The global population swells by billions every decade, and the amount of people needing medical treatment climbs wildly in turn. Human processing ability alone is no longer sufficient to sift through and service a worldwide patient list numbering in the hundreds of millions; we are outnumbering ourselves. Doctors and medical staff, tasked with mastering an ever-more advanced arsenal of treatment techniques, must now look to technological innovation to manage the flow of patients, and ensure adequate care for all.
The field’s tech leaders recognize a dire need for efficiency. Facing the chore of sorting infinite amounts of medical data, many innovators believe that perhaps the most effective approach is to teach that data to sort itself. This is the beauty of machine learning: using the logic of statistical analysis, computers can actually be taught to transcend their programming; they can learn to identify patterns, make decisions, and tailor accurate predictions from blocks of input data. Coupled with the transformative power of telehealth, machine learning has the potential to personalize and optimize healthcare like never before, and usher modern medicine into a future once imagined only by the most optimistic science fiction.
One vocal advocate for uniting the fields of machine learning and digital health is Dr. Yulun Wang, a leader in surgical robotics and founder of pioneering telehealth company InTouch. Wang believes that “machine learning will soon be integrated cohesively into healthcare delivery through telehealth so that big data sets can be gathered and analyzed to improve global care. It will also improve individual care by matching the specifics of a patient’s diagnosis and treatment plan to millions of comparable cases.” He predicts that by 2050, much of healthcare will be digitally deployed, providing patients a virtual experience as routine and simple as today’s online banking. See a previous post by me on the proliferation and future of TeleMedicine.
Such predictions seem overwhelmingly feasible, considering the myriad ways machine learning is already speeding the implementation and adoption of digital health. Frontlining healthcare’s machine learning initiative is IBM Watson Health, who has developed a supercomputer capable of using cognitive computing algorithms to “read” 200 million pages of text in three seconds, analyze it, and draw relevant medical conclusions. The skill of systems like the Watson computer will be invaluable in connecting into individually effective treatment the obscure but important discoveries buried within the world’s estimated 150+ billion gigabytes of medical data.
Machine learning deployed through digital health is proving its worth not only as a research tool, but a lifesaving method of diagnostic prevention. To demonstrate this potential, the Stanford Machine Learning Group recently teamed with digital health provider iRhythm Technologies to develop a deep learning algorithm that can pinpoint twelve types of abnormal heart rhythms. Researchers analyzed a data set provided by over 30,000 patients (over five times larger than standards set by previous cardiac studies) to compose an AI “neural network” with detection abilities similar to those of AI currently utilized in speech recognition and computer vision. Data can be gathered through the Zio, iRhythm’s wearable biosensor, and analyzed to diagnose irregular heartbeats.
And the capabilities don’t stop there; artificial intelligence startup Enlitic has lead the way in the creation of computers capable of detecting cancer in patients. Despite not receiving a classical Ivy-League med-school education, the computers build by Enlitic utilize machine learning, comparing the charts, symptoms, scans and vitals of patients against millions of others, to correctly diagnose a patient. Where other doctors may misread a chart, Enlitic’s AI-radiologists are able to use hard data and numbers to ensure the scans are read and interpreted quickly, efficiently, and objectively. In short, computers like the ones developed by Enlitic could not only change the digital health landscape, they could save lives.
Picture a medical futurescape where patients digitally consult doctors to seek symptom relief. Instantly, each nuanced detail of a patient’s medical history, and current physical state is examined across the entirety of human medical research via intelligent supercomputer, producing individually tailored treatment insights capable of not only treating, but curing nearly every malady. Such daydreams may appear overly fantastic and utopian, until we compare what machine learning can presently do to the rate at which its technology is advancing.
So, how will pharmaceutical providers adapt to this future? Product information must be easily machine-digestible and, of course, have the adequate data and value proposition to justify its appropriate usage. Pharma may not just be marketing to patients and prescribers, but also the machines behind them, serving up answers and recommendations. It will be interesting to see how the industry continues to evolve to meet this rapidly changing treatment paradigm.