Using industry standards for low power sensor integration and new microfluidic standards for OoC connection and control, we develop the basis for an open technology platform for the digital twin principle where sensors and in vitro assays are combined and provide additional data about patient-specific response.
DIGIPREDICT will develop ambulatory wearable for early COVID-19 symptoms detection as well as the next generation of Vasculature- and Heart-on-chip systems.
When the body detects a viral infection, it begins producing large amounts of cytokines, which are proteins that signal the immune response. But in some COVID-19 patients it results in a ‘cytokine storm,’ an excessive immune-system response that can lead to seriously damaging cardiovascular dysfunction. Our goal is to be able to detect the first signs of these storms and track them in real-time, which would be a major step forward in the treatment of high-risk patients.
To do so, biomarkers sensors will be designed and fabricated to collect analytes in sweat and detect biomarkers for cytokine storm and cardiovascular system. This requires the combination of scientific and technical excellence in multiple disciplines.
The core of the DIGIPREDICT digital twin will be a smart patch with integrated technology for collecting a range of medical data, such as blood oxygen levels, breathing rate and body temperature.
The patches will also include nanosensors linked to an artificial-intelligence (AI) smartphone application in order to continually track specific biomarkers that indicate a cytokine storm may be brewing.
Ascilion will design and manufacture chips with arrays of ultra-sharp, hollow microneedles in monocrystalline silicon using micro electro mechanical system (MEMS) technology.
A typical microneedle chip (10x10 mm2) will contain about one hundred 0.5 mm high microneedles where the bores or internal microneedle lumens are connected in a fluidic capillary system designed to interface to biomarker sensors on the back side of the chip.
Such chipset will be developed by IMEC-NL to convert the output of the IS-FET sensor into a digital signal that can then be transferred to a compact electronic assembly developed by IMEC-NL, which in turn can transmit the information over a radio link to an external receiver.
Respiratory activity (RA) monitoring is vital for pulmonary impaired patients. Compact strain sensors offer excellent advantages for their operation as standalone, easy to use, low cost sensors. Ultra-small form factors of carbon nanotube devices (CNT) have unique advantages over larger, composite materials.
The high gauge factors of specific CNTs in the high strain regime make them excellent material candidates for achieving high sensitivity of respiratory activity monitoring. ETHZ will develop such individual CNT devices that will be integrated onto skin compatible polymers to perform RA monitoring.
A battery operated and wireless enabled wearable patch will be developed by IMEC-NL for the sensing of respiratory-related physiological symptoms. This wearable will target measuring respiratory parameters (rhythm), oxygenation (SpO2), cardiac (heart rate) and contextual information (skin temperature). The patch will be able to collect both un-processed data (such as temperature, optical and biopotential data) and output of embedded algorithms (respiration rhythm and oxygenation).
The integration into a multi-modal sensory system and the testing of a wearable version of microphones for body sounds will be performed by EPFL, and the information collected by this wearable will be transmitted wirelessly and made available for in-hospital processing and interpretation.
Organs-on-Chips (OoC) are microfluidic devices in which living tissues are cultured under well-controlled and dynamic conditions. In the chips, tissue function is continuously monitored, both in stable conditions and when they are challenged with disease stimuli.
By controlled integration of patient-specific aspects (e.g. their stem cells, their blood plasma, or their biomarker profiles), OoC can be regarded as ‘physical avatars’ of particular patients. This means that the response of the organ-on-chip to a disease stimulus or therapeutic intervention will be analogous to the response of a particular patient.
Heart-on-Chip and Vessel-on-Chip models will be developed by integrating human stem cell-derived heart tissues and vascular tissues in microfluidic chips. The chips will include electrical pacing, active blood-like perfusion and three-dimensional organization to realistically model the situation in the human body.
By using microfluidic ‘transistors’, UTWENTE will build a platform that can automatically control hundreds of individual heart-on-chip and vessel-on-chip microdevices. The platform will be used to expose the chips to disease-like episodes with dynamic biomarker values (e.g. heart rates, inflammatory cytokine levels). The resulting data will be used to build a predictive map of disease progression.
CMOS-based microelectrode array (MEA) chips will be implemented for the Heart-on-Chip application. These high-density MEA chips will offer high spatial and temporal resolution for electrical activity recording and pacing of the cells in the OoC devices.
IMEC-BE will also design and fabricate a new type of chip consisting of flexible mesh-like electrode arrays to serve as a barrier measurement device for heart-on-chip and vasculature-on-chip.
The Machine Learning (ML) algorithms and predictive models that build up the Digital Twin will be deployed in a distributed fashion, running most of the computation required to calculate high-level pathophysiological parameters directly in the sensor devices. Edge computing delivers advanced real-time processing capabilities of sensed data, and it can be combined with cloud-assisted analytics. Both together facilitate coordinated and multi-layer learning for faster and personalized decision-making at the edge.
Novel ultra-low-power System-on-Chip (SoC) architectures will enable this and the training of complex models based on the biomarkers' dynamics possible, enriched through federated learning with medical records available at the hospital. The whole system improves the responsiveness to predict the trajectory of the disease while preserving the patient's privacy.
Data evolution and prediction from physiological time-series analysis, causality and ML models will be communicated through interactive graphical user interfaces and comprehensive representations of clinical outcomes. Visualization tools will be developed by EPFL to organize and display the available information in a way that can be processed and analyzed quickly and intuitively.
In the design process the main requirements will be to: (i) provide innovative and informative data representations by using colors, symbols, and interactions; (ii) give answers faster, allow to understand patterns of change quickly; (iii) highlight things that would rather go unnoticed (trends, dependencies); (iv) allow interactivity with information displayed, investigate cause-effect relationships.
DIGIPREDICT is designed to be in permanent exchange with the scientific community, industry and patients. EPFL, as coordinator and dissemination manager, will coordinate the DIGIPREDICT scientific dissemination and training activities.
SCIPROM leads the dissemination to the general public and IMEC-NL will oversee the exploitation of the results. EPOS will lead the community building and ensure the DIGIPREDICT strives to set up an interdisciplinary network across Europe centered on Digital Twins in strict compliance with the European legislation and ethics requirement.