For example, confirmation of a cancer diagnosis is getting much easier. The New York Times reported that blood tests -- known as "liquid biopsies" -- have now been shown to generally match the results of a tumor biopsy. The blood tests look for DNA fragments from the tumor that signal its presence. The liquid biopsies are useful for both detecting the presence of a tumor and its ongoing monitoring.
The current generation of tests are not perfect, with as many as 15% of tumors not generating enough DNA to be detected, but they do offer the advantage of not requiring an invasive procedure. The expectation is that testing will get even better, allowing for earlier detection of more cancers, although the problems of both false positives and detection of tumors that would not end up harming the patient if left untreated remain.
The FDA has just given approval for one such test, and numerous companies are vying for position in the space -- STAT reports that some 38 companies in the US alone are working in it. With direct-to-consumer lab tests becoming more of an option, we may soon have more cancer diagnoses, with more decisions required about what needs to be treated versus just monitored, and, indeed, what is "cancer," since some tumors grow only very slowly or not at all.
Still, if we have a choice -- and we soon will in many cases -- most of us would probably opt for a blood test rather than a biopsy.
The concept of what constitutes a "test" itself is undergoing some broadening. Instead of an actual test, liquid or otherwise, Microsoft reported that they may be able to use your search history help determine if you have cancer, even before you or your doctor realize it. They think their approach can be extended to other serious illnesses, helping catch them at earlier stages.
The Microsoft researchers used search logs to identify people whose queries strongly suggested they'd been diagnosed with pancreatic cancer. Then they worked backwards on those people's search histories to determine what kind of queries were associated with the subsequent diagnosis. There may not have been any actual diagnostic tests or procedures done yet, possibly just searches on symptoms, but they found "signals" in the search history.
The method isn't perfect, but the researchers claim: "We find that we can identify 5 to 15 percent of cases while preserving extremely low false positive rates." They believe their false positives are on the order of 1 in 100,000, which compares very favorably to most actual diagnostic tests.
Gosh, if Bing can do that, imagine what Google could do
The researchers were quick to note that their research was only a proof of concept and that Microsoft has no plans to develop a product based on the discovery (which seems odd!). Instead, they wanted to get the medical community "excited" about the approach in hopes that it could end up being used as an early warning system for serious diseases.
Microsoft researchers had previously used search data to help identify unreported prescription drug side effects, and there was also the Google effort to use search data to help predict flu epidemics, which ended up not quite working out as planned, although the approach is far from dead. These examples all illustrate that artificial intelligence can use Big Data to find things out about our health that we might never have otherwise found, or found out too late.
One of these days you may be typing in a search and Cortana politely but firmly might suggest that you better see a doctor for a specific potential diagnosis. It might even set up a Skype call with a specialist for you.
Some of the most intriguing changes to testing relate not to diagnostic tests, but to clinical trials. Researchers at the Lawrence Livermore National Laboratory are developing a "human-on-a-chip" that they hope could predict the impact of drugs, viruses, or toxins on humans. The chip, known as iCHIP, essentially models the workings of human biological systems -- currently the central nervous system, peripheral nervous system, the blood-brain barrier, and the heart. Actual cells, such as neurons, are seeded on a chip, and microelectrodes can do real-time monitoring of electronic signals from them.
The hope is that use of iCHIP will lessen or even eliminate the need for animal or even human testing. Use of animals for testing (particularly for cosmetics) has become highly controversial, even banned in some cases. Congress is close to passing a bill that would encourage the use of alternatives to animal testing.
There are both ethical and efficacy questions with animal testing, so iCHIP may be an idea whose time has come. Indeed, iCHIP's principal investigator boldly says: "The ultimate goal is to fully represent the human body," integrating all the biological systems together to create a "complete testing platform."
Chimpanzees everywhere may be breathing a sigh of relief.
If Microsoft is using Big Data to take the place of many standard diagnostic tests, a company called Insilico Medicine is trying to do the same with clinical trials. They use deep learning and artificial intelligence to predict how a drug will impact human cells, rather than relying on animal trials.
Alex Zhavoronkov, Insilico's CEO, believes that:
"...animal testing is not very representative of what the human outcome will be...We need something better, and something better is creating a virtual human to simulate the activity of many drugs on many tissues at once. That can only be done using really deep data."Creating a "virtual human" or putting all the human biological systems on chips are ambitious goals, especially given our increased recognition of both the complexity and importance of our microbiome (whose cells outnumber ours). On the other hand, no one is particularly happy with the cost and timeliness of clinical trials, animal or human, and identifying health issues sooner and less invasively is certainly desirable.
OK, so maybe we're not getting a tricoder anytime soon, but, all in all, we're making real progress on testing.