Broadband Near-Infrared Spectroscopy (bNIRS) has emerged as a promising technique for non-invasive monitoring of the redox state of cytochrome-c-oxidase (CCO), a key enzyme in cellular energy production. While early work focused on linear approaches based on the modified Beer-Lambert Law (MBLL), recent decades have seen substantial diversification in both instrumentation and computational strategies. To capture this evolution, we conducted a systematic review following PRISMA guidelines across PubMed, Web of Science, ScienceDirect (limited to the journal NeuroImage), and IEEE Xplore, identifying 35 studies that reported novel hardware or algorithmic approaches to CCO reconstruction. Hardware developments ranged from broadband lamps and supercontinuum lasers to LED and CMOS-based miniaturised systems, reflecting a trade-off between spectral coverage, portability, and sensitivity. Algorithmic innovations encompassed refinements of MBLL, diffusion theory, stochastic Monte Carlo modelling, and emerging machine learning methods, each addressing challenges of scattering, spectral overlap, and low signal-to-noise. Despite progress, the field remains limited by variability in instrumentation, standardised validation protocols, and the inherent weakness of the CCO signal relative to haemoglobin. We conclude that advancing bNIRS toward robust, clinically relevant metabolic monitoring will require integration of wearable system design, high-performance computational modelling, and shared benchmarking datasets. This review provides a structured synthesis of hardware and algorithmic advances, highlighting the underlying physics that govern light-tissue interaction and reconstruction, and identifying key directions for future research at the intersection of optical modelling, biomedical engineering, and translational neuroscience.
Ward et al. (Tue,) studied this question.