Welcome to the Neurotech Society Journal Club, where we dive deep into the cutting-edge academic papers driving innovation in the world of neurotechnology. Our sessions are designed for members to engage with the latest research spanning brain-computer interfaces, neural engineering, neuromodulation, and beyond. Whether you're an expert in the field or just starting, the Journal Club is a collaborative space to explore, discuss, and challenge ideas, helping everyone expand their knowledge of the fascinating world of neurotech.
We believe in building a community-driven experience, which is why we encourage you to take an active role in shaping our discussions. Have a paper you think the society should explore? Submit it using the form at the bottom of this page! Whether it’s a ground-breaking study or a topic you're curious about, we value your suggestions and look forward to seeing what you bring to the table.
Our Journal Club operates through our dedicated 'Journal Club & Resources' WhatsApp group within the ICL Neurotech community, making it easy to stay connected and informed. Sessions are held in a hybrid format, with in-person meetings for those at Imperial, complemented by virtual participation via Zoom. This ensures accessibility for all members, no matter where you are!
This paper demonstrates advancements in decoding visual perception from brain activity using generative and foundational AI systems. The authors compare functional Magnetic Resonance Imaging (fMRI) and magnetoencephalography (MEG) for their respective strengths in brain decoding: fMRI offers high spatial resolution but low temporal resolution (~0.5 Hz), while MEG captures high temporal resolution (~5000 Hz), making it more suitable for real-time applications. They introduce an MEG-based decoding model, integrating pretrained image embeddings, an MEG module trained end-to-end, and a pretrained image generator. Their results highlight three key findings: (1) the MEG decoder achieves 7X improvement in image retrieval over traditional linear decoders, (2) decoding with the DINOv2 model, a recent foundational image model, best captures late brain responses, and (3) MEG can decode high-level visual features, while fMRI performs better for low-level features. These findings mark progress toward real-time decoding of visual processes in the brain, though results remain preliminary.
TAGS: MEG, fMRI, Brain Decoding, Visual Perception, Neuroimaging, Generative AI, Temporal Resolution, High-level Visual Features, Image Retrieval, Real-time Brain Decoding, DINOv2, Foundational Models
This paper dives into the fascinating challenge of how to measure brain activity at an incredibly detailed level across the entire brain, something current techniques can’t fully achieve. It explores the limitations of existing methods—like electrodes not capturing enough neurons or light scattering through brain tissue—and highlights how these physical barriers are holding us back. But it doesn't stop there. The authors suggest exciting possibilities, like embedding sensors directly into brain tissue and using wireless transmission, to overcome these obstacles and revolutionize how we map brain activity.
TAGS: Brain Mapping, Neural Recording, Optical Methods, Electrical Recording, Magnetic Resonance, Molecular Sensors, Wireless Data, Brain Activity
This paper tackles the exciting challenge of harnessing living neurons for computing, moving beyond traditional artificial neural networks to what’s called "wetware computing" with organoids. The researchers developed a groundbreaking platform that enables large-scale experiments on neural organoids, which can survive over 100 days, continuously monitoring and stimulating them. The system automates tasks like fluid changes and allows researchers to run complex, 24/7 experiments remotely using a Python-based API, gathering massive amounts of data. This innovation opens up new possibilities for exploring biological computing and advancing AI through biological systems.
TAGS: Wetware Computing, Organoid Intelligence, Neural Organoids, Electrophysiology, Artificial Intelligence, Bio-AI, Remote Research, Neuroplatform, Deep Learning, Reinforcement Learning, Biological Computing
This paper highlights the unique role of fMRI in bridging different areas of neuroscience, from cognitive to clinical, by providing non-invasive access to the human brain in action. While fMRI has already deepened our understanding of brain function across various cognitive and behavioral states, the authors argue that its real potential lies in uniting typically siloed subfields of neuroscience. By summarizing the strengths and limitations of fMRI and showcasing successful studies, the paper lays out a roadmap for using fMRI to foster greater interdisciplinary collaboration and push neuroscience into a more integrated future.
TAGS: fMRI, Cognitive Neuroscience, Systems Neuroscience, Clinical Neuroscience, Computational Neuroscience, Brain Imaging, Interdisciplinary Research, Human Brain Function, Neuroimaging, Brain Mapping