Artificial intelligence is revolutionizing our world in many unimaginable ways. At the verge of the Fourth Industrial Revolution, humanity is currently witnessing the first steps made by machines in reinventing the world we live in. And while we keep debating about the potential drawbacks and benefits of substituting humans with intelligent, self-learning machines, there's one area where AI's positive impact will definitely improve the quality of our lives: the health care industry.
Machine learning algorithms can process unimaginable amounts of info in the blink of an eye. And they can be much more precise than humans in spotting even the smallest detail in medical imaging reports such as mammograms and CT scans.
The company Zebra Medical Vision developed a new platform called Profound, with algorithm-based analysis of all types of medical imaging reports that is able to find every sign of potential conditions such as osteoporosis, breast cancer, aortic aneurysms and many more with a 90 percent accuracy rate. And its deep learning capabilities have been trained to check for hidden symptoms of other diseases that the health care provider may not have been looking for in the first place. Other deep learning networks even earned a 100 percent accuracy score when detecting the presence of some especially lethal forms of breast cancer in biopsy slides.
Computer-based analysis is so much more efficient at (and less costly than) interpreting data or images than humans, that some have even argued that in the future it could become unethical not to substitute AI in some professions such as radiologists and pathologists! (For more on IT in medicine, see The Role of IT in Medical Diagnosis.)
Electronic Medical Records (EMRs)
The impact of electronic medical records (EMRs) on health information technology is one of the most controversial topics of debate of the last decade. According to some studies they represent a turning point in improving quality of care while increasing productivity and timeliness as well. However, many health care providers found them cumbersome and difficult to use, leading to substantial technology resistance and widespread inefficiency. Could the newer AI-driven software come to the rescue of the many doctors, nurses and pharmacists fumbling every day with the unwieldy clunkiness of EMRs?
One of the biggest issues with this new health care technology is that it forces clinicians to spend way too much of their precious time performing repetitive tasks. AI can easily automate them, however, for example by using speech recognition during a visit to record every detail while the physician talks with the patient. Charts can and will include much more detailed data that could be collected from a variety of sources such as wearable devices and external sensors, and the AI will feed them directly into the EMR.
But moving forward from the first step of data collection, when enough relevant info is correctly understood and extrapolated by deep learning algorithms, it can be used to help improve quality of care in a lot of ways. It can enhance patients’ adherence to treatment and reduce preventable events, or even guide doctors via predictive AI analytics in treating high-cost, life-threatening conditions. Just to name a practical example, a recent study published in the JAMA Network found how the big data extracted from EMRs and digested by an AI at the University of California, San Francisco Health helped with the treatment of potentially lethal Clostridium difficile (C. diff) infections.
And it's easy to see how much medical record data mining is going to be the next “big thing” in health care, when none other than Google launched its own Google DeepMind Health project to improve the speed, quality and equity of access to care.
Clinical Decision Support (CDS)
Another interesting example of deep learning can help machines make better decisions than their human counterparts is the proliferation of clinical decision support (CDS) tools.
These tools are usually built into the EMR system to assist clinicians in their work by suggesting the best treatment course, warn of potential dangers such as pharmacological interactions or previous conditions, and analyze even the slightest detail in a patient’s health record.
An interesting example is MatrixCare, a software house that was able to integrate Microsoft's famous AI Cortana in their tool used to manage nursing homes. The potent analysis capabilities of the machine learning engine strengthened the decision-making ability of the support tools incommensurably.
“One doctor can read a medical journal maybe twice a month,” explained CEO John Damgaard, “Cortana can read every cancer study published in history before noon and by 3 p.m. is making patient-specific recommendations on care plans and improving outcomes.”
CDS also brings forward the argument that machines are able to communicate with each other much better than humans do. In particular, different medical devices can all be connected to the internet just like any other internet of things (IoT) device (wearables, monitors, bedside sensors, etc.), and to the EMR software as well. Interoperability is a critical issue of modern health care as delivery of care fragmentation is a major cause of inappropriate treatment and increased hospitalizations. When led by smart AI, the various EMR platforms become able to “talk” to each other through the internet, increasing cooperation and collaboration between different wards and even different health care facilities.
Developing a new drug through clinical trials is often a very costly affair. Not just in terms of time (we're talking about decades) and dollars invested (the costs may easily reach up to several billion dollars), but human lives as well. Many new pharmaceuticals require, in fact, many years of additional testing on real-world subjects during the so-called postmarketing period, and it's not so uncommon that many serious (or even deadly) side effects are discovered many years after a medication has been launched.
Once again, efficient supercomputer-fueled AI can root out new drugs from a database of molecular structures that no human could ever dare to analyze. A prominent example is Atomwise's AI, which was able to predict two drugs that could put a stop to the Ebola virus epidemic. In less than one day, their virtual search was able to find two safe, already existing medicines that could be repurposed to fight the deadly virus. The best part is that they found a way to effectively react to a pandemic emergency just by scanning through drugs that had already been marketed to patients for years, proving their safety. (To learn more about how technology is guiding drug development, see Big Data's Influence in Medicine and Pharmaceuticals.)
A Leap into the Future
Some of the most amazing technologies are not ready yet, being nothing more than just prototypes, but their implications are so breathtaking that they're still worth mentioning.
One of these is precision medicine, a really ambitious discipline that uses deep genomics algorithms to scan through a patient's DNA looking for mutations and anomalies that could be linked to diseases such as cancer. People like Craig Venter, one of the fathers of the Human Genome Project, are currently working on a new generation of computational technologies that can predict the effects of any genetic alteration, paving the road to individualized treatments and early detection of many preventable diseases.
A Word to the Wise
As excited as we may be because of the huge potential of introducing AI to health care, it is important that we understand its limitations. Using AI in medicine is not devoid of risks, although many of them will be easily overcome once we get accustomed to it.
The maxim “do no harm” is critical to establish some ethical standards that would act as boundaries. Today we're invested in the responsibility of building the framework upon which the future generations will make their decisions.