Many times after my talks, people ask me whether algorithms could theoretically be better at making a diagnosis than doctors. With my doctor’s cap on, I must defend the art of medicine. But as a medical futurist I need to tell my honest views. Making a diagnosis is an art. We humans are not engineering products, and therefore measuring a few parameters and tweaking a few knobs will not diagnose and cure our diseases.
Part of the art of diagnosis is looking at the whole human being. The way the patient walks, speaks, smells, or thinks are all important to the final diagnosis.
The other part of a correct diagnosis is learning and gathering whatever information is relevant to the patient’s case. Being a physician entails a commitment towards life–long learning. Doctors frequently dive into biomedical databases of peer–reviewed papers to find the information they need. Keeping up to date with the latest literature in our field of interest is normal routine in the life of a physician.
The first part is what improves with experience, requires creativity and instinct. Technology might never replace that. Wearable devices and home monitoring services can measure a lot of parameters, but the physician’s impression on meeting a patient is irreplaceable.
The other part is where everybody fails. No doctor in the world can be perfectly up–to–date on recent medical advances as there are 25 million papers on Pubmed.com. No doctor can be certain they have found all the pertinent information their case requires. To do so, we do need help from the world of technology.
Computers practicing medicine
In 1996, the IBM supercomputer Deep Blue challenged Garry Kasparov, the reigning world chess champion. Kasparov won, and headlines around the world celebrated the triumph of humanity over the computer. IBM used the experience from the match to improve Deep Blue’s algorithms, and asked for a re–match in 1997. Kasparov lost to Deep Blue this time, 2.5 to 3.5. Kasparov argued that if he had had access to the same databases as the computer, he could have won the match.
Based on Kasparov’s suggestion a new form of chess match format was introduced in Spain in 1998. Advanced Chess players played against one another by using chess software. The human player decides the move, but the human–software pairing is considered a team. This is a winning combination of human creativity and the power of computing.
But there have been no such digital aids in medicine – until now. Watson, IBM’s new supercomputer, aims to fill this gap in medicine. After it beat two highly skilled players in the television quiz show Jeopardy, US clinics started testing its application in the practice of medicine. The advantage Watson offers is an ability to comb through patient records, English textbooks, and millions of medical papers in existing databases – in seconds instead of decades. Its algorithms arrive at diagnostic suggestions, and assign probable success rates to them. In the end the treating physician makes the decision, assisted by all pertinent information gleaned by Watson.
Two features of Watson are noteworthy. It employs natural language processing, meaning it can understand written and spoken language. It also uses deep question–and–answer technology. It can enter into conversation and learn more during it. Given that medical practices use different electronic medical records, Watson must be able to understand both structured and unstructured data. Some physicians take notes about a diabetes patient, mentioning “diabetes” or “T1D” in the clinical summary. Understanding natural language means distinguishing between a note that is important and one that is not in a given context.
IBM Watson at the oncology clinic
The MD Anderson Cancer Center was one of the first leading clinics that announced it would start using Watson. In oncology, mountains of studies are published every day, and every patient’s case is unique due to cancer’s ability to randomly mutate. Finding information relevant to a particular patient’s case is difficult no matter how much experience the treating physician has. When Watson goes over the patient’s case, it comes up with the list of suggestions for treatment and assigns a confidence value between very low and very high. MD Anderson researchers evaluated Watson’s success rates and found it to be very efficient. Because physicians rate the suggestions that Watson comes up with, it improves with each case.
In May 2015, a collaboration between Watson, Epic – a software company focusing on electronic health records – and the Mayo Clinic began. Epic has 350 customers that exchange over 80 million medical records annually. The Mayo Clinic has more than 1 million patient visits a year and conducts at least a thousand clinical trials at any one time. Using Watson to analyze that huge amount of data or answer the questions of patients seems like a good step forward. One Mayo Clinic oncologist called Watson’s potential to provide clinical trial information wherever is needed crucial.
A supercomputer becomes a cognitive computer when it tries to reproduce the behavior of the human brain through artificial intelligence. Human physiology is so complex that we can benefit from computers that mimic the way we think and pose questions. Such computers improve by learning just as physicians do – except they improve more quickly.
So can a computer uncover relevant information to a patient’s case better? Absolutely. Can it diagnose more accurately than a physician? There is no reason to believe it won’t be able to, although we’ll have to wait years – maybe even decades – for that level of accuracy. Would such algorithms replace doctors then? I highly doubt it. Their role will change but they will always be needed. If you ask patients whether they want to be treated by a computer or a person, the majority will choose the person. We are social beings. We need to discuss our health issues not only for the sake of receiving proper treatment, but also because words alone can heal.
And there is currently no algorithm or smartphone app for empathy or understanding. Don’t doubt that there will be some that mimic the way we provide empathy, but this is not the direction of current developments.
For the first time in history, making a decision about a patient’s case will not be based on the lucky uncovering of crucial information from the haystack of medical databases.. With cognitive computing, physicians will be able to focus on helping their patients, instead of keyboards and monitors.
The practice of medicine remains an art, but its colors can be mixed by an algorithm if the painter is still human.