Philosophers-Stone is a inference tool that converts a single-channel overnight sleep EEG into a quantitative index of brain health.
It applies a validated, peer-reviewed multi-cohort deep-learning model trained on 36,000 sleep recordings to estimate cognitive performance, disease likelihoods, and mortality-related physiological patterns.
The tool outputs both a single Brain Health Score and a 1024-dimensional latent embedding suitable for research and biomarker discovery.
If you use or reference this tool, please cite the peer-reviewed study:
Ganglberger, W., Sun, H., Turley, N., ... & Westover, M. B. (2026).
Brain Health from Sleep EEG: A Multicohort, Deep Learning Biomarker for Cognition, Disease, and Mortality.
NEJM AI, 3(3), AIoa2500487.
Available here.
- Sleep scientists
- Neurologists and dementia researchers
- Aging and cognitive-decline investigators
- Psychiatry researchers
- Data scientists working with physiological signals
- Clinical-trial teams exploring EEG-based biomarkers
- Brain Health Score (single interpretable metric)
- 1×1024 latent brain-health embedding (AI-derived sleep features)
- Predictions for cognition, disease risk, and mortality-related physiology
- Optional outputs: spectrograms and per-recording JSON summaries
- Python 3.10+
- Conda recommended
Recommended setup:
./run_sample.sh
This creates the recommended environment if needed, downloads the model checkpoint from Hugging Face, and runs the included sample files.
Philosophers-Stone first uses PHILOSOPHER_MODEL_FILE when set. In a source
checkout it then looks for ./model_files/SleepPhilosophersStone.ckpt;
package installs fall back to ~/.cache/philosophers-stone/model_files/.
If the checkpoint is missing, explicit inference calls can auto-download it from:
https://huggingface.co/wolfgang-ganglberger/philosophers-stone
A CSV with columns:
filepath(absolute path to EEG file)age(years)sex(0=female, 1=male)
Philosophers-Stone accepts single-channel overnight EEG in HDF5 (.h5) or EDF (.edf) format. Preferred channel: C4-M1.
| Format | Requirements |
|---|---|
| HDF5 (.h5) | - Dataset: signals/c4-m1 (1-D float array, full night) - Attributes: sampling_rate=200, unit_voltage="uV" - Extra channels/annotations ignored - Manifest uses absolute paths |
| EDF (.edf) | - Must contain a C4-M1 channel (label variants allowed) - Any sampling rate accepted; auto-resampled to 200 Hz with anti-aliasing |
Sample full-night EEG data is included under ./sample-data/.
python philosopher.py \
--manifest_csv phi_manifest.csv
Or use the one-command conda bootstrap:
./run_sample.sh
Optionally, you can use Docker; see the Docker ReadMe.
For applications that want to call Philosopher's Stone directly:
pip install -e .
The package exposes an array-based API:
from phi_utils.philosopher_utils import Config, infer_brain_health
result = infer_brain_health(
eeg_uv,
fs_hz=200,
age=65,
sex=1,
file_id="study-001",
cfg=Config(),
)
eeg_uv must be one overnight EEG channel in microvolts. The API does not
write output files; callers decide where to persist returned scores, latent
features, and optional stage probabilities.
-
Summary CSV (
phi_out/phi_results.csv) Columns include:file_id, filepath, age, sex, brain_health_score, total_cognition_score, fluid_cognition_score, crystallized_cognition_score, lhl_1…lhl_1024 -
Latent embedding (
lhl_1…lhl_1024) A 1024-dimensional vector summarizing brain-health-relevant EEG patterns. -
Optional JSON files under
phi_out/json/ -
Optional spectrograms under
phi_out/figures/
- Use a GPU if available
- Keep
batch_size=1
CC BY-NC 4.0 — Attribution-NonCommercial 4.0 International.
See the LICENSE file for details.