Skip to content

CHANGELOG.md

Latest commit

 

History

History
64 lines (56 loc) · 2.93 KB

CHANGELOG.md

File metadata and controls

64 lines (56 loc) · 2.93 KB

Changelog

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

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

[0.2.0] - 2025-02-22

Added

  • GPU backends (backends.py): Tiered NumPy → CuPy → JAX with automatic fallback
  • Welford streaming statistics: O(1) memory for mean/variance during bootstrap
  • Disparity filter (core.py): Serrano et al. (2009) multiscale backbone extraction
  • sc_observed field in BootstrapResult: non-bootstrapped SC for comparison
  • Graph analysis (graph_analysis.py):
    • Rich club coefficients Φ(k) with degree-preserving null models
    • Small-world propensity (Muldoon et al. 2016)
    • Hub detection with Guimerà-Amaral classification (provincial/connector/kinless)
    • Communicability via matrix exponential with bootstrap CIs
  • Statistical inference (inference.py):
    • Network-Based Statistic (NBS) with FWER permutation control
    • Threshold-Free NBS (TFNBS, Baggio et al. 2018)
    • Edge-wise permutation testing with FDR/Bonferroni correction
    • Global metric permutation testing
    • Connectome-Predictive Modeling (CPM) with bootstrap aggregating
    • Partial Least Squares (PLS) brain-behavior with bootstrap ratios
  • Along-tract profiling (along_tract.py):
    • Per-node bootstrap CIs on tract profiles
    • Group comparison with cluster-based permutation correction
    • Bundle membership stability via bootstrap RecoBundles
  • Voxel-level bootstrap (voxel_bootstrap.py):
    • Wild bootstrap for DTI (Rademacher/Webb distributions, HC2/HC3)
    • Residual bootstrap for CSD (Jeurissen et al. 2011)
    • Full voxel-bootstrap connectome pipeline
  • Storage (storage.py):
    • HDF5 save/load for all analysis types
    • BIDS-compatible export (BEP017/BEP038)
  • Extended visualizations (viz_extended.py):
    • NBS results, rich club curves, hub cartography
    • Along-tract profiles, CPM predictions, communicability, PLS

[0.1.0] - 2025-02-01

Added

  • Initial release as sars.tractogram_bootstrap module
  • Core bootstrap engine (core.py):
    • StreamlineAssignment, EdgeStats, BootstrapResult data structures
    • MRtrix3 file loading (tck2connectome -out_assignments)
    • Synthetic assignment generation from SC matrices
    • Weighted tractogram bootstrap with running statistics
    • Edge reliability classification (robust/present/fragile/spurious)
  • Community detection (community.py):
    • Probabilistic community detection across bootstrap samples
    • Co-assignment matrices and consensus partition
    • Node stability metrics
    • Graph metrics with bootstrap CIs (density, strength, modularity, efficiency, transitivity)
  • Visualization (viz.py):
    • SC uncertainty maps (mean, std, CV)
    • Edge classification matrices
    • Community results (co-assignment, stability)
    • Graph metric distributions
    • Clinical correlation plots