Data

All 85 runway recordings analysed in the study are released under CC BY 4.0, together with the annotations required to reproduce the analyses.

Data deposit

The recordings are deposited on Zenodo under doi:10.5281/zenodo.23286364 (CC BY 4.0), with MD5 checksums in md5sums.txt.

Contents of the deposit

• Recordings: one archive per recording (<M-D-YYYY>_ratN.zip, approximately 400 MB each), containing the raw pressure packets from the four boards with the camera-frame timestamps on the same clock (MATLAB .mat) and the overhead video (MP4).
• Annotations (annotations.zip): camera-to-sensor homographies; blinded manual paw labels (FL/FR/HL/HR) for 81 recordings; manually selected walking windows and the corresponding review decisions; manually reviewed contacts over entire recordings, including standing; and body mass by rat and date.
• Demonstration package (demo_2-24-2026_rat4.zip, 415 MB): one complete recording organized as in the full deposit, for the demonstration workflow.

Processed outputs are not deposited; they are regenerated by the analysis code. Once extracted into one folder, the deposit is organized as follows:

Path	Contents
<M-D-YYYY>/runway/ratN/runway_tactile_data_*.mat	raw pressure packets, with the camera-frame timestamps
<M-D-YYYY>/runway/ratN/camera_*.mp4	overhead video
annotations/recordings.csv	the 85 analysed recordings (paths relative to the data folder)
annotations/homographies/	camera-to-sensor projective transform, one per recording
annotations/blind_labels/	manual FL/FR/HL/HR labels, made blind to the automated assignment
annotations/mobile_bout_windows.csv	manually selected walking windows
annotations/cop_trajectory_corrections.csv	review decisions (retain or trim) that produced those windows
annotations/corrected_contacts.csv	manually reviewed contacts over entire recordings
annotations/mass.csv	body mass by rat and date

Study design

Five male Lewis rats (12–16 weeks, 390–430 g), three with left sciatic nerve pinch and two sham-operated, were recorded before surgery and over the following two weeks. Each recording is a session of approximately 90 s on the runway with M = 200, threshold 700 and digital-potentiometer setting 100. The raw packets are the direct output of the readout boards: for every frame, the taxels measured and their 12-bit values, including readings below threshold (see How it works and the firmware README).

Data format

In each .mat file, BoardN_timestamps, BoardN_pixel_indices and BoardN_pixel_values contain one entry per frame for each board, BoardN_board_ids the hardware identifier of the board, and cameraK_frame_timestamps the time of every video frame on the same clock. A taxel index i is row i // 32, column i % 32 for i < 2048, and row (i - 2048) // 32, column (i - 2048) % 32 + 32 otherwise. Boards are placed by hardware ID 4, 3, 2, 1 from left to right along the runway. site_tools/ta_frames.py in the repository reconstructs pressure frames in this way in Python.