Cognitive vs. Cancer – a pattern for application of cognitive computing

From Healthcare IT News

In this recent announcement, IBM and the American Cancer Society are teaming to use IBM Watson to improve the accuracy and personalization of advice given to cancer patients.

This is an example of a powerful pattern for using cognitive computing to its best advantage: providing evidence-based insights and guidance from a base of knowledge that has more information than any one person could comprehend, and that adds new information faster than any person could hope to assimilate.

Medicine is a very active area of research – best guess is that medical researchers are publishing well over a 1 million peer-reviewed articles per year. According to the National Institutes of Health, cancer research gets about 5% of the funding across all research topics so it’s safe to say that there are around 50,000 cancer-related research articles per year.

It’s also safe to say that it’s more than any oncologist is going to be able to read in a year, let alone retain, let alone evaluate, let alone put into practice.

But thanks to natural language processing (NLP) which lets a computer read and comprehend writing, Watson can read written language. And thanks to natural language tools like annotators that let human experts train the NLP how to interpret the medical language in a research article, Watson can read (ingest) 50,000 cancer-related papers each year with no trouble at all. And it will retain everything that it reads, so with some guidance from medical experts to understand the relative quality and importance from different sources, authors, etc., Watson can accumulate the knowledge from every publication worth knowing.

The real power then kicks in: Watson has already been trained to help oncologists determine treatment options for their patients (see Watson Oncology at the MSK site). When new information is ingested from new medical research, Watson can slot the information into its internal worldview (its corpus). Then the next time there is a request for a treatment option, Watson will consider the newly ingested research along with everything else it already knows in forming its evidence-based response – a piece of advice to a doctor or a cancer patient. Qualified medical professionals can provide feedback on the medical quality of the responses (think “Rate this Response”) and the machine learning aspect of cognitive computing takes the feedback on and tunes its worldview accordingly.

Practically speaking, you have a computer that can provide advice and guidance based on a) evidence from all of the information that it has ingested – all of which it remembers and can mobilize in the appropriate context, and b) continuous learning from its training interactions with human experts, and its advisory interactions with the people asking for advice.