NeurOCNN: A Neural-Operator-Based Model for Physiological Time Series

Neural operators have become a central tool in scientific machine learning for learning discretization-consistent solution operators, achieving strong results on partial differential equation (PDE) benchmarks. Physiological time series, however, are highly nonstationary and dominated by localized transient events, properties that can challenge both PDE-oriented neural operators and conventional deep models. We propose NeurOCNN, a neural-operator-based model for physiological signals that learns a robust function-to-label mapping. NeurOCNN integrates continuous-time, spline-parameterized convolutions to capture localized morphology with Fourier projection pooling for variable-to-fixed dimensional mapping, thereby enabling robust, discretization-invariant inference under zero-shot sampling-rate shifts. Empirically, NeurOCNN outperforms standard neural-operator baselines, achieves performance comparable to state-of-the-art methods, and maintains stable accuracy across multiple previously unseen evaluation sampling rates. The code is available at https://github.com/Idsl-group/NeurOCNN.

Wearable sensors and medical devices record physiological signals such as brain activity, eye movement, and heart rhythms. However, different devices may record similar types of signals at different sampling rates. This creates an important challenge: a model trained using data from one device may not work as reliably when used with data from another device. This paper introduces NeurOCNN, a new artificial intelligence model designed to analyze physiological signals more reliably across different recording settings. Rather than depending on one fixed sampling rate, NeurOCNN focuses on meaningful patterns as they occur over real time. This allows the model to remain stable even when a signal is recorded at a sampling rate it has not seen during training. The model was tested on several datasets of physiological signals, including brain, eye, and heart signals. It performed as well as, or better than, many existing methods and was especially strong when the test data used sampling rates that were different from those used during training. These results suggest that NeurOCNN could help make physiological signal analysis more reliable across different devices and real-world healthcare settings.