Associated manuscript: Yifan Zhao, Tao Jin, and Bistra Iordanova. Saturation power dispersion analysis for the imaging of amide proton transfer (APT) and nuclear Overhauser effect (NOE).

1. Overview

This dataset contains phantom and in vivo rat MRI data for investigating amide proton transfer (APT) and nuclear Overhauser effect (NOE) through saturation-power dispersion analysis. The measurements were acquired on a Bruker 9.4 T MRI scanner. The analysis uses the apparent relaxation ratio (ARR) to separate a narrow APT or NOE component from broader background contributions.

The release contains CSV tables of normalized signals and ARR, phantom image datasets with .hdr, .img, and .mat companion files, in vivo MATLAB image arrays and ADC maps, fitted parameter maps, MATLAB data and analysis scripts, and region-of-interest (ROI) definitions. The scientific descriptions and acquisition parameters below follow the associated manuscript. File descriptions and counts refer to the contents of this upload folder.

2. Authors and contact

• Yifan Zhao — Department of Radiology, University of Pittsburgh, Pittsburgh, Pennsylvania, USA.
• Tao Jin — Department of Radiology, University of Pittsburgh, Pittsburgh, Pennsylvania, USA. Corresponding author: taj6@pitt.edu.
• Bistra Iordanova — Department of Bioengineering, University of Pittsburgh, Pittsburgh, Pennsylvania, USA.

3. Files and folder structure

All 11 CSV files are located in the dataset root. The current image and ROI data are stored in the Phantom/ and In_vivo/ directories. Folder names containing 2dseq are retained, although the supplied images now use .hdr/.img pairs with companion .mat files where provided.

Data_to_uplaod/
├── README.md
├── B1_dispersion_tabular.csv
├── B1_Disp_data_phantom__BSA_diff_pH_3.4ppm.csv
├── B1_Disp_data_phantom__Diff_labeile_proton_-3.6ppm.csv
├── B1_Disp_data_phantom__Diff_labeile_proton_3.6ppm.csv
├── B1_Disp_data_phantom__EWA_diff_pH_3.6ppm.csv
├── Full_Z_spectrum_tabular.csv
├── Full_Z_spectrum_different_labile_proton__Sheet1.csv
├── in_vivo_tabular.csv
├── in_vivo_data__-3.6ppm.csv
├── in_vivo_data__3.6ppm.csv
├── in_vivo_data__6ppm.csv
├── Phantom.zip/
│   ├── Phantom_Different_labile_proton/
│   │   ├── B1_disp/
│   │   │   ├── Original_2dseq_data/
│   │   │   └── ROIs/
│   │   └── Full_spectrum_different_Power/
│   │       ├── Original_2dseq_data/
│   │       └── ROIs/
│   └── Phantom_Different_pH/
│       ├── Nic/
│       │   ├── Original_2dseq_data/
│       │   └── ROIs/
│       └── EWA/
│           ├── Origianl_2dseq_data/
│           └── ROIs/
└── In_vivo.zip/
    ├── Original_2dseq_data/
    │   └── <dataset_id>/
    │       ├── ADC_map.hdr / ADC_map.img / ADC_map.mat
    │       ├── Original_circle2_<dataset_id>_slice1.mat
    │       ├── Original_circle2_<dataset_id>_slice2.mat
    │       ├── 36ppm/
    │       │   ├── ARR_36ppm_<parameter>_map_slice<slice>_<dataset_id>_circle2.*
    │       │   └── S_sat_to_ARRmap_36ppm.m
    │       ├── n36ppm/
    │       │   ├── ARR_n36ppm_<parameter>_map_slice<slice>_<dataset_id>_circle2.*
    │       │   └── S_sat_to_ARRmap_n36ppm.m
    │       └── 6ppm/
    │           ├── ARR_6ppm_<parameter>_map_slice<slice>_<dataset_id>_circle2.*
    │           └── S_sat_to_ARRmap_6ppm.m
    └── ROIs/
        ├── ROI_Stroke_Normal/
        │   ├── Slice1/
        │   └── Slice2/
        └── Whole_brain_ROI/

The in vivo tree shows a repeated subject layout. <dataset_id> is one of 20241216, 20250212, 20250224, 20250228, 20250303, or 20250310; <slice> is 1 or 2. Map extensions represented by .* are .hdr, .img, and .mat. Slash-separated extensions in the tree represent separate companion files.

For numerical analysis, begin with the three _tabular.csv files. Each has a single header row and one measurement per subsequent row. Other CSV files use sample blocks with variable labels and blank separator rows. They provide corresponding measurements in a different layout and should not be counted as additional independent observations.

4. MRI acquisition summary

These are manuscript acquisition parameters; exported or processed image dimensions may differ. The in vivo files distinguish two slices.

The deposited dispersion tables contain 15 saturation-field amplitudes, in µT:

0.30, 0.40, 0.50, 0.65, 0.80, 1.00, 1.25, 1.50, 1.80, 2.20, 2.60, 3.00, 3.40, 3.80, 4.20.

In vivo dispersion measurements were acquired at +3.6, −3.6, and +6 ppm, with 12, 8, and 4 averages per power, respectively. The manuscript's enumerated power list additionally includes 2.4 µT while describing 15 values; this value is absent from the supplied tables.

5. Phantom data

5.1 Samples and experiments

The different-labile-proton experiment used five samples prepared in 1× phosphate-buffered saline (PBS), measured at room temperature:

Z-spectra were acquired from −6 to +6 ppm at unevenly spaced offsets, using 0.5, 1.0, 2.1, and 4.2 µT saturation amplitudes. Power-dispersion measurements were acquired at +3.6 and −3.6 ppm.

For the pH-series experiments, the manuscript describes 150 mM Nic at +3.4 ppm and 15% EWA at +3.6 ppm, each at pH 6.25, 6.50, 6.75, 7.00, and 7.25. Both series were prepared in PBS and measured at room temperature. The +3.4 ppm CSV filename and table labels retain BSA; the corresponding ROI files in Phantom_Different_pH/Nic/ROIs/ are named ROI_Nic_ph625.roi, ROI_Nic_pH650.roi, ROI_Nic_pH675.roi, ROI_Nic_pH700.roi, and ROI_Nic_pH725.roi.

Phantom ROIs were selected in regions with B₀ inhomogeneity below 0.05 ppm.

5.2 Power-dispersion tables

B1_dispersion_tabular.csv contains 300 measurements: five sample types at each of two offsets, plus two five-point pH series, each sampled at 15 powers.

Blank composition/preparation fields indicate that a separate label was not supplied. The sample and source labels provide additional context.

The four B1_Disp_data_phantom__*.csv files preserve block layouts:

Rows labeled w(uT) or Power(uT) contain the saturation amplitudes; S_sat/S_0 or S_sat/S0 contain normalized signals; and ARR contains the corresponding ARR values. Values above 1 for normalized signal and negative ARR values are retained as supplied.

5.3 Z-spectrum tables

Full_Z_spectrum_tabular.csv contains 165 measurements: five samples × 33 RF offsets. Its columns are sample_label, RF_offset_ppm (ppm), and S_sat_over_S0 (dimensionless).

Full_Z_spectrum_different_labile_proton__Sheet1.csv organizes these measurements into five blocks. Sample headers occur in rows 1, 6, 11, 16, and 21, followed by RF-offset and normalized-signal rows. Measurements occupy columns B–AH.

The CSVs contain one spectrum per sample and do not specify its B₁ amplitude. The associated image folder contains four power-coded image datasets with basenames S_CEST_050cw_norm, S_CEST_100cw_norm, S_CEST_210cw_norm, and S_CEST_420cw_norm. Each dataset comprises matching .hdr, .img, and .mat files.

5.4 Image and ROI files

Each image dataset is supplied as a matching .hdr, .img, and .mat set. The phantom folders therefore contain 35 image datasets (105 companion files) and 20 ROI files in total. Image files are located in the Original_2dseq_data/ subfolders, except for the EWA pH series, which uses the supplied spelling Origianl_2dseq_data/.

The dispersion and Nic pH-series image basenames follow S_CEST_<power>cw_norm, with codes such as 030, 100, and 420 indicating 0.30, 1.00, and 4.20 µT. In both folders, the 0.80 µT dataset is named S_CEST_080cw_norm9; retain the complete basename when opening its companion files. The EWA pH-series image dataset is named S_CEST_CW_average and is supplied as one three-file set. Its dimensions and the available power information are described below.

5.5 Image dimensions and third-dimension ordering

The first two dimensions are the 64 × 64 spatial image matrix. The third dimension indexes RF offsets or saturation powers as specified below; it does not represent anatomical slices. Frame indices are one-based.

Different-labile-proton B₁ dispersion: each saturation power has a normalized S_sat/S₀ image stack of size 64 × 64 × 5. The five frames correspond to RF offsets in this exact order:

Full Z-spectrum: four saturation powers were acquired: 0.5, 1.0, 2.1, and 4.2 µT. Each power has an image stack of size 64 × 64 × 33. The 33 frames correspond to the following RF offsets, in ppm, in the order stored in the image:

[6, 5.5, 5, 4.5, 4.2, 4, 3.8, 3.6, 3.4, 3.2, 3,
 2.8, 2.6, 2.3, 2, 1.9, 1.7, 1.5, 1, 0.5, 0,
 -0.5, -1, -1.5, -2, -2.5, -3, -3.5, -4, -4.5, -5, -5.5, -6]

The image-stack order runs from +6 to −6 ppm. The Z-spectrum CSVs list offsets in ascending order, so match measurements by RF offset rather than by column/frame position.

EWA pH series: S_CEST_CW_average was acquired at a fixed RF offset of +3.6 ppm. The confirmed series contains 15 saturation powers, corresponding to an intended data array of 64 × 64 × 15. The third dimension indexes these powers in the following order, in µT:

[0.3, 0.4, 0.5, 0.65, 0.8, 1, 1.25, 1.5, 1.8, 2.2, 2.6, 3.0, 3.4, 3.8, 4.2]

File consistency note: the current .hdr file still declares 64 × 64 × 18. The acquisition description above uses the 15 powers confirmed by the data provider. The stored image/header should be reconciled with this 15-power series before frame-based analysis; no additional power values are assigned to the three extra declared frames.

Nic pH series: 15 saturation powers were acquired, using the same 0.3–4.2 µT list above. Each power has an image stack of size 64 × 64 × 2. Frames 1 and 2 correspond to RF offsets +2.6 ppm and +3.4 ppm, respectively. The +3.4 ppm pH-series measurements therefore correspond to the second frame of each power-specific image.

6. In vivo data

6.1 Animal model and ROI selection

Six female Sprague-Dawley rats underwent suture-induced permanent MCAO in the left hemisphere. Imaging was performed 2–3 hours after occlusion. Animal experiments were approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh.

ADC maps were acquired using spin-echo EPI with b = 5 s/mm² along one axis and b = 1200 s/mm² along six directions. Stroke ROIs encompassed the infarcted region on ADC maps; reflection across the brain midline defined the contralateral normal ROI.

6.2 In vivo tables

in_vivo_tabular.csv contains 540 measurements: six dataset identifiers × three RF offsets × two regions × 15 powers.

The three in_vivo_data__*.csv files contain the corresponding offset-specific values in block format. Each has 57 rows and 16 columns. Column A contains identifiers and variable names; columns B–P hold the 15 measurements.

Following each identifier, rows contain RF offsets, B₁ amplitudes, normalized signals for Normal and Stroke, and ARR for Normal and Stroke, in that order. Blank rows separate subject blocks. The tabular and block-layout in vivo values match.

6.3 Subject image and analysis files

Each folder in In_vivo/Original_2dseq_data/ contains:

• ADC_map.hdr, .img, and .mat: an ADC map and its companion files.
• Original_circle2_<dataset_id>_slice1.mat and ...slice2.mat: MATLAB files containing image variables for each anatomical slice. Each image variable has dimensions 64 × 64 × 3. The third dimension contains RF offsets in the exact order [−3.6, +3.6, +6.0] ppm: frame 1 is −3.6 ppm, frame 2 is +3.6 ppm, and frame 3 is +6 ppm. Anatomical slice identity is given by the filename, not by the third array dimension.
• Three offset folders containing fitted parameter maps and a MATLAB analysis script for that offset.

The supplied in vivo and phantom images are stored as .hdr/.img pairs with associated MATLAB files. Directory names containing 2dseq are retained from the data organization; they do not indicate the current image file format.

Across all six subjects, there are 120 fitted ARR parameter maps, each with matching .hdr, .img, and .mat files; six ADC map sets; 12 slice-specific MATLAB data files; and 18 MATLAB scripts. Preserve companion files with matching basenames.

6.4 ROI files

In_vivo/ROIs/ROI_Stroke_Normal/ contains 22 .roi files: 10 in Slice1/ and 12 in Slice2/. Filenames identify normal or stroke tissue and the dataset identifier. For 20250303, stroke and normal ROIs are supplied only for slice 2 because no stroke region was observed in slice 1.

In_vivo/ROIs/Whole_brain_ROI/ contains 13 ROI files. Twelve files provide slice-1 and slice-2 ROIs for codes 1216, 0212, 0224, 0228, 0303, and 0310, matching the month/day portions of the six dataset identifiers.

7. ARR and fitted parameters

With normalized saturated signal Z = S_sat/S₀ and B₁ in µT, manuscript Eq. (5) defines:

ARR = (1/Z − 1) / B₁²

ARR is expressed in µT⁻². Fitted amplitudes represent component values extrapolated to B₁ = 0. Widths are half widths at half maximum (HWHM), in µT, along the saturation-field-amplitude axis.

At +6 ppm, the manuscript uses a single background Lorentzian. The supplied filenames label this component Amp1 and width1; these names do not imply an APT component at +6 ppm.

8. Software and use

CSV files can be opened with spreadsheet software or standard data-analysis tools. Import dataset_id as text. Blank separator rows in block-layout CSVs are not observations.

The included .m files are MATLAB scripts. Inspection identifies calls to fit and fittype, requiring the Curve Fitting Toolbox, and to the image-loading helper load_nii. Resolve these dependencies and review the scripts' input-loading sections before running them. The distributed in vivo source arrays are now in the Original_circle2_*.mat files. Any script that still loads the removed image-stack files must be adapted to read these MATLAB variables with the RF-offset order described in Section 6.3; changing a pathname alone is not sufficient. The scripts were inspected for documentation but were not executed. Exact software versions and a complete dependency installation procedure are not supplied in the manuscript.

The .hdr/.img image pairs can be opened using ImageJ with a compatible image-format reader or plugin. Keep the matching .hdr and .img files in the same folder. Use MATLAB-compatible software for .mat files. The third-dimension frame order for phantom images is documented in Section 5.5. Apply .roi files using software that supports their ROI format, checking subject, slice, image orientation, and dimensions before analysis.

9. Funding, reuse, and citation

This work was supported by National Institutes of Health grants RF1NS116450 and R01NS134810.

Please cite the associated manuscript by Yifan Zhao, Tao Jin, and Bistra Iordanova, titled Saturation power dispersion analysis for the imaging of amide proton transfer (APT) and nuclear Overhauser effect (NOE), together with the dataset's repository citation when available.