Multiplex imaging is essential for understanding complex tumor-immune interactions in cancer treatment, yet traditional analysis methods are either laborious or often fail to accurately capture the spatial and functional heterogeneity of tumors. Nucleai has developed a deep learning-based multiplex imaging analysis pipeline that enhances tumor profiling accuracy by integrating automated cell segmentation, deep learning-based protein quantification, and automated cell typing and spatial architecture characterization. This pipeline calculates complex spatial features, capturing the distribution of cell phenotypes and tumor-immune interactions across the tissue.
We applied Nucleai’s pipeline to an ultra-deep multiplexed immuno-fluorescence panel of 44 proteins, assessing immune and metabolic cell states association with clinical outcome in NSCLC patients treated with immune checkpoint blockade (ICB). 14 cell types were assigned, and were further subclassified to 45 cell subtypes by clustering of metabolically and immunologically relevant proteins. Five distinct tumor clusters were identified, each varying in cellular composition, immune and metabolic states, clinical outcomes and spatial biomarkers associated with immunotherapy response. A multivariate ICB outcome predictor based on tumor cluster-specific biomarkers demonstrated high accuracy in an independent test set, predicting clinical benefit and survival outcomes effectively.
These findings highlight the value of a robust multiplex imaging analysis pipeline and the potential of advanced spatial analysis to guide personalized therapeutic strategies in NSCLC, ultimately improving patient outcomes.