Fluorescence imaging enables multiscale profiling of therapeutic antibody delivery across entire surgical specimens

Abstract Antibody delivery to solid tumors is often assumed to be uniform, yet this has not been systematically evaluated in human tissues. Optically labeled antibodies such as panitumumab-IRDye800CW (Pan800) remain detectable in FFPE specimens, enabling direct assessment of in vivo drug distribution. We aimed to validate macroscopic (21–337 μm/pixel) fluorescence measurements against microscopy (0.3 μm/pixel) and to characterize multi-scale heterogeneity of Pan800-derived fluorescence using a high-throughput, slide-free fluorescence imaging approach. A total of 946 FFPE blocks from seventeen patients who received systemic Pan800 were analyzed using near-infrared fluorescence imaging. Selected blocks were sectioned for fluorescence microscopy validation. Fluorescence intensities were compared between cancer-positive and -negative blocks, decalcified versus non-decalcified specimens, and across blocks within and between patients. Across resolutions (21–337 μm), maximum fluorescence intensity (FIₘₐₓ) discriminated cancer-positive from -negative blocks, while low-resolution scanning enabled rapid acquisition (~ 1 min/block). Macroscopic fluorescence signals correlated with microscopy at the section level (R 2 = 0.73), within patients (R 2 = 0.66), and across patients (R 2 = 0.66). Decalcification significantly reduced fluorescence at the block level (median FI max : 875 vs 287; P < 0.0001), and this reduction was confirmed in patient-level paired analyses when stratified by cancer status (P < 0.05). After excluding decalcified samples, FIₘₐₓ successfully differentiated cancer-positive from cancer-negative blocks (AUC = 0.78), with higher accuracy in lymph nodes than primary tumors. Marked intra- and inter-patient heterogeneity in Pan800-derived fluorescence was observed. Macroscopic fluorescence correlates with microscopic measurements and pathologically confirmed cancer presence, validating a high-throughput method for Pan800 distribution in clinical specimens. Signal heterogeneity precludes inference from single sections, supporting multi-block evaluation and a hierarchical workflow integrating screening with targeted microscopy. Clinical trial registration: NCT#04511078, first posted on 08/12/2020, https://clinicaltrials.gov/study/NCT04511078 ; NCT#05945875, first posted on 07/14/2023, https://clinicaltrials.gov/study/NCT05945875 .