) and show that longer PMIs do not compromise image quality and can still yield sufficient detail in brain tissues to produce reliable images. Thus, our study demonstrates that diffusion imaging can be conducted using human cadavers with extended PMI, making it possible to conduct studies in regions where immediate processing is not feasible.

Our findings also advocate a more nuanced understanding of PMI's effect. Norms between compulsory and fully rigid times on one hand and freedom from any limitation on the other could be reconsidered, as we successfully used a specimen with a 7-day PMI (174 h) stored in formalin without compromising the quality of the diffusion images. Our findings present the possibility of using human ex vivo samples that might otherwise have been considered inappropriate to extend the research potential in neuroanatomical studies.

The cadaveric brainstem-cerebellum specimen used in our study underwent formalin fixation at 7 days (174 h) postmortem, followed by long-term formalin storage using standard practices in tissue preservation consistent with the methods outlined by Kim et al. (2021), Massey et al. (2012), Miller et al. (2011), and Tafoya et al. (2017). Our next steps were implemented to accommodate the unique attributes of our sample and our advanced imaging technique. Obtaining cadaveric head-brain specimens with a low PMI is often challenging. In many cases, cadaveric tissues supplied to anatomical laboratories originate from commercial anatomic tissue suppliers and not from intrainstitutional opportunities (i.e., patients who have died in the hospital). Even with in-house anatomic tissue gift or donation programs, the availability of tissues may not be optimal, and such tissues may harbor pathology or disease processes that disrupt the normal anatomy. Thus, acquisition of appropriate tissue often requires special arrangements. The main advance is that cadaveric specimens are often not acquired in an optimal time frame; most cadaveric brain specimens used in neurosurgical anatomical dissection laboratories have been preserved and immersed in formalin solutions, and optimally acquired specimens are infrequently available at best and often associated with considerable cost (>$2,000 each). We have a wealth of experience (>30 years) in neurosurgical anatomical studies (Mignucci-Jiménez et al., 2024; On et al., 2024). We chose a specimen with a relatively long PMI to indicate that, with proper cadaver preservation, custom MRI techniques, and accurate postprocessing, detailed imaging data can be collected even with longer storage times for specimens that will also undergo dissection studies. In addition, at the time, the cadaveric head studied was an intact specimen and in good condition for imaging and correlative anatomic study by dissection. This does not mean that specimens with shorter PMIs would not be usable in the future or for similar studies. For this study, this particular specimen met all the conditions of an optimal head-brain specimen for imaging and dissection.

Specifically, the specimen was rehydrated in a normal saline solution for 7 days prior to imaging. Although not extensively documented, this step is crucial for restoring the tissue's natural hydration, potentially influencing the MRI signal characteristics. Next, on the day of imaging, the specimen was placed in a custom-designed 3D-printed container tailored to fit a 7T MRI pore. This step signifies an advancement over traditional sample holders, such as test tubes or syringe cylinders, because it allows for a more precise fit and minimization of motion artifacts, a concern highlighted in previous studies (Dyrby et al., 2011). Our container featured 2 outlet ports, each connected to a 3-way stopcock filled with Fomblin, one of the inert fluorinated oils commonly used for their MR invisibility and artifact reduction properties (Iglesias et al., 2018). Nonradiosensitive spacers were strategically placed to conserve the amount of inert solution used, an efficiency not commonly reported in existing protocols. Before imaging, we introduced a vacuum degassing process for 30 min, followed by continuous vibration. This approach, which is not explicitly detailed in prior literature, effectively enhanced air bubble removal, a crucial step to avoid image distortion due to magnetic susceptibility differences.

Our protocol leverages an established methodology while introducing innovative steps to optimize specimen preparation for high-resolution MRI (Adil et al., 2021). This approach addresses key challenges such as motion artifacts, air bubble interference, and tissue hydration, thereby enhancing the quality and reliability of the imaging data. Our work thus demonstrates the potential of customized specimen preparation methods in advancing neuroimaging research.

Our novel MRI protocol for cadaveric brain specimens draws upon foundational principles for DTI and diffusion kurtosis imaging yet introduces specific modifications to enhance imaging precision.

Our extended anatomic imaging cycle, with a duration of 3 h and 28 min and a high resolution of 78.125 × 78.125 × 500 mm3, is designed to capture the intricate details of the brainstem and cerebellum. This approach aligns with Dyrby et al. (2011), who emphasize the importance of detailed imaging for accurate fiber reconstruction. The extended duration and high resolution align with the recommendations of Henriques et al. (2020) for capturing subtle anatomical variation. The DTI protocol, lasting 48 h and 48 min with a b-value of 3,000 s/mm² across 60 directions, significantly exceeds the typical b-value ranges suggested for ex vivo imaging. This higher b-value, as supported by Maffei et al. (2022) and Jones et al. (2020) is crucial for enhancing the contrast and angular resolution to enable more precise mapping of complex fiber architectures.

We aimed to strike a balance between high-resolution imaging and scan duration, and with the sequences used, we achieved these ultrahigh-resolution anatomical imaging results in a relatively convenient 48 h and 48 min. We chose our TR/TE to limit the overall stress put on our gradient equipment. Reducing the number of targeted slices would shorten the scan time, but the duration also depends on the sample; scanning a larger sample would require longer scan times. Increasing the resolution or diffusion directions will also lengthen scan time. In other studies, we are doping the specimen in a solution with gadolinium, which causes an increase in the scan time; however, our results are still under analysis. In summary, scan time depends on many factors. Other studies may achieve high-resolution imaging with longer scan durations. Although this protocol has already provided us with high-resolution imaging, longer scan durations using gadolinium for contrast enhancement may be required in future studies.

We used a deterministic algorithm with an advanced technique for identifying and tracking neural pathways in DSI Studio for tractography. The specific parameters set for fiber tracking, including angular thresholds and track lengths, ensured accuracy and relevance in the pathways identified. This approach to tractography, influenced by the work of Schilling et al. (2018), Grisot et al. (2021), and Yendiki et al. (2022), demonstrates our ability to provide detailed and reliable interpretations of neural structures using our protocol.

Ultrahigh-resolution anatomic MRI and DTI of formalin-fixed ex vivo human brain have been previously studied by Paxinos et al. (2023) and Naidich et al. (2009). They successfully obtained detailed anatomical images using different protocols (Table 3). However, they encountered challenges such as longer scan times and artifacts. To address these challenges, we captured ultrahigh-resolution anatomical images while optimizing scan times and minimizing artifacts (Table 3). Our protocol differs from those protocols in several important perspectives (Paxinos et al., 2023; Naidich et al., 2009). First, in terms of specimen preparation, the specimen in our study had a PMI of 174 h, followed by a 6-month formalin fixation. The specimen was then rehydrated in normal saline for 7 days before imaging, which may help wash out free formalin to optimize tissue quality for imaging. Instead of using saline perfusion before formalin fixation, as is used in other protocols (Paxinos et al., 2023), we opted for passive rehydration with normal saline after fixation, 1 week before imaging. Additionally, we used a 4% formalin solution rather than the 10% and 15% solutions used in other protocols (Paxinos et al., 2023; Naidich et al., 2009). We have also performed a degassing protocol to decrease the artifacts (Figure 1).

Similar to other MRI study acquisitions, we used a 7T preclinical scanner (Paxinos et al., 2023). Although higher field strengths can offer better signal-to-noise ratios (Naidich et al., 2009), differences in diffusion directions can offset these advantages. After DTI, we used the fast low-angle shot sequence for anatomical imaging to reduce the sequence interval instead of other sequences used in other protocols (Adil et al., 2021; Naidich et al., 2009). This combination enabled us to achieve high-resolution anatomical and DTI data, optimizing the visualization of fine neuroanatomical structures with a shorter scan time.

Our protocol included 60 diffusion directions, providing a robust DTI and tractography dataset and allowing for accurate mapping of white matter tracts. As seen in other protocols, increasing the number of diffusion directions produces more detailed tractography results but causes extremely long scan times (Paxinos et al., 2023). Regarding voxel size, our protocol achieved an in-plane resolution of 0.55 mm with a slice thickness of 0.5 mm, offering intricate detail for DTI and anatomical imaging. Using finer voxel sizes causes extended scan times (Paxinos et al., 2023). Thus, we selected a voxel size that balances detail with practical scanning duration.

We aimed to strike a balance between high resolution and reasonable scan times. Our protocol offers advanced diffusion imaging and tractography with 60 diffusion directions, providing detailed anatomical imaging without the impractical scan times associated with other protocols. As such, our protocol may be an optimal choice for detailed neuroanatomical studies, particularly for presurgical trajectory planning studies, considering both time efficiency and resolution. Furthermore, although we used a specimen with a relatively long PMI in our study, we still obtained ultrahigh-resolution imaging results and intricately visualized the crucial anatomical structures and white matter tracts of the brainstem-cerebellum complex. However, further studies with more samples, particularly with short PMIs, are needed to refine and optimize the protocol.

Although we performed manual segmentation in our study, other options exist, such as automatic image segmentation with the aid of artificial intelligence (Mofatteh, 2021; Khalili et al., 2019). Even though consulting the neuroanatomy atlases significantly improves precision, automatic segmentation using artificial intelligence algorithms may decrease the risk of subjectivity and potential errors. On the other hand, manual segmentation depends on the interpreter's neuroanatomical knowledge and may be time-consuming. Automatic segmentation can address this issue with a success rate similar to that of physicians (Mofatteh, 2021; Dolz et al., 2016; Yamashita et al., 2008). The automatic segmentation methods are mainly preferred in the preoperative planning of high-risk brainstem tumor surgery and in localizing the epileptogenic zone in epilepsy surgery (Mofatteh, 2021; Kassahun et al., 2014).

In this study, a single specimen was used, which limits the generalizability of our findings. Future studies with a larger number of specimens are necessary to enhance the robustness of the results. Although we could obtain detailed neuroanatomical images using a specimen with a relatively long PMI, it will be essential to include specimens with shorter PMIs in future studies to assess potential differences in tissue quality and imaging outcomes. Moreover, studies that match PMI with tissue quality and MRI data are needed to establish the optimum PMI without compromising tissue morphology.