Periodic Reporting for period 1 - InterictalSleepDetct (Non-invasive detection of interictal epileptiform discharges (IEDs) in the mesial temporal lobe (MTL) during sleep)
Période du rapport: 2024-04-01 au 2025-09-30
In this project, we set out to finalize development and validation of a detection tool that allows identifying pathological electrical brain activities during sleep and work towards its commercialization. In our first milestone, we established that special sleep EEG that includes zygomatic EEG sensors (“zEEG”) placed over the cheekbones, together with special machine-learning based algorithms, can identify some ‘epileptic-like’ sharp abnormal electrical brain activities that occur in the medial parts of the temporal lobe – regions deep in the brain that were believed to be largely inaccessible with non-invasive approaches. In phase 1 / first milestone (M1) we established detection of these events in epilepsy patients. We show that zEEG can accurately detect a subset of pathological events occurring deep in the brain in temporal lobe regions, that zEEG outperforms standard EEG sensors in this task, that automatically detected rates can distinguish epilepsy patients from healthy individuals without visual inspection, and that this detection predicts cognitive impairment. In phase 2 / second milestone (M2P) we further show that the rate of pathological events is elevated in aging, further increased in mild cognitive impairment patients (representing early Alzheimer’s), and correlates with poorer individual cognition. We also developed business tools (third milestone M3) and performed IP evaluation and expansion (M4, M5). Accordingly, the institutions’ patent committees notified us on July 2025 that they have agreed to join continue for next phase of our PCT. We are now finalizing milestone six (M6) to focus on business partners and collaboration.
Memory consolidation depends critically on coordinated communication between brain regions during sleep such as the cerebral cortex and the hippocampus. This nocturnal dialogue is highly vulnerable to electrophysiological disturbances, particularly interictal epileptiform discharges (IEDs) arising from the hippocampus and medial temporal lobe. Such pathological events fragment the healthy coordination during sleep supporting memory, and are increasingly recognized across a spectrum of neurological and neurodevelopmental conditions—including epilepsy, prodromal Alzheimer’s disease (AD), amnestic mild cognitive impairment (aMCI), ADHD, and autism. Yet despite their broad relevance, the ability to measure these deep-brain disturbances non-invasively has remained extremely limited. For example, existing early Alzheimer’s biomarkers (imaging or blood assays) cannot capture abnormal electrophysiology in memory circuits; invasive intracranial methods are not scalable; and standard scalp EEG lacks sensitivity to medial temporal activity. As a result, the field lacks a practical biomarker of hippocampal electrophysiological integrity during sleep, leaving a major gap in understanding how sleep disturbances contribute to cognitive decline.
This project aims to fill this critical gap by introducing and validating a first-of-its-kind, noninvasive biomarker of medial temporal lobe epileptiform discharges during sleep, derived from a zygomatic EEG (zEEG) machine-learning framework. By leveraging simultaneous intracranial and zygomatic recordings, independent validation cohorts, and application to aging and early-stage Alzheimer’s disease, the project seeks to quantify pathological deep-brain activity in a manner that is sensitive, scalable, and clinically meaningful. The overarching objectives are three-fold: (1) to establish the physiological validity and precision of zEEG-based detection of hippocampal and medial temporal IEDs; (2) to determine how these sleep-related events map onto cognitive dysfunction across individuals; and (3) to demonstrate the utility of this method in aging and Alzheimer’s disease, where early electrophysiological disturbances may precede detectable structural or molecular changes.
Memory consolidation depends critically on coordinated communication between brain regions during sleep such as the cerebral cortex and the hippocampus. This nocturnal dialogue is highly vulnerable to electrophysiological disturbances, particularly interictal epileptiform discharges (IEDs) arising from the hippocampus and medial temporal lobe. Such pathological events fragment the healthy coordination during sleep supporting memory, and are increasingly recognized across a spectrum of neurological and neurodevelopmental conditions—including epilepsy, prodromal Alzheimer’s disease (AD), amnestic mild cognitive impairment (aMCI), ADHD, and autism. Yet despite their broad relevance, the ability to measure these deep-brain disturbances non-invasively has remained extremely limited. For example, existing early Alzheimer’s biomarkers (imaging or blood assays) cannot capture abnormal electrophysiology in memory circuits; invasive intracranial methods are not scalable; and standard scalp EEG lacks sensitivity to medial temporal activity. As a result, the field lacks a practical biomarker of hippocampal electrophysiological integrity during sleep, leaving a major gap in understanding how sleep disturbances contribute to cognitive decline.
This project aims to fill this critical gap by introducing and validating a first-of-its-kind, noninvasive biomarker of medial temporal lobe epileptiform discharges during sleep, derived from a zygomatic EEG (zEEG) machine-learning framework. By leveraging simultaneous intracranial and zygomatic recordings, independent validation cohorts, and application to aging and early-stage Alzheimer’s disease, the project seeks to quantify pathological deep-brain activity in a manner that is sensitive, scalable, and clinically meaningful. The overarching objectives are three-fold: (1) to establish the physiological validity and precision of zEEG-based detection of hippocampal and medial temporal IEDs; (2) to determine how these sleep-related events map onto cognitive dysfunction across individuals; and (3) to demonstrate the utility of this method in aging and Alzheimer’s disease, where early electrophysiological disturbances may precede detectable structural or molecular changes.
Scientific and technical aspects of the activities performed include:
A. Data collection and analysis for model training, testing, and independent validation in patients with mesial temporal lobe epilepsy and in healthy age-matched controls.
B. Data collection and analysis for model application in an aging cohort including individuals with amnestic Mild Cognitive Impairment (MCI), early Alzheimer’s disease (AD), and matched controls.
Our main outcomes include establishing three key findings:
1. Sleep zygomatic EEG signal features accurately detect a subset of interictal epileptiform discharges, capture local neuronal activity, and outperform standard scalp EEG.
2. Automatically detected discharge rates distinguish mesial temporal epilepsy patients from controls without visual inspection and predict cognitive impairment.
3. Event rates are elevated in aging, further increased in aMCI, and correlate with poorer individual cognition.
Research supported by the PoC grant yielded two scientific publications, a manuscript published and another manuscript currently under review: “Annotated interictal discharges in intracranial EEG sleep data and related machine learning detection scheme”. Scientific Data. Published: 18 December 2024; and “Sleep zygomatic EEG detects medial temporal epileptiform biomarkers of mild cognitive impairment. under review.
A. Data collection and analysis for model training, testing, and independent validation in patients with mesial temporal lobe epilepsy and in healthy age-matched controls.
B. Data collection and analysis for model application in an aging cohort including individuals with amnestic Mild Cognitive Impairment (MCI), early Alzheimer’s disease (AD), and matched controls.
Our main outcomes include establishing three key findings:
1. Sleep zygomatic EEG signal features accurately detect a subset of interictal epileptiform discharges, capture local neuronal activity, and outperform standard scalp EEG.
2. Automatically detected discharge rates distinguish mesial temporal epilepsy patients from controls without visual inspection and predict cognitive impairment.
3. Event rates are elevated in aging, further increased in aMCI, and correlate with poorer individual cognition.
Research supported by the PoC grant yielded two scientific publications, a manuscript published and another manuscript currently under review: “Annotated interictal discharges in intracranial EEG sleep data and related machine learning detection scheme”. Scientific Data. Published: 18 December 2024; and “Sleep zygomatic EEG detects medial temporal epileptiform biomarkers of mild cognitive impairment. under review.
The expected impacts of this project are substantial.
Our findings show that we can now measure abnormal medial temporal lobe activity during sleep in a reliable and noninvasive way, and that this activity is directly related to problems with thinking and memory. This breakthrough has important consequences for both basic science and clinical care. For researchers, it provides a new window into human hippocampal activity without the need for brain implants, allowing studies that connect detailed brain-circuit mechanisms with real human behavior. The same approach could be adapted to detect other important brain signals during sleep, such as those involved in forming memories, opening new opportunities to study learning, emotions, and decision-making.
Clinically, this work answers a major unmet need. Current early Alzheimer’s tests—like PET or MRI scans, spinal fluid tests, or blood biomarkers—cannot capture this abnormal electrical activity in memory-related brain circuits, and they are expensive, invasive, or lack fine detail. Our method offers a practical alternative that can detect this activity during sleep and could be used at home or for long-term monitoring. Beyond aging and epilepsy, this approach may also be useful for conditions such as traumatic brain injury, ADHD, and autism, where hidden abnormal brain activity could help guide diagnosis or treatment.
Presentation of business plan to the technology transfer office at TAU and to Joy Ventures, as well as analysis of freedom-to-operate, yielded the following main insight: commercialization of this technology likely requires to combine it with an intervention (e.g. drug, stimulation device) to go beyond diagnostics-only and be attractive for investment. Accordingly, we are beginning further research into incorporating our technology within a closed-loop system that also employs stimulation.
Our findings show that we can now measure abnormal medial temporal lobe activity during sleep in a reliable and noninvasive way, and that this activity is directly related to problems with thinking and memory. This breakthrough has important consequences for both basic science and clinical care. For researchers, it provides a new window into human hippocampal activity without the need for brain implants, allowing studies that connect detailed brain-circuit mechanisms with real human behavior. The same approach could be adapted to detect other important brain signals during sleep, such as those involved in forming memories, opening new opportunities to study learning, emotions, and decision-making.
Clinically, this work answers a major unmet need. Current early Alzheimer’s tests—like PET or MRI scans, spinal fluid tests, or blood biomarkers—cannot capture this abnormal electrical activity in memory-related brain circuits, and they are expensive, invasive, or lack fine detail. Our method offers a practical alternative that can detect this activity during sleep and could be used at home or for long-term monitoring. Beyond aging and epilepsy, this approach may also be useful for conditions such as traumatic brain injury, ADHD, and autism, where hidden abnormal brain activity could help guide diagnosis or treatment.
Presentation of business plan to the technology transfer office at TAU and to Joy Ventures, as well as analysis of freedom-to-operate, yielded the following main insight: commercialization of this technology likely requires to combine it with an intervention (e.g. drug, stimulation device) to go beyond diagnostics-only and be attractive for investment. Accordingly, we are beginning further research into incorporating our technology within a closed-loop system that also employs stimulation.