A new machine learning algorithm devised by scientists at the University of California San Francisco could help doctors diagnose multiple sclerosis (MS) earlier, by analysing healthcare records and identifying symptoms earlier, a study has found.
There is quite a bit of research which shows people with MS can have symptoms of the condition years before they are diagnosed. A big obstacle to diagnosis is that many early symptoms are not MS-specific and relatively mild, so the connection is not always made.
The researchers in this study cited previous research which showed that in the three years leading up to diagnosis, many people with MS have more appointments with urologists and psychiatrists, plus receive a higher amount of musculoskeletal, genito-urinary, or hormone-related prescriptions.
“These findings hint that underlying biological signals must be present months or even years before diagnosis,” the researchers wrote, adding that “information from these specialist visits could be pivotal in uncovering differences” between people with and without MS.
Earlier diagnosis is the golden ticket for researchers, because it means people could get MS treatment sooner and better outcomes further down the line.
Previously researchers have used machine learning to analyse electronic healthcare records. The computer generates algorithms to complete goals set for the data – such as separating people with MS and people without.
The computer is only looking for patterns in the data, however, and it doesn’t take into account the actual meaning of the data. This often results in it not being possible to know which specific factors a computer is using to decide.
In this latest study, the data from electronic health records was condensed into a single biomedical knowledge graph called a SPOKE. These graphs were then used to identify SPOKEsigs – health-related signatures relating to different biological processes of genes.
The computer uses these SPOKEsigs to sort people with MS from those without, and researchers could see which ones were most important for making decisions.
Further statistical test showed that the accuracy of the algorithm will improve the more additional healthcare data is incorporated.
Source: MS-UK 10 January 2022