bio-pulse-pack / README.md
solsticestudioai's picture
Add Bio-Pulse sample (10K sequences) with README, SCHEMA, parquet, JSONL
00b456e verified
---
license: cc-by-4.0
task_categories:
- tabular-classification
- time-series-forecasting
language:
- en
tags:
- synthetic
- healthcare
- digital-health
- wearables
- remote-patient-monitoring
- precision-medicine
- physiological-telemetry
- early-warning-detection
- clinical-decision-support
- time-series
- hipaa-safe
pretty_name: Bio-Pulse Physiological Telemetry Pack
size_categories:
- 10K<n<100K
configs:
- config_name: default
data_files:
- split: train
path: bio_pulse_sample.parquet
---
# Bio-Pulse Physiological Telemetry Pack (Sample)
**A synthetic wearable-sensor physiological telemetry dataset for early-warning detection, digital-health model training, and remote patient monitoring research.** Each row is a complete physiological event sequence — from baseline vitals through a progressing clinical scenario (drug interaction, septic cascade, nocturnal hypoglycemia) — with phenotype context, causal-chain labels, and anomaly-signature detection metadata.
Built by [SolsticeAI](https://www.solsticestudio.ai/datasets) as a free sample of a larger commercial pack. 100% synthetic. No real patient, wearable, or clinical data — fully HIPAA-safe: no PHI, no identifiers, no re-identification risk.
## What is included
| File | Rows | Format | Purpose |
|---|---:|---|---|
| `bio_pulse_sample.parquet` | 10,000 | Parquet | Columnar, typed, best for analytics |
| `bio_pulse_sample.jsonl` | 10,000 | JSON Lines | Streaming / LLM training friendly |
**This sample:** 10,000 physiological event sequences, balanced across 4 criticality tiers and 3 diagnostic classes.
**Criticality classes:** `low`, `medium`, `high`, `life_threatening` (~2,500 each)
**Diagnostic classes:** `Drug_Interaction_Toxicity`, `Septic_Shock_Cascade`, `Nocturnal_Hypoglycemia` (~3,300 each)
**Patient phenotypes:** `pediatric_asthmatic`, `active_longevity`, `high_stress_corporate`, `sedentary_diabetic`, `geriatric_frailty`
**Sensor assets covered:** Continuous Glucose Monitor, Smart Watch v3, O2 Ring, Neural Patch v2
## Record structure
Each record is one physiological event sequence with 7 top-level fields:
| Field | Type | Contents |
|---|---|---|
| `schema_version` | string | Pack schema version (`1.0.0-bio-pulse-sample`) |
| `event` | struct | `id`, `trace_id`, `timestamp`, `criticality`, `outcome`, `confidence` |
| `patient_context` | struct | `phenotype`, `age_group`, `existing_conditions[]` |
| `medical_logic` | struct | `diagnostic_class`, `causal_chain[]`, `prediction_window_minutes`, `tracked_biomarkers[]` |
| `vitals_telemetry` | list<struct> | Ordered sensor readings: `timestamp`, `sensor_id`, `sensor_asset`, `event_name`, `vitals` (hr, glucose, spo2, temp_c, cortisol_ugdl, systolic_bp, qt_interval_ms), `patient_id` |
| `detection_logic` | struct | `signature`, `anomaly_score`, `baseline_deviation` |
| `simulation` | struct | `synthetic`, `model`, `fidelity`, `target_diagnoses[]` |
See [SCHEMA.md](./SCHEMA.md) for the full nested field breakdown.
## Why this dataset is useful
Most public physiological datasets are either highly-scoped single-signal recordings (MIMIC ICU, PhysioNet challenges) or flat aggregated reports. This pack is shaped around what digital-health and wearable-AI teams actually need to train predictive monitoring models:
- Multi-signal sequences rather than single-vital recordings
- Balanced criticality tiers across low → life-threatening
- Causal-chain labels that connect trigger events to outcomes (not just end-state labels)
- Phenotype-aware variability so models don't overfit one population
- Explicit prediction-window fields for early-warning benchmarks
- Cross-sensor coverage (CGM, smart watch, pulse ox, wearable patch)
- Synthetic-by-design — safe to share across vendors, regulators, and research teams
## Typical use cases
- Early-warning detection model training
- Wearable-device AI and firmware validation
- Remote patient monitoring (RPM) pipelines
- Clinical decision support prototypes
- Digital-health analytics and dashboards
- Phenotype-aware risk stratification
- LLM fine-tuning on physiological narratives + reasoning
- Benchmarking anomaly scores and false-alarm reduction
## Quick start
```python
import pandas as pd
import pyarrow.parquet as pq
df = pq.read_table("bio_pulse_sample.parquet").to_pandas()
# Criticality distribution (balanced)
print(df["event"].apply(lambda e: e["criticality"]).value_counts())
# Average anomaly score by diagnostic class
df["dx"] = df["medical_logic"].apply(lambda m: m["diagnostic_class"])
df["anomaly"] = df["detection_logic"].apply(lambda d: d["anomaly_score"])
print(df.groupby("dx")["anomaly"].mean().round(3))
# Pull the QT-interval trajectory for one episode
row = df.iloc[0]
traj = [(v["timestamp"], v["vitals"]["qt_interval_ms"]) for v in row["vitals_telemetry"]]
print(traj)
```
Streaming form:
```python
import json
with open("bio_pulse_sample.jsonl") as f:
for line in f:
episode = json.loads(line)
# one physiological sequence per line
```
## Responsible use
This dataset is intended for **digital-health research and model development**. It contains synthesized physiological sequences and clinical-scenario labels — it does **not** contain real patient records, PHI, wearable-device telemetry, or identifiable data of any kind. It is HIPAA-safe by construction (no covered data is present). Models trained on this data must be independently validated against real clinical data before any deployment in patient care.
## License
Released under **CC BY 4.0**. Use freely for research, prototyping, education, and commercial development with attribution.
## Get the full pack
This Hugging Face repo is a **10K-episode sample**. The production pack scales to 1M+ episodes, wider phenotype and age coverage, additional diagnostic classes (atrial fibrillation, asthma exacerbation, hypertensive crisis, post-op recovery, stroke precursors), richer biomarker panels, device-specific signal-fidelity variants, parquet + JSONL + FHIR-aligned formats, and buyer-specific configurations.
**Self-serve (Stripe checkout):**
- [**Sample Scale tier — $5,000**](https://buy.stripe.com/7sY5kD2j85QTfSb5lfeEo03) — ~25K records, one subject, 72-hour delivery.
**Full pack + enterprise scope:**
- [www.solsticestudio.ai/datasets](https://www.solsticestudio.ai/datasets) — per-SKU pricing across Starter / Professional / Enterprise tiers, plus commercial licensing, custom generation, and buyer-specific variants.
**Procurement catalog:**
- [SolsticeAI Data Storefront](https://solsticeai.mydatastorefront.com) — available via Datarade / Monda.
## Citation
```bibtex
@dataset{solstice_bio_pulse_pack_2026,
title = {Bio-Pulse Physiological Telemetry Pack (Sample)},
author = {SolsticeAI},
year = {2026},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/solsticestudioai/bio-pulse-pack}
}
```