Ed. Note: This is the third installment in a three-part report exploring IBM’s so far two-year project to create the ultimate diagnostic and individualized treatment engine for lung, breast, and blood cancer patients using its unique Watson natural language cognitive engine. The first part defined the challenge and discussed why the Watson team has taken it on. The second part looked at what Watson is and the complexity of the challenges it is designed to meet.

Healthcare is a natural fit for Watson, says Watson Project CTO, IBM Fellow, & VP Rob High, because it is experiencing “this enormous flood of valuable information, and the reality is that doctors don’t have the time to consume it all. So we can make a huge difference by providing them access to the most relevant information for their patients … without requiring that they spend hours and hours reading all the available material.”

That means that instead of following guidelines designed for the average person, the doctor can focus on the most relevant information, and possible diagnoses and treatments, rated with a percentage of probability that each is the correct diagnosis for that patient, based on that patient’s unique chemistry, genetics, epigenetics, environmental conditions, and other aspects of their physiology.

“Inside the healthcare industry we focused on oncology because the variance in treatment is so broad and the efficacy in that treatment is so profound that the difference we can make is truly substantial.”

And this is a truly interactive approach between Watson, the doctor, and patient. “Say Watson comes up with a guideline and treatment but only has a 34% confidence in this approach and a 29% confidence in some alternative treatment.” Watson will define what further information would help it refine its analysis and recommend the medical test that would provide that information, allowing the doctor to order the right test to get the right data. Applying that across the full patient population allows the hospital to get the optimal use from its test and treatment resources to maximize patient outcomes while reducing cost.

The Watson Vision
IBM has spent two years so far working with subject matter experts in cancer to build up the context that it needs to make that right decision. Part of that process involves running actual cancer patient records and having those top oncologists review Watson’s results against their knowledge and the actual patient treatment history. Then they go back to strengthen the context and statistical aspects of the system to improve those results. That process is still ongoing.

However, the team has now deployed Watson at Wellpoint to help with its resource allocation. So Watson is beginning to enter the production phase of the project.
And at Cleveland Clinic, the IBM Research team is both modeling how medical students learn and letting those students see how Watson reaches its conclusions. That, says High, “helps students understand the patterns and strategies for how Watson learns things that they might also use. And that’s reciprocal – Watson’s also learning from the students.”

The team has started with lung, breast, and blood cancers. But over time they plan to broaden that focus to include other cancers and then other major diseases with complex treatment regimens. And they envision applying Watson to other aspects of medicine as well. Payers can use Watson to speed up and increase accuracy of pretreatment authorizations. And eventually, Watson will also be applied to research and discovery.

“One of the things clinical research organizations want to do is compare the relevance of their research results to standard guidelines. Using Watson to infer results from patient medical information helps them to quickly identify alternative treatments in the standard literature and make comparisons to the research results they are producing.”

Treatment-Planning-as-a-Service
IBM is delivering Watson’s services as sort of a “Treatment-Planning-as-a-Service” from a central, HIPAA-compliant IBM data center. “Today IBM does not retain any private information in Watson, which gives us more flexibility to provide the service,” High says. Over time, and particularly where highly regulated identifiable patient information needs to be kept in Watson’s database, IBM anticipates making Watson available on-premise at major institutions as a remotely managed system, managed by IBM.

“Absolutely this is Big Data,” High says. “There’s an enormous flood of information that is being produced every day that is relevant medicine. Part of IBM Watson’s capability is that it can go into an incredibly large corpus of knowledge like that and find answers to the question you’re asking and handle it efficiently in reasonable response times, where ‘reasonable’ is measured in seconds or minutes rather than the days, weeks, or even months that a human researcher would need.”