IISc: New algorithm to study brain connectivity

Indian Institute of Science (IISc) have developed a new Graphics Processing Unit (GPU) based machine learning algorithm.

About the New Algorithm

  • The new Algorithm is called RaAl-LiFE (Regularized, Accelerated, Linear Fascicle Evaluation).
  • It can help scientists in understanding and predicting connectivity between different regions of brain, in a better manner.
  • It can rapidly analyse the enormous data generated from diffusion Magnetic Resonance Imaging (dMRI) scans of human brain.
  • Using the new algorithm, team was able to evaluate dMRI data 150 times faster as compared to existing state-of-the-art algorithms.
  • This study was published in the journal ‘Nature Computational Science’.

Brain Connectivity

Millions of neurons are fired in the brain every second. It generates electrical pulses, which then travel across neuronal networks, from one point of brain to another, by connecting cables or axons. These connections are required for computations that the brain performs. Understanding the brain connectivity is significant to uncover brain-behaviour relationships at scale.

Conventional approaches vs new research

Conventional approaches to study brain connectivity are invasive and use animal models. On the other hand, dMRI scans provide a non-invasive method to study brain connectivity. Axons, connecting different areas of the brain acts as its information highways. This is because, bundles of axons are shaped like tubes, through which water molecules move along their length.  dMRI allows scientists in tracking this movement, in a bid to create a comprehensive map of fibres network across the brain. This map is called a connectome.

Diffusion-weighted magnetic resonance imaging

dMRI is the use of specific software and MRI sequences that generates images from resulting data. It uses the diffusion of water molecules for generating contrast in MR images. The dMRI allows mapping of diffusion process of molecules in biological tissues.




Latest E-Books