Table of Contents

• Contact Information
• Related Publication
• Dates of Data Collection
• Funding
• Recommended Citation
• Subject Keywords
• Dataset Overview
• Brief Methods
• Abbreviations and Labels
• Description of the Data and File Structure
• Folder Descriptions
• File Naming Patterns
• File Descriptions
• Data Dictionary
• Relationship Between Source, Processed, Converted, and Summary Files
• Relationship to Manuscript Figures and Tables
• Recommended Software
• Missing, Excluded, or Unused Data
• Quality Control
• Access, Licensing, and Sensitive Information
• MRI Facility Acknowledgment
• AI Assistance Disclosure
• Dataset Versions

---

Contact Information

Principal Investigators

Carlos Rinaldi-Ramos
J. Crayton Pruitt Family Department of Biomedical Engineering; Department of Chemical Engineering
University of Florida
carlos.rinaldi@ufl.edu

Lakiesha Williams
J. Crayton Pruitt Family Department of Biomedical Engineering
LWilliams@bme.ufl.edu

Dataset Lead / Alternate Contact

Hayden J. Good (first author)
University of Florida
haydengood34@ufl.edu

---

Related Publication

Good HJ, Tabatabaei M, Alzamora E, Eswaran N, Ali SR, Velazquez-Albino AC, Simon ZD, Febo M, Williams LN, Rinaldi-Ramos C. (2026). Sensitive Detection of In Vivo Traumatic Brain Injury Foci using Magnetic Particle Imaging. Nanoscale. https://doi.org/10.1039/D6NR02497A

Manuscript status: published online (volume and page numbers not yet assigned)

---

Dates of Data Collection

• Nanoparticle synthesis and characterization: 2024
• Experiment 1 (Longitudinal MPI/CT): 2024
• Experiment 2 (MPI/CT): 2024
• Experiment 2 (MRI): 2024–2025

---

Funding

This work was supported by the National Institute for Neurological Disorders and Stroke (NINDS) of the National Institutes of Health (NIH) under award number R21NS125089. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

A portion of the MRI work was performed in the McKnight Brain Institute at the National High Magnetic Field Laboratory's Advanced Magnetic Resonance Imaging and Spectroscopy (AMRIS) Facility, which is supported by National Science Foundation Cooperative Agreement DMR-2128556 and the State of Florida. This work was also supported in part by NIH award S10 RR025671 for MRI/S instrumentation.

---

Recommended Citation

Good HJ, Tabatabaei M, Alzamora E, Eswaran N, Ali SR, Velazquez-Albino AC, Simon ZD, Febo M, Williams LN, Rinaldi-Ramos C. (2026). Data for: Sensitive Detection of In Vivo Traumatic Brain Injury Foci using Magnetic Particle Imaging [Dataset]. Dryad. https://doi.org/10.5061/dryad.6t1g1jxdz

---

Subject Keywords

Nano-technology, Medical engineering, magnetic particle imaging, Traumatic brain injury, Magnetic resonance imaging, Nanoparticles, Nanotechnology, iron oxide nanoparticles, computed tomography, Longitudinal imaging, murine model

---

Dataset Overview

This dataset supports the manuscript "Sensitive Detection of In Vivo Traumatic Brain Injury Foci using Magnetic Particle Imaging." The study evaluates magnetic particle imaging (MPI) as a noninvasive tool for detecting and characterizing acute traumatic brain injury (TBI) in a murine model. Iron oxide nanoparticles (IONPs, PEG-coated) were administered intravenously after closed-head controlled cortical impact (CCI) injury, and their biodistribution was tracked using MPI.

The repository contains:

• Particle characterization data: TEM particle diameters, dynamic light scattering (DLS), magnetometry (M–H curve and Langevin–Chantrell fit) and magnetic particle relaxometry (RELAX) data for the IONPs used in the study.
• Imaging hardware and registration: STL files for the custom MPI/CT animal bed and CT insert, and the bed-fiducial CT and MPI scans and the CT-to-MPI transforms used for MPI–CT co-registration.
• Experiment 1 — Longitudinal MPI/CT: raw 3D MPI and CT DICOM images collected at seven timepoints over 7 days from 6 mice (3 TBI, 3 control).
• Experiment 2 — Multimodal MPI/CT/MRI: 3D MPI and CT data from 6 IONP+ mice (3 TBI/IONP+, 3 Control/IONP+) and 11.1 T MRI data from 12 mice (the 6 IONP+ mice plus 3 TBI/IONP– and 3 Control/IONP–) at a single 24 h timepoint.
• Figure source data: CSV files containing the values plotted in Figures 4B, 5D and 5E and Supplementary Figures 3–6.
• Histology: immunofluorescence (IHC) images of sagittal brain sections at 7 days post-injury (Iba1, GFAP, Hoechst): whole-section images, the magnified fields and the single-field acquisitions behind Figure 2 and Supplementary Figure 2.

Repository files include raw/native DICOM, Bruker MRI, and CSV data, along with summary characterization files and documentation.

---

Brief Methods

Animal Model

All animal procedures were conducted at the University of Florida under an approved Institutional Animal Care and Use Committee (IACUC) protocol (#202300000294). Rodent husbandry and care adhered to the University of Florida's Standard of Care for Rodents policy throughout the study. The University of Florida is accredited by AAALAC International (AAALAC #00023), holds an Animal Welfare Assurance from OLAW (D16-00244), and maintains USDA registration (58-R-0003).

All animals were male C57BL/6 mice aged approximately 6 months. TBI was induced using a controlled cortical impact (CCI) device under the following parameters: 3 mm impact tip, 2.6 mm depth, 300 ms dwell time, 6.0 m/s velocity. Surgery was closed-head (no craniotomy). Control animals received anesthesia but no impact.

Nanoparticle Administration

IONPs were administered at a dose of 100 µg Fe in 100 µL (1 mg Fe/mL; the nanoparticle stock was diluted in 1× sterile saline) as a single intravenous (tail vein) injection approximately 15 minutes after TBI induction (for TBI animals) or at the equivalent time (for controls). IONP– animals (Experiment 2 only) received an injection of the vehicle (1× sterile saline) without nanoparticles.

MPI Acquisition (Experiments 1 and 2)

MPI images were acquired using a MOMENTUM field-free line (FFL) scanner (Magnetic Insight, Inc.) in 3D full field-of-view mode ("Std 3D" scan mode) with the software development kit (SDK) enabled. Acquisition used standard settings (0 cm offset, 21 projections). The scanner software exports each 3D volume as two DICOM series, a native reconstruction and a sharpened reconstruction; the deposited series are the sharpened reconstructions, which are the volumes used for all quantification. Images are provided in DICOM format.

CT Acquisition

CT images were acquired using an IVIS Spectrum CT instrument (PerkinElmer) co-located with the MPI scanner for anatomical reference and image registration. CT files are provided in DICOM format. In Experiment 1, one CT image was acquired per animal per timepoint. In Experiment 2, one CT image was acquired per animal at the 24 h timepoint.

MPI–CT Image Registration (both experiments)

Animals were transferred between the MPI and CT scanners on the removable bed top (Supplementary Figure 1), and the CT volume of each scan session was aligned to its MPI volume with a fixed CT-to-MPI transform determined from the bed itself. The bed carries four fiducial capillaries, each holding a small sample of MPI tracer mixed with CT contrast between two caps; the capillaries were imaged without an animal in both modalities in 2024 (Registration/Registration_CT.dcm, Registration/Registration_MPI/), and the transforms were obtained in 3D Slicer (version 5.8.0 for Experiment 1 and 5.8.1 for Experiment 2, as recorded in the analysis scene files) by landmark registration of the four fiducials, applied to the CT so that it aligns with the MPI volume.

Two transforms were used. Registration/CT_to_MPI_2.tfm, an affine fiducial registration, was used for the Experiment 1 TBI animals and for all Experiment 2 animals. Registration/CT_to_MPI_1.tfm, an earlier near-rigid fiducial registration from the same bed scans, was used for the Experiment 1 control animals (C1–C3), whose analysis was completed before the affine transform was made. Evaluated on this repository's bed scans — MPI fiducial positions taken as the signal maxima and CT positions as the centroids of the tracer-containing capillary segments — CT_to_MPI_2 reproduces the four fiducial pairs with a root-mean-square residual of 0.18 mm (maximum 0.26 mm) and CT_to_MPI_1 with 0.86 mm (maximum 1.16 mm). The two transforms place the same CT point within 0.34 mm of each other at the head region of interest, within about 0.6 mm (2–3 MPI voxels of 0.25 mm) at the centre of the MPI field of view, and within 1.6 mm at its edges. The different transform used for the Experiment 1 controls therefore displaces their 3 mm-diameter quantification sphere by one to three voxels relative to the TBI animals, which has no bearing on the reported signal dynamics or conclusions.

Coordinate frame of the MPI volumes. All deposited 3D MPI series have the same geometry (241 × 241 × 480 voxels of 0.249 × 0.249 × 0.25 mm). The analyses were performed with the MPI volumes in a frame in which the slice with Instance Number 0 lies at z = −64.875 mm and successive slices lie at decreasing z (in 3D Slicer's RAS convention: origin (−29.8755, 29.8755, −64.875) mm, IJK-to-RAS direction matrix diag(1, −1, −1)). The DICOM headers of the same series place slice k at z = −64.875 + 0.25 k mm, so a reader that positions slices by their Image Position (Patient) values — including the current 3D Slicer loaders — obtains the volume mirrored in z relative to the analysis frame (origin (−29.8755, 29.8755, 54.875) mm, same direction matrix); x and y are identical in both frames. The two CT-to-MPI transforms are defined with respect to the analysis frame. To reproduce the registration from the deposited files, load the MPI DICOM series, apply Registration/MPI_DICOM_to_analysis_frame.tfm (z′ = −z − 129.75 mm; a reflection, equal to its own inverse) to the MPI volume, and apply CT_to_MPI_2.tfm (or, for the Experiment 1 control animals, CT_to_MPI_1.tfm) to the corresponding CT volume. The CT DICOM files are in the frame given by their headers and need no correction. This procedure was verified in 3D Slicer 5.8.1 on the bed scans (with the MPI volume under the reflection and the CT under either transform, the four fiducials coincide) and on animal scans from both experiments loaded beside the original analysis scenes: the deposited MPI series under the reflection are voxel-for-voxel the analyzed volumes, the deposited transforms are those of the scenes, and the quantification segments of the scenes applied to the deposited volumes reproduce the deposited signal values.

MPI Quantification (Experiment 1)

MPI signal in the head was quantified with the 3D Slicer Segment Statistics module within a spherical region of interest of 3 mm diameter (1.5 mm radius) placed at the impact site (nominally 907 voxels of 0.249 mm, 14.1 mm³; the realized count varies slightly with the sphere's placement on the voxel grid and is given for every measurement in the deposited tables). Total MPI signal is the mean voxel signal in the ROI multiplied by the number of voxels, i.e. the sum of the voxel values, in scanner arbitrary units (a.u.). For Figure 4B, each animal's total signal was divided by the mean of the three TBI animals' first-timepoint (t0) totals (72.871 a.u.), so the TBI group averages 1 at t0. One-phase decay models, Y = (Y0 − Plateau)·exp(−K·t) + Plateau with K > 0, unweighted, were fit to each group in GraphPad Prism (version 11.1.0) with the three animals entered as replicates at the group-mean scan time of each timepoint. The fitted half-lives are 23.9 h (95% CI 11.0–48.8 h) for the TBI group and 2.1 h (95% CI 0.77–7.0 h) for the control group; the groups differ significantly (extra sum-of-squares F test, F(3,36) = 95.4, p < 0.0001).

MRI Acquisition (Experiment 2)

MRI was performed at the University of Florida Advanced Magnetic Resonance Imaging and Spectroscopy (AMRIS) Facility on an 11.1 T scanner with Bruker ParaVision 6.0.1 (Bruker BioSpin). Per subject, a T2-weighted RARE anatomical scan, a 3D multi-gradient-echo (MGE) scan for T2* mapping, and two diffusion-weighted scans (46- and 23-direction shells) were acquired; the acquisition directory of each is listed in MRI/Description.csv. The RARE and MGE acquisitions are deposited; the diffusion-weighted acquisitions were not used in the study and are not deposited (see Missing, Excluded, or Unused Data). For TBI/IONP+ animal M2 the MGE scan was repeated because of excess artifact, likely from motion; the animal was checked and rescanned, and the repeat scan (directory 13) was used.

MRI data are provided in native Bruker/ParaVision format. T2* values were extracted from the MGE data in ITK-SNAP (version 4.2.2) for four regions of interest, with brain regions identified by reference to the mouse brain atlas of Paxinos and Franklin (The Mouse Brain in Stereotaxic Coordinates, 2nd ed., Academic Press, 2001). The ipsilateral ROI was drawn over the centre of the lesion, with the number of slices covered set by the extent of the lesion; the contralateral ROI was drawn in non-lesioned tissue on the opposite side with matched size and slice coverage; the same procedure was applied to the control animals. A control cortical ROI was placed over the somatosensory cortex, covering barrel field and trunk regions. Its location was determined by comparing the MRI cortical ROI to the same region on an atlas, in a location just dorsal to the corpus callosum at the level of the dorsal hippocampus. The muscle ROI served as a tissue-control comparison: because ROIs cannot be moved in ITK-SNAP, it was drawn anew over visually identifiable muscle away from the injury site, at approximately the same slice coverage and size (number of voxels) as the brain ROI. The reported values are the mean T2* (ms) of the voxels within each ROI; ROIs were drawn separately for every animal. The IONP– animals were imaged by MRI only; they did not undergo MPI or CT.

Statistical Analysis (Experiment 2)

Statistics for Figure 5 were computed in GraphPad Prism (version 11.1.0) and the results sheets are deposited as CSV files alongside the data. For Figure 5D, an ordinary one-way ANOVA across the four groups (n = 3 each) was run separately for each region; this is the primary test. Tukey's multiple comparisons test over all six group pairs was run as the secondary, post-hoc test, which is used to locate differences only when the ANOVA is significant. No ANOVA reached significance (p = 0.054–0.887), so no post-hoc comparison is reported in the manuscript. The full Tukey table (24 comparisons) is nevertheless deposited for completeness (Fig5D_Tukey_multiple_comparisons.csv), with the two pairs discussed in the manuscript marked; it contains one comparison with an adjusted p below 0.05 (muscle, Control/IONP+ vs TBI/IONP–, p = 0.038), which is not interpreted because its ANOVA was not significant. The muscle region was analyzed as a tissue control. For Figure 5E, pSNR and total MPI signal of the TBI/IONP+ and Control/IONP+ groups were compared by unpaired two-tailed t tests (pSNR: t(4) = 9.08, p = 0.0008; total signal: t(4) = 9.93, p = 0.0006). A t test of mSNR, which is not plotted in the manuscript, is included for completeness (p = 0.0006). In the Prism file the Figure 5E data are entered on a grouped table with one animal per row, so Prism reports that "only the mean values were analyzed"; because each row holds a single value, this is identical to a t test on the three animals per group.

IONP Characterization

Dynamic Light Scattering (DLS): Hydrodynamic size distributions were measured in DI water using a Brookhaven Instruments ZetaPALS with the particle sizing option. Intensity-, volume-, and number-weighted size distributions are reported as histograms in logarithmic bins, as output by the instrument software, together with lognormal fits (unimodal for the number-weighted distribution, bimodal for the volume- and intensity-weighted distributions). The number-weighted lognormal fit gave a median hydrodynamic diameter D_A = 44.5 nm with ln σ_A = 0.095 (reported as 0.10 in the manuscript). Fit parameters for all three weightings are listed in the Data Dictionary.

Magnetic Particle Relaxometry (RELAX): RELAX measurements were acquired using the MOMENTUM scanner (Magnetic Insight, Inc.) in relaxometry mode. Five technical replicate measurements were collected from the same particle sample (Fe mass: 8.711 µg Fe per measurement, 10 µL volume). The point spread function (PSF) was recorded as signal (arbitrary units) vs. applied field (T) for both positive and negative field sweeps. For Supplementary Figure 6, each scan's signal was divided by the sample Fe mass (0.008711 mg) to give specific signal (a.u./mg Fe), and the five scans were averaged point by point. The deposited CSV files contain the raw PSF data; header metadata fields include fwhm and amplitude, which are reported directly by the MOMENTUM instrument (not derived by post-processing).

Transmission electron microscopy (TEM): Iron oxide cores were imaged after synthesis (17 images at 120 kx). The projected area of each particle was measured with a MATLAB script, and the diameter reported for each particle is the equivalent-circle diameter, d = √(4A/π). 460 particles were measured; their arithmetic mean ± standard deviation is 21.7 ± 2.9 nm (reported as 21.8 ± 2.9 nm in the manuscript). The histogram in Supplementary Figure 3 uses 2 nm bins centered on even diameters (e.g., the "20 nm" bar counts particles with 19 ≤ d < 21 nm) and is reproduced exactly by the deposited diameters.

Magnetometry: Magnetization vs. applied field was measured at room temperature on a Quantum Design MPMS-3 magnetometer over ±7 T (sample LE061224_LTE051524_2, PEG-coated particles made from core batch LTE051524, the same cores used for the RELAX measurements). Magnetization is normalized by the mass of iron in the sample, computed assuming magnetite. The data were fitted with a Langevin–Chantrell model (Langevin function weighted by a lognormal distribution of magnetic diameters), giving a median magnetic diameter of 20.82 nm with ln σ = 0.2295, a saturation magnetization of 119.8 A·m²/kg Fe and a calculated magnetic fraction φ = 0.9985 (R² = 0.999). The fitted M–H curve and the fitted magnetic-diameter distribution are deposited with the measured data.

Histology

Brains harvested 7 days post-injury were fixed, cryoprotected (10% then 30% sucrose in PBS, 24 h each at 4 °C), frozen in isopentane on dry ice, embedded in OCT and cryosectioned sagittally at 20 µm. Sections were blocked (3% goat serum, 0.3% Triton X-100 in PBS, 2 h), incubated overnight at 4 °C with anti-GFAP (Sigma-Aldrich AB5541, astrocytes) and anti-Iba1 (Wako 019-19741, microglia), then with goat anti-chicken IgY–Alexa Fluor 633 (GFAP) and goat anti-rabbit IgG–Alexa Fluor 488 (Iba1), counterstained with Hoechst and mounted in VECTASHIELD. Images were acquired on a Keyence BZ-X810 fluorescence microscope, one channel at a time, as stitched whole-section images and as single-field acquisitions of the injury region; the deposited images are the microscope's pseudo-coloured exports (Hoechst blue, Iba1 green, GFAP red), one file per channel plus, for the magnified fields, the microscope's composite. Scale bars in the manuscript figures: 100 µm (Figure 2), 1 mm, 50 µm and 100 µm (Supplementary Figure 2A, C, D).

---

Abbreviations and Labels

Abbreviations

Animal Identifiers

Longitudinal Timepoints (Experiment 1 only)

---

Description of the Data and File Structure

The repository contains DICOM image files, Bruker MRI data directories, and CSV files. The top-level structure is:

/
├── README.md
├── Repository_Abstract.md
├── Bed_STL/
│   ├── Bed_Top.stl
│   ├── Bed_Bottom.stl
│   └── CT_Insert.stl
├── Registration/
│   ├── Registration_CT.dcm             ← bed phantom CT with fiducials
│   ├── Registration_MPI/               ← bed phantom MPI DICOM series (480 slices)
│   ├── CT_to_MPI_1.tfm                 ← earlier near-rigid CT→MPI transform (Exp 1 controls)
│   ├── CT_to_MPI_2.tfm                 ← affine CT→MPI transform (Exp 1 TBI, all Exp 2)
│   └── MPI_DICOM_to_analysis_frame.tfm ← reflection: MPI DICOM frame → analysis frame
├── Histology/                          ← IHC images, one file per channel (+ composite for magnified fields)
│   ├── Fig2_Sham_section/              ← sham whole section (Fig 2 top row): Fig2_Sham_section_{Hoechst,Iba1,GFAP}.tif
│   ├── Fig2_Sham_inset/                ← sham magnified field (Fig 2 second row): ..._{Hoechst,Iba1,GFAP,Composite}.jpg
│   ├── Fig2_TBI_section/               ← TBI whole section (Fig 2 third row): Fig2_TBI_section_{Hoechst,Iba1,GFAP}.tif
│   ├── Fig2_TBI_inset/                 ← TBI magnified field (Fig 2 bottom row): ..._{Hoechst,Iba1,GFAP,Composite}.jpg
│   ├── SuppFig2_TBI_overview/          ← TBI whole section, full resolution (Supp Fig 2A–B): SuppFig2_TBI_{Hoechst,Iba1,GFAP}.tif
│   ├── SuppFig2_TBI_panelC_field/      ← single-field acquisition, Supp Fig 2C: ..._{Hoechst.tif,Iba1.jpg,GFAP.jpg,Composite.jpg}
│   └── SuppFig2_TBI_panelD_field/      ← single-field acquisition, Supp Fig 2D: ..._{Hoechst,Iba1,GFAP,Composite}.jpg
├── Particle_Characterization/
│   ├── DLS_SuppFig4_histogram.csv
│   ├── DLS_SuppFig4_lognormal_fits.csv
│   ├── TEM_SuppFig3_particle_diameters.csv
│   ├── TEM_SuppFig3_histogram.csv
│   ├── Magnetometry_SuppFig5_MH_measured.csv
│   ├── Magnetometry_SuppFig5_LangevinChantrell_fit_curve.csv
│   ├── Magnetometry_SuppFig5_magnetic_diameter_distribution.csv
│   ├── Magnetometry_SuppFig5_fit_parameters.csv
│   └── RELAX/
│       ├── RELAX_PSF_specific_signal_SuppFig6.csv
│       └── 2024-06-14_[time]_0[23-27]_LE06012LTE051524/  ← 5 replicate folders
│           ├── relaxometer_positive.csv
│           └── relaxometer_negative.csv
├── Experiment1_LongitudinalStudy/
│   ├── Fig4B_MPI_signal_by_animal.csv
│   ├── Fig4B_one_phase_decay_fit_parameters.csv
│   ├── Fig4B_one_phase_decay_fit_curve.csv
│   ├── CT/
│   │   ├── C1/ through C3/             ← t0–t6 CT per control animal
│   │   └── M1/ through M3/             ← t0–t6 CT per TBI animal
│   └── MPI/
│       ├── Control/
│       │   └── C1/ through C3/         ← t0–t6 MPI series per control
│       └── Injured/
│           └── M1/ through M3/         ← t0–t6 MPI series per TBI animal
└── Experiment2_MPI_MRI/
    ├── Fig5D_T2star_by_region.csv
    ├── Fig5D_one_way_ANOVA_results.csv
    ├── Fig5D_Tukey_multiple_comparisons.csv
    ├── Fig5E_MPI_pSNR_total_signal.csv
    ├── Fig5E_unpaired_t_test_results.csv
    ├── CT/                             ← one .dcm per IONP+ animal (C1–C3, M1–M3)
    ├── MPI/                            ← DICOM series per IONP+ animal
    │   ├── C1/ through C3/
    │   └── M1/ through M3/
    └── MRI/                            ← Bruker/ParaVision native format, all 12 animals
        ├── Description.csv
        ├── README.txt
        └── 20[date]_[ID]_MPI_TBI_[animal]_1_1/  ← 12 subject directories

Note: IONP– animals (PB_C1–PB_C3, PB_M2–PB_M4) appear only in the MRI directory. This cohort was imaged by MRI only; no MPI or CT was acquired for these animals.

---

How the files are packaged on Dryad

Dryad presents a dataset as a flat list of files, with this README held separately by Dryad. To keep the number of files manageable and the folder structure intact, the image data are uploaded as nine zip archives, while the summary CSV files and the STL files are uploaded as individual files so that they can be previewed and downloaded on their own. Every entry inside an archive carries its full path from the repository root (for example Experiment1_LongitudinalStudy/MPI/Injured/t0_M1/...), so extracting all nine archives into one folder and placing the loose files in the folders named below reproduces the tree shown above exactly. Each archive was verified after creation against the source files (entry names, sizes and CRC-32 checksums; no missing, extra or differing entries).

Individual (unzipped) files, with the folder they belong to in the tree above:

---

Folder Descriptions

Bed_STL/

STL files for the custom 3D-printed animal bed (Supplementary Figure 1). The animal is imaged under anesthesia on Bed_Top.stl mounted on Bed_Bottom.stl in the MPI scanner; the bed top is then detached and mounted on CT_Insert.stl using guide pegs for CT imaging, so the animal is not repositioned between modalities. All three are binary STL files (Bed_Top: 173,820 triangles; Bed_Bottom: 8,152; CT_Insert: 3,108). STL files carry no unit information; dimensions are in mm.

Registration/

Files used for MPI–CT co-registration in both experiments. See "MPI–CT Image Registration" under Brief Methods for how they are used and for the coordinate frame.

Registration_CT.dcm — CT scan (IVIS Spectrum CT, 2024) of the animal bed carrying four fiducial capillaries that are visible in both MPI and CT. Not an animal scan.

Registration_MPI/ — 3D MPI DICOM series (480 slices, BedwFids_0p5cm_0000.dcm to _0479.dcm; sharpened reconstruction) of the same bed and fiducials, acquired on the MOMENTUM scanner in 2024 in the same scan mode as the animal images. The four fiducial signal maxima in this volume coincide with the fiducial landmarks used for both registrations.

CT_to_MPI_1.tfm — Earlier near-rigid CT-to-MPI transform from the same bed scans (the 3D Slicer node ct to mpi of the Experiment 1 control analysis scenes), used for the Experiment 1 control animals (C1–C3). The version applied in the analysis, deposited here, carries the rotation elements rounded to two decimal places relative to the original fit (at most 0.5 mm effect at the edges of the MPI field of view). Fiducial residual 0.86 mm RMS.

CT_to_MPI_2.tfm — Affine CT-to-MPI transform (ITK text format, AffineTransform_double_3_3: 3 × 3 matrix followed by a translation, ITK/LPS "from parent" convention) obtained by landmark registration of the four fiducials; the 3D Slicer node CT_to_MPI_New of the analysis scenes. Used for the Experiment 1 TBI animals and all Experiment 2 animals; the Experiment 2 analysis scene carries a copy of it with the rotation elements rounded to two decimal places, which displaces the CT by at most 0.5 mm within the MPI field of view (0.26 mm at the head signal) and does not affect the MPI signal statistics; the full-precision transform is deposited. Fiducial residual 0.18 mm RMS. Load in 3D Slicer with Add Data and apply to a CT volume in the Transforms module.

MPI_DICOM_to_analysis_frame.tfm — Linear transform (matrix diag(1, 1, −1), translation (0, 0, −129.75) mm) that maps an MPI volume loaded from the deposited DICOM series, with slices positioned by their header coordinates, into the frame in which the analyses were performed and in which the two CT-to-MPI transforms are defined. Apply it to the MPI volume (never to the CT) before applying CT_to_MPI_2 or CT_to_MPI_1 to the CT. It is its own inverse.

Histology/

Immunofluorescence images of 20 µm sagittal brain cryosections at 7 days post-injury, acquired on a Keyence BZ-X810 microscope (see Brief Methods, Histology). 25 files in seven folders. Channels are named in the file names: Hoechst (nuclei, blue), Iba1 (microglia, Alexa Fluor 488, green) and GFAP (astrocytes, Alexa Fluor 633, red); Composite is the microscope's three-channel merge. Every single-channel file is the microscope's pseudo-coloured export stored as an 8-bit RGB image in which only the channel's colour component carries signal; the channel files of a folder are pixel-aligned and overlay to the composite shown in the manuscript. No intensity adjustment or annotation (scale bars, dashed circles and boxes) is included — those were added when the figures were assembled.

• SuppFig2_TBI_overview/SuppFig2_TBI_{Hoechst,Iba1,GFAP}.tif — full-resolution stitched image of the whole sagittal section of a TBI animal, about 23,290 × 9,460 px (Supplementary Figure 2A; panel B is the cerebellar region of the same images).
• SuppFig2_TBI_panelC_field/ and SuppFig2_TBI_panelD_field/ — two single-field acquisitions (1,920 × 1,440 px, microscope field IDs XY01_00021 and XY01_00014) at the cerebellar impact site of the same section, shown in Supplementary Figure 2C (the yellow-boxed region; the panel is a further magnification within the field) and 2D (the cyan-boxed region). Three channel files plus the composite each; the panel C Hoechst channel is a TIFF, the other files are the microscope's JPEG exports.
• Fig2_TBI_section/Fig2_TBI_section_{Hoechst,Iba1,GFAP}.tif — whole sagittal section of a TBI animal at reduced resolution (5,300 × 2,035 px); the Figure 2 TBI top-row panels are its cerebellar region. This is a separate stitched image at lower resolution, not a downsampled copy of the Supplementary Figure 2 overview.
• Fig2_Sham_section/Fig2_Sham_section_{Hoechst,Iba1,GFAP}.tif — whole sagittal section of a sham animal at reduced resolution (5,431 × 2,306 px); the Figure 2 sham top-row panels are its cerebellar region.
• Fig2_TBI_inset/ and Fig2_Sham_inset/ — the magnified fields of Figure 2 (TBI 2,048 × 2,048 px, sham 4,096 × 4,096 px), three channel files plus the composite each, as JPEG snapshots exported from the microscope viewer.

Formats: TIFF files are single-page, 8-bit RGB, lossless Deflate compression (the microscope's uncompressed exports were recompressed without any change to the pixel data; decoded pixels are identical to the originals, 687 MB → 47–108 MB per overview file); JPEG files are deposited exactly as exported. All open in ImageJ/Fiji, Python (Pillow, tifffile), MATLAB and standard image viewers. The image files carry no calibrated pixel size; the scale bars in the manuscript figures give the scale.

Particle_Characterization/

Contains characterization data for the IONPs used in both experiments.

DLS_SuppFig4_histogram.csv — Instrument-reported size histograms (32 logarithmic bins from 30 to 314 nm) for the number-, volume-, and intensity-weighted distributions, each normalized to its maximum. Source data for the bars in Supplementary Figure 4.

DLS_SuppFig4_lognormal_fits.csv — Lognormal fits to the three histograms, evaluated at 501 diameters from 30 to 325 nm. Source data for the fitted lines in Supplementary Figure 4. The fits were evaluated at 10,000 points in the authors' spreadsheet; every 20th point is deposited.

TEM_SuppFig3_particle_diameters.csv — Equivalent-circle diameter of each of the 460 particles measured by TEM, with the image number (1–17) and the particle index assigned by the measurement script. Source data for the histogram in Supplementary Figure 3.

TEM_SuppFig3_histogram.csv — The 2 nm-bin histogram of those diameters as plotted in Supplementary Figure 3 (bin center, bin edges, particle count).

Magnetometry_SuppFig5_MH_measured.csv — Measured magnetization vs. applied field (123 points, in measurement order), the blue circles of Supplementary Figure 5.

Magnetometry_SuppFig5_LangevinChantrell_fit_curve.csv — The fitted Langevin–Chantrell M–H curve evaluated at 678 field values, the red line of Supplementary Figure 5.

Magnetometry_SuppFig5_magnetic_diameter_distribution.csv — The lognormal distribution of magnetic diameters obtained from the fit, evaluated at 241 diameters from 1 to 403 nm and scaled to a maximum of 1.

Magnetometry_SuppFig5_fit_parameters.csv — Sample identifier and the fitted parameters (φ, median magnetic diameter, ln σ, saturation magnetization) with units and definitions, plus the R² reported in the manuscript.

RELAX/ — Five replicate RELAX measurements of the same particle sample, plus the Supplementary Figure 6 source data. See below.

Particle_Characterization/RELAX/

The five replicate acquisition folders are named by timestamp, replicate index (_023 through _027), and exam label:

2024-06-14_12-22-59_023_LE06012LTE051524/
2024-06-14_12-24-59_024_LE06012LTE051524/
2024-06-14_12-26-46_025_LE06012LTE051524/
2024-06-14_12-28-41_026_LE06012LTE051524/
2024-06-14_12-30-47_027_LE06012LTE051524/

Each folder contains only the raw PSF data files:

• relaxometer_positive.csv — Raw PSF data, positive field sweep (field in T, signal in a.u.; 3201 data points).
• relaxometer_negative.csv — Raw PSF data, negative field sweep (field in T, signal in a.u.; 3201 data points).

The sample Fe mass for RELAX measurements was 8.711 µg Fe (10 µL volume at 0.8711 mg Fe/mL). This mass is used to calculate specific signal (a.u./mg Fe) and is the normalization basis for the PSF data shown in Supplementary Figure 6.

RELAX_PSF_specific_signal_SuppFig6.csv — Source data for Supplementary Figure 6, which shows all five scans (white lines) and their average (black line) for both sweep directions. Derived from the ten raw CSVs above by converting field from T to mT, dividing each scan's signal by 0.008711 mg Fe, and taking the point-by-point arithmetic mean of the five scans. All five scans share an identical field grid. Generated with Python (NumPy) during repository preparation.

Experiment1_LongitudinalStudy/

Contains longitudinal CT and MPI data from 6 mice (3 TBI: M1–M3; 3 control: C1–C3) across 7 timepoints (t0–t6, spanning about 30 min to 7 days after impact for TBI animals or after IONP injection for controls). This experiment corresponds to Figures 3 and 4 and Supplementary Figure 7 in the manuscript.

Fig4B_MPI_signal_by_animal.csv — Segment statistics of the 3 mm-diameter head ROI for each animal at each timepoint (42 rows: 6 animals × 7 timepoints): voxel count, volume, mean, maximum, total signal, normalized total signal, pSNR and mSNR, with the path of the source MPI DICOM series. The normalized values are the points plotted in Figure 4B.

Fig4B_one_phase_decay_fit_parameters.csv — Best-fit parameters, 95% profile-likelihood confidence intervals and goodness-of-fit statistics of the one-phase decay fits to each group, as reported by GraphPad Prism.

Fig4B_one_phase_decay_fit_curve.csv — The fitted curves and their 95% confidence bands (the solid and dotted lines of Figure 4B), as tabulated by Prism at 1,000 points from 0 to 175 h.

CT/ — CT DICOM files organized by animal. Each animal folder (C1–C3, M1–M3) contains one .dcm file per timepoint named t#_[ID]_CT.dcm.

MPI/ — Longitudinal MPI 3D DICOM series (sharpened reconstruction). Data are organized by group (Control/, Injured/) and then by animal, with one subfolder per timepoint (t0–t6, named t#_[ID]). Each timepoint folder contains a numbered DICOM series (t#_[ID]_Std_3D_0000.dcm to _0479.dcm) representing a single 3D MPI volume. All files in a timepoint folder must be loaded together as a DICOM series.

Excluded animal and labels in the MPI DICOM files. One Experiment 1 TBI animal was excluded because of a compromised tail-vein injection; no data from that animal are deposited. Because the scanner had already assigned labels to that animal and to the scans acquired after it, the labels recorded by the instrument during Experiment 1 do not match the animal numbering used in the manuscript. For consistency with the manuscript, the file names and the Study Description (0008,1030) and Series Description (0008,103E) header fields of every deposited MPI series (both experiments and the registration scan) were set before deposit to the manuscript's animal and timepoint labels (for example t1_M2_Std_3D), replacing the scanner's acquisition-time labels and operator comments. No other header field and no pixel data were modified; the scanner project name (Patient's Name), series number, acquisition date and time (Content Date/Time), operator and all DICOM unique identifiers are as written by the instrument, and the Rinaldi-Ramos Lab retains a table linking each deposited series to its original scanner files. The MPI DICOM files are the sharpened reconstructions exported by the MOMENTUM scanner.

Experiment2_MPI_MRI/

Contains CT, MPI, and MRI data from 12 mice at a single 24 h timepoint (after impact for TBI animals, after injection for controls). Groups: TBI/IONP+ (M1–M3), Control/IONP+ (C1–C3), TBI/IONP– (M1–M3; MRI directories PB_M2–PB_M4, see below), and Control/IONP– (PB_C1–PB_C3). This experiment corresponds to Figure 5 in the manuscript.

TBI/IONP– labels: MRI directories PB_M2, PB_M3, and PB_M4 correspond to TBI/IONP– animals M1, M2, and M3 as labeled in the manuscript and in Fig5D_T2star_by_region.csv.

Fig5D_T2star_by_region.csv — T2* values in four regions for all 12 animals, with the path of each animal's MRI folder.

Fig5E_MPI_pSNR_total_signal.csv — Segment statistics of the 3 mm-diameter head ROI for the six IONP+ animals (voxel count, volume, mean, maximum, SD, total signal), the background standard deviation, and the derived pSNR and mSNR, with the paths of the source MPI series and CT file. pSNR and total signal are the values plotted in Figure 5E.

Fig5D_one_way_ANOVA_results.csv — One-way ANOVA table for each of the four T2* regions (F, degrees of freedom, p, R², sums of squares, Brown–Forsythe test of equal variances), one row per region, as computed by GraphPad Prism from Fig5D_T2star_by_region.csv.

Fig5D_Tukey_multiple_comparisons.csv — Tukey multiple-comparisons results for every pair of groups in every region (24 rows): mean difference, 95% confidence interval, q statistic and Tukey-adjusted p value. The column comparison_described_in_manuscript marks the two pairs discussed in the manuscript.

Fig5E_unpaired_t_test_results.csv — Unpaired t-test results for pSNR, total MPI signal and mSNR (TBI/IONP+ vs Control/IONP+), one row per metric: group means and SDs, mean difference with 95% confidence interval, t, df, p value and the F test of equal variances, as computed by GraphPad Prism from Fig5E_MPI_pSNR_total_signal.csv.

CT/ — One CT DICOM file per IONP+ animal: C1.dcm, C2.dcm, C3.dcm, M1.dcm, M2.dcm, M3.dcm. IONP– animals did not undergo CT.

MPI/ — 3D MPI DICOM series (sharpened reconstruction) for IONP+ animals only (C1–C3, M1–M3). Each animal folder contains a numbered DICOM series representing a single 3D volume. IONP– animals did not undergo MPI.

MRI/ — Bruker/ParaVision 6.0.1 native format MRI data for all 12 animals (RARE and MGE acquisitions; see Missing, Excluded, or Unused Data for the diffusion-weighted scans). Contains:

• Description.csv — Table mapping each subject's Bruker directory to subject ID, group, treatment, and sequence directory numbers. The directories listed in its DtiEpi_46_dir and DtiEpi_23_dir columns are not present in the deposit (diffusion-weighted scans, not used in the study); the RARE_dir and MGE_dir directories are.
• README.txt — AMRIS facility notes, including the acknowledgment wording requested by the facility.
• 12 subject directories named 20[YYYYMMDD]_[ID]_MPI_TBI_[animal]_1_1/ (or MBI_TBI for three Control/IONP– directories — naming typo; see note below).

Note on directory names: Three Control/IONP– MRI directories use MBI_TBI in place of MPI_TBI. The affected directories are 20250217_090236_MBI_TBI_PB_C1_1_1, 20250217_105309_MBI_TBI_PB_C2_1_1, and 20250217_122740_MBI_TBI_PB_C3_1_1. MBI is a naming error made at the scanner (confirmed by the MRI co-author); MPI_TBI was intended. The directories are kept under their original instrument-generated names.

Within each subject directory, Bruker acquisition subdirectories are numbered (see Description.csv). The pdata/ subdirectory within each acquisition contains the reconstructed image (2dseq) and associated metadata.

---

File Naming Patterns

CT and MPI DICOM (Experiment 1)

t[#]_[ID]_[modality].dcm               ← CT: one file per animal per timepoint
t[#]_[ID]/                             ← MPI: folder holding one 3D DICOM series
t[#]_[ID]_Std_3D_[####].dcm            ← MPI: one slice of that series

CT and MPI DICOM (Experiment 2)

[ID].dcm                                ← CT: one file per animal
[ID]_Std_3D_24h_[####].dcm             ← MPI: one slice of a 3D DICOM series

RELAX Scan Folders

[YYYY-MM-DD]_[HH-MM-SS]_0[##]_[ExamID]/

---

File Descriptions

---

Data Dictionary

Particle_Characterization/DLS_SuppFig4_histogram.csv

Particle_Characterization/DLS_SuppFig4_lognormal_fits.csv

Fit parameters (lognormal distribution with median diameter D and geometric width ln σ; subscript A is the main population and B the minor population; Φ_A is the fraction of the main population):

Particle_Characterization/TEM_SuppFig3_particle_diameters.csv

Particle_Characterization/TEM_SuppFig3_histogram.csv

Particle_Characterization/Magnetometry_SuppFig5_MH_measured.csv

Particle_Characterization/Magnetometry_SuppFig5_LangevinChantrell_fit_curve.csv

Particle_Characterization/Magnetometry_SuppFig5_magnetic_diameter_distribution.csv

Particle_Characterization/Magnetometry_SuppFig5_fit_parameters.csv

Long format: one row per parameter with columns parameter, value, units, definition. Parameters: sample_id; phi (calculated magnetic fraction, dimensionless); Dm_median (median magnetic diameter, nm); ln_sigma (natural-log geometric standard deviation of the lognormal distribution); Ms (saturation magnetization, A·m²/kg Fe, iron mass computed assuming magnetite); R_squared (0.999, as reported in the manuscript).

relaxometer_positive.csv / relaxometer_negative.csv

These files have a comment-header section (lines beginning with #) followed by 3201 data rows.

Header fields (from # comment lines):

Data columns:

Field range: −0.160 T to +0.160 T (−160 mT to +160 mT), 3201 equally spaced points.

To compute specific signal: divide signal (a.u.) by sample Fe mass in mg. For this dataset, Fe mass = 0.008711 mg (8.711 µg Fe).

Particle_Characterization/RELAX/RELAX_PSF_specific_signal_SuppFig6.csv

3201 rows; one row per field point.

Experiment1_LongitudinalStudy/Fig4B_MPI_signal_by_animal.csv

Negative values (control animals at late timepoints) arise from the image reconstruction algorithm and correspond to background signal; they are retained as exported.

Experiment1_LongitudinalStudy/Fig4B_one_phase_decay_fit_parameters.csv

Experiment1_LongitudinalStudy/Fig4B_one_phase_decay_fit_curve.csv

Experiment2_MPI_MRI/Fig5D_T2star_by_region.csv

Because animal_id repeats across groups, animal_id plus group (or mri_source_folder) uniquely identifies a row.

Experiment2_MPI_MRI/Fig5E_MPI_pSNR_total_signal.csv

Experiment2_MPI_MRI/Fig5D_one_way_ANOVA_results.csv

One row per T2* region (cortex, muscle, ipsilateral, contralateral). All values are as reported by GraphPad Prism, rounded to five significant figures.

Experiment2_MPI_MRI/Fig5D_Tukey_multiple_comparisons.csv

One row per group pair per region (6 pairs × 4 regions). Units of T2* differences are ms, as in Fig5D_T2star_by_region.csv.

Experiment2_MPI_MRI/Fig5E_unpaired_t_test_results.csv

One row per metric (psnr, msnr, total_signal_au), each tested TBI/IONP+ vs Control/IONP+ (n = 3 per group). Units follow the metric: dimensionless for pSNR and mSNR, a.u. for total signal.

---

Relationship Between Source, Processed, Converted, and Summary Files

---

Relationship to Manuscript Figures and Tables

---

Recommended Software

---

Missing, Excluded, or Unused Data

The following data were acquired but are not deposited:

• Diffusion-weighted MRI: two diffusion-weighted (DTI-EPI) acquisitions per subject (46- and 23-direction shells; Bruker directories listed in the DtiEpi_46_dir and DtiEpi_23_dir columns of MRI/Description.csv) were acquired in the same sessions as the deposited RARE and MGE scans but were not analysed for the study and are not deposited; they are retained internally and available from the corresponding author.
• Retained internally: complete scanner project directories (MOMENTUM), RELAX instrument PDFs and parameter files, and intermediate 3D Slicer scene files are kept in the Rinaldi-Ramos Lab archive and are available from the corresponding author on request.
• Excluded animal: One TBI animal in Experiment 1 was excluded because of a compromised tail-vein injection (see "Excluded animal and labels in the MPI DICOM files"); no data from this animal are deposited.

---

Quality Control

• MPI background noise: The background standard deviation used for signal-to-noise ratios in both experiments was 0.008547 a.u., the average voxel standard deviation of 3D MPI scans of the empty scanner bore acquired with the same scan settings as the animal images.
• RELAX: Five replicate measurements were collected for particle characterization. FWHM agreement across replicates: positive sweep 8.74 ± 0.22 mT; negative sweep 8.71 ± 0.19 mT (mean ± SD). Results consistent across replicates indicate stable instrument performance.
• MPI DICOM series: Each 3D MPI acquisition is a complete DICOM series that should be loaded as a single volume.

---

Access, Licensing, and Sensitive Information

The deposited files do not contain human participant identifiers, protected health information, password restrictions, or third-party copyrighted material. All animal studies were conducted in accordance with IACUC protocol #202300000294, approved by the University of Florida Institutional Animal Care and Use Committee.

Data are provided under the license terms used by Dryad.

---

MRI Facility Acknowledgment

The MRI data were acquired at the McKnight Brain Institute at the National High Magnetic Field Laboratory's Advanced Magnetic Resonance Imaging and Spectroscopy (AMRIS) Facility. In keeping with the facility's guidance, we ask that publications using the MRI data acknowledge the facility and its support with the following statements. This is a request, not a condition of use; the data are released under the CC0 waiver.

"A portion of this work was performed in the McKnight Brain Institute at the National High Magnetic Field Laboratory's Advanced Magnetic Resonance Imaging and Spectroscopy (AMRIS) Facility, which is supported by National Science Foundation Cooperative Agreement DMR-2128556 and the State of Florida."

"This work was supported in part by an NIH award, S10 RR025671, for MRI/S instrumentation."

---

AI Assistance Disclosure

An AI language model (Claude, Anthropic) was used to assist with drafting and organizing this README based on author-provided repository contents, file descriptions, metadata, and methods information. The dataset authors reviewed and edited the README and are responsible for the accuracy of all file descriptions, variable definitions, and metadata.

---

Dataset Versions