We are searching data for your request:
Upon completion, a link will appear to access the found materials.
Study: Evidence of cancer from search engine data
Microsoft researchers found in a study that search engine queries can identify signs of certain cancers among users. However, the aim of the scientists is not to diagnose the disease online, but to encourage people at risk to see a doctor.
Cancer diagnosis through search engine queries
Even if health experts repeatedly warn against self-diagnoses on the Internet, this hardly prevents anyone from looking for the possible causes of complaints online. Pancreatic cancer does not appear in the 50 most frequently searched diseases on the net, but typical symptoms such as abdominal pain or back pain do. However, it was precisely this type of cancer, the first signs of which are often unspecific and which is therefore often recognized too late, was the focus of a study by Microsoft researchers. They showed that pancreatic carcinomas can be detected months before the medical diagnosis by evaluating inquiries to Microsoft's search engine Bing.
Anonymous data evaluated
Eric Horvitz, Ryen White and John Paparrizos worked with a record of 6.4 million inquiries, according to a report from Technology Review. They report on the results in the medical journal "Journal of Oncology Practice". According to them, they filtered the searches from the anonymized data, in which they could make a clear diagnosis, for example "I was diagnosed with pancreatic cancer" or "Why did I get cancer in the pancreas?" Based on this, they searched the search histories several months back for signs that indicate the disease. In addition to abdominal pain and unexplained weight loss, this also includes dark urine. The scientists also included risk factors such as alcohol addiction and obesity.
Low number of false alarms
“We discovered patterns in the search queries that can predict the future occurrence of search queries that lead to a clear diagnosis,” the authors explain. According to the information, they succeeded in early detection in five to 15 percent of the cases. The “extremely low error rates of 1 in 100,000” are also impressive. However, Ryen White emphasized on the Microsoft blog that online diagnosis of the disease was not the goal: "The goal is to encourage those who are at the highest risk to see the doctor."
Psychological effects of the warnings have not been clarified
The team reportedly wanted to initiate a discussion among doctors. Even if the basic feasibility could be demonstrated, there are no plans to send a Microsoft cancer alert to the screen in the near future. Nevertheless, the scientists said: "We are excited to apply the analysis to other devastating and difficult-to-diagnose diseases." But even if web-based screening is simple and inexpensive, according to the researchers, the psychological effects of such warnings are far from clear. (ad)