Changes in version 0.2.0.9000                      

  - Corrected cube dimensions, masks, wavelength/FWHM alignment, spatial
    plotting, TIFF/ENVI pixel order, truncated ENVI input and
    unsupported integer exports.
  - SG smoothing now returns full-window center bands consistently
    across explicit backends; derivatives use wavelength units.
    Calibration validates reference compatibility and bad-pixel repair
    uses immutable valid neighbors.
  - Replaced generated hemoglobin fitting references with a pinned
    tabulated source. Beer-Lambert fitting requires an explicit domain
    and separates coefficients from concentrations requiring optical
    pathlength. Default StO2 is a research model fraction. The former
    ratio method errors; relative indices are unscaled.
  - Added equality-constrained nonnegative unmixing with convergence
    diagnostics, optional residual storage and chunk equivalence.
    PCA/MNF respect masks and rank limits and expose reusable
    transforms.
  - Repaired explorer initialization, stale results, paired uploads,
    TIFF metadata entry and complete archive downloads; added
    cancellable background computations.
  - Added processing recipes, resumable manifest analysis, reusable
    classifier prediction, group-separated evaluation and group-level
    uncertainty summaries.
  - Expanded exact-value regressions, browser and minimal-dependency
    validation, reference sensitivity reports, and performance
    measurements. Clinical claims remain outside the evidence supplied
    by synthetic tests. See MIGRATION.md for breaking changes and
    planning/implementation-status.md for release evidence.

                        Changes in version 0.1.0                        

Initial release

Data I/O

  - hs_read_envi(): Read ENVI .hdr + binary hyperspectral cubes
    (BSQ/BIL/BIP).
  - hs_read_tiff(): Read multi-channel TIFF files.
  - hs_read_cubert(): Read Cubert .cu3s session files via Python bridge
    (optional).
  - hs_read_cube(): Auto-detect format and dispatch to correct reader.
  - hs_write_envi(), hs_write_tiff(), hs_export_png(): Export functions.

Calibration

  - hs_calibrate(): Full calibration workflow (dark + white reference).
  - hs_fix_bad_pixels(): Statistical bad pixel detection and
    interpolation.

Preprocessing

  - hs_smooth(): Savitzky-Golay smoothing and derivatives.
  - hs_snv(): Standard Normal Variate correction.
  - hs_msc(): Multiplicative Scatter Correction.
  - hs_absorbance(): Reflectance to absorbance conversion.
  - hs_continuum_removal(), hs_resample().

Biomedical Indices

  - hs_sto2(): Tissue oxygen saturation (500-815 nm).
  - hs_npi(): Near-infrared perfusion index (655-910 nm).
  - hs_thi(): Tissue hemoglobin index (530-825 nm).
  - hs_twi(): Tissue water index (adapted for Cubert range).
  - hs_ndi(): Generic normalized difference index.
  - hs_clinical_indices(): Compute all available indices.

Analysis

  - hs_pca(), hs_mnf(): Dimensionality reduction.
  - hs_umap(): UMAP spectral embedding (optional, requires uwot).
  - hs_sam(): Spectral Angle Mapper classification.
  - hs_classify_svm(), hs_classify_rf(): Machine learning
    classification.
  - hs_unmix_nnls(): Linear spectral unmixing via NNLS.
  - hs_beer_lambert(): Chromophore concentration fitting.

Visualization

  - autoplot.hsi_cube(): Plot RGB, single band, or spectra.
  - hs_plot_clinical(): TIVITA-style clinical panel display.
  - hs_plot_index(): Pseudocolor index map with clinical palette.
  - hs_plot_spectra(): Spectral profile plots with mean +/- SD.
  - scale_color_wavelength(), theme_hsi().

Shiny Application

  - hs_run_app(): Interactive 6-tab HSI explorer.

Data

  - hs_example_cube(): Synthetic 30x30 tissue cube (430-910 nm, 61
    bands).
  - hs_simulate_cube(): Configurable synthetic cube generator.
  - hs_chromophore_data(): Published HbO2/Hb extinction coefficients.