In order to foster new cross-disciplinary collaborations, the EPFL Presidency launched in 2019 a call for proposals for an Interdisciplinary Seed Fund. Six projects were selected in 2020.
This project aims to integrate the most-advanced artificial intelligence technologies into cutting-edge clinical applications to understand and treat neurological deficits.
Professors Courtine and Bloch (G-LAB UPCOURTINE) recently founded NeuroRestore, a center that combines preclinical and clinical research programs to understand and treat motor deficits after spinal cord injury, stroke, and Parkinson disease. In particular, they are developing neuromodulation therapies delivered in the brain or spinal cord to alleviate motor deficits in patients with spinal cord injury and Parkinson disease. The personalization of these therapies requires high-resolution recordings of movement patterns, muscle activity, and responses to stimulation parameters. Currently, these procedures involve time-consuming recordings, complex analyses, and expensive technologies. These limitations hinder the widespread applications of these therapies.
Recent work from Professor Alahi’s group (VITA) has shown the possibility of extracting joint positions in real time through conventional cameras and of reconstructing the body posture with high fidelity in a noisy environment. Initially designed for detection of pedestrians and social interactions by smart robots, this approach is well-suited for high-accuracy monitoring of motor functions and gait analyses for widespread deployment and acquisition of large-scale databases. Using only conventional cameras, this strategy is ideal for monitoring and controlling motor deficits outside of the laboratory environment, offering disruptive opportunities to detect deficits and titrate neuromodulation therapies in conventional hospital settings and even patients’ home.
The brain computations responsible for making decisions are challenging to unravel. Scientists have started to peel away the layers of complexity surrounding such processes by recording neuronal activity and interneuronal communications in model organisms, such as the nematode worm, Caenorhabditis elegans. These efforts have revealed that a specific set of neurons, ‘interneurons’, are responsible for decision making. However, we lack pinpoint methods to perturb these neurons to understand their specific functions. Reactive natural products —created in most, if not all, organisms as a result of conserved metabolic processes— regulate the functions of a host of proteins. Many of these reactive small molecules are elevated in areas of high metabolic activity, particularly neurons. Indeed, adduction of several reactive natural products to neuronally-localized proteins is linked to their general regulation and also to the onset of neuronal diseases in certain contexts. These small molecules thus offer ideal ways to chemically control neuronal functions that could give real-world relevant insight into how decision-making processes work. However, the capricious nature of these reactive molecules renders traditional methods limited in harnessing their full potential.
This project combines state-of-the-art techniques to map neuronal networks and circuits (Professor Rahi’s laboratory, LPBS), focusing on specific interneurons and chemical-biology methods to tame metabolites and related reactive drugs and shepherd them to specific cell types, or specific proteins therein (Professor Aye’s laboratory, LEAGO), in Caenorhabditis elegans.
This new collaboration seeks to control functions of specific neurons using precision chemical-biology tools, 4D imaging, and network modeling, so that individual decision-making events can be tied to specific proteins, cells or subcellular regions. This project is aimed at developing new means to control neuronal cell functions and understanding neuronal circuits and specific roles of specific reactive natural metabolites and drugs impacting those circuits.
Being able to image the morphology, connectivity and real-time activity of a complex set of neurons and, ultimately, of a living brain, is arguably one the most ambitious goals of neuroscience and neuro-imaging. Continuous progress in this direction is providing new fundamental knowledge about the hierarchical organization of in-vitro networks and slices of mammalian brain, where information is encoded and distributed from individual neurons and synapses up to macroscopic areas showing coordinated patterns.
This project is aimed at developing a new platform for imaging of neuronal activity consisting of nanostructured diamond containing nitrogen-vacancy centers, which will act as remote quantum sensors of the magnetic and electric fields generated by varying membrane potentials linked to neuronal activity. This platform will be designed so as to be compatible with advanced existing tools such as multi-electrode arrays, genetically encoded voltage and calcium indicators, and optogenetics stimulation, in order to allow for a multi-modal study of neural networks.
All microscopic currents related to neuronal activity (action potentials and ion channels) imply the existence of an extremely small magnetic field, quickly fading away with distance from the cell. Similarly, the membrane potential is associated with an electric field, rapidly screened by the intercellular ionic medium away from the membrane. The proposed approach consists of culturing neurons directly on top of a nanostructured diamond surface to ensure close proximity to nitrogen-vacancy (NV) centers, while optimizing the nanophotonic architecture for efficient NV fluorescence excitation and collection.
On the long term, this project aims to lay the ground for a disruptive real-time neuro-imaging technique, which will have the potential to answer open clinical questions, such as detecting early signs of neurodegenerative diseases in neuron networks.
This project offers an innovative device for wireless neurostimulation. The wireless device is an array of thousands freestanding, ultra-small, and individually addressable CMOS-pixels, embedded into a conformable mesh for easy surgical placement onto the human visual cortex. Each CMOS-pixel is an active element able to receive a wireless signal and convert it into an electrical pulse to stimulate the nervous system.
Visual prostheses are used to revert blindness: a medical condition affecting more than 39 million people worldwide (World Health Organisation). Cortical visual prostheses (CVP) might address every form of blindness regardless of its origin. The first attempts in history to revert blindness with implantable prostheses were using cortical stimulators. Despite this large potential, CVPs have not reached a significant clinical development because of technical limitations, such as a lack of mechanical compliance, complex surgical approaches related to the wiring from the neural interface to the implantable pulse generator (IPG), and low resolution due to the limited number of electrodes that can be addressed with wires and implantable IPGs.
The proposed approach will overcome, in a single step, all the above-mentioned issues limiting the clinical development of CVPs. It will provide a medical therapy to a very large proportion of the blind population worldwide. This approach will also find applications in every implantable stimulation device currently in medical use, such as deep brain stimulators (for Parkinson disease, epilepsy, obsessive-compulsive behavior, and addiction), spinal cord stimulators (for motor recovery and pain relief), cochlear implants (to revert deafness), pacemakers (for cardiac dysfunctions), etc.
This research aims to develop a new immersive environment in order to support novel human interaction with billions of objects, enabling us to engage with the world of Big Data.
Big Data is indisputably increasing in all fields of research. In astrophysics, with the upcoming gigantic observatories such as the Square Kilometer Array Observatory (SKAO) , the Vera Rubin Observatory (formerly LSST) , or the Cherenkov Telescope Array (CTA) , scientists will have to deal with Exa-Byte levels of data ! In parallel to the Universe observations, state-of-the-art cosmological numerical simulations have reached a complexity today whereby several billion mass elements can be included with the complex astrophysics and up to several trillions when only Gravity is included. Chief among the different challenges of Big Data, is its visualisation. While machine learning and neural networks can assist scientists in data exploration, the visualization of data remains vital for researchers to undertake accurate analysis and interpretation. Combining the skills of LASTRO in cosmological observations and simulations, the immersive interactive visualisation experience of eM+, and the IIG’s expertise in embodied interactions, this project aims to develop new methods that will enable the exploration of gigantic surveys and simulations via virtual reality environments. Exploring means traveling not only through positions but also through the velocity fields, as well as accessing the complete physical contents of simulations. The project will also enable users to visualise stars and galaxies and to access their fascinating physical properties.
 1 Exa Byte = 1018 bytes
Hagfeldt, Anders (Uppsala University)
Advances in fundamental particle physics more and more evidently become limited by the available technological solutions for the next generation of particle detectors and accelerators. Inevitable challenges are due to requirements on detector radiation hardness and granularity, both of which are necessary to cope with high flux of incoming particles, coupled with the need for large dimensions.
A standard material of choice for charged particle and photon detection is silicon, which has relatively low flexibility in the choice of the electrical and optoelectronic properties and which at the same time requires high standards for its handling. Complicated manufacturing processes translate into the high cost of the detectors employing silicon, and hence traditional silicon usage is not possible on the very large scale required by the new high-energy physics frontiers.
Now for the first time, after decades of silicon dominance, a new class of materials – perovskites – has appeared which could radically change the detector industry approach for high-energy physics and related fundamental sciences. So far, the numerous studies carried out around the globe, and at EPFL in particular, have concentrated on the industrial application of perovskites for solar cell production with priorities dictated by the high photovoltaic efficiency and low handling costs.
The very same arguments hold for the usage of perovskites in particle detectors. This application case adds another set of criteria to be optimized as compared to the solar cells. However, the very fact that it is conceivable to engineer a material according to its desired features opens a unique possibility for a breakthrough in particle detection. EPFL, which brings together a world-leading expertise for solar cells development and particle-detector R&D know-how, is the best place to find the most suitable perovskite candidates for high energy physics purposes, to prove their applicability with dedicated measurements, to make use of them in actual projects, and finally to prove and create a new paradigm in particle detectors.