by Susan Eustis
MD Anderson sees data access as key to achieving personalized cancer therapy care delivery.
MD Anderson is the leading cancer diagnostic and treatment center worldwide, rivaled by Sloan Kettering and others, but clearly better than any by several measures. MD Anderson has research on pediatric brain cancers regarding genomic sequencing, hedgehog pathways, p53 degradation and EGFR signaling. The results of the research have been published in a variety of respected scientific publications.
Translating those research advances to clinically proven therapies is daunting. MD Anderson and IBM’s Watson supercomputer are collaborating to seek to learn how to be able to apply pattern recognition to the analysis of questions and to sort patterns into key data components for the purpose of delivery of personalized medicine. The collaboration between the research specialists and the computer specialists is in a manner that is designed to let clinicians interrogate multiple large databases to select the most likely answer to a question relating to the delivery of personalized cancer treatment.
As IBM Watson is trained by the MD Anderson experts, much as a medical student would be trained by the leading medical practitioners, Watson is learning how to process data in a manner that supports rendering a diagnosis. The aim is to keep Watson up to date with the latest research findings so that it can support clinicians quickly and effectively. Watson is different from the best medical student, it remembers more, and it remembers better.
At a special public seminar at Rice University with the explicit purpose of discussing the potential real-life applications of Watson, it was explained that the computer can help physicians interrogate large sets of diagnostic or outcomes data as they pertain to the questions the clinician has about a given patient. This could then be used to train Watson as a clinician support tool.
This goes far beyond data mining. It applies pattern-recognition technology to very large data sets to determine the diagnosis that best fits the patient data. Based on its analysis, Watson could potentially determine a treatment plan that best fits unique disease characteristics. Watson can take a list of symptoms and organize those as key data components. A Watson query of data is conducted. Watson selects a diagnosis. Watson then selects and prioritizes likely treatments.
Specialists at MD Anderson are working to use Watson to facilitate the critical leap needed to achieve truly personalized medicine. The success of Watson as a system, depends on what data hospital organizations are willing to release to the database. More data is one aspect of the issue, quality of data is a far more significant issue, MD Anderson can provide data that can be trusted. There is enough of the trusted data to be significant, and there is enough data to permit comparisons.
This is what is needed so that the clinicians can be sure of the trustworthiness of an answer provided by Watson. Physicians can base decisions on years of patient data — the data in mind about their own patients. That is a valuable combination of data and experience. But physicians can not know all the research, sometimes their data is outdated, sometimes they just have not heard of the latest advances. Thus enters Watson to compliment what the physician or group practice knows.
MD Anderson is leading the way in developing personalized therapies for cancer. The data available to Watson can be in a format that hides the personal medical information of a patient. Training Watson is like training a medical student, you want the computer to make decisions based on the most up to date, accurate research available.