Biological Computing as a Translational Platform for Delirium and Adaptive Network Failure: A Perspective from Anesthesia and Critical Care Medicine

Delirium remains one of the most pervasive yet least understood forms of acute brain dysfunction in perioperative and intensive care medicine, associated with substantial short-term mortality, prolonged intensive care unit stays, and persistent long-term cognitive impairment in survivors. Despite decades of clinical observation, no single experimental model has bridged the mechanistic gap between molecular neurochemistry, network-level electrophysiology, and the clinical phenomenology of acute brain failure. Animal paradigms incompletely recapitulate human cortical dynamics, whereas bedside neurophysiological monitoring provides only an indirect, downstream representation of the underlying network state. Recent advances in biological computing and organoid intelligence have introduced platforms in which living human-derived neuronal networks, cultured on microelectrode arrays and coupled to computational interfaces, exhibit measurable adaptive electrophysiological behavior and, in some configurations, acquire simple task-related functions through closed-loop feedback. In this perspective, such systems may offer a conceptually informative, although structurally reductive, experimental substrate for mechanistically dissecting delirium, conceived as a paradigmatic failure of adaptive network function. This perspective outlines three perturbational paradigms relevant to anesthesia and intensive care: pharmacological challenge, inflammatory challenge, and a combined-hit model designed to approximate the hypoactive delirium of the septic, critically ill patient. A complementary paradigm is then introduced, based on pharmacologically perturbing acquired network-level functions, and proposed as a substrate for in vitro cognitive pharmacodynamics. Recovery trajectories, network resilience, and the long-term horizon of patient-specific induced pluripotent stem cell-derived cultures for personalized risk stratification are also discussed. The substantial structural, biological, and ethical limitations of current platforms are explicitly acknowledged. These systems lack laminar cortical architecture, thalamo-cortical reciprocal loops, and ascending neuromodulatory systems central to anesthetic pharmacology and to several leading neuropathogenetic theories of delirium. Nonetheless, biological computing systems may eventually serve as a translationally meaningful intermediate layer between molecular neuroscience and whole-brain clinical physiology, with delirium serving as a paradigmatic and clinically urgent test case.