Analysis of Watson for oncology and clinicians' treatment recommendations for patients with breast cancer in Korea: A single center experience.

Background: Various clinical applications have been attempted using artificial intelligence (AI) clinical decision support system (CDSS), and it has become a starting point for personalized cancer treatment. We aimed to identify the degree of agreement between the AI-CDSS, Watson for Oncology (WFO), and the clinician in treatment recommendations for Korean breast cancer patients and to provide guidelines for future improvement.

Methods: One hundred and eighty-three breast cancer patients who underwent treatment at the Pusan National University Hospital between January 1, 2016 and May 31, 2017 were enrolled in this study. The concordance between WFO's and clinicians' treatment recommendations were examined according to various factors.

Results: WFO gave the same treatment option recommendations as clinicians in 74 (40.4%) of the cases. According to the logistic regression, the difference in recommendation concordance between stage I and stage III was statistically significant (P = 0.004), and there was no difference among other factors.

Conclusion: The concordance of treatment recommendations was low overall. However, this is largely attributable to the differences of medical insurance system and healthcare environment between the United States and Korea. In the future, region-specific features should be considered or reflected during the development of AI-CDSS.