Artificial intelligence in urologic oncology: the actual clinical practice results of IBM Watson for Oncology in South Korea

Abstract

Background

Artificial intelligence (AI) is changing our life, including the medical field. Repeated machine learning using big data made various fields more predictable and accurate. In medicine, IBM Watson for Oncology (WFO), trained by Memorial Slone Kettering Cancer Center (MSKCC), was first introduced and applied in 14 countries worldwide.

Our study was designed to assess the feasibility of WFO in actual clinical practice. We aimed to investigate the concordance rate between WFO and multidisciplinary tumor board (MTB) in Urologic cancer patients.

Materials and methods

We reviewed retrospectively collected data for consecutive patients who underwent WFO and MTB simultaneously in the diagnosis of urologic malignancy before determining further treatment between August 2017 and September 2020. We compared the recommendation of the AI system, WFO (IBM Watson Health, Cambridge, MA), with the opinion of MTB for further managing all patients diagnosed with urologic malignancies such as prostate, bladder, and kidney cancer.

Results

A total of 55 patients were enrolled in our study. The number of patients with prostate cancer was 48. The number of bladder and kidney cancer patients was 5 and 2, respectively. The overall concordance rate between WFO and MTB was 92.7%. Three patients could not suggest proper treatment options using WFO, and the recommended choice of WFO was not feasible in the Korean Health Insurance Review and Assessment Service.

Conclusions

The decision of WFO showed a high concordance rate with a multidisciplinary tumor board for urologic oncology. However, some recommendations of WFO were not feasible in actual practice, and WFO still has some points to improve and modify. Interestingly, applying WFO is likely to facilitate a multidisciplinary team approach.