Artificial Intelligence
Consulting
Development ServicesMicrosoft researchers have recently proved that the analysis of large volumes of user web / mobile search queries could in certain cases help detect pancreatic cancer before it's been diagnosed in a patient.
The results of the study by Eric Horvitz and Ryen White from Microsoft and John Paparrizos from the Columbia University was published in the The Journal of Oncology Practice on June 7, 2016. The study goal was to understand whether or not the analysis of web search queries could help detect that something is wrong with the user before the official diagnosis has been established.
To identify trends, the scientists applied a reverse engineering approach. They identified users who used Bing to search for information about pancreatic cancer and then explored their earlier search queries to detect any premonitory symptoms.
The study found that user search for cancer symptoms was the key marker of cancer detection. While scientists don't disclose their analytics system details, they claim it is able to distinguish random queries created out of curiosity or for research purposes from those created out of concern and fear for own physical condition (alleged symptoms) and those created by people who already have disease symptoms.
According to Horvitz, the study was inspired by a phone call with his friend who described him his symptoms. A single call was enough for the scientist to establish a very disappointing diagnosis and advise that his friend visits a doctor. A few months later this guy died of pancreatic cancer.
A percentage of patients who can live with this diagnosis up to five years is very low. The main reason of a high death rate is a late detection of the disease. The scientists believe their technology can help increase the percentage of patients who can live with pancreatic cancer up to five years from today's 3% to 5% - 7% in the near future. At the same time, this technology can significantly decrease expenditures for medical research by filtering out and excluding people with false symptoms.
All search queries used in the study were anonymous; therefore, the researchers were unable to reach out to users who were identified as those with clear cancer symptoms.
Today's medical community is very skeptical about the efficiency of the method used in the study. However, the researchers are going to dispel skepticism by converting Cortana into a medical consultant who will make appropriate recommendations to the user whose search queries have been identified as "containing cancer symptoms".
One of the researchers, Dr. Ryen White, has recently been appointed as Head of Microsoft's new eHealth division.
By the way, Google already conducted a similar research back in 2009. The company was going to use search queries analysis to predict to the flu epidemics. However, the study results weren't as reliable as expected.
In 2013, Microsoft specialists had used search queries analysis to detect side effects of medications before they were found by FDA.
