This is the latest in an occasional series of posts about A New Era of Computing. IBM envisions a monumental shift over the coming years: computing will be ubiquitous and machines will learn from their interactions with data and humans–essentially programming themselves. This quantum leap will be enabled by advances in artificial intelligence, data analytics, computing systems and nanotechnology. It will result in a smarter, better planet.
What if any doctor in the world had access to the expertise of the best doctors in the world when choosing among treatment options for an individual patient? That’s the vision that’s driving a small group of scientists at IBM Research – Haifa.
Their work is getting it’s first real-world tryout with Fondazione IRCCS Istituto Nazionale dei Tumori, a public health institute in Milan, Italy, which specializes in the study and treatment of cancer. The IBM Research project, called Cli-G, for clinical genomics research, is a biomedical analytics system that uses machine learning, among other technologies, to provide physicians with treatment advice tailored for an individual. The Cli-G system combines a wide variety of data, including statistical records of the outcomes of particular treatments, clinical and genetic information about the particular patient and the expertise of top physicians. “We’re incorporating the knowledge of experts–all the things the physician brings in from past experience,” says Boaz Carmeli, the IBM researcher who leads the project.
Cli-G is a cousin to IBM Watson, the deep question-and-answer technology that beat two past-champions at the Jeopardy! TV quiz show and which is now being used in healthcare and financial services. Both are learning systems. While Watson focuses primarily on gathering unstructured textual information from published sources, Cli-G is aimed at gathering specific types of information. Carmeli and his colleagues have coined a term, Evicase, to describe the way they structure information. It’s a combination of evidence-based medicine, which is statistical analysis of treatment outcomes, with case-based reasoning, which is knowledge gathered by studying the best practices of top physicians. Scientists in IBM Research are exploring how Cli-G and Watson could be used to compliment one another.
The project grew out of a long-term effort by IBM’s Haifa scientists to develop a network that every party involved in healthcare delivery can tap into to share information. Cli-G adds sophisticated analytics to the network. The technology uses all of the information available to predict the most likely outcomes for a particular patient for various treatment options. Then, based on Evicase, it recommends what it considers to be the best treatment.
The Istituto plans on using Cli-G in two ways. In addition to the decision-support tool for its physicians, the organization will be able to get an aggregate view of patient care–enabling it to evaluate the performance of various departments or teams, and using this knowledge to make changes that could improve results. Already, the Haifa researchers have provided the Istituto with analysis of some of its clinical treatment data, which identifies situations where physicians deviated from standard treatment guidelines and tracks the outcomes.
This engagement is just the first of what the Haifa researchers hope will be several similar partnerships with healthcare organizations. They hope to branch out beyond cancer treatment to other important disease areas including HIV/AIDs and hypertension. “There’s a tremendous amount of care data around the world, but it’s not being used,” says Carmeli. “Now we can use it and provide better care for patients. There are exciting possibilities.”
We’re still a long way from being able to provide a global expert system for physicians, but we’re taking important first steps toward that goal.