AI-Enabled On-Chip Holographic Imaging for Culture-Free, Single-Fungus-Resolved Antifungal Susceptibility Testing

Invasive fungal infections are among the most time-critical clinical conditions, yet antifungal susceptibility testing (AFST) remains fundamentally constrained by culture-dependent, population-averaged workflows that require 48-72 h. Here, we introduce a culture-free and rapid AFST platform based on dynamic, single-fungus-resolved phenotypic profiling enabled by AI-powered lens-free holographic imaging. The platform integrates ∼10 mm2 field-of-view on-chip lens-free holographic imaging with a low-cost, easy-to-operate microfluidic chip and a U-Net-based reconstruction and segmentation pipeline, enabling parallel, time-resolved profiling of hundreds of individual fungal cells and their morphodynamic behaviors. Dynamic fungal growth, hyphal elongation, and morphological transitions are quantified to generate multiparameter phenotypic fingerprints, which are interpreted by a machine-learning model for rapid susceptibility classification within 6 h. To demonstrate the clinical translational capability, we applied it directly to clinical urine, swab, and blood-culture specimens without conventional fungal cultivation and isolation and achieved 95.8% category agreement and 4.2% minor error with the clinical reference standard. By shifting susceptibility testing from delayed, population-averaged end points toward rapid, single-fungus-resolved dynamic phenotyping, this work provides a scalable, culture-free platform for rapid phenotypic AFST with potential applications in point-of-care and resource-limited settings.