The use of artificial intelligence and machine learning is altering the face of medicine, with the potential to help improve overall medical diagnosis in an enormous way.
While investors are flocking to the field, there remain concerns for the validation of the technologies, exacerbated by cost and access to data, and the simple basic understanding of how computers reach certain conditions.
Regina Barzilay, a professor at the Massachusetts Institute of Technology whose work focuses on natural language processing—training computers to understand and internalize human speech—has a group that is currently collaborating with MassachusettsGeneral Hospital, applying their knowledge and expertise in machine learning and artificial intelligence to the enhancement of cancer diagnosis and treatment. The researchers are investigating whether computers have the capability to detect signs of breast cancer in mammograms before humans, and whether machine learning can prompt doctors to utilize extensive quantities of available data in order to make more personalized decisions regarding treatment. Dr. Constance Lehman, a professor of radiology at Harvard MedicalSchool and chief of breast imaging at Mass General, believes that machines can indeed read mammograms—with the potential to “open up a whole revolution in health care.”
Dr. Eric Topol, director of the Scripps Translational Science Institute, states that the potential is “perhaps the biggest in any type of technology we’ve ever had in the field of medicine,” as computers have the inherent ability to transcend human intelligence. Moreover, technological giants including IBM’s Watson, Alphabet, and Philips, with pharmaceutical companies and startups, are investing in the market for artificial intelligence in health care. The grow this projected to rise by 40 percent per year, reaching $6.6 billion in 2021.
Many of the earliest applications are projected to be in the diagnosis of disease. Pathologists at Harvard Medical School, MIT, and CalTech are attempting to improve cancer diagnosis through images by training computers to assess digital slides, and ultimately learn how to differentiate cancerous cells. Their results indicate that the combination of computer and pathologist reduced error rate by 85percent.
Researchers are also using machine learning to make correlations and connections in data that the human eye is unable to see. A team at Mt. SinaiHospital in New York has developed a system that combs through de-identified health data in the hospital system, and combines information in a multitude of ways. The ‘deep learning approach’ attempts to utilize an enormous amount of information and data from health records, in order to enhance disease diagnosis. Results published in last year’s journal Nature state that the team was able to improve the prediction of diseases ranging from schizophrenia to severe diabetes.
Data also points to the applications of artificial intelligence in medicine to span to consumer health, including the potential for data from wearables to improve health care in any location. CEO of Philips Frans Van Houten suggests that sensor technology, through picking up personalized data, can subsequently be used to create more personalized treatment plans. “Through artificial intelligence and machine learning, we can take these massive amounts of data and interpret what is going on, and then get to the first and right treatment.”
*General Electric’s healthcare division has recently announced a partnership with the corporate parent of two of Harvard University’s teaching hospitals, with the goal of developing artificial intelligence products for medicine. Chief executive of GE Healthcare points to the creation of solutions that get “a better clinical and economic outcome,” likely by leveraging the company’s dominant position in medical imaging into a new ownership of medical AI.