Dr. Brandon Beaber Video Review of SuStain
Dr. Brandon Beaber features a video overview of the SuStain publication in Nature Communications which provides both an accessible explanation of its technology and impact, as well as offers a professional examination of its findings from a clinical perspective. Dr. Baaber is a “board-certified neurologist with subspecialty training in multiple sclerosis and other immunological diseases of the nervous system. He is a partner in the Southern California Permanente Medical Group and practices in Downey, California (South Los Angeles). ” His YouTube channel posts weekly videos focusing on various aspects of MS.
Here are a few excepts from this week’s post on the SuStain research:
“This is a very interesting article on identifying multiple sclerosis subtypes using unsupervised machine learning and MRI data from various clinical trials. So instead of relying on our subjective judgement with clinical phenotypes maybe we should look at the MRI and be more objective.”
“I have to give credit to first author, Arman Eshaghi, who’s a neuroscientist from the United Kingdom, for this excellent publication.”
“I’ll give an analogy to a chess computer. Back in the day, people used to design chess computers and say, ‘the pawn is worth 1, the bishop is worth 3, the rook is 5, your goal is to checkmate the king, maximise your evaluation function’.” Now they’re saying…’the goal is to checkmate the king, you [the computer] figure out what to do. You tell us what the intermediate goals are’…That’s exactly what [this paper] does. They let the machine group people with MS into categories of machine subtypes based on their MRI scans.”
“You can also have injury to normal appearing white matter. So even in an area where there is no lesion, there’s evidence of pathological changes both in multi-modal MRI and in autopsy studies, biopsy sections, and things like that. Now you can see this to some extent on conventional MRI though not with the naked eye…It’s sort of a subtle black hole that humans can’t see, but computers can.”
“Now this is the most interesting thing which is response to treatment…The interesting thing is that not everyone benefited [from the drugs]…Maybe we could use this type of data to determine who would benefit from a particular treatment…I think this is really the beginning of designer treatments giving a specific treatment to a specific individual based on their phenotype of multiple sclerosis.”