yahya021/bold5000-stpp

BOLD5000 CSI1 STPP Benchmark Dataset

A benchmark-ready Spatio-Temporal Point Process (STPP) dataset derived from the BOLD5000 fMRI dataset (CSI1 subject, unfiltered BOLD signal), following the official split semantics of the Neural STPP paper.

---

Dataset Description

Each record represents a sequence of neural activation events detected in a 3D fMRI scan volume. Events are defined as voxels where the z-scored BOLD signal crosses a threshold (rises above 6.0 standard deviations) between consecutive time frames.

The dataset covers 1 subject (CSI1), 10 fMRI runs, each split into 10 time-windowed segments, yielding 100 sequences total.

---

Source Format

Raw data was stored in a NumPy .npz archive (bold5000_csi1_stpp.npz).
Each key is a 6-digit string of the form SSRRNN, where:

• SS = session index (always 01 here)
• RR = run index (01–10)
• NN = segment index within the run (00–09)

Each key maps to a (N, 4) float64 array with columns [t, x, y, z].

---

Sequence Unit

One .npz key = one sequence (one 10-segment time window of one fMRI run).
No new windowing or segmentation was applied.

---

Event Schema

Field	Type	Description
t	float	Temporal position within the segment, normalized to [0, 10], dequantized with uniform noise
x	float	Voxel x-index (range ≈ 0–106), dequantized with uniform noise
y	float	Voxel y-index (range ≈ 0–106), dequantized with uniform noise
z	float	Voxel z-index (range ≈ 0–81), dequantized with uniform noise

Note: This dataset has 4 event fields (t, x, y, z), unlike the standard 3-field (t, x, y) STPP datasets. The spatial coordinates are 3D voxel indices in MNI-like scanner space, not geographic coordinates.

Values are exported as-is — no normalization applied.

---

File Structure

bold5000-stpp/
├── train.jsonl        # 70 sequences (runs 01–07)
├── val.jsonl          # 10 sequences (run 08)
├── test.jsonl         # 20 sequences (runs 09–10)
├── splits.json        # {"train": [...], "val": [...], "test": [...]}
├── dataset_meta.json  # Task/schema metadata
└── README.md

---

JSONL Row Schema

{
  "sequence_id": "010100",
  "events": [
    {"t": 0.530, "x": 12.4, "y": 55.2, "z": 30.8},
    {"t": 1.201, "x": 44.9, "y": 71.0, "z": 12.3},
    ...
  ]
}

Example (Python)

import json

with open("train.jsonl") as f:
    for line in f:
        seq = json.loads(line)
        sid    = seq["sequence_id"]   # e.g. "010100" (sess=01, run=01, seg=00)
        events = seq["events"]
        t = [e["t"] for e in events]
        x = [e["x"] for e in events]
        y = [e["y"] for e in events]
        z = [e["z"] for e in events]

---

Source & License

• Source data: BOLD5000 — a large-scale fMRI dataset (Chang et al., 2019). Data is shared under the Creative Commons Attribution 4.0 (CC BY 4.0) license.
• Processing code reference: facebookresearch/neural_stpp (MIT License)