“Can a computer algorithm replace my doctor?”
Not yet and probably not ever… On the other hand, you probably wouldn’t want to be cared for by a doctor who doesn’t embrace the latest technology. Machine learning sounds like the epitome of hi-tech (and it is). But what, in plain English, is medicine based machine learning?
When it comes to neurologists, neuroradiologists and neurosurgeons, machine learning involves feeding huge databases into a computer. These might be pathology reports, MRI images, blood tests… the list goes on and on.
The next step involves someone who knows how to code: he or she writes a computer program, which sifts through the data, tirelessly seeking useful patterns. Most algorithms are a variant of Artificial Neural Networks, which draw their inspiration from the mammalian brain.
When the computer flies solo, it's called unsupervised machine learning. Unsupervised machine learning is OK and can yield some impressive results.
If unsupervised machine learning is considered the Minor Leagues, supervised machine learning is the Major League of Artificial Intelligence. Supervised machine learning relies on an expert in the field to provide the seeds of knowledge and pattern recognition. The programmer then constructs algorithms based on convolutional neural networks, recurrent neural networks or adversarial neural networks. By employing such a strategy machine learning becomes deep learning.
“OK, that all makes sense… but what’s it good for?”
Alzheimer’s Disease (AD)- Deep Learning has been applied to brain imaging (such as MRI). It may help distinguish patients who develop AD (which worsens over time) from patients who have mild problems thinking and remembering (which will remain stable).
Autism Spectrum Disorder (ASD) and Attention Deficit Hyperactivity Disorder (ADHD)- Deep Learning has been applied to brain imaging and clinical data (results of physical and mental examination). It may help identify brain networks (connections between neurons (brain cells)) and brain regions as well as physical and mental signs, which improve the diagnosis of ASD and ADHD as well as suggest avenues of treatment.
Epilepsy- Deep Learning has been applied to the mountains of electroencephalogram (EEG, brain wave) data, which has been collected by innumerable epileptic patients over the years. It may help predict the likelihood, timing and frequency of future seizures, perhaps allowing for novel treatments or prevention.