New Breakthroughs in Brain Mapping: Mouse, Fly, and Human AI Models скачать в хорошем качестве

New Breakthroughs in Brain Mapping: Mouse, Fly, and Human AI Models

In this video, we dive into the cutting-edge world of Connectomics and Computational Neuroscience. We explore how scientists are using AI to map the wiring diagrams of entire brains and predict neural activity with unprecedented accuracy. From the complete wiring diagram of the fruit fly to foundation models of the mouse visual cortex, we are closer than ever to understanding the physical hardware of the mind. [Paragraph 2: What’s Covered / Chapter Breakdown] We break down major datasets and AI models revolutionizing the field: • The Fruit Fly (Drosophila) Brain: Discover FlyWire, the first complete neuronal wiring diagram of an adult fly brain, which required an estimated 33 person-years of proofreading. We look at how this connectome reveals sensorimotor processing and the flow of information across 78 brain regions. • The Zebrafish (ZAPBench): Explore ZAPBench, a benchmark for predicting whole-brain activity in larval zebrafish. Learn how researchers record over 70,000 neurons simultaneously using light-sheet microscopy and use video models (like 3D U-Nets) to forecast brain activity. • The Mouse Visual Cortex: A look at the MICrONS project and new "Foundation Models" that can predict how thousands of neurons respond to natural videos, static images, and noise, acting as a "digital twin" of the visual system. • Human Brain Fragment: A tour of the H01 dataset—a petavoxel fragment of the human cerebral cortex reconstructed at nanoscale resolution, revealing 50+ synapses on individual neurons and strange "unidentified cortical objects". • C. elegans (OpenWorm): An update on the OpenWorm project, which aims to simulate the biology of a worm in software and even robotic bodies. [Paragraph 3: The Tech Stack] Learn about the AI and imaging tech making this possible: • Electron Microscopy (EM) & Light-Sheet Imaging: How we capture data from the scale of synapses (nanometers) to whole-brain activity. • AI Architectures: How Transformers (like QuantFormer and POCO) and Flood-Filling Networks are used for segmentation and forecasting neural dynamics. [Paragraph 4: Attribution & Licenses] This video references open-access research published under the CC BY 4.0 License. Key studies include: • FlyWire & Drosophila Connectome: Dorkenwald et al. (2024), Scheffer et al. (2020), Shiu et al. (2024). • Zebrafish & ZAPBench: Lueckmann et al. (2025), Immer et al. (2025). • Mouse Foundation Models: Wang et al. (2025). • Human Cortex: Shapson-Coe et al. (2021). • OpenWorm: Sarma et al. (2018).