Skip to contents

hyperspectR 0.3.0

CRAN preparation

hyperspectR 0.2.0.9001 (development candidate, not released)

  • Retry UMAP’s identified older-R PCA dependency failure with base SVD while preserving embedding settings and propagating unrelated errors. Compare canonical filesystem paths in the cross-platform TIVITA resume regression.
  • Preserve cube quality metadata through band selection and spatial subsets; reject ambiguous fields and malformed coordinate vectors. Resampling retains source-grid quality flags separately from current-grid data.
  • Validate ENVI metadata and missing values, preserve int32 input, reject lossy float32 overflow, and stage TIFF exports before replacing existing files. TIFF output explicitly uses float64. Cubert measurement indices must be whole.
  • Preserve calibration invalidity, avoid integer subtraction overflow, stabilize spectral angles, and validate derivative orders before conversion.
  • Check ordered preprocessing when reusing fitted models; validate batch output schemas and checksums before resuming. See MIGRATION.md for saved-model and batch-output changes.
  • Align explorer image/spectra downloads with the displayed selection, support unknown wavelengths as band indices, and restore launcher state on exit.
  • Preserve scientific coverage warnings, anchor wavelength colors to physical wavelengths, and correct image scaling for small and large finite ranges.
  • Require viridisLite >= 0.4.0 for the documented palettes, repair standalone first-use instructions, and align citation metadata with the development state.

hyperspectR 0.2.0.9000

  • Standardize the named v5 package logo across README, documentation pages and available application headers, with matching light and dark artwork.

  • 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.

hyperspectR 0.1.0

Initial release

Data I/O

Calibration

Preprocessing

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

Visualization

Shiny Application

Data