Hyperspectral Imaging and AI Detect Cellular Oxidative Stress

By Akshatha ChandrashekarReviewed by Susha Cheriyedath, M.Sc.Apr 20 2026

Hyperspectral imaging combined with AI identifies oxidative stress in red blood cells via spectral signatures. The non-invasive method achieves high accuracy, enabling early disease detection and advancing personalized diagnostics.

A new study in Nature Communications Medicine reports an optical diagnostic strategy that combines hyperspectral imaging (HSI) with artificial intelligence (AI) to detect oxidative stress (OS) in red blood cells. The researchers show that OS-induced biochemical changes alter membrane optical scattering properties, which are detectable using HSI.

The proposed method accurately distinguishes children with Autism Spectrum Disorder (ASD) from neurotypical controls, highlighting its strong potential for early detection and personalized diagnostics.

Bridging Optical Diagnostics and Oxidative Stress

Oxidative stress plays a major role in many diseases by disrupting cellular structure and function, particularly in red blood cells. RBC membranes contain high levels of polyunsaturated fatty acids, which makes them especially vulnerable to oxidative damage. This damage alters membrane organization, mechanical properties, and protein activity. Despite its importance, current methods for detecting oxidative stress remain invasive, costly, or insufficiently sensitive for early-stage changes.

Unlike conventional imaging, HSI captures both spatial and spectral information, generating a unique “spectral fingerprint” for each pixel. Previous studies have shown the value of HSI in biomedical imaging, its use for detecting oxidative stress at the cellular membrane level remains limited.

To address this gap, the study develops a standardized HSI-based framework to detect oxidative stress in RBC membranes. It combines lipidomic analysis to validate biochemical changes with AI models to interpret complex spectral data. The approach aims to establish optical signatures of RBC membranes as reliable indicators of systemic oxidative stress and disease-related alterations.

Hyperspectral Imaging and AI-Driven Analysis

The study uses a multi-step experimental and computational framework that combines optical imaging, biochemical analysis, and machine learning. Researchers collect blood samples using EDTA as an anticoagulant and analyze them within five hours to ensure stability. They place a small droplet (2 µL) on a glass slide and image it using hyperspectral dark-field microscopy, capturing spectral data in the 400–1000 nm range.

Using the Spectral Angle Mapper (SAM) algorithm, researchers identify and map eight distinct spectral signatures (endmembers) across cell membranes. These signatures reflect variations in membrane composition and structure. To model oxidative stress, they treat samples with hydrogen peroxide (H₂O₂) at controlled concentrations. This treatment induces membrane lipid oxidation without causing cell lysis. In parallel, lipidomic analysis using gas chromatography quantifies changes in fatty acid composition and links them to spectral features.

For clinical validation, researchers analyze RBC samples from 31 neurotypical children and 27 children with ASD. Each sample produces multiple hyperspectral images, creating a large dataset. They use artificial neural networks (ANNs), combined with the TWIST optimization system, to classify subjects based on spectral patterns. Additional tools, such as Auto-Contractive Maps, help visualize relationships between variables and identify key spectral markers.

Optical Signatures of Oxidative Stress

The study demonstrates that hyperspectral imaging can detect subtle structural and biochemical changes in RBC membranes. In healthy samples, researchers identify eight consistent spectral signatures, showing high reproducibility across individuals. These signatures reflect the organization of membrane lipids and proteins, as well as light-scattering properties.

Download the PDF of this page here

Upon oxidative treatment, the analysis reveals significant shifts in spectral distributions. Specific endmembers show marked increases or decrease, indicating structural reorganization of the membrane. A decrease in polyunsaturated fatty acids and an increase in saturated fatty acids confirm oxidative damage and membrane remodeling. This establishes a direct link between spectral features and biochemical alterations.

On applying the method to clinical samples, the method reveals similar spectral patterns in children with ASD. Several spectral signatures change in the same direction as in the oxidative stress model, suggesting shared underlying mechanisms. One of the spectral components emerges as a key indicator of oxidative damage, showing strong correlation with lipid composition and membrane organization.

Researchers further validate these findings by measuring Na⁺/K⁺-ATPase activity, a membrane-bound enzyme. The study reports a significant reduction in enzyme activity in ASD samples, consistent with oxidative stress–induced membrane dysfunction. Importantly, this activity correlates with specific spectral signatures, reinforcing the biological relevance of the optical data.

AI analysis achieves high classification performance, with over 93% accuracy, sensitivity, and specificity in distinguishing ASD from neurotypical subjects. The results indicate that hyperspectral data contain rich diagnostic information that machine learning can effectively extract. The combination of optical imaging and AI thus enables robust, non-invasive detection of disease-associated cellular changes.

Toward Non-Invasive Optical Diagnostics

This study establishes HIS as a powerful tool for detecting oxidative stress at the cellular level. It detects biochemical and structural changes using only a small blood sample, marking a significant advancement in diagnostic technology.

The findings highlight the potential of HSI for early detection of oxidative stress–related conditions, including neurodevelopmental disorders such as ASD. Future studies with larger and more diverse populations will help validate and refine this approach. Beyond ASD, researchers can extend this framework to a wide range of diseases linked to oxidative stress, including cardiovascular and metabolic disorders.

The integration of lipidomics, optical imaging, and AI creates a comprehensive platform for personalized medicine. It supports continuous monitoring of cellular health and may guide targeted interventions to restore membrane integrity and antioxidant balance. Overall, this study marks a significant step toward label-free, rapid, and cost-effective optical diagnostics. It opens new opportunities to translate biophotonic technologies into routine clinical practice.

Journal Reference

Vartian, R., Sansone, A., et al. (2026). AI-based autism identification from hyperspectral imaging detection of oxidative stress in pediatric red blood cells. Communications Medicine 2026. DOI: 10.1038/s43856-026-01581-y,  https://www.nature.com/articles/s43856-026-01581-y

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.