Advancing precision oncology through hNQO1-activatable NIR-II probes: Integrating molecular imaging with artificial intelligence

Traditional imaging modalities often lack the molecular specificity and spatial resolution required for real-time tumor visualization, particularly in complex surgical settings. This narrative review examines the emerging class of human NAD(P)H oxidoreductase 1 (hNQO1)-activatable fluorescent probes operating within the near-infrared-II (NIR-II, 1000–1700 nm) and near-infrared-III (NIR-III, 1700–2500 nm) windows, which offer transformative potential for tumor imaging and surgical guidance. hNQO1, overexpressed in various malignancies including pancreatic, lung, and breast cancers, while minimally present in normal tissues, serves as an ideal biomarker for selective probe activation. However, its residual expression in certain normal tissues such as the kidney and vascular endothelium necessitates careful probe design to minimize off-target activation. This review systematically explores the design principles underlying hNQO1-activatable probes, emphasizing structural features that enable enzyme-specific fluorescence activation with minimal background interference. Chemical synthesis strategies, including quinone propionate-based self-immolative linkers and advanced bioconjugation techniques, are discussed alongside optimization approaches to enhance probe stability, sensitivity, and biocompatibility. The review highlights clinical applications in real-time tumor visualization, accurate margin delineation, early cancer detection, and therapeutic monitoring. Significantly, the integration of artificial intelligence (AI) and deep learning frameworks for automated image analysis, tumor segmentation, and radiomics-based feature extraction represents a critical advancement, transforming qualitative fluorescence signals into quantitative, clinically actionable data. Crucially, AI-derived quantitative biomarkers extracted from hNQO1-activated fluorescence, such as spatial heterogeneity indices, activation kinetics, and texture features, can be linked to clinical endpoints including treatment response prediction, recurrence risk stratification, and overall survival, providing a data-driven foundation for personalized oncology decision-making. Translational challenges including regulatory pathways, pharmacokinetics, safety considerations, model validation requirements, and real-world clinical deployment frameworks are addressed, alongside future perspectives on theranostic applications and multimodal imaging integration. Together, hNQO1-activatable NIR-II probes combined with intelligent computational systems represent a paradigm shift toward data-driven, high-precision cancer care. • hNQO1-activatable probes enable highly selective tumor imaging in NIR-II • Probes activate specifically in cancer cells overexpressing hNQO1 enzyme • NIR-II imaging provides deep tissue penetration with minimal scattering • AI-driven segmentation enables precise real-time tumor margin detection • Deep learning transforms fluorescence signals into quantitative biomarkers