Periodic Reporting for period 3 - ConnectToBrain (Connecting to the Networks of the Human Brain)
Période du rapport: 2022-09-01 au 2024-02-29
Non-invasive modulation of the brain with transcranial magnetic stimulation (TMS) has demonstrated good results and has been approved widely for clinical use and partly insurance coverage. However, with the current technology, the locus (position) in the brain of the stimulation can only be changed slowly and real-time information of the electrical brain state is not taken into account, limiting the therapies to a single brain site and open-loop stimulation. The ConnectToBrain project will develop for clinical use a transformative multi-locus closed-loop TMS technique, with 3 main areas of development:
1) Multi-locus TMS (mTMS) coil array, power and control electronics, and user interface, the array covering most of the cortical mantle and allowing real-time high-precision control over the locus, direction, intensity, and timing of the stimulation pulses.
2) Real-time analysis of brain activity and connectivity by using information from high-density electroencephalography (hd-EEG) and structural and functional magnetic resonance imaging (MRI) for brain-state-dependent and closed-loop stimulation.
3) Demonstration of feasibility, safety, and efficacy of the developed techniques and methods, and their therapeutic utility in dysfunctional brain networks in brain disorders such as Alzheimer’s disease and motor stroke.
By the end of the project, we expect to have developed new technology capable of correcting dysfunctional brain networks in several brain disorders with better therapeutic efficacy than current state-of-the-art techniques and the potential to induce a major paradigm shift in therapeutic neuromodulation. If successful and widely transferred for clinical use, ConnectToBrain technology and methodology will lead to a substantial reduction of suffering and economic burden caused by brain disorders.
1) Multi-locus TMS (mTMS) coil array, power and control electronics, and user interface, the array covering most of the cortical mantle and allowing real-time high-precision control over the locus, direction, intensity, and timing of the stimulation pulses.
2) Real-time analysis of brain activity and connectivity by using information from high-density electroencephalography (hd-EEG) and structural and functional magnetic resonance imaging (MRI) for brain-state-dependent and closed-loop stimulation.
3) Demonstration of feasibility, safety, and efficacy of the developed techniques and methods, and their therapeutic utility in dysfunctional brain networks in brain disorders such as Alzheimer’s disease and motor stroke.
By the end of the project, we expect to have developed new technology capable of correcting dysfunctional brain networks in several brain disorders with better therapeutic efficacy than current state-of-the-art techniques and the potential to induce a major paradigm shift in therapeutic neuromodulation. If successful and widely transferred for clinical use, ConnectToBrain technology and methodology will lead to a substantial reduction of suffering and economic burden caused by brain disorders.
The corona crisis slowed down most aspects of the work. As a result, the project duration was extended by 12 months. Additional delays were due to unforeseen technical challenges to engineer the device and to produce documentation to conform to European medical device regulations (MDR). A significant additional delay of more than a year was due to getting regulatory approval in compliance with MDR for the already installed device to be used for human studies at EKUT.
At AALTO, we first experimented with the previously built 2-coil mTMS device to help develop closed-loop algorithms for ConnectToBrain multi-coil systems. We then developed, built, thoroughly tested, and put to experimental use the planned 5-channel prototype. The technical development required more work than initially anticipated. Consequently, we had to hire more people in the hardware and software groups to speed up the project. An improved 5-channel prototype, with novel details, including comprehensive safety features and compatibility for hospital use, was delivered at EKUT in late 2022 and at UdA in September 2023. We have started designing the next generation of devices, first a 12-channel mTMS system. We have developed solutions for TMS-compatible EEG and algorithms to suppress EEG signal artifacts and noise. We have also introduced a new EEG cap design for reducing TMS-related artifacts in EEG. Also, we have made a major breakthrough towards algorithmically guided, closed-loop TMS. Namely, our new Bayesian algorithm can change stimulation parameters based on either electromyographic (EMG) or EEG responses, automatically finding optimal stimulation parameters such as those that maximize the motor evoked potential (MEP) in a target muscle.
At EKUT, we have completed proposed experiments on brain-state-dependent TMS in simple 2-node motor networks in healthy subjects. We confirmed our hypothesis that effective connectivity between the the two nodes is most strongly expressed if probed when the local µ-rhythm in the two nodes is in-phase rather than out-of-phase. We have completed experiments to test the Communication Through Coherence theory, one of the central neurobiological hypotheses of the ConnectToBrain project, by repetitively stimulating the two nodes at specific phase relations of the ongoing µ-rhythm for plasticity induction. The findings overall were non-significant. . In addition, we have completed several studies that elaborate on principles of signal-to-noise ratio for real-time estimation of the phase in EEG-recorded oscillations, and on causal decoding of EEG data for estimation of cortical excitability at the individual level. We have extended the application of real-time analysis of EEG for brain-state-dependent TMS to the alpha- and theta-rhythms of prefrontal cortex. We have completed data acquisition in a proof-of-principle experiment of closed-loop TMS in a 2-node motor network in healthy subjects with conventional TMS. The experiments were preceded by extensive simulation studies to demonstrate feasibility of deep reinforced learning algorithms. The findings demonstrate the capability of the deep machine-learning algorithm to fully automatically identify the phase of the µ-rhythm associated with the highest or lowest excitability of the 2-node motor network in a given individual, which is breakthrough in the C2B project. We will validate these findings using the EKUT 5-channel mTMS device.
At UDA, we have worked towards establishing algorithms to assess whole-brain connectivity in real time from EEG data. We have demonstrated that a reliable connectivity can be estimate in real-time with a data analysis window of about 5 cycles. Using data acquired at EKUT, we found a relationship between mu rhythm functional connectivity in the motor network and Motor Evoked Potential (MEP). In a larger cohort, we identified, using Hidden Markov Models, fast large-scale brain states lasting few hundreds of milliseconds with a specific spatial, temporal and spectral fingerprint and variable relation to MEP amplitude. Using motor cortex stimulation data from the AALTO 2-coil mTMS system, we showed that connectivity between the left Supplementary Motor Area (SMA) and left and right M1 is modulated by the stimulation orientation over the left SMA. Finally, we ran an experiment using the 5-coil mTMS system in AALTO, exploiting the possibility to stimulate different locations. In parallel, we developed a novel algorithm for cross-frequency coupling based on multidimensional data. We also collected a large multimodal dataset to exploit the strength of machine learning techniques with the goal to identify fast-dynamic large-scale brain states that predict corticospinal excitability. Overall, the results of these studies are crucial for assessing the role of connectivity in brain-state-dependent stimulation at multiple sites by mTMS.
At AALTO, we organized the BrainSTIM conference in May 2020 (virtual) and in June 2023 (physical, in Espoo, Finland), and the 8th and 9th Science Factories: TMS–EEG Summer Schools and Workshops in September 2022 and May 2023. In addition, all 3 sites together have presented special scientific sessions on the ConnectToBrain theme, for instance, in the Basic and Clinical Multimodal Imaging Conference (“BaCI”, virtual, October 2021), in the Brain Stimulation Conference (Charleston, SC, USA, December 2021), in 32nd International Congress of Clinical Neurophysiology (Geneva, CH, September 2022), 22nd International Biomagnetism Conference (Birmingham, UK, August–September 2022), the annual meetings of the German Society of Clinical Neurophysiology in March 2023 in Hamburg and 2024 in Frankfurt.
At AALTO, we first experimented with the previously built 2-coil mTMS device to help develop closed-loop algorithms for ConnectToBrain multi-coil systems. We then developed, built, thoroughly tested, and put to experimental use the planned 5-channel prototype. The technical development required more work than initially anticipated. Consequently, we had to hire more people in the hardware and software groups to speed up the project. An improved 5-channel prototype, with novel details, including comprehensive safety features and compatibility for hospital use, was delivered at EKUT in late 2022 and at UdA in September 2023. We have started designing the next generation of devices, first a 12-channel mTMS system. We have developed solutions for TMS-compatible EEG and algorithms to suppress EEG signal artifacts and noise. We have also introduced a new EEG cap design for reducing TMS-related artifacts in EEG. Also, we have made a major breakthrough towards algorithmically guided, closed-loop TMS. Namely, our new Bayesian algorithm can change stimulation parameters based on either electromyographic (EMG) or EEG responses, automatically finding optimal stimulation parameters such as those that maximize the motor evoked potential (MEP) in a target muscle.
At EKUT, we have completed proposed experiments on brain-state-dependent TMS in simple 2-node motor networks in healthy subjects. We confirmed our hypothesis that effective connectivity between the the two nodes is most strongly expressed if probed when the local µ-rhythm in the two nodes is in-phase rather than out-of-phase. We have completed experiments to test the Communication Through Coherence theory, one of the central neurobiological hypotheses of the ConnectToBrain project, by repetitively stimulating the two nodes at specific phase relations of the ongoing µ-rhythm for plasticity induction. The findings overall were non-significant. . In addition, we have completed several studies that elaborate on principles of signal-to-noise ratio for real-time estimation of the phase in EEG-recorded oscillations, and on causal decoding of EEG data for estimation of cortical excitability at the individual level. We have extended the application of real-time analysis of EEG for brain-state-dependent TMS to the alpha- and theta-rhythms of prefrontal cortex. We have completed data acquisition in a proof-of-principle experiment of closed-loop TMS in a 2-node motor network in healthy subjects with conventional TMS. The experiments were preceded by extensive simulation studies to demonstrate feasibility of deep reinforced learning algorithms. The findings demonstrate the capability of the deep machine-learning algorithm to fully automatically identify the phase of the µ-rhythm associated with the highest or lowest excitability of the 2-node motor network in a given individual, which is breakthrough in the C2B project. We will validate these findings using the EKUT 5-channel mTMS device.
At UDA, we have worked towards establishing algorithms to assess whole-brain connectivity in real time from EEG data. We have demonstrated that a reliable connectivity can be estimate in real-time with a data analysis window of about 5 cycles. Using data acquired at EKUT, we found a relationship between mu rhythm functional connectivity in the motor network and Motor Evoked Potential (MEP). In a larger cohort, we identified, using Hidden Markov Models, fast large-scale brain states lasting few hundreds of milliseconds with a specific spatial, temporal and spectral fingerprint and variable relation to MEP amplitude. Using motor cortex stimulation data from the AALTO 2-coil mTMS system, we showed that connectivity between the left Supplementary Motor Area (SMA) and left and right M1 is modulated by the stimulation orientation over the left SMA. Finally, we ran an experiment using the 5-coil mTMS system in AALTO, exploiting the possibility to stimulate different locations. In parallel, we developed a novel algorithm for cross-frequency coupling based on multidimensional data. We also collected a large multimodal dataset to exploit the strength of machine learning techniques with the goal to identify fast-dynamic large-scale brain states that predict corticospinal excitability. Overall, the results of these studies are crucial for assessing the role of connectivity in brain-state-dependent stimulation at multiple sites by mTMS.
At AALTO, we organized the BrainSTIM conference in May 2020 (virtual) and in June 2023 (physical, in Espoo, Finland), and the 8th and 9th Science Factories: TMS–EEG Summer Schools and Workshops in September 2022 and May 2023. In addition, all 3 sites together have presented special scientific sessions on the ConnectToBrain theme, for instance, in the Basic and Clinical Multimodal Imaging Conference (“BaCI”, virtual, October 2021), in the Brain Stimulation Conference (Charleston, SC, USA, December 2021), in 32nd International Congress of Clinical Neurophysiology (Geneva, CH, September 2022), 22nd International Biomagnetism Conference (Birmingham, UK, August–September 2022), the annual meetings of the German Society of Clinical Neurophysiology in March 2023 in Hamburg and 2024 in Frankfurt.
We have already progressed beyond the state of the art in three major ways.
1) We have constructed a multi-locus TMS device, with 5 overlapping coils: one round, two figure-of-eight, two cloverleaf-shaped. When activated simultaneously, these coils can produce a focal electric field in the brain. By adjusting the relative intensities of the 5 coil currents, the focal point of stimulation in the cortex can be electronically defined; these stimulation points can be changed in less than a millisecond, allowing spatially distributed sequences of TMS pulses that were not possible before. This work has involved the design of the coils, power electronics, control electronics, navigation, patient support systems, software architecture and user interface design, all of which include novel features that will allow safe and efficient use of the technology.
2) We have developed algorithms for analyzing electromyography EMG and EEG data for brain-state-dependent and closed-loop brain stimulation. We have developed algorithms to predict the phase of EEG-recorded oscillations at sensor and source levels in real time, to deliver TMS pulses at desired phases of the oscillatory activity. We have also developed algorithms that remove noise and artifacts from the EEG signal, in particular eye-movement and TMS-evoked muscle artifacts. Furthermore, we have developed algorithms for determining time-dependent changes of functional connectivity in the brain based on EEG; this will allow to perform state-dependent and closed-loop stimulation based on connectivity information. Moreover, we have developed closed-loop rTMS based on deep reinforced learning algorithms to enhance connectivity in nodes of the motor network.
3) We have demonstrated with existing one-coil technology that TMS efficacy for induction of long-term plasticity in motor cortex can be enhanced by triggering the pulses at the right phase of alpha-frequency EEG signals. We have demonstrated with mTMS that one can find cortical representation areas without moving the transducer and with a much smaller number of pulses than with traditional methods.
By the end of the project, we expect to have developed new technology capable of correcting dysfunctional brain networks in several brain disorders with better therapeutic efficacy than current state-of-the-art techniques and the potential to induce a major paradigm shift in therapeutic neuromodulation. If successful and widely transferred for clinical use, ConnectToBrain technology and methodology will lead to a substantial reduction of suffering and economic burden caused by brain disorders.
1) We have constructed a multi-locus TMS device, with 5 overlapping coils: one round, two figure-of-eight, two cloverleaf-shaped. When activated simultaneously, these coils can produce a focal electric field in the brain. By adjusting the relative intensities of the 5 coil currents, the focal point of stimulation in the cortex can be electronically defined; these stimulation points can be changed in less than a millisecond, allowing spatially distributed sequences of TMS pulses that were not possible before. This work has involved the design of the coils, power electronics, control electronics, navigation, patient support systems, software architecture and user interface design, all of which include novel features that will allow safe and efficient use of the technology.
2) We have developed algorithms for analyzing electromyography EMG and EEG data for brain-state-dependent and closed-loop brain stimulation. We have developed algorithms to predict the phase of EEG-recorded oscillations at sensor and source levels in real time, to deliver TMS pulses at desired phases of the oscillatory activity. We have also developed algorithms that remove noise and artifacts from the EEG signal, in particular eye-movement and TMS-evoked muscle artifacts. Furthermore, we have developed algorithms for determining time-dependent changes of functional connectivity in the brain based on EEG; this will allow to perform state-dependent and closed-loop stimulation based on connectivity information. Moreover, we have developed closed-loop rTMS based on deep reinforced learning algorithms to enhance connectivity in nodes of the motor network.
3) We have demonstrated with existing one-coil technology that TMS efficacy for induction of long-term plasticity in motor cortex can be enhanced by triggering the pulses at the right phase of alpha-frequency EEG signals. We have demonstrated with mTMS that one can find cortical representation areas without moving the transducer and with a much smaller number of pulses than with traditional methods.
By the end of the project, we expect to have developed new technology capable of correcting dysfunctional brain networks in several brain disorders with better therapeutic efficacy than current state-of-the-art techniques and the potential to induce a major paradigm shift in therapeutic neuromodulation. If successful and widely transferred for clinical use, ConnectToBrain technology and methodology will lead to a substantial reduction of suffering and economic burden caused by brain disorders.