The study of connectomics, which aims to map the complex network of animal brains, is growing rapidly. In just ten years, it has evolved from its early stages to a discipline that holds the promise of unravelling the mysteries of cognition and the physical basis of brain disorders like Alzheimer’s. Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Harvard’s Samuel and Lichtman Labs are using advanced electron microscopes combined with the analytical power of machine learning.
Unlike traditional electron microscopy, their integrated AI, called “SmartEM,” acts like a brain, learning from the specimen as it acquires images. This intelligent system focuses on relevant pixels at nanoscale resolution, mimicking how animals explore their surroundings. SmartEM aids connectomics in swiftly examining and reconstructing the intricate network of synapses and neurons in the brain with precision.
The crucial integration of hardware and software involves embedding a GPU into the microscope’s support computer. This allows machine-learning models to run on the images, directing the microscope to areas of interest identified by the AI. This approach mirrors human eye control, facilitating a rapid understanding of the images.
The lead architect of SmartEM, Yaron Meirovitch, compares the process to how humans focus on crucial points when looking at a face or reading a book. This novel microscope concept accelerates the examination and reconstruction of brain segments, making it more efficient and cost-effective than traditional methods.
The historical context of neuroscience is highlighted, starting with Santiago Ramón y Cajal’s Nobel Prize-winning work over a century ago. The field has evolved from studying simple organisms like C. elegans to exploring more complex brains like those of zebrafish and mice. The article emphasizes the challenges of managing vast amounts of data, with mapping the mouse brain alone requiring a staggering thousand petabytes.
The researchers tested SmartEM on octopus brain tissue, reducing imaging time significantly and enabling synapse-level circuit analysis. The team envisions a future where connectomics is affordable and accessible, allowing a broader range of research institutions to contribute to neuroscience. They aim to apply the technology in hospitals for pathology studies, making the process more efficient. The research, supported by the NIH BRAIN Initiative, was presented at the 2023 International Conference on Machine Learning (ICML) Workshop on Computational Biology in collaboration with scientists from Thermo Fisher Scientific.
Stay up-to-date on the latest tech and auto news and insights! Follow us on Google News, Instagram, X (formerly Twitter), Pinterest, and Facebook for daily updates.
