Few ideas in the last decade have provoked as much excitement, or as much confusion, as the introduction of artificial intelligence (AI) in oncology.
From the first moment we announced our plans to apply our Watson technology to help oncologists, we were met with a stark dichotomy of emotion. The headlines ran the spectrum from hype (your next doctor might be a robot!) to cynicism (5 reasons AI in healthcare will fail).
Today, five years into the journey to help improve cancer treatment through data, analytics and AI, while we’re still very much in the early stages, I’m happy to report that the real-world progress is far more encouraging than either of those early storylines would suggest.
In fact, not only is AI being used to support physicians in the delivery of cancer care today, it is producing quantifiable results while charting a course for the future.
Several dozen examples of this progress were on display at the 2019 American Society for Clinical Oncology (ASCO) Annual Meeting, an annual gathering of the world’s leading oncologists and cancer researchers.
Based on our review of these abstracts, and our involvement with many of the studies presented at ASCO, we are learning how best to apply this technology. Hint: robo-oncologists are not the answer.
RELATED: Verily’s India project reveals the work of applying AI to real-world clinical care
What is happening, today, though, is that AI is helping physicians manage volumes of data too great to be processed by the human mind. The practical application of AI in image recognition (radiology, pathology, dermatology) is well known. This technology can also assist a physician who might not recall the latest literature pertaining to a particular clinical scenario by offering treatment options for consideration along with curated medical literature.
AI can also be used to digest the eligibility criteria of thousands of clinical trials and match patients to appropriate trials. The technology also can be utilized to rapidly annotate the results of tumor genome sequencing, identifying potential therapeutic options that are personalized to the patient.
Data at ASCO demonstrated the technology could potentially be used to facilitate shared decision-making between physician and patient, to enrich a multi-disciplinary tumor board discussion, for physician education and potentially for remote consultation as well.
This is significant progress. What started as a moonshot with a monumental goal to bring AI into cancer care and an uncharted path to achieve it now has a clear direction forward.
RELATED: Industry Voices—Beyond the hype: Here’s how oncologists view the future impact of AI
As an example of this tangible progress, consider the results of a study presented at ASCO, which analyzes 1,000 breast, lung and colorectal cancer patients in India. The study found that when a multidisciplinary tumor board—a group of physicians in different specialties who review and discuss treatment options for patients—was presented with AI-derived treatment options from Watson for Oncology, they changed their treatment decisions in 13.6% of cases based on the information provided.
The reasons for these changes?
In 55% of cases, the AI provided recent evidence for newer treatment options; in 30% of cases, the new options were more personalized and in 15% of cases the technology surfaced new insights from genotypic and phenotypic data and evolving clinical experiences.
This is such an important set of findings because it shows exactly where AI technology can add the greatest value by complementing existing medical processes and bringing new information into the treatment decision process that might not otherwise have been included.
Similarly, a study of hematological malignancies conducted at Hallym University College of Medicine in South Korea found that Watson for Genomics identified clinically actionable insights that were missed by manual interpretation in 33% of cases.
And, in another study with the Oncology Department at Beijing Chaoyang Integrative Medicine Emergency Medical Center, researchers found that the addition of AI-powered treatment options into the patient consultation process helped to improve overall patient confidence in treatment plans.
RELATED: IBM Watson Health touts recent studies showing AI improves how physicians treat cancer
Across these varied applications of AI to real-world cancer care around the world, the technology is becoming increasingly defined and understood—not as a replacement for doctors or a magic cure machine but as a backstop to help caregivers and patients consider treatment options and a productivity tool to deliver insights quickly.
It’s still early in the evolution of AI in oncology and there will still be challenges ahead. But the results we’ve seen over the past year offer the most compelling evidence yet of the important role this technology is already playing in improving the way physicians choose to treat cancer globally and the massive potential to achieve so much more.
Nathan Levitan, M.D. is the chief medical officer for IBM Watson Health Oncology and Genomics.