Watson Health set its sights on applying cognitive/AI on a wide range of healthcare domains including evaluating radiology images, unlocking genomic data to improve diagnoses and treatment decisions, assisting physicians in developing treatment plans for cancer patients and enabling population health. Those skeptics in the healthcare industry that challenged IBM's ability to democratize Watson should take note that Watson for Oncology is being deployed in mainstream hospitals and is actively being used by oncologists across the world.
When Watson Health was conceived and launched just about two years ago they embarked on an aggressive path of acquisition and sought partners to co-develop specific applications. The shared objective was using Watson's ability to digest and learn medicine from a range of medical data including the results of research, images and pathology slides, genomic sequences, clinical guidelines and other sources and serve up recommendations and insights on the best clinical action for a physician to consider. Academic medical centers were the first to be exposed to Watson and most notably Memorial Sloan Kettering and Cleveland Clinic partnered with IBM to develop solutions in cancer care. Watson Health has fought skepticism as the industry waited for the democratization of Watson beyond the Jeopardy! gameshow victory. That promise is emerging as oncologists across the globe are using Watson for Oncology to regularly assist them in determining the optimal treatment for cancer patients.
IBM Watson utilizes a broad range of cognitive/AI technologies, from unstructured text analytics, question and answer processing and knowledge base creation to machine learning and automated reasoning based on statistical techniques. These technologies are present in Watson for Oncology.
It is virtually impossible for physicians to stay abreast of advances in medical care due to an explosion of medical information. Currently, approximately 50,000 oncology research papers are published annually, (1) and by 2020 medical information is projected to double every 73 days(2), outpacing the ability of humans to keep up with the proliferation of medical knowledge. Watson for Oncology was trained at Memorial Sloan Kettering (MSK) requiring Watson to ingest data from more than 300 medical journals, more than 200 textbooks, and nearly 15 million pages of text. It is no wonder that it took five years from Watson's debut on Jeopardy for Watson for Oncology to go mainstream.
Part of the initial skepticism about mainstream adoption of Watson was the fear on the part of the medical community that Watson would replace physicians. This misperception exists with the application of artificial intelligence applied across all industries. While artificial intelligence is used to automate routine and repeatable tasks it will not replace the knowledge worker. Automation has always led to changes in worker duties and responsibilities, but for the most part, automation has brought more jobs as well as more leisure time. The washing machine and dryer, the vacuum cleaner and the microwave oven are just some of the inventions that have created more opportunities than they have taken away. Cognitive/AI technology is the ultimate in automation. It will inevitably change workers jobs, but it will also provide more capabilities and data-driven guidance in augmenting how a worker performs their jobs. Workers will get better and more productive as AI agents provide advice, recommendations and guidance.
The misunderstanding in health about AI taking away autonomy and jobs still exists as industry representatives continue to refer to Watson as making decisions, which is not correct. In the case of Watson for Oncology its recommendations provide insights about different treatment options and provides oncologists with information regarding drug options and administration instructions. Watson ranks the evidence-based treatment options, linking to peer reviewed studies and clinical guidelines. Its machine-learning capability means it continuously learns, gaining in knowledge over time. IBM reports that physicians that are using Watson for Oncology embrace the power it provides to review clinical records and establish evidence-based treatment options for their patients based on the most current medical literature. Physicians by nature look for evidence based advice and guidance and as such are relying on Watson.
IBM conducts concordance studies that evaluate Watson's recommendations against those of individual physicians and tumor boards. The results are promising and documented in a recently released study authored by S.P. Somashekhar, MD, an oncologist and chairman of the Manipal Comprehensive Cancer Center, Manipal Hospitals, in Bengaluru, India. To assess the "accuracy" of Watson for Oncology and Manipal’s multidisciplinary tumor board – a group of 12 to 15 oncologists who meet weekly to review cases at the hospital – Somashekhar and colleagues studied the cases of 638 breast cancer patients who had been treated at Manipal Hospitals. Ninety percent of Watson Oncology’s recommended standard treatment and for consideration recommendations were in accord with those of the tumor board. In an initial, preliminary analysis, the degree of concordance was 73 percent overall, but this was influenced by comparing decisions made at different points in time. In both cases, concordance varied depending on the type of breast cancer. In some cases due to Watson’s access to newly published research results and to the complexity of some cases its results varied from tumor boards.
IBM reports that physicians are using Watson for Oncology in a variety of ways; as part of their pre-visit work-up to determine a treatment plan, to validate a proposed treatment plan or to explore options for new treatment plans when a patient's condition has changed. Doctors are also using Watson for Oncology as part of the patient's visit to demonstrate how the recommended treatment plan was determined, and as a tool used by Tumor Boards as part of their review of individual cases. IBM also believes that Watson for Oncology will in the future be used as part of the training curriculum for oncologists. Physicians access Watson for Oncology through a web browser (Watson for Oncology is cloud-based software as a service), and depending on the environment physicians can access Watson for Oncology from their Electronic Medical Record system.
Watson for Oncology is currently in use in seven health systems world-wide that represent 39 hospitals. U.S. adoption is limited to one US hospital and by an organization offering second opinions. When asked about the limited penetration in the U.S., IBM responded that given the high level of sophistication and specialization in cancer care available in the U.S., Watson for Oncology was best suited to regional or rural hospitals where oncologists treat a wide variety of cancers and may have limited subspecialty expertise. Globally, access to physicians with deep knowledge in specific cancer is more limited which heightens the need for Watson for Oncology. Watson for Oncology is now available to assist clinicians in developing treatment plans for breast, lung, colorectal, cervical, ovarian and gastric cancers. IBM and MSK plan to train Watson on at least 9 additional cancer types this year, covering nearly 80 percent of the worldwide incidence of cancer. As Watson for Oncology becomes trained on more and more cancer types, increased adoption in the U.S. and globally is expected.
The growing footprint of Watson for Oncology and the early positive results being documented bode well for continued democratization of Watson in healthcare.
This blog was co-authored by David Schubmehl, Research Director, Cognitive/AI Systems and Content Analytics, IDC.
(1) PubMed, accessed at pubmed.com
(2)Densen, Peter, Challenges and Opportunities Facing Medical Education 2011. Transactions in the American Clinical and Climatological Association. Accessed at https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3116346/