ADAPTIVE QUANTUM NEURAL NETWORKS FOR PREDICTING SEPSIS ONSET AND HEART RATE FROM AUTONOMIC PHYSIOLOGICAL TIME SERIES

Neural regulation of cardiovascular and autonomic function generates complex physiological time series that reflect brain–body interactions. Predicting these dynamics is clinically important for identifying early autonomic dysregulation in conditions such as sepsis, where neuroimmune and cardiovascular control can become compromised. We propose the Quantum Neural Information Optimization Network (QNION), an adaptive quantum machine learning framework designed to model noisy physiological signals by jointly optimizing quantum data encoding and variational circuit structure. QNION selects from four encoding schemes and incrementally grows circuits from a gate pool of rotation and entangling operations, enabling systematic exploration of representations while constraining circuit depth to mitigate optimization pathologies. We evaluated QNION on two clinically relevant tasks: prediction of sepsis onset from multivariate physiological time series and continuous heart rate forecasting. Across repeated trials, QNION outperformed classical deep learning baselines including convolutional neural networks, bidirectional long short-term memory networks, and transformer-based models. For sepsis onset prediction, QNION achieved a mean RMSE of 0.21 compared to 0.37–0.40 for classical baselines, and a mean MAE of 0.04 compared to 0.14–0.16. For heart rate forecasting, QNION achieved a mean RMSE of 2.5 compared to 2.9–3.4, and a mean MAE of 0.8 compared to 1.8–2.4. These results suggest that adaptive quantum architectures may provide advantages for modeling nonlinear autonomic dynamics in noisy physiological datasets, and offer a framework for evaluating when quantum approaches improve prediction performance in neurophysiological time-series modeling.