A device-restricted C. elegans whole-brain recording dataset for ASCENT: annotation-free neuron tracking benchmark

This dataset accompanies the manuscript “Annotation-free whole-brain neuron tracking via transformer-based self-supervised learning” by Han et al. The manuscript introduces ASCENT, an Annotation-free Self-supervised Contrastive Embeddings for 3D Neuron Tracking method.

It provides the in-house whole-brain (head ganglion) recording of C. elegans used for benchmarking and analysis in the ASCENT study, along with the corresponding ground-truth and predicted neuron detections and tracks.

The recording captures neuronal activity across the head ganglion of a C. elegans (strain ZM9624, expressing pan-neuronal GCaMP6 and mNeptune) confined within a microfluidic device that permits controlled chemical stimulation. Worms were subjected to repeated cycles of 60s buffer followed by 30s E. coli OP50 supernatant stimulus. Imaging was performed on a Bruker Opterra II swept-field confocal microscope using a 40× 0.75 NA air objective and a Photometrics Evolve 512 Delta EM-CCD camera. The dataset consists of 1,100 volumetric frames (2 × 9 × 512 × 512; channels × z × y × x) acquired at 3.3 Hz with a spatial resolution of 1.5 μm (z) and 0.243 μm (xy).

Ground-truth neuron tracks were generated by manually proofreading and correcting an initial set of tracks generated by ZephIR (https://github.com/venkatachalamlab/ZephIR). These ground-truth and ASCENT-predicted detections and tracks are provided in CSV format compatible with Napari and standard cell-tracking frameworks.

Files included

1. InHouse_Opterra_CH_RG_Day1_OP50Stim_1100frames.h5
   • Raw volumetric imaging data
   • Structure:

     root/
       ├── t0/c0  (Z × Y × X)
       ├── t1/c0
       ...

   • c0: mNeptune, c1: GCaMP6
2. InHouse_Opterra_coords_GT.csv
   • Ground-truth neuron detections.
   • Columns: object_id,t,z,y,x
3. InHouse_Opterra_coords_PRED.csv
   • ASCENT-predicted neuron detections.
   • Columns: object_id,t,z,y,x
4. InHouse_Opterra_tracks_GT.csv
   • Ground-truth neuron tracks (Napari-compatible).
   • Columns: TrackID,ObjectID,t,z,y,x
5. InHouse_Opterra_tracks_PRED.csv
   • ASCENT-predicted neuron tracks (Napari-compatible).
   • Columns: TrackID,ObjectID,t,z,y,x

Usage notes