IBM and Memorial Sloan-Kettering Cancer Center are partnering to apply IBM's Watson artificial intelligence technology to cancer diagnosis and therapy selection.
The partners hope to create a system, tentatively dubbed Watson Enabled Diagnosis and Treatment Adviser for Oncology, to integrate patients' clinical information, molecular and genomic data, and clinical oncology practices from the literature in order to advise medical professionals on the best treatments for their patients, Josko Silobrcic, a medical scientist at IBM, told BioInform this week.
This oncology application will be one of a series of commercial healthcare applications dubbed Watson for Healthcare that IBM plans to begin rolling out in 2013, he said.
IBM kicked off its efforts to bring Watson into clinics last February when it announced that it was combining the system's deep question answering, natural language processing, and machine-learning capabilities with Nuance's speech recognition and Clinical Language Understanding solutions to help physicians diagnose and treat patients (BI 2/18/2011).
At that time, IBM also said that it was working with collaborators at Columbia University and the University of Maryland to figure out how Watson can best help in the clinical setting as well as to incorporate some healthcare-specific adaptations into the system (BI 3/4/2011).
The system under development at MSKCC will use Watson's computational and natural language processing abilities as well as MSKCC's clinical expertise and cancer information to create an evidence-based decision support system that will help doctors create personalized cancer diagnostics and make treatment recommendations for their patients that are based on current evidence, the partners said.
These reports will include detailed records of the data and evidence that were used to make the recommendations, IBM said.
The partners have already begun developing the first applications, which will focus on treatment regimens for lung, breast, prostate, and colorectal cancers based on information in MSKCC’s electronic health record, which includes genomic and molecular information, clinicians' notes and observations, and relevant published literature, Silobrcic said.
He said that IBM has begun developing algorithms that can parse resources such as oncology journals, textbooks, and clinical trial databases in order to extract specific oncology terms and then identify relationships between these terms.
Further, the partners are working on algorithms that will allow users to run queries and are also determining what kinds of information Watson needs to continue to train itself, he said.
The goal is to create an oncology application that will be able to look at patient data as well as evidence-based medical information from the literature and make correlations between the two that will serve as the basis for personalized treatment options, Silobrcic said.
For example, if a particular biomarker is present, the tool will look for diagnostic and treatment evidence pertinent to that biomarker in other patients with similar genetic profiles as well as in the literature, he explained.
The tool will also be able to ask for additional information that will enable it to improve its treatment suggestions. For example, it could ask for additional parameters that would allow it to select which of two suggested chemotherapy treatments would be better suited in a particular case, he said.
IBM expects to begin testing the oncology version of Watson for Healthcare later this year with an undisclosed group of beta testers.
IBM has not disclosed its marketing plans for the oncology solution nor has it set a price for the system at present, Silobrcic said.
Have topics you'd like to see covered in BioInform? Contact the editor at uthomas [at] genomeweb [.] com.