Quantitative Fluorescence Imaging with SI-Traceable Radiance: Radiometric Fluorescence Characterization with a Calibrated Solid-State Emitter

SignificanceFluorescence imaging remains largely qualitative and device-specific, limiting reproducibility and intersystem comparisons. Advancing toward quantitative imaging requires a radiometric characterization framework that provides SI-traceable units while explicitly addressing the interdependent factors that govern image formation. AimEstablish and evaluate a radiometric characterization framework that converts device-native counts to SI-traceable imaged radiance ({micro}W{middle dot}cm-2{middle dot}sr-1) and aggregate fluorescence yield (sr-1) while accounting for interdependent factors that influence fluorescence image formation. ApproachA calibrated Lambertian solid-state radiometric emitter target (RET) was combined with a three-step radiometric framework consisting of the Radiance Transfer Curve (RTC), the Radiance Imaging Transform (RIT), and the Fluorescence Imaging Transform (FIT). The RTC establishes system responsivity as a function of radiance; the RIT applies this calibration at the pixel level to map digital counts to SI-traceable radiance; and the FIT performs pixel-wise excitation normalization of the RIT image to produce an aggregate fluorescence yield image. The framework was tested through distance and aperture invariance, digital-vs-physical ROI analyses, RTC acquisition, and application of the RIT and FIT to an ICG concentration target and a breast lumpectomy phantom. ResultsThe RET exhibited Lambertian behavior, with no significant dependence of the measured radiance on distance or aperture; imager responsivity (R{lambda}) also remained invariant within uncertainty across working distances and f-numbers. Digitally masked ROIs reproduced R{lambda} obtained with matched physical apertures, enabling ROI and pixel-level radiance transfer. RTCs acquired over 49 radiances captured sensor and processing nonlinearities. The RIT provided a per-pixel mapping from counts to radiance ({micro}W{middle dot}cm-2{middle dot}sr-1). Applying pipeline-specific RTCs, the RIT and FIT reconciled large discrepancies across RAW, 8-bit, and log10 image processing pipelines, yielding closely aligned radiance-concentration curves and improved SBR/SNR/CNR agreement. In a breast lumpectomy phantom, FIT produced SI-traceable aggregate fluorescence yield (sr-1) images and absolute contrast metrics in an anthropomorphic geometry. ConclusionsThe combined framework converts device-native counts into SI-traceable radiance and aggregate fluorescence yield at the pixel level, providing a practical basis for reproducible quantitative fluorescence imaging. The feasibility results across distance/aperture tests, ROI analyses, image pipelines, and phantom imaging indicate readiness for broader evaluation. Future work will establish formal uncertainty budgets and assess robustness across devices, geometries, and excitation conditions to support adoption as a quantitative reporting standard.