Two scientists from Microsoft’s Redmond Hill research lab in Washington have just shown how using a person’s Internet search history could lead to the early detection of cancer. Last week, the Journal of Oncology Practice published a feasibility study by Dr. Ryen White and Dr. Eric Horvitz of the Microsoft research team, showing that it is possible to detect pancreatic cancer in an individual simply by analyzing their Internet search history – meaning their cancer could be detected even before it was diagnosed.
It’s not uncommon these days for people to turn to the web in search of health-related information or to look for answers about symptoms they’re experiencing. Searchers are often concerned about the development of a new symptom, and they type descriptive terms about their problems into search engines without a second thought. Dr. White and Dr. Horvitz, the lead researchers in this study, decided to use this information to see if they could harness all of the data generated from search histories to find out whether it held any clues to diagnosing cancer early.
Finding Symptom Patterns in Search History
The researchers began by looking at anonymized searches on Bing (Microsoft’s search engine) that indicated someone had been diagnosed with pancreatic cancer. From there they worked backward to locate earlier queries that showed that the Bing user had actually been experiencing symptoms of pancreatic cancer before they were officially diagnosed. By analyzing the search histories, they effectively stitched together little clues that meant nothing on their own, but in aggregate pointed to a diagnosis.
This is significant because Internet search histories could potentially be used as a type of early warning system or red flag for health signs that might otherwise be missed. While doctors routinely take medical histories, which includes asking patients about their symptoms, such verbal conversations are prone to forgetfulness or even miscommunication, if for example a family member is speaking for the patient rather than the patient him or herself. Search history, however, includes all of the symptoms that a person searches for, whether they remember searching for it or not.
The study’s authors reported that they were able to identify approximately 5 to 15 percent of pancreatic cancer cases with a false positive rate as low as 1 in 100,000. A low false positive rate is extremely important since it can help to measure how successful the search algorithm really is.
A New Way to Detect Cancer?
So what does this discovery mean for the future? Although the results of this study are promising, Dr. White acknowledges that the field of health-related data generated from search histories is still new territory for the medical profession, and at present Microsoft does not have any plans to develop any products linked to this finding.
However, the team hopes the encouraging findings from this research will excite the broader medical community and generate discussion about how such a screening methodology might be used. One suggestion is that it might involve analyzing anonymized data and designing a method for people to opt-in to receive a notification about health risks, either directly or through their doctors, in the event that the algorithms pick up a pattern of search queries that could signal a health concern.
The hope is that this type of technology can also lead to breakthroughs in the early detection of other types of cancer such as mesothelioma. Early detection of mesothelioma is the best way to improve prognosis, and it allows treatment can be started sooner, giving less chance of the cancer spreading to other parts of the body.
Because the symptoms of mesothelioma (like pancreatic cancer) are nonspecific and develop slowly over a long period of time, it is not hard to imagine an unknown sufferer searching online for symptoms like shortness of breath, weight loss or fatigue. Without a doctor present to connect all the dots these symptoms might otherwise go unnoticed, but as this study shows, that doesn’t always have to be the case.
Although this research was limited, it’s clear to see the benefits of intertwining health and technology and if these promising results can be refined and reproduced, we might soon see our search engines playing a much more significant role in the battle against cancer.