Rapid assessment of patient derived cancer organoids using label-free imaging and an automated analysis pipeline

Background: Tumor heterogeneity presents a major challenge in effective cancer treatment, particularly in colorectal cancer (CRC), by limiting the efficacy of therapies and driving resistance. Patient-derived cancer organoids (PDCOs) have emerged as powerful preclinical models that faithfully recapitulate the genomic, morphological, and metabolic profiles of primary tumors. However, current methods for rapidly and reproducibly assessing PDCOs are limited. Label-free imaging methods are a promising tool to measure organoid level heterogeneity and rapidly screen drug response in PDCOs. However, manual analysis of wide-field optical redox images is inefficient and laborious for large-scale drug screens. Here, we developed an automated pipeline for PDCO segmentation, single-PDCO tracking, and background correction in autofluorescence images. Methods: Wide field optical redox imaging (WF ORI) provided organoid-level measurements of treatment response without labels or additional reagents by measuring the autofluorescence intensity of the metabolic co-enzymes NAD(P)H and FAD, and the optical redox ratio, defined as the fluorescence intensity of [NAD(P)H/NAD(P)H+FAD], was used to measure the oxidation-reduction state of multiple CRC PDCO lines. Development of leading-edge analysis tools, isolating the ORI measurement to a 32μm region at the outer edge of the PDCOs, helped to maximize the sensitivity and reproducibility of treatment response measurements using WF ORI in CRC PDCOs. The automated pipeline includes segmentation using a fine-tuned Cellpose model, automated single-PDCO tracking over time via custom python code, and background correction. Glass's delta (G∆) is used to measure the PDCO treatment effect size. Results: Leading-edge analysis improves sensitivity to redox changes in treated PDCOs (G∆ = 1.462 vs G∆ = 1.233). Automated segmentation, when compared to manual masks, achieved mean Dice scores ≥0.8, indicating high reproducibility. Additionally, automated PDCO tracking accuracy exceeded 94% by two metrics, recall and Jaccard index, when compared to manual tracking. Importantly, the automated pipeline resolves single-PDCO responses over time with comparable sensitivity to drug treatment with over 127× faster processing time compared to the manual process. Conclusion: Overall, we demonstrate that combining PDCOs with accessible imaging and analysis techniques enables high-throughput detailed evaluation of tumor heterogeneity and therapeutic response.