Deep Learning-Powered Scalable Cancer Organ Chip for Cancer Precision Medicine

Functional precision oncology complements genomic approaches by directly testing treatment options on patient-derived models. However, existing platformssuch as patient-derived xenografts (PDXs) and patient-derived organoids (PDOs), face major barriers in clinical use due to technical challenges, including limited standardization, high costs, long assay times, scalability constraints, and incomplete recapitulation of the patient tumor microenvironment (TME). Here, we present a scalable, low-cost Organ Chip (OC) platform fabricated entirely from thermoplastics via injection molding. Leveraging a patented channel geometry and surface treatment, the device achieves barrier-free hydrogel confinement through capillary pinning without porous membranes, micropillars, or other barrier structures. This automation-compatible platform supports tissue-specific extracellular matrices and co-culture through versatile perfusion modes, with robust imaging compatibility. We demonstrate its feasibility for drug sensitivity testing using multiple cell lines and patient-derived primary cells, with imaging-based phenotypic profiling for accurate quantification of drug responses, closely aligning with clinical outcomes. Additionally, we integrated a deep learning-based image translation model that predicts fluorescence staining from bright-field images. This approach enables longitudinal, label-free phenotypic analysis with higher sensitivity than conventional endpoint staining. Together, this integrated cancer OC system overcomes key technical challenges and offers a promising framework for functional precision oncology through high-throughput, patient-relevant drug testing.