A spectral audit framework reveals task-dependent aperiodic reliance across EEG and ECG deep learning
arXiv:2606.08583v1 Announce Type: new Abstract: Deep learning on physiological time series is interpreted through domain-specific features -- oscillatory rhythms in EEG, morphological complexes in ECG -- yet these signals sit atop a broadband aperiodic 1/f-like envelope that covaries with arousal, age, and pathology. We introduce a spectral audit framework combining aperiodic/periodic decomposition, phase-preserving Fourier interventions, sham controls, and simulation validation. Aperiodic reliance was task-dependent and architecture-general: across six neural architectures, flattening drops e
The paper introduces a novel framework for interpreting deep learning models applied to physiological time series, addressing a critical need for explainability in 'black box' AI systems in healthcare and neuroscience.
A strategic reader should care because improving the interpretability of AI in medical diagnostics and brain-computer interfaces can accelerate adoption, build trust, and unlock new application areas for AI in sensitive domains.
This framework changes how researchers and practitioners can audit and understand the specific features (oscillatory vs. aperiodic) deep learning models prioritize from complex physiological data, moving beyond opaque performance metrics.
- · AI explainability researchers
- · Medical AI developers
- · Neuroscience research
- · Healthcare diagnostics
Improved trust and regulatory acceptance for AI-driven medical devices and diagnostic tools utilizing physiological signals.
Faster development and deployment of personalized medicine solutions based on sophisticated analysis of individual physiological data.
Potential for new therapeutic interventions or diagnostic markers derived from a deeper understanding of how AI interprets fundamental biological signals.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG