HypIX

HypIX enables visual exploration and machine-learning-based analysis of hyperspectral image (HSI) datasets to detect spatial, spectral, and temporal patterns for environmental monitoring and organism stress-response studies (e.g., Desmophyllum pertusum).

Key Features:

• Exploratory data analysis and information visualization: Integrates methods for selecting regions of interest (ROI) across 2D spatial coordinates, time points, and spectral wavelengths to investigate HSI data.
• Machine-learning dimensionality reduction: Applies advanced machine learning techniques to reduce HSI dimensionality while preserving relevant spectral–temporal information.
• Temporal-series analysis workflows: Implements two workflows for time-series HSI—morphology-based filtering for detecting morphological changes over time and embedded-driven response analysis for identifying spectral responses tied to experimental conditions.
• Spectral change detection: Detects shifts in spectral signatures over time to characterize stress responses and assess individual tolerance levels.
• Reproducibility: Analysis outputs have demonstrated good inter-observer agreement as reported in the referenced study (PMID: 35939502).

Scientific Applications:

• Environmental monitoring of marine ecosystems: Time-series HSI analysis to monitor and assess health and stressors in marine habitats.
• Cold-water coral stress-response studies: Analysis of spectral signatures of Desmophyllum pertusum exposed to particle stressors using HSI collected from pre-exposure to six weeks post-exposure (PMID: 35939502).
• Individual tolerance assessment: Tracking temporal spectral changes to evaluate differential tolerance among samples subjected to varying particle types and concentrations.

Methodology:

Uses machine-learning-based dimensionality reduction, ROI selection across spatial, temporal, and spectral dimensions, morphology-based filtering and embedded-driven response analysis workflows, and temporal/spatial pattern and spectral-shift identification on hyperspectral imaging time-series.