EPFL: Unveiling the Potential of the GPU-accelerated Signed Distance Field Technique in computational neuroscience

The intricacies of the brain, a marvel of complexity, encompass billions of neurons, astrocytes, and a network of elaborate blood vessels. Unraveling the structure and interactions within this intricate organ is paramount in the realm of neuroscience.

A powerful technique that has emerged in recent years is the application of signed distance fields (SDFs) to construct realistic shapes directly from raw data, eliminating the need for meshes. These shapes are dynamically generated at rendering time from simple primitives such as spheres, cylinders, cones, etc.

The signed distance field technique excels in generating high-fidelity graphics that faithfully represent the complexities of biological structures. By encoding distance information for each point in space, this technique facilitates the creation of intricate and realistic 3D reconstructions directly from raw data (.h5 or .swc).

A notable advantage of employing signed distance fields is their capacity to enable interactive exploration. Researchers can seamlessly navigate through reconstructed 3D models of neurons, astrocytes, and blood vessels, gaining a dynamic and immersive understanding of their spatial relationships. This interactive feature enhances the exploration process, fostering a more intuitive and insightful analysis of the data.

The utilization of the signed distance field technique to visualize and explore neurons, astrocytes, and blood vessels signifies a substantial advancement in the field of neuroscience. This potent approach, originally developed with insights from Inigo Quilez's blog, provides high-quality graphics that empower researchers to study the intricacies of the brain's architecture.

The implementation of the signed distance field technique, initially available in the Blue Brain BioExplorer CPU back-end, underwent a transformative upgrade. Over the last two days, I successfully ported the technique to GPU using the CUDA programming language and the NVIDIA OptiX ray-tracing framework. This enhancement not only accelerates the visualization process but also opens new horizons for real-time, high-performance exploration of intricate neural structures. This collaborative fusion of cutting-edge technologies ensures that the exploration of neural intricacies remains at the forefront of scientific discovery, propelling our understanding of the brain into new dimensions.

• Blue Brain BioExplorer: https://github.com/BlueBrain/BioExplorer

• Inigo Quilez - articles- distance functions: https://iquilezles.org/articles/distfunctions/

• SDF Tracing visualization: https://www.shadertoy.com/view/lslXD8

 

EPFL: Interactive visualization of the Fusion Reactor Tokamak plasma

In my dedicated efforts to advance nuclear fusion, I embarked on a journey to translate conceptual ideas into tangible actions. This mission commenced with the creation of a test dataset, meticulously crafted with insights derived from engaging discussions within our team. To expedite the visualization and exploration of this data, I turned to the remarkable open-source application known as BioExplorer.

BioExplorer, originally developed for biological datasets, proved to be an invaluable asset in our venture. While its roots lay in the realm of biology, its adaptability and versatility allowed it to seamlessly integrate with our nuclear fusion research. This application, once primarily designed for biological data, had evolved to accommodate and process a wide range of scientific datasets, transcending its initial domain.

In the context of nuclear fusion, the core challenge lies in modeling the behavior of superheated plasma, the fuel for this revolutionary energy source. To visually represent this complex phenomenon, I envisioned the plasma as an array of particles, with each particle intricately characterized by its position and direction. The direction vector of each particle not only represented its position but also served as a direct indicator of the particle's charge, a crucial factor in the fusion reaction.

Efficiency was paramount in this ambitious undertaking, and BioExplorer played a pivotal role. This open-source application, known for its adaptability, provided a user-friendly platform to visually explore and analyze the scientific datasets, irrespective of their origin. It allowed us to gain valuable insights and a deeper understanding of the plasma's behavior, further propelling our research forward.

To ensure real-time visualization of the plasma's behavior, I harnessed the formidable processing power of Graphics Processing Units (GPUs). These GPUs were thoughtfully optimized with a tree-based acceleration structure, making them an ideal choice for handling the intricate calculations required for real-time visual representation of the plasma's behavior.

Perhaps the most astonishing aspect of this project was its efficiency. What might seem like a Herculean task was accomplished in no more than two days of relentless work. This achievement not only underscores the power of modern computational tools but also highlights the immense potential of nuclear fusion as a clean and virtually limitless energy source, hinting at a future where our energy needs may be met sustainably and efficiently, with the invaluable assistance of open-source applications like the versatile and adaptable BioExplorer.

Check out the BioExplorer source code and the Python Jupyter Notebook used to create the image above.

CERN: Particles Composition and Interactions Using the Nuon Model

Drawing inspiration from René Brun's remarkable publication titled "Particles Composition and Interactions Using the Nuon Model" I embarked on a venture to visualize the magnetic fields engendered by sub-particles within the realm of collisions.

 

Although the data I employed differs from that presented in the paper, I found the concept of moving away from the conventional depiction of particles as simple spheres and instead illustrating their genuine magnetic fields to be captivating.

 

 

With the computational capabilities readily accessible on ordinary consumer PCs today, it is now possible to calculate these fields in real-time, opening up novel avenues for visualizing and exploring the intricacies of the sub-atomic world.

 

Try it for yourself!! https://particle-physics.netlify.app

Full source code: https://github.com/favreau/NuonModelVisualizer

HARVARD: The Harvard Brain

So proud to appear in the Harvard Brain Spring 2023 issue.

Studies In Silico: An Interview With Cyrille Favreau On EPFL’s Blue Brain Project, by Lara Ota, Buse Toksöz, and Kei Hayashi

http://www.theharvardbrain.com/spring-2023-8203lara-ota-buse-toksoumlz-and-kei-hayashi.html

EPFL: Blue Brain BioExplorer goes RTX!

Exciting news! I just released the version 1.6.0 of the Blue Brain BioExplorer. It now goes #NVDIARTX with OptiX backend compatibility, #AI denoiser, and new stereo camera! High-quality rendering of scientific datasets and #VR use-cases with pure ray-tracing.

Open source code: https://github.com/BlueBrain/BioExplorer

EPFL: Machine Learning and understanding the role of blood glucose levels in the severity of COVID-19

On the third anniversary of the first lock-down, we look back at how Machine Learning helped reveal the role of blood glucose levels in the severity of COVID-19.

With access to enough open data, imagine what other problems could be tackled.

So proud we made it! 😀

For the 3D visualization part of the movie, I developed the open-source Blue Brain Explorer, in collaboration with Emmanuelle Logette for the scientific part. Check it out:

https://github.com/BlueBrain/BioExplorer

Watch the documentary:

Read more: https://lnkd.in/dW7m6Ea

#scivis #ScientificVisualization #ScientificExploration #RayTracing #covid19 #sarscov2 #covid19pandemic

 

CERN : Visual scientific exploration at Blue Brain, and beyond

by Cyrille Favreau (EPFL), Evelyne Schmid Osborne (EPFL)

 

Friday Feb 17, 2023, 11:00 AM → 12:00 PM Europe/Zurich

503/1-001 - Council Chamber (CERN)

Scientific exploration relies on building software that combines data integration, analysis and interactive visualization to build, modify and navigate through large scientific datasets. For this, Blue Brain built and open-sourced the Blue Brain BioExplorer. The Blue Brain BioExplorer was originally developed to answer key scientific questions related to the Coronavirus as a use case and to deliver a visualization tool. Today, the BioExplorer allows reconstructing, visualizing, exploring and describing in detail the structure and function of highly-detailed biological structures such as molecular systems, neurons, astrocytes, blood vessels, and more.

 

The BioExplorer is built as an extension of Brayns, the official underlying and generic rendering platform that was designed to easily adapt to all fields of science.

 

In this very visual talk, we will present EPFL's Blue Brain Project, and explain how we could architecture and build the application that is now being used to produce high quality and high fidelity media, as well as interactive and immersive experiences of the digital reconstruction of the mouse brain.

We will finally discuss the impact of using real raw data for science communication and dissemination.

CERN reference: https://indico.cern.ch/event/1253917/