Hyperspectral Imaging Techniques in Biomedical Applications

Hyperspectral imaging (HSI) acquires multidimensional data capturing both spatial and spectral information across tens to hundreds of contiguous wavelength bands. In biomedical applications it offers a non-invasive, label-free means to probe tissue physiology, morphology and composition. Light–tissue interactions including absorption, reflection and scattering vary with biochemical constituents and microstructure, giving rise to unique spectral fingerprints. HSI modalities range from widefield camera systems to fibre-optic probes and endoscope-compatible devices, covering visible to near-infrared regimes. Data processing combines pre-processing, dimensionality reduction, spectral unmixing and advanced machine-learning or deep-learning algorithms to extract diagnostically relevant features. Applications span real-time surgical guidance for tumour delineation, detection of microcirculatory impairments in critically ill patients, digital histopathology for computational slide analysis, and assessment of organ viability in transplant contexts. Recent advances in sensor miniaturisation and computational throughput have accelerated translation from bench to bedside. Standardisation of acquisition protocols and the emergence of open spectral datasets have facilitated robust algorithm development and benchmarking. HSI promises to enhance diagnostic precision, reduce reliance on exogenous contrast agents and improve intraoperative decision-making, with potential impact on global health through cost-effective, rapid, point-of-care solutions.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent studies have demonstrated the utility of automated machine-learning pipelines applied to skin hyperspectral data for bedside monitoring of microcirculation in critically ill patients. By analysing reflectance signatures at the palm and fingertips, classifiers distinguished healthy controls from sepsis patients with an area under the receiver operating characteristic curve of 0.92, highlighting HSI’s potential for rapid sepsis screening. In neurosurgical applications, a benchmark employing k-fold cross-validation on in vivo hyperspectral brain images achieved a median macro F1-score of over 70% for delineating high-grade, low-grade and metastatic tumours, underscoring the value of combined spectral and spatial processing frameworks for intraoperative tumour detection. The release of a large-scale surgical dataset comprising 5,758 annotated porcine organ images acquired over 500–1,000 nm has established a standardised resource for algorithm development; this openly accessible collection enables comparative studies of spectral variability across organs, imaging angles and individuals, thereby expediting the translation of HSI methods to clinical practice.

Hyperspectral Imaging Techniques in Biomedical Applications publication trend

The graph below shows the total number of articles in hyperspectral imaging techniques in biomedical applications across all publications each year (not limited to Nature Index journals).

Technical terms

Hyperspectral imaging: An imaging modality that collects and processes information from across the electromagnetic spectrum, producing a three-dimensional dataset combining spatial and spectral dimensions.

Hypercube: The data structure generated by hyperspectral imaging, comprising two spatial axes and one spectral axis representing intensity values at each wavelength.

Spectral signature: The characteristic pattern of reflectance or absorption as a function of wavelength, unique to specific tissue types or biochemical compounds.

Machine learning: A suite of computational techniques that enable automated analysis of complex hyperspectral data, including classification, segmentation and feature extraction based on training from labelled samples.

References

1. A proof of concept for microcirculation monitoring using machine learning based hyperspectral imaging in critically ill patients: a monocentric observational study. Critical Care (2024).
2. Hyperspectral imaging benchmark based on machine learning for intraoperative brain tumour detection. npj Precision Oncology (2023).
3. HeiPorSPECTRAL - the Heidelberg Porcine HyperSPECTRAL Imaging Dataset of 20 Physiological Organs. Scientific Data (2023).