Advantages

- A new type of biomarker that can capture intermittent and short‑duration brain activities
- Enables simultaneous analysis of not only local activities (e.g., frontal or occipital regions) but also brain network connectivity and interactions
- Allows rapid extraction of intrinsic frequency components from EEG data
- Applicable not only to EEG but also to various types of time‑series signals, with potential utility for identifying diverse oscillatory phenomena

Technology Overview and Background

Intrinsic frequencies in brain activity refer to the frequencies of electrical oscillations naturally generated by neural cells and neural networks. They exist across multiple spatial scales, from single neurons to whole‑brain networks, and span a wide range from ultra‑low frequencies around 0.05 Hz to several hundred hertz. Intrinsic frequencies change with brain state, such as wakefulness versus sleep or rest versus cognitive engagement, and disease‑specific alterations have been reported in conditions including dementia, Parkinson’s disease, epilepsy, and chronic pain.
However, real brain activity consists of multiple intrinsic frequencies that are superimposed and exhibit non‑stationary dynamics, with state transitions and decay over time. Conventional Fourier‑based approaches assume stationarity of the underlying signal, making it difficult to accurately capture subtle shifts and changes in intrinsic frequencies that occur in real neural dynamics.
Researcher’s group has refined a brain‑signal decoding based on Dynamic Mode Decomposition (DMD) and established a method to extract continuous distributions of intrinsic frequencies from complex EEG time‑series data, identifying these distributions as promising new biomarkers of brain activity.

Patents and Publication

- PCT/JP2024/036766(national phase: Japan, US)
https://patentscope2.wipo.int/search/ja/detail.jsf?docId=WO2025109908
- Fukuma R, et al. arXiv [Preprint]. 2025 Jul 14. arXiv:2507.10145. doi:10.48550/arXiv.2507.10145.
https://arxiv.org/abs/2507.10145

Principal Investigator & Academic Institution

Prof. Takufumi YANAGISAWA (Department of Neuroinformatics, Graduate School of Medicine, The University of Osaka)

Data

- In EEG comparisons between healthy controls and patients with dementia (Alzheimer’s disease [AD] and frontotemporal dementia [FTD]), our method revealed a clear shift in the intrinsic frequency distribution in FTD patients. Using this distribution as features, we achieved higher classification accuracy than conventional Fourier‑based features, and notably reduced misclassification between AD and FTD.
- In EEG comparisons between healthy controls and patients with Parkinson’s disease (PD), our method detected a subtle shift of the alpha‑band peak frequency in PD patients from 12–13 Hz down to 8–9 Hz. Using the intrinsic frequency distributions as features again yielded better classification performance than conventional Fourier‑based analysis.

Expectations

We are seeking industrial partners interested in the development of this analysis method for brain‑activity monitoring and disease diagnosis. We also welcome collaboration, including integrating this technology with existing EEG devices and measurement platforms owned by partner companies. If you have any interest, we would be pleased to provide additional technical information and to arrange a meeting with the investigators to discuss potential next steps.