Visible-light spectroscopy (VLS) has emerged as a promising technique for monitoring
localized microvascular hemoglobin oxygen saturation in clinical tissues. While earlier
systems demonstrated the feasibility of measuring oxygen saturation in superficial tissues
using fiber-optic probes, the integration of artificial intelligence (AI) has the potential to significantly improve diagnostic capability. This study presents an AI-assisted visible-light spectroscopy framework designed to enhance real-time detection of tissue hypoxia and tumor microenvironment abnormalities. Machine learning models were developed to classify
spectral signatures obtained from biological tissues and identify patterns associated with
ischemia and tumor hypoxia. Experimental results demonstrate improved accuracy and
robustness compared with traditional spectral fitting approaches. The proposed system
provides a promising platform for real-time clinical diagnostics during endoscopy, surgery,
and interventional procedures.
