Periodic Reporting for period 1 - RELIEVE (First Closed-loop non-Invasive Seizure Prevention System)
Okres sprawozdawczy: 2023-04-01 do 2024-03-31
The goal of Project RELIEVE is to build the very first non-invasive effective closed-loop monitoring and intervention system for brain- related disorders. The outcomes can be used to treat or manage various psychiatric and neurological disorders. We do this by pushing the technological boundaries in two separate domains. (1) AI domain: We develop a class of mathematical and algorithmic tools for brain data based on the recently developed mathematical theories that are potentially useful for the real-time monitoring of the brain data. Such advancements have shown promise in robotics and self-driving cars and have high potential to work efficiently in highly dynamic environments where personalization and low computation power is a must. This makes such algorithms runnable using brain data such as EEG (electric activity on the head surface) on ordinary wearable devices. (2) Neurostimulation domain: We combine two characteristics of the ultrasound waves in stimulation and imaging of the nervous system to build the first smart-navigated wearable ultrasound patch. We also choose the vagus nerve as the target for neurostimulation as one of the most promising sites to interact with the nervous system with proven implications for a large spectrum of neurological and psychiatric disorders (e.g. dementia, depression, epilepsy, etc.). We call this unit 'WU-VNS' standing for non-invasive wearable ultrasound vagus nerve stimulation. In the next three years, we will use epilepsy as the first use case of the developed technologies to train and test a closed-loop system. For this, the AI monitors the brain through a patch that records EEG in addition to other physiological measures such as heart rate and motion. Upon prediction of a forthcoming anomaly (seizure in this case) by the AI unit, the neurostimulation module activates and stimulates the vagus nerve non-invasively. During this process also a so-called 'active learning' happens in which the AI learns from the reactions of the nervous system to the stimulation protocol and can fine-tune the protocol for future interventions. To achieve this, we have designed a complex phase-based development and testing plan: The first two generations (Gen. 1 and 2) are the intermediate versions of the full system and act in open loops validating each of the AI and neurostimulation subsystems. These two generations already have high potentials to independently turn into medical device products with large market needs. Ultimately, in Gen. 3 we close the loop by integrating the AI and the WU- VNS and consequently validate the efficacy and the usability of the system.
Within the first year:
The technical outcomes so far in the first year are as follows: (1) We have successfully built such an EEG device taking into account the final regulatory concerns as a medical device; (2) We successfully built an AI already with higher-than-expected accuracy. Testing the algorithms has so far been done on the (limited accessible already existing data) from the epilepsy monitoring units (EMUs); (3) We have also succeeded embedding such an AI (with same complexity level) on a similar device with reasonable computational power, memory, and battery usage; (4) We have managed to engage and include the opinions of various stakeholders including the patients, their family members and the clinicians for a patient-centric design.
In terms of outcomes to make such solutions available for the patients, we have achieved the following: (1) Taking significant steps to build a technical dossier for the clinical testing of the device; (2) Preparing the regulatory dossier and engaging the regulatory bodies for ultimate approval of the device both in Europe (CE) and in the US (FDA); (3) Constructing a rigid business case for the device in various geographies taking into account various stakeholders interests.
The technical outcomes so far in the first year are as follows: (1) We have successfully built such an EEG device taking into account the final regulatory concerns as a medical device; (2) We successfully built an AI already with higher-than-expected accuracy. Testing the algorithms has so far been done on the (limited accessible already existing data) from the epilepsy monitoring units (EMUs); (3) We have also succeeded embedding such an AI (with same complexity level) on a similar device with reasonable computational power, memory, and battery usage; (4) We have managed to engage and include the opinions of various stakeholders including the patients, their family members and the clinicians for a patient-centric design.
In terms of outcomes to make such solutions available for the patients, we have achieved the following: (1) Taking significant steps to build a technical dossier for the clinical testing of the device; (2) Preparing the regulatory dossier and engaging the regulatory bodies for ultimate approval of the device both in Europe (CE) and in the US (FDA); (3) Constructing a rigid business case for the device in various geographies taking into account various stakeholders interests.
So far, we have developed a reduced montage EEG system which is more energy efficient and probably the first in kind with the aim of ultra-long recordings. It is also the very first brain monitoring system that is equipped with an embedded AI in it. Furthermore, the AI in this device has demonstrated beyond the state of the art performance in terms of specificity and sensitivity for seizure prediction. ABT has already made one patent on the whole concept of the system and its methods. Furthermore, the clinical studies have already started before the expected time due to a mitigation process in which we divided the clinical testing into two distinct sections one on the validation of the AI only using data from the standard of care. The second part is testing the developed device's signal quality in the patients in first-in-human trials.