No tumor is the same. Personalized medicine promises to find the best individual therapy for each patient. Based on mutations in the tumor cell genome, the tumor may qualify and thus determine which therapy would be most promising.
In the future, doctors at the University Hospital of Geneva (HUG) will receive Watson's support for Genomics. The IBM Watson Health artificial intelligence (AI) -based tool aggregates data from a vast database of clinical studies and clinical studies used to analyze the genome of a patient's tumor. according to a report for doctors, HUG said Wednesday.
HUG is the first European university hospital to use Watson for Genomics. This will help doctors at HUG offer more personalized cancer treatment, said Rodolphe Meyer, deputy chief information officer at HUG. Outside Europe, especially in the US, this tool is already being used in some places.
AI aims specifically to save time looking for the best possible treatment strategy: The tool will produce a report within ten minutes, a manual search will take about 160 hours, said Nathan Levitan of IBM Watson Health.
The company has developed a variety of AI-based personalized medicine tools. Watson criticized oncology for making false or even dangerous suggestions about therapy. This tool is intended for learning by analyzing text from medical literature and medical records and for developing therapeutic strategies.
However, the data proved to be very complex for AI analysis. Especially in patient records, the information is not always clear how much AI should be needed.
On top of that, the obstacle was that KI's "mindset" based on statistics for data search and weight data differs from that of an experienced medical professional. The latter sometimes finds relevant information for therapeutic decisions in a subordinate clause or notes that his patient belongs to a small group of special cases in which therapy is effective. However, for AI with statistics, special cases are not relevant.
In the case of Watson for Genomics, the case is somewhat different: gene data is far more structured than patient data. Gene mutations are infallible or not. Through natural language processing, the tool reviews specialist literature and clinical trials for specific information about these mutations, potentially drawing attention to new drugs or clinical trials that have just begun and may be of benefit to the patient.
In a study published in the journal Oncologist in 2017, Watson for Genomics CI provided valuable additional information to approximately 30 percent of the approximately 1,000 patients who avoided expert mutation analysis and the development of a 2018 therapeutic strategy in the journal Frontiers in Medicine, confirmed this: In an analysis of nearly 200 patients, the tool suggested 88 of the 104 treatment options also listed by specialists, but provided an additional 225 treatment options.