BATBern53 - EPFL - Blue Brain and related technical challenges

Scientific exploration has increasingly shifted from traditional storage-centric approaches to compute-centric methods, leveraging advanced software for data integration, analysis, and interactive visualization of vast scientific datasets. To address this need, Blue Brain developed and open-sourced Blue Brain Brayns, a technical platform for the visualization of in-silico datasets, which later gave birth to the Blue Brain BioExplorer. Initially created to tackle key scientific questions related to the Coronavirus and to deliver a powerful visualization tool, the BioExplorer now excels in reconstructing, visualizing, exploring, and detailing the structure and function of biological systems, from molecular systems up to the full brain. In this highly visual presentation, I will demonstrate how I architected and built the application that EPFL's Blue Brain Project utilizes to generate high-quality, high-fidelity images and movies, along with interactive and immersive experiences of the digital reconstruction of the rodent brain. By prioritizing computational efficiency over traditional storage, we enable real-time manipulation and exploration of large datasets. Additionally, I will present a few examples of how AI is a game changer, solving problems that were almost impossible to solve with classic techniques. These examples will highlight the transformative impact of AI in enhancing image quality, speeding up data processing, and uncovering new scientific insights that were previously beyond our reach

1. Blue Brain Project

3. Single neurons

4. EPFL Single neurons

5. Microcircuit "Blue Brain Project

8

7. EPFL Brain Regions Blue Brain Project

8. Brain Regions

Blue Brain Project

(CB Bran ProgcuEPFL 2005-2022 Al ight reserved. Data visualisation: Cie Fareau/ Bue Brain BoExperer

9. Blue Brain Project

10. EPFL

Pl o EN

Blue Brain Project

+5 reserved, Data vieuasaton Orte Fasa / Bue Bran BoExpirer

(Eu Bran ProgcVEPRL 2005-2022 Au

11. EPFL BrainR eg ions Blue Brain Project

Thalamus Somatosensory Cortex en Vascular
2331K neurons +1.7M neurons
+40M synapses +478 synapses - OO)
2015 : 2021
Microcircuit Hippocampus CA1 Neocortex
231K neurons +460K neurons +107M neurons
37M synapses 800M synapses 888 synapses

Digitally reconstructing and simulating the brain

12. Brain Simulation Blue Brain Project

13. AY

15. 4

“Blue Brain Project

4) Molecular Systems

16. EPFL Blue Brain BioExplorer timeline Blue Brain Project

Neurons & Astrocytes

Be
Harvard Brain

Interview
Python notebooks over WS Metabolism Cosmology
Optix GPU engine

OSPRay CPU engine (et) Enzyme reactions
ET) Scientific visualization and exploration platform

2015 u 2019 2020 2021 2022 2023
Phase 1: Engineering Phase 2: Scientific visualization and visual exploration

Neuromodulation

AE) ton channels

17. EPFL Technical use-cases Blue Brain Project

18. EPFL Technical use-cases

20. Science Outreach Blue Brain Project

3
3
=
E
2
E
5
o
ES
3
RE >
es > S :
3 entre’ neurosciences“
E innovations artistiques
pr a
~ e, >

re SS

Centre Pompidou - Paris - February 2020

21. Storage vs Compute duebrinproject

SS

22. EPFL Storage vs Compute

New method for visualizing particle collisions
(Collaboration with René Brun)

23. Storage vs Compute

NAS

Blue Brain Project

Interactive visualization of plasma in Tokamak «¿3

uw:

ud en Gl ri

7

(CB Bran ProjecvEPAL 2005-2022 Al oh reserved, Data visuatsaton: Ole Favreau /Bve Brain Boiler

24. AGIMRIes > Visual Studio Code HR. ICAA

am" Storage vs Compute "sie Spain Project °

25. Use of Al TAI Y JB im 0 es ri Project

26. Berk

Image 2 Image
A new paradigm in concept storage

Image description
Computer Vision

> >

The image shows a group of red and

Prompt

white links arranged in a pattern. The twink of ore cits
links are primarily red, with some white Blood va
‘ones interspersed throughout. The es
colors ofthe links are vibrant and stand ee
ut against the background.
The composition ofthe links creates a N
igh fideli visually appealing design, with the red E
High fidelity inns forming asking contrast against à
scientific dataset the white ones. The overall image gives El
E

off a sense of order and structure, with
the links neatly arranged. The colors of
the links are deep red and bright white,
adding to the visual interest of the image.

ran Project