Ultrasound neural interface enabling acute spinal cord monitoring and blood flow modulation

Kerensky, Max 1 ; Krishnan, Siddharth1; Liang, Ruixing 1 ; Kempski Leadingham, Kelley2; Quiroz, Victor 1 ; Punnoose, Joshua 2 ; Smit, Constantin2; Routkevitch, Denis 1 ; Kumar, Avisha1; Manbachi, Amir 2

Affiliations

Contributions

Max Kerensky: Writing – original draft, writing – review & editing, conceptualization, investigation, methodology, data curation, supervision, software, project administration. Siddharth Krishnan: Writing – original draft, writing – review & editing, investigation, methodology, data curation, software, conceptualization, project administration, supervision. Ruixing Liang: Writing – original draft, writing – review & editing, conceptualization, investigation, methodology, data curation, supervision, project administration, software. Kelley Kempski Leadingham: Writing – original draft, writing – review & editing, investigation, methodology, software, supervision, project administration, data curation, conceptualization. Victor Quiroz: Writing – original draft, conceptualization, investigation, writing – review & editing, methodology, data curation, project administration. Joshua Punnoose: Conceptualization, investigation, methodology, software. Constantin Smit: Methodology, conceptualization, investigation. Denis Routkevitch: Conceptualization, investigation, writing – review & editing, methodology, software. Avisha Kumar: Writing – original draft, methodology. A. Daniel Davidar: Writing – original draft, writing – review & editing, conceptualization, investigation, methodology, project administration, supervision. Leonard Steger: Writing – original draft, software. Carly Weber-Levine: Conceptualization, investigation, writing – review & editing, methodology, data curation. Kelly Jiang: Conceptualization, investigation, writing – review & editing, methodology. Meghana Bhimreddy: Investigation, writing – review & editing, methodology, data curation. Stuart Bauer: Investigation, data curation, methodology. Angelica Lopez: Investigation, writing – review & editing, methodology, data curation, software. Patrick Kramer: Writing – original draft, investigation, writing – review & editing, methodology, data curation. Kayla Robinson: Methodology, software. Arjun Menta: Conceptualization, investigation, writing – review & editing, methodology, data curation. Abdul Karim Ghaith: Writing – original draft, writing – review & editing, conceptualization, investigation, methodology, data curation, supervision, software. Andrew Hersh: Conceptualization, writing – original draft, writing – review & editing, investigation, methodology, data curation, supervision. Haley Abramson: Conceptualization, investigation, writing – review & editing, methodology, data curation. Ali Mohammadabadi: Investigation, software. John Theodore: Writing – original draft, investigation, data curation. Osamah Zmily: Investigation, writing – review & editing, data curation. Ritvik Jillala: Writing – review & editing, investigation. Vikas Vattipally: Investigation, writing – review & editing. David Kopala: Software, conceptualization. Siavash Shirzadeh Barough: Writing – review & editing, methodology, software, data curation. Mohammad Hossein Amini: Supervision, methodology. Jonathan Maltz: Writing – original draft, writing – review & editing, conceptualization, investigation, methodology, data curation, software, supervision. Francisco Chavez: Methodology, software, investigation. Kyle Morrison: Supervision, methodology, writing – review & editing. Laura Roy: Conceptualization, software, writing – review & editing. Ian Suk: Conceptualization, software. Ryan Lee: Writing – review & editing, data curation. Betty Tyler: Supervision, investigation, writing – review & editing. Mark Luciano: Supervision. Nitish Thakor: Conceptualization, writing – review & editing, methodology, supervision. Joshua Doloff: Investigation, writing – review & editing, methodology, supervision, project administration. Nicholas Theodore: Writing – original draft, conceptualization, project administration, investigation, writing – review & editing, methodology, supervision. Amir Manbachi: Conceptualization, writing – review & editing, writing – original draft, investigation, methodology, data curation, supervision, project administration, software

Published Oct 01, 2026 on Dryad. https://doi.org/10.5061/dryad.nk98sf891

Data files

Oct 01, 2026 version files 8.93 MB

MUSIC_FUS_data.zip

8.91 MB
README.md

20.20 KB

Abstract

The days following decompression surgery for acute spinal cord injury (SCI) are a critical period for postoperative monitoring. After wound closure, clinicians lack on-demand access to local spinal cord morphology and hemodynamics, and care therefore remains guided largely by systemic surrogates rather than direct postoperative insight into the injured cord. We present the Multimodal Ultrasound System for Injury Care (MUSIC), a temporary epidural ultrasound neural interface placed during standard decompression surgery to enable postoperative B-mode imaging, Doppler-derived hemodynamic monitoring, and future device-integrated ultrasound actuation. Across simulations, benchtop, human cadaveric, and porcine studies, MUSIC enabled on-demand visualization of cord structure, hematoma tracking, detection and longitudinal tracking of localized Doppler flow signatures attributable to the sulcal artery region, and, in healthy pigs, proof-of-concept focused ultrasound modulation of blood-flow velocity. MUSIC establishes a temporary postoperative acoustic interface that provides spinal cord information to inform monitoring, future management, and localized intervention strategies after SCI.

Dryad DOI: https://doi.org/10.5061/dryad.nk98sf891

Associated manuscript: Ultrasound neural interface enabling acute spinal cord monitoring and blood flow modulation
Journal: Science Advances
Manuscript ID: aeg2342

Dataset summary

This Dryad dataset contains the experimental spectral-Doppler time-series data used for the focused-ultrasound (FUS) blood-flow-velocity (BFV) analyses in the associated study of the Multimodal Ultrasound System for Injury Care (MUSIC).

The study evaluated whether MUSIC-delivered focused ultrasound could modulate Doppler-derived spinal cord blood-flow velocity in healthy porcine experiments. Across three porcine experiments, FUS parameters were varied, including input voltage, pulse repetition frequency (PRF), pulse/burst duration, and insonation duration. The analyses compared FUS-off spectral-Doppler signals immediately before and after FUS delivery. FUS-on intervals were generally excluded from the main pre/post analysis because the FUS pulses could generate spectral interference.

The uploaded archive is:

MUSIC_FUS_data.zip

It contains two organizational views of the same underlying data:

Original/: the canonical source organization, preserving the chronological experimental sections.
Further_class/: byte-identical copies of the Original/ MATLAB files reorganized by experimental parameter grouping for convenient downstream analysis.

Important: Further_class/ does not contain additional independent observations. It is a reorganized copy of the source data in Original/. The archive contains 38 source .mat files in Original/ and 78 reorganized copies in Further_class/. Hash comparison confirms that every .mat file in Further_class/ is identical to one of the 38 source files in Original/.

The 38 source files define 30 repeated FUS sessions across three porcine experiments. A FUS session is represented by two adjacent FUS-off recordings: the end of one recording is used as the pre-FUS endpoint and the beginning of the next recording as the post-FUS endpoint.

Relationship to the manuscript

These data support the proof-of-concept FUS/BFV analyses described in the main manuscript and Supplementary Materials, including the parameter-sweep and pre/post-FUS analyses associated with main Figure 5 and Supplementary Figures S9-S12.

The associated study interprets these measurements as Doppler-derived local hemodynamic measurements. They should not be interpreted as direct measurements of vessel diameter, volumetric blood flow, vasodilation, tissue perfusion, or therapeutic efficacy.

Description of the data and file structure

Top-level structure

MUSIC_FUS_data/
├── Original/
│   ├── 0307/
│   │   └── Replicate/
│   ├── 0627/
│   │   ├── a.5KHz Control Section/
│   │   ├── b. Duty Cycle Control Section/
│   │   └── c. Long FUS Section/
│   ├── 0729/
│   │   ├── a. MAX Section/
│   │   ├── b. Replicate Section/
│   │   └── c. Long FUS Section/
│   └── FUS6.mat
└── Further_class/
    ├── 3.7/
    ├── 6.27/
    ├── 6.27duty/
    ├── 6.27voltage/
    ├── 7.29/
    └── 7.29duration/

The labels 0307, 0627, and 0729 identify the three porcine experimental sessions in the archived analysis workflow. The corresponding parameter-organized labels are 3.7, 6.27, and 7.29.

`Original/`: canonical chronological source data

The Original/ directory contains the 38 source MATLAB files used to construct the analysis dataset.

Folder	Number of `.mat` files	Number of adjacent pre/post FUS sessions	Experimental context
`0307/Replicate/`	5	4	Replicate FUS trials from the first porcine experiment. Parameter-organized copies divide these into 1-kHz and 5-kHz groups at 60 V and a `.07ms` setting.
`0627/a.5KHz Control Section/`	7	6	Input-voltage/pressure-dose section from the second porcine experiment. Parameter-organized copies group adjacent trials at 20 V, 30 V, and 45 V with a 5-kHz and `.07ms` setting.
`0627/b. Duty Cycle Control Section/`	7	6	Pulse/burst-duration or duty-dose section from the second porcine experiment. Parameter-organized copies group adjacent trials at `.15ms`, `.2ms`, and `.4ms` settings with 1 kHz and 40 V.
`0627/c. Long FUS Section/`	2	1	Single longer-duration FUS comparison from the second porcine experiment.
`0729/a. MAX Section/`	7	6	Experimenter-labeled `MAX` section from the third porcine experiment. Parameter-organized copies include `1kHz 20V .35ms` and `5kHz 20V .4ms` groups and selected 2-minute comparisons.
`0729/b. Replicate Section/`	7	6	Replicate FUS trials from the third porcine experiment. Parameter-organized copies divide the sequence into 1-kHz and 5-kHz groups at 20 V and a `.07ms` setting.
`0729/c. Long FUS Section/`	2	1	Long-duration FUS comparison from the third porcine experiment; the reorganized duration folder labels this comparison `10min`.
`Original/FUS6.mat`	1	—	Source recording used as the final boundary recording for the 0627 voltage-series reorganization. It is paired with the preceding recording when forming the final voltage-comparison session.

Total: 38 source MATLAB files defining 30 adjacent pre/post FUS sessions.

How adjacent files define a FUS session

Within each chronological section, files are ordered baseline/recovery recordings around FUS applications.

For a sequence such as:

FUSoff0.mat
FUSoff1.mat
FUSoff2.mat

the analysis treats:

the end of FUSoff0.mat as the pre-FUS endpoint for session 1;
the beginning of FUSoff1.mat as the post-FUS endpoint for session 1;
the end of FUSoff1.mat as the pre-FUS endpoint for session 2; and
the beginning of FUSoff2.mat as the post-FUS endpoint for session 2.

Thus, n + 1 ordered FUS-off recordings define n adjacent FUS sessions.

Some legacy source files in Original/ are named FUS0.mat, FUS1.mat, etc. Their parameter-organized copies in Further_class/ are used as FUS-off boundary recordings in the analysis. Users should therefore rely on the chronological ordering and the parameter-organized grouping rather than interpret the legacy FUS prefix as indicating that the archived time series itself represents a FUS-on interval.

`Further_class/`: parameter-organized copies

Further_class/ contains no new measurements. It reorganizes copies of the source .mat files so that adjacent recordings corresponding to a particular parameter comparison are located together.

`Further_class/3.7/`

3.7 1kHz 60V .07ms/: three ordered recordings defining two adjacent sessions.
3.7 5kHz 60V .07ms/: three ordered recordings defining two adjacent sessions.

These files are copies of Original/0307/Replicate/.

`Further_class/6.27/`

This directory provides grouped parameter sequences from the second porcine experiment:

6.27 5kHz 20V .07ms/
6.27 5kHz 30V .07ms/
6.27 5kHz 45V .07ms/
6.27 1kHz 40V .15ms/
6.27 1kHz 40V .2ms/
6.27 1kHz 40V .4ms/
6.27 Long FUS Section/

The numeric strings in these folder names are experimental-setting labels:

kHz: pulse repetition frequency (PRF) label, in kilohertz;
V: FUS system input-voltage setting, in volts;
ms: pulse/burst-duration setting, in milliseconds.

`Further_class/6.27voltage/`

This folder expands the voltage-series data into individual adjacent pre/post pairs:

20-V trials: folders labeled 20V .07ms and 20V .07ms_1;
30-V trials: folders labeled 30V .07ms and 30V .07ms_1;
45-V trials: folders labeled 45V .07ms and 45V .07ms_1.

Each two-file folder defines one adjacent FUS session. The suffixes _1 and _2 are replicate/trial identifiers and do not represent physical units.

`Further_class/6.27duty/`

This folder expands the pulse/burst-duration series into individual adjacent pre/post pairs:

.15ms and .15ms_2;
.2ms and .2ms_2;
.4ms and .4ms_2.

Each two-file folder defines one adjacent FUS session. These comparisons were used to investigate dose dependence related to pulse/burst-duration and duty-cycle settings.

`Further_class/7.29/`

This directory provides parameter-organized copies from the third porcine experiment:

7.29 1kHz 20V .07ms/
7.29 5kHz 20V .07ms/
7.29 Max 1kHz 20V .35ms/
7.29 Max 5kHz 20V .4ms/
7.29 Short FUS 2min/
7.29 Short FUS 2min_2/
7.29 Long FUS Section/

MAX is retained as the experimenter-defined section label. The actual numeric experimental-setting labels are preserved in the folder names.

`Further_class/7.29duration/`

This folder isolates the duration comparison:

7.29 Short FUS 2min_1/: two recordings defining a 2-minute comparison session.
7.29 zLong FUS 10min/: two recordings defining a 10-minute comparison session.

Two additional .mat files at the root of 7.29duration/ are duplicate copies of the same source recordings used in the short-duration comparison and do not constitute additional sessions.

MATLAB file contents and variable dictionary

Each .mat file contains one primary one-dimensional spectral-Doppler envelope time series plus an extraction-threshold scalar and MATLAB object metadata.

Two naming conventions occur because the source data were exported in different experimental sessions. US_chunk and US1 represent the same logical type of signal; similarly, thresh and thresh1 represent the same logical type of saved extraction parameter.

Variable	Type / shape	Units or coding	Meaning
`US_chunk`	1 × N numeric array (`float64`)	Hz	Extracted spectral-Doppler envelope/frequency-shift time series. This is the primary analysis signal in files using the `US_chunk` naming convention.
`US1`	1 × N numeric array (`float64`)	Hz	Same logical signal type as `US_chunk`, used in a subset of source files created under a different export naming convention.
`thresh`	1 × 1 numeric scalar (`float64`)	Unitless extraction parameter	Threshold value saved with files using `US_chunk`. In this archive values range from 0.05 to 0.775. It is retained for provenance; the archived Python pre/post analysis does not use this scalar as an analysis outcome.
`thresh1`	1 × 1 numeric scalar (`float64`)	Unitless extraction parameter	Same logical type as `thresh`, paired with files using `US1`.
`None`	MATLAB object metadata	Not an analysis variable	Serialization metadata associated with the MATLAB datetime object saved during export (`t_US_chunk`). It is not used in the archived Python BFV analysis and should not be interpreted as a physiological variable.

Across the canonical Original/ source files:

26 files store the primary signal as US_chunk;
12 files store the primary signal as US1.

The length of the primary signal varies by recording because recording duration varied across experimental sections. In the canonical source files, array lengths range from 5,886 to 25,836 samples.

Sampling and time windows

The archived analysis workflow uses 1,000 samples for the 20-second pre-FUS and post-FUS comparison windows, corresponding to an exported envelope sampling rate of approximately 50 samples/s (50 Hz).

For each FUS session:

pre-FUS window: the final 1,000 samples (20 s) of the preceding FUS-off recording;
post-FUS window: the first 1,000 samples (20 s) of the following FUS-off recording.

For selected supplementary analyses, expanded approximately 30-second FUS-off windows were used as described in the manuscript.

Interpretation of the primary signal

US_chunk/US1 contains the extracted spectral-Doppler envelope expressed as Doppler frequency shift in Hz. The manuscript uses these spectral-Doppler data to assess relative and normalized changes in Doppler-derived blood-flow velocity across pre- and post-FUS conditions.

The archived frequency-shift series should not be interpreted as a direct measurement of volumetric blood flow or tissue perfusion.

File-name and folder-name interpretation

Label/pattern	Interpretation
`0307`, `0627`, `0729`	Internal experimental-session/date labels corresponding to the three porcine experiments.
`3.7`, `6.27`, `7.29`	Parameter-organized versions of the same three experimental-session labels.
`FUSoffN`, `offN`	Ordered FUS-off baseline/recovery recording. `N` indicates position in the experimental sequence.
`FUSN` in selected `Original/` folders	Legacy source filename. In the reorganized analysis structure, these files are used as FUS-off boundary recordings.
`1kHz`, `5kHz`	PRF setting label in kilohertz.
`20V`, `30V`, `40V`, `45V`, `60V`	FUS system input-voltage setting in volts.
`.07ms`, `.15ms`, `.2ms`, `.35ms`, `.4ms`	Pulse/burst-duration setting label in milliseconds.
`2min`, `10min`	FUS insonation-duration label in minutes.
`_1`, `_2`	Replicate/trial grouping suffix; no physical unit.
`Replicate`	Experimental section containing repeated FUS trials.
`Duty Cycle Control Section`	Experimental section used to compare pulse/burst-duration and related duty-dose settings.
`Long FUS Section`	Experimental section used for longer-duration FUS comparison.
`MAX Section` / `Max`	Experimenter-defined section label retained from the source dataset; numeric settings are provided separately in parameter-organized folder names.

Missing/invalid values and analysis handling

The .mat files do not use a dedicated textual missing-value code.

In the archived Python analysis workflow:

zero-valued points in the extracted envelope can be treated as invalid/missing envelope samples;
invalid points are converted to missing values and interpolated;
extreme outliers and abrupt signal jumps may be removed/interpolated as part of preprocessing;
smoothing and pre/post window extraction are then applied for the relevant analyses.

Because exact preprocessing differs by analysis/figure, users seeking to reproduce the manuscript should use the archived publication-version software listed below rather than reimplement preprocessing solely from this README.

Experimental context and analysis logic

The focused-ultrasound experiments were conducted in three healthy porcine experiments. During each experimental section, FUS parameters were varied to evaluate whether dose-related settings were associated with changes in local Doppler-derived BFV.

The principal experimental variables included:

FUS system input voltage;
pulse repetition frequency (PRF);
pulse/burst-duration or duty-related settings; and
insonation duration.

A typical multi-session section consisted of repeated approximately 2-minute FUS sessions separated by FUS-off baseline/recovery recordings. Selected sections included a longer approximately 10-minute FUS exposure. Recovery time was provided between experimental sections.

The main pre/post analysis uses only FUS-off windows immediately before and after therapy. This avoids the harmonic interference that can be present in spectral-Doppler recordings during FUS delivery.

For a session-level percentage-change calculation, the archived software computes:

100 × (mean_post - mean_pre) / mean_pre

where mean_pre and mean_post are calculated from the selected pre- and post-FUS windows.

These repeated sessions are within-animal observations and should not be treated as 30 independent animals.

Sharing/access information

Permanent experimental-data archive

Dryad:
https://doi.org/10.5061/dryad.nk98sf891

Permanent publication-version software archives

MUSIC FUS Analysis
Zenodo: https://doi.org/10.5281/zenodo.22907737
GitHub development repository: https://github.com/HEPIUSLAB/MUSIC_FUS_Analysis

HEPIUS MUSIC Imaging
Zenodo: https://doi.org/10.5281/zenodo.22908234
GitHub development repository: https://github.com/HEPIUSLAB/HEPIUS_MUSIC_imaging

MUSIC Mobile App
Zenodo: https://doi.org/10.5281/zenodo.22908477
GitHub development repository: https://github.com/HEPIUSLAB/MUSIC_APP

HEPIUS Lab GitHub organization:
https://github.com/HEPIUSLAB

Code/software

The software used to organize, preprocess, analyze, and visualize the FUS spectral-Doppler data is archived permanently in the MUSIC FUS Analysis Zenodo record:

https://doi.org/10.5281/zenodo.22907737

The archived software includes Python source code and Jupyter notebooks. The analysis code recognizes either US_chunk or US1 as the primary spectral-Doppler envelope variable and implements the pre/post pairing logic described above.

The GitHub repository is maintained as a development resource; the Zenodo DOI should be used when citing the publication-version software.

Licensing

Data deposited in Dryad are released under Dryad's applicable CC0 waiver.

Software is licensed separately according to the license stated in each Zenodo software record.

Abbreviations

BFV: blood-flow velocity
FUS: focused ultrasound
MUSIC: Multimodal Ultrasound System for Injury Care
PRF: pulse repetition frequency
RI: resistivity index
SCI: spinal cord injury