Brain Simulation: the future of personalized medicine

Simulating a human brain sounds like a plot from a science-fiction movie or the latest crazy “moonshot project” from Mark Zuckerberg. In fact, brain simulation is already here! It’s one of the fastest growing fields of neuroscience – and you can even try it at home on your laptop!

Brain simulation is not about mind reading or creating artificial intelligence but first and foremost a revolutionary tool for understanding how our brain works and how to deal with its diseases like stroke, epilepsy or neurodegenerative diseases like Alzheimer’s or Parkinson’s.

These diseases are still not understood very well and therapeutic success is varying a lot from patient to patient. And yet, they affect 150 million people worldwide and cause yearly expenses of 200 billion Euros for treatment, management and caretaking.

However, practical brain simulation faces four formidable challenges:

  1. Examination:

    It’s incredibly hard to measure and examine what’s happening inside a living, active brain. Obviously, opening the skull and poking in needle sensors is not a desirable option. Non-invasive examination methods like EEG, fMRI, DTI, MEG and PET either require expensive hardware or have limited spatial/temporal resolution. Overcoming such weaknesses by simultaneous measuring is limited by basic physics (wearing a cap with metal electrodes inside a fast moving magnetic field can fry your scalp).

  2. Individualization:

    Although the general anatomy of the human brain is known fairly well on the level of molecules, cells and organs, every human brain has a unique shape, size and connectivity. That’s why one person can compose an entire symphony in their mind while others already struggle to hum along with a melody. Therefore, a useful brain simulation needs to implement the individual characteristics of a person.

  3. Size:

    The human brain contains about 86 billion brain cells (aka neurons), each with an average of 1,750 connections to other neurons (aka synapses). Simulating such a network with molecular accuracy would require 1 billion terabytes just for storage. For comparison, the size of the entire internet is approx. 15 billion terabytes.

  4. Computation:

    On a rough scale, simulating the computational capacity of only one human brain (median estimate: 10^18 FLOPS) would require the entire global computing resources currently used for Bitcoin mining. But computational capacity alone still wouldn’t suffice: A human brain packs this capacity into an incredibly small volume (excelling in signal throughput), using very little energy (about 13 W) and has a completely different “hardware architecture” than our silicon-based von-Neumann-computers.

The Brain Simulation Section at Charité

The Brain Simulation Section at the Charité Universitätsmedizin Berlin, led by Prof. Dr. Petra Ritter, has been at the forefront of network-based brain simulation, computational neuroscience and its clinical application for more than 10 years. Its research activities cover the entire process of brain simulation:

  • Acquiring multi-modal brain sensor data
  • Developing automated processing pipelines for sensor data
  • Creating computational modeling algorithms for molecular processes up to mesoscopic populations and macroscopic connectomes
  • Writing simulation software for various devices, from smartphones to HPC clusters
  • Running simulations for individual patients to obtain diagnostic clues, test pharmaceutical and surgical interventions and consult clinicians about optimal outcomes
  • Cataloging and sharing large cohorts of brain models to explore relations between structural network features, biophysical parameters and empirical brain states
  • Disseminating brain simulation to society by providing real-time brain computer interfaces for personal health, gaming and interactive art installations