Skip to main content

ALTEA

Autonomous Learning for Tomographic Ensembles and Attributes

A reproducible, provenance-tracked Python pipeline for automated quality control, segmentation and 3-D morphometry of tomographic image stacks such as FIB-SEM. Built with porous and heterogeneous materials in mind (MOFs, battery electrodes, catalysts, membranes, rock), but general to any two-phase or multi-phase volumetric acquisition.

CI License: MIT


Why

Analysing FIB-SEM tomography today typically involves two costly, poorly reproducible steps: an operator manually discards a large fraction of acquired slices (e.g. 500–600 acquired, 300–400 kept) by visual inspection, and the image treatment is tuned by eye per sample "like Photoshop", with no single procedure that applies across samples. ALTEA turns both into deterministic, config-driven, versioned operations, and records exactly what was done to every volume.

ALTEA is a reproducible orchestration layer, not a single model. Every stage is explicit, every parameter lives in a config file, and every run emits a provenance record (software and dependency versions, input data hash, per-stage parameters and metrics). The same input always yields the same result — and you can prove it.

Pipeline

load → quality control → drift correction → preprocessing → segmentation → morphometry
Stage What it does
io Load FIB-SEM stacks (multi-page TIFF or slice directories) with voxel spacing
qc Per-slice sharpness / brightness / drift / curtaining scores; reproducible, threshold-driven slice selection (replaces manual curation)
align Rigid drift correction by phase cross-correlation
preprocess Deterministic denoising and contrast normalization — every parameter recorded
segment Pluggable backends: classical (Otsu, watershed) and learned (random-forest pixel classifier; optional deep U-Net)
morphometry Porosity, specific surface area, pore-size distribution, connectivity/percolation, geometric tortuosity — all voxel-size-aware
acquire Convergence analysis: how few slices suffice for a target accuracy (basis for cost-aware acquisition)

Install

pip install altea            # core
pip install "altea[deep]"    # + optional deep-learning segmentation backends
pip install "altea[dev]"     # + test / lint tooling

Quickstart

No data required — run the synthetic demo:

altea demo --output runs/demo

Or on your own stack:

altea run --input stack.tif --config configs/fibsem_default.yaml --output runs/sample1

In Python:

from altea import Pipeline
from altea.datasets import make_porous_volume, add_acquisition_artifacts

clean, ground_truth = make_porous_volume(porosity=0.35)
raw = add_acquisition_artifacts(clean, blur_slices=(8,), charge_slices=(15,))

results = Pipeline.from_yaml("configs/fibsem_default.yaml").run(raw, output_dir="runs/demo")

m = results["morphometry"]
print(m.porosity, m.specific_surface_area, m.tortuosity)
print(results["qc_report"].summary())      # which slices were dropped, and why

Segmentation backends

Backends share one interface and self-register, so switching is a one-line config change:

segment:
  backend: pixel_rf      # otsu | watershed | pixel_rf | unet
  params: {sigmas: [1, 2, 4], n_estimators: 200}

Add your own by subclassing SegmentationBackend and decorating with @register_backend. Nothing else in the pipeline changes.

Reproducibility

Every run writes provenance.json:

{
  "altea_version": "0.1.0",
  "dependency_versions": {"numpy": "...", "scikit-image": "..."},
  "input_hash": "sha256:...",
  "config": {...},
  "stages": [{"name": "qc", "params": {...}, "metrics": {...}}, ...]
}

Roadmap

  • Cost-aware autonomous acquisition. The acquire module already quantifies the accuracy-vs-slice-count trade-off. The next step is an active policy that decides on the fly how many sections to acquire to reach a target uncertainty, reducing destructive beam time.
  • Deep segmentation backends (2-D U-Net / nnU-Net wrapper) under altea[deep].
  • Sparse-view reconstruction of discarded slices.

Status

Alpha. Interfaces may change before 1.0.

Citation

If you use ALTEA, please cite it via CITATION.cff.

License

MIT — see LICENSE.

Metadata

Release files for altea 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for altea 0.1.0
File Size Uploaded Provenance Actions
altea-0.1.0.tar.gz 34.6 kB Without provenance Details for altea-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for altea 0.1.0
File Interpreter ABI Platform Actions
altea-0.1.0-py3-none-any.whl Python 3 none any Details for altea-0.1.0-py3-none-any.whl

Total release size: 73.6 kB

Release files / altea-0.1.0.tar.gz

Download URL altea-0.1.0.tar.gz
Size 34.6 kB
Tags Source
SHA-256 checksum
How to use checksums
76f2ea47fe649e2a19c886a2fbeb1fdaf457e0bc5b2fd8b2f027ed04cf741d1f
BLAKE2b-256 checksum
How to use checksums
4134b6cedc575be2c7c4dfcb5000a217c07c54fb200ff91bf6df4d7ce2bf1a4e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.10.8

Release files / altea-0.1.0-py3-none-any.whl

Download URL altea-0.1.0-py3-none-any.whl
Size 39.0 kB
Tags Python 3
SHA-256 checksum
How to use checksums
a61f16bf7fc587f5292cdb673347ae5c1fe50a6d517fed287a94153592985444
BLAKE2b-256 checksum
How to use checksums
f1071de9027e91e3b574bb7b781069e7ef5435456aaf59ec733a9f439df53215
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.10.8

Security for altea 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page