Changelog

Source: NEWS.md

MycoHalo 0.1.0

Complete rewrite. The earlier grayscale-threshold prototype could not separate grey Zymoseptoria colonies from the bacterial colony and its halo and did not separate several cultures in one dish.

  • Colour-calibrated analysis in CIELAB with flat-field illumination correction and optional grey-card calibration (calibrate_color(), analyze_plate(reference = )).
  • Texture-aware object detection, so that strongly melanized colonies on dark agar are found.
  • Fungus / bacterium / halo classification by a layout-initialised, noise-aware colour-direction mixture (classify_pixels()), or by a supervised classifier (train_classifier(), annotate_plate()).
  • Separation of cultures sharing one dish: plate layouts (plate_layout()), distance-transform colony cores, seeded watershed, satellite detection and half-contrast edge placement (segment_colonies()).
  • Measurements: size, shape, melanization index with spatial block-bootstrap CIs, ImageJ-compatible grey value, side-specific melanization facing the bacterium, growth inhibition (PIRG), gaps to bacterium and halo, halo geometry and colour (measure_colonies()), edge-to-centre profiles (colony_profiles()).
  • plot_melanization_map(): 3D shaded melanization landscape and 2D MI map.
  • make_metadata(), layout_by_treatment() and correct_facing_bias() for experiment-level workflows; cast-shadow detection; local half-contrast edge placement; semi-supervised (seed-clamped) EM; calibrated on real Zymoseptoria control and confrontation photographs.
  • Batch processing with metadata and per-plate layouts (analyze_plates()), QC figures (plot_qc()), profile plots (plot_profiles()).
  • Plate-aware statistics with mixed models (compare_melanization()).
  • Plate simulator with ground truth (simulate_plate()) and a validation suite (inst/validation/).
  • No Bioconductor dependency: image kernels (exact Euclidean distance transform, connected components, watershed, separable filters) are implemented in C++ (Rcpp).