Skip to content

CHANGELOG.md

Latest commit

 

History

History
84 lines (67 loc) · 3.49 KB

CHANGELOG.md

File metadata and controls

84 lines (67 loc) · 3.49 KB

Changelog

All notable changes to AtlasPatch will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.1.0] - Unreleased

Added

Slide and Patient Encoding

  • encode-slide command for slide embeddings appended into slide_features/<encoder> inside the canonical per-slide H5
  • encode-patient command for manifest-driven patient embeddings written under patient_features/<encoder>/<case_id>.h5
  • Slide encoder registry and built-in slide encoders: titan, prism, moozy
  • Patient encoder registry and built-in patient encoder: moozy

Packaging and Release

  • Optional extras for slide and patient encoder stacks: titan, prism, moozy, slide-encoders, and patient-encoders
  • GitHub Release to PyPI trusted publishing workflow in publish.yml

1.0.0 - 2025-02-03

Added

Core Pipeline

  • SAM2-based tissue segmentation finetuned on ~35,000 diverse WSI thumbnails
  • Four-checkpoint pipeline architecture:
    • detect-tissue: tissue detection with mask visualization
    • segment-and-get-coords: segmentation + patch coordinate extraction
    • process: full pipeline with feature embedding
    • --save-images: optional patch image export
  • HDF5 output format with coordinates (coords) and features (features/<encoder>)
  • Slide passport metadata in H5 files (vendor, MPP, staining info)
  • Per-slide lock files for safe parallel job execution
  • MPP override via CSV for slides with missing/incorrect metadata

Feature Extractors (66)

  • Natural image backbones: ResNet (18/34/50/101/152), ConvNeXt (tiny/small/base/large), ViT (B/L/H)
  • DINOv2: small, base, large, giant
  • DINOv3: vits16, vitb16, vitl16, vith16_plus, vit7b16 (+ SAT variants)
  • Pathology encoders: UNI v1/v2, Phikon v1/v2, Virchow v1/v2, GigaPath, CHIEF-CTransPath, Midnight, OpenMidnight, MUSK, PathOrchestra, H-optimus-0/1, H0-mini, CONCH v1/v1.5, Hibou B/L
  • Lunit models: ResNet50 (BT/SwAV/MoCov2), ViT-S (patch16/patch8 DINO)
  • CLIP variants: RN50/101/50x4/50x16/50x64, ViT-B/L
  • Medical CLIP: PLIP, MedSigLIP, Quilt (B-32/B-16/B-16-PMB), BiomedCLIP, OmiCLIP
  • Custom encoder plugin system via --feature-plugin

Visualization

  • Tissue mask overlays
  • Contour visualization
  • Patch grid overlays
  • Configurable output directory structure

CLI & Configuration

  • atlaspatch CLI with subcommands
  • Configurable patch size, target magnification, step size
  • Device selection for segmentation and feature extraction
  • Batch size controls for segmentation and feature extraction
  • Precision options (float32, float16, bfloat16)
  • Fast mode (default) for high-throughput processing
  • Content filtering options (white/black thresholds)
  • Recursive directory processing
  • Skip existing / force overwrite modes

HPC Support

  • SLURM job script templates (jobs/atlaspatch_patch.slurm.sh, jobs/atlaspatch_features.slurm.sh)
  • Multi-GPU support via separate device flags
  • Configurable worker counts for parallel processing

Documentation

  • Comprehensive README with usage examples
  • Pipeline checkpoint diagrams
  • Feature extractor reference table
  • FAQ section for common issues
  • Issue templates (bug report, feature request)
  • Pull request template

Fixed

  • Contour scaling from mask resolution to thumbnail
  • H-optimus transformation pipeline
  • MUSK model registration