Organoid-based imaging is moving into real-time tracking as researchers combine patient-derived pancreatic tumor organoids with deep learning to watch CAR T cells attack living tumor tissue. The approach uses a deep learning algorithm, OrganoIDNet, to automatically track how CD318-targeting CAR T cells destroy tumor regions over time. By running the readouts inside an organoid model derived from individual patients, the platform aims to reduce reliance on endpoint measurements and make it easier to compare cell-killing kinetics across CAR designs or patient samples. That matters for practical optimization of CAR T activity while limiting the need for repeated in vivo experiments. The work also signals a broader shift toward “closed-loop” translational testing—using live microtumor models plus automation—to tighten the feedback cycle between engineering choices and functional tumor response.