Depth estimation from wavelength-dependent fluorescence spread profiles across NIR-I and NIR-II

Wide-field fluorescence-guided imaging is inherently depth ambiguous. Multispectral imaging across NIR-I and NIR-II may encode depth because longer-wavelength fluorescence undergoes less diffuse broadening in tissue.

The study aims to determine whether wavelength-dependent radial spread profiles of subsurface indocyanine green (ICG) fluorescence can be used to estimate depth from a single excitation wavelength.

ICG phantoms (1 to $50\mu M$ ) were imaged through ex vivo chicken breast under 785 nm excitation using four emission bands spanning NIR-I and NIR-II. Radial spread profiles from high-dynamic-range images formed a multispectral dictionary, and depth was estimated with an inverse model evaluated leave-one-thickness-out across 2.2 to 6.0 mm under concentration-matched and concentration-randomized conditions, randomized exposure subsets, and spectral-band ablations, and compared with analytical and empirically calibrated scalar dual-emission fluorescence-ratio baselines of Leblond et al. Two ICG-localized sentinel lymph nodes from a breast cancer specimen were analyzed ex vivo.

Longer-wavelength channels produced more compact spread profiles, and full width at 20% peak intensity increased linearly with tissue thickness for all filters ( $R^2=0.94$ to 0.95). The NIR-II-only subset (940 and 1070 nm) yielded sub-millimeter root mean square error at every held-out thickness under both concentration conditions (0.32 to 0.66 mm), whereas the analytical and empirical scalar-ratio baselines yielded 1.58 to 5.24 and 1.13 to 2.74 mm. Clinical sentinel lymph nodes preserved the wavelength-dependent profile narrowing, with phantom-to-clinical normalized-width RMSE of 3.3% to 8.5%.

Wavelength-dependent fluorescence spread profiles across NIR-I and NIR-II can encode depth information for subsurface ICG targets, providing a wide-field, single-fluorophore, single-excitation approach compatible with emerging multispectral surgical imaging platforms.