Can We Decode FMRI datasets using Neural Networks? (Arguably for Ailurophiles)
Spoiler Alert, the answer is yes, but It’s not so simple, this article will show you why.
Neural Networks, the big guns of modern-day Machine learning! From Google’s Deepmind learning how to run, albeit like you would when Mum shouts at you for leaving the dishes in the sink, to Deepdream giving Van Gogh competition on abstract art, one cannot help but be impressed by the progress Neural Network technology has made from the age of physical perceptrons to the amalgam of rectification functions in nodes with weights corrected by backpropagation to make them more accurate. One thing I have to say is,
Dear humans, worry not, AI is not yet at a level where it will take over you, give it one more decade, then maybe you can start worrying, but till then, let us look at one task Neural Networks can do brilliantly, decoding FMRI data to identify the stimuli that invoked them.
FMRI decoding
Traditional FMRI data processing entails exposure of the subjects to stimuli and monitoring the Blood Oxygen Levels variation in the brain as a result of this exposure. This variation leads to the alteration of blood hemoglobin (It is weakly ferromagnetic as it contains Iron) magnetic properties, which…
