Utilizing NVIDIA IndeX Platform for Nanoscale Brain Imaging

2019-11-23

At the 2019 Supercomputing Conference in Denver, researchers from MIT, Harvard University, UC Berkeley, and other organizations showcased their research findings - nanoscale brain imaging. The neural circuits in the brain are composed of structures spanning seven orders of magnitude, which are further composed of thousands of different protein types. While expansion microscopy has allowed imaging of densely labeled brain tissue at near-millimeter scale with nanoscale resolution, it lacks the contrast to distinguish specific proteins and the speed to image multiple specimens. On the other hand, lattice light-sheet microscopy provides molecular contrast but lacks the resolution for dense neural tracing or precise localization of specific molecules within submicron-scale structures. For large-volume imaging, super-resolution fluorescence microscopy rapidly bleaches fluorophores and lacks the speed for effective whole-brain or whole-cortex imaging of multiple specimens. To address these challenges, researchers combined two imaging techniques. Expansion microscopy creates an expanded optical transparent body model of fluorescent samples while preserving the original relative distribution of their fluorescent labels. Then, lattice light-sheet microscopy images the model in three dimensions with minimal photobleaching, with a speed sufficient to image the entire fruit fly brain or the entire region of the mouse cortex in about 2 to 3 days, with effective resolution for multiple labels. With its high speed, nanoscale resolution, and the ability to utilize genetic targeting, cell-type specificity, and protein-specific fluorescent labeling, ExLLSM bridges the gap between the high throughput of conventional optical pipelines in neuroanatomy and the corresponding ultra-high resolution. The core of this work is the combination of expansion microscopy (ExM) and lattice light-sheet microscopy (ExLLSM), using a high-speed nanoscale molecular microscope to capture super-resolution images of neural circuits. Visualizing such large microscope data has always been a computational challenge for researchers, who have been forced to create static view videos instead of being able to browse the data. This hinders scientific discoveries and requires many hours to generate preprocessed visual effects. In a demonstration at the NVIDIA booth, 1.3 TB of multi-channel ExLLSM data displayed the 3D spatial relationships between essential proteins responsible for many brain functions, including neural connectivity and signal regulation. To visualize the data, the team used the NVIDIA IndeX platform, a 3D volume visualization SDK that allows scientists and researchers to visualize and interact with massive datasets in real-time, regardless of data size. NVIDIA IndeX provides instant visual feedback, insights, and determinism using an NVIDIA GPU cluster. In this case, it includes six NVIDIA DGX-1 systems, each with multiple NVIDIA V100 GPUs. The researchers stated, "This technology should enable statistically rich large-scale studies of neurodevelopment, sexual dimorphism, stereotypy, and structure-function relationships of behavior or neural activity, all with molecular contrast."

RECENT AI NEWS

RECENT AI TOOLS