Medical digital twins are increasingly proposed as tools for modeling, prediction, and control of physiological systems, yet their educational treatment often overlooks fundamental limitations imposed by biological delay, sensing constraints, and safety‑critical design trade‑offs. This instructional laboratory introduces students to the concept of an AI‑enabled medical digital twin through a control‑systems lens, emphasizing how physiological transport delays can render conventional biological feedback insufficient for disturbance rejection. Using intracranial cerebrospinal‑fluid (CSF) pulsation as a motivating application, students observe how delayed mechanical forcing synchronized with the cardiac cycle can generate persistent disturbances that cannot be effectively suppressed by feedback alone. The lab develops a simplified digital twin in MATLAB and Simulink that models a delayed physiological plant subjected to a cardiac‑synchronous disturbance. Rather than simulating ECG morphology, cardiac activity is abstracted as ECG‑derived timing information, represented by a continuously evolving cardiac phase variable with slow heart‑rate variability. This phase serves as a predictive reference signal for an adaptive feedforward controller based on a least‑mean‑squares (LMS) adaptive filter. Students explore how the adaptive controller learns the timing relationship between cardiac activity and downstream mechanical forcing, allowing anticipatory disturbance cancellation that overcomes biological delay. Through guided simulations, students analyze why feedback alone fails in delayed physiological systems, how feedforward adaptive control exploits timing correlations, and how learned parameters encode electromechanical coupling rather than signal amplitude or morphology. The lab explicitly highlights ethical and engineering constraints relevant to implantable and safety‑critical devices, including size, weight, power, thermal limits, sensing abstraction, and algorithmic simplicity. Original clinical imaging data used to motivate the disturbance model are intentionally abstracted to respect data‑use and privacy considerations, reinforcing responsible biomedical system design. This laboratory supports upper‑division undergraduate and early graduate courses in biomedical engineering, control systems, and cyber‑physical systems. Learning outcomes align with ABET student outcomes related to modeling, system analysis, design under constraints, and ethical responsibility. By integrating digital twins, adaptive control, and physiological realism within a carefully scoped abstraction, the lab provides students with a concrete framework for understanding how AI‑enabled feedforward control can augment, but not replace, biological regulation in medical systems.
Icaro dos Santos (Faculty)