Accurate identification of brain states and pathological transitions remains a central challenge in clinical neuroscience, particularly in epilepsy management. Current neuroimaging modalities provide complementary but incomplete solutions: electroencephalography (EEG) offers high temporal resolution but poor spatial localization, while functional magnetic resonance imaging achieves fine spatial detail at the cost of portability and temporal precision (1). Functional ultrasound imaging (fUSI) has emerged as a promising alternative, offering depth-resolved hemodynamic maps with spatial resolution of 200–400 μm, millisecond-scale sensitivity, and bedside portability, without contrast agents or invasive procedures (2). This work investigates the potential of fUSI for classifying epileptic brain states, with particular focus on the preictal phase, i. e. the transitional period preceding seizure onset. Clinically critical yet notoriously difficult to detect, this phase can last minutes to over an hour with high inter- and intra-patient variability. Reliable identification could enable timely therapeutic intervention and advance real-time seizure forecasting. Our framework integrates simulated and real electrophysiological data. Synthetic neural signals simulating interictal, preictal, and ictal states were generated using an excitatory-inhibitory population model. Real intracranial EEG recordings were drawn from the HUP dataset (3) (34 drug-resistant epilepsy patients), preprocessed at 500 Hz, and segmented into 3-second windows. Synthetic fUSI signals were derived through convolution with empirically established hemodynamic response functions (4). Classification used a 1D convolutional neural network with three convolutional layers, ReLU activations, and max-pooling. On simulated data, fUSI achieved accuracy up to 0.86 for preictal detection; on real recordings, performance remained stable (~0.78–0.79), comparable to sEEG-based models. These results suggest that fUSI-derived vascular signals carry sufficient information to discriminate epileptic brain states, including the elusive preictal phase. Future work will incorporate multimodal EEG–fUSI fusion, larger cohorts, and experimental validation toward robust clinical seizure prediction. (1) ABREU, R.; LEAL, A.; FIGUEIREDO, P., Frontiers in human neuroscience, 2018 (2) DEFFIEUX, T., et al. Current opinion in neurobiology, 2018 (3) BERNABEI, J. M., et al. Brain, 2022 (4)NUNEZ-ELIZALDE, A. O., et al. Neuron, 2022.

Assessing the Potential of fUSI for Seizure Detection and Preictal Identification / Gambosi, B., Buda, C., Toschi, N., Astolfi, L.. - (2026). (2nd summer school of the IMPRS on Cognitive NeuroImaging Lipzig, Germany ).

Assessing the Potential of fUSI for Seizure Detection and Preictal Identification

Benedetta Gambosi
;
Christian Buda;Laura Astolfi
2026

Abstract

Accurate identification of brain states and pathological transitions remains a central challenge in clinical neuroscience, particularly in epilepsy management. Current neuroimaging modalities provide complementary but incomplete solutions: electroencephalography (EEG) offers high temporal resolution but poor spatial localization, while functional magnetic resonance imaging achieves fine spatial detail at the cost of portability and temporal precision (1). Functional ultrasound imaging (fUSI) has emerged as a promising alternative, offering depth-resolved hemodynamic maps with spatial resolution of 200–400 μm, millisecond-scale sensitivity, and bedside portability, without contrast agents or invasive procedures (2). This work investigates the potential of fUSI for classifying epileptic brain states, with particular focus on the preictal phase, i. e. the transitional period preceding seizure onset. Clinically critical yet notoriously difficult to detect, this phase can last minutes to over an hour with high inter- and intra-patient variability. Reliable identification could enable timely therapeutic intervention and advance real-time seizure forecasting. Our framework integrates simulated and real electrophysiological data. Synthetic neural signals simulating interictal, preictal, and ictal states were generated using an excitatory-inhibitory population model. Real intracranial EEG recordings were drawn from the HUP dataset (3) (34 drug-resistant epilepsy patients), preprocessed at 500 Hz, and segmented into 3-second windows. Synthetic fUSI signals were derived through convolution with empirically established hemodynamic response functions (4). Classification used a 1D convolutional neural network with three convolutional layers, ReLU activations, and max-pooling. On simulated data, fUSI achieved accuracy up to 0.86 for preictal detection; on real recordings, performance remained stable (~0.78–0.79), comparable to sEEG-based models. These results suggest that fUSI-derived vascular signals carry sufficient information to discriminate epileptic brain states, including the elusive preictal phase. Future work will incorporate multimodal EEG–fUSI fusion, larger cohorts, and experimental validation toward robust clinical seizure prediction. (1) ABREU, R.; LEAL, A.; FIGUEIREDO, P., Frontiers in human neuroscience, 2018 (2) DEFFIEUX, T., et al. Current opinion in neurobiology, 2018 (3) BERNABEI, J. M., et al. Brain, 2022 (4)NUNEZ-ELIZALDE, A. O., et al. Neuron, 2022.
2026
File allegati a questo prodotto
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/1778694
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact