When it comes to data curation, the problem isn’t the rise of Big Data, but the haphazard way data often present themselves.
That’s how John Quackenbush characterized the issue in a panel Wednesday morning at the MedCity CONVERGE conference in Philadelphia on practical applications of artificial intelligence (AI) and machine learning (ML) in oncology. Quackenbush, the director of the Center for Cancer Computational Biology at the Dana-Farber Cancer Institute in Boston, was referring to the difficulties of faced by curators when they had to go into clinical trial protocol pages and take down the studies’ entry criteria manually due to the inconsistent way they were written. “I like to characterize it not as a Big Data problem, but as a messy data problem,” he said.
Moderator Ayan Bhattacharya, who serves as advanced analytics specialist leader at Deloitte Consulting, noted that health management organizations, health plans and others have been investing in technology to assist curation that had previously been the work of human editors. But that technology has to work, Quackenbush responded, comparing technology that doesn’t work to long waits he has encountered trying to schedule appointments at the Apple Store.
All the same, AI promises to significantly help clinicians perform their jobs, particularly as they find themselves drowning in data. “Right now, for every hour I spend talking to patients about their healthcare plans, I spend two hours behind a computer,” said panelist Dr. Tufia Haddad, chair of breast cancer medical oncology and of information technology at the Mayo Clinic in Rochester, Minnesota. That time behind a computer screen, she said, involves a lot of data mining. Meanwhile, the doubling of knowledge in oncology can make it hard to keep up even for academic oncologists focused on specific cancer types, let alone community oncologists who must deal with solid tumors and hematological cancers across the spectrum, she said. That is where AI can come in handy, she said, but it’s not without resistance from oncologists. “When I go to my colleagues and talk about this technology … there’s immense skepticism because, isn’t technology what got us into this mess?”
Still, the goal of AI is to augment people rather than replace them, Quackenbush said, adding that if one proposes to replace radiologists, the radiologists will fight them, but proposing augmentation is a win-win. Arterys head of corporate development Carla Leibowitz said, for example, that her company’s cloud computing and AI system for medical imaging isn’t telling physicians what to do, but mapping out how patients similar to theirs have responded to therapy. The workforce will change significantly and have to evolve with the roles of physicians, who will have to be trained in how to work with AI, she said.
Photo: Alaric DeArment, MedCity News