Doctors have just learned something new about postoperative spinal cord repair for spinal cord injuries. A link was found between the success of long-term recovery and high blood pressure during surgery. This may seem like insignificant news, but it can help millions of patients get on their feet faster. The most interesting thing is how this discovery was made. It was done by a computer program.
The new software carried out a meta-analysis of long-forgotten neurological studies – and using topological data analysis was able to identify patterns that interested specialists. The material for the meta-analysis was the results of studies that were written off as useless 20 years ago. After all, at that time there were no advanced machine learning systems, such as Ayasdi has now provided for the University of California at San Francisco, where a group of neurologists and statisticians is working. The functionality of the program is described in a scientific article published yesterday in the journal Nature Communications. “What seemed like nonsense turned out to be very valuable information,” says
Adam Ferguson, Principal Investigator at the Brain and Spinal Injury Center at the University of California, San Francisco, and co-author of the research paper. How valuable this information is has yet to be explored in clinical trials, but history already raises some interesting questions. For example, does it make sense to conduct new clinical trials if you can re-analyze the results of old ones? And one more thing: should the raw data from “unsuccessful” studies be published in the open access, even if their result was not a scientific article or some kind of discovery.
Spinal cord injury is one area of medicine that hardly develops. There have been no significant discoveries in this area for 20 years. Spinal injuries are complex and therefore more difficult to understand than other systems. The researchers decided to once again analyze the array of old archived data in search of some hidden relationships.
To do this, they used topological data analysis, a technique that applies the concepts of geometric topology to find hidden patterns in large datasets. The program found many patterns, most of which were known to physicians, including the ineffectiveness of various drugs. However, the harmful effect of high blood pressure on the long-term recovery of a patient was something no one knew.
Neurologists say that they would hardly have been able to detect such a pattern using conventional methods, without the use of data mining. There are too many variables to consider.
The inventor of topological data analysis is Stanford mathematician Gunnar Carlsson and co-author of the scientific work. He is now the president of Ayasdi, which he founded to use topological data analysis and machine learning technologies at the same time. Ayasdi software analyzes data without human input, does not require theories or direction of work.
Before the spinal cord, the topological data analysis technique made another medical discovery: it helped detect a unique mutation in breast cancer that has a high survival rate.