IDDF2019-ABS-0095 Can artificial intelligence support the clinical decision making for hepatocellular carcinoma?

Abstract

Background Artificial intelligence (AI) is emerging as a revolutionary technology with the power to transform healthcare. IBM Watson for Oncology (WFO), as the first AI clinical decision support system (CDSS) with a cognitive-support approach for therapy selection, has been investigated about its impact on clinical decision making in some cancer types and showed potential benefit in cancer therapy. However, the use of WFO in hepatocellular carcinoma (HCC) has not been reported. Methods A cross-sectional retrospective study was performed to evaluate the degree of recommended treatment concordance between WFO and multidisciplinary team (MDT) for 550 HCC patients diagnosed between Jan 2013 and Jun 2013 at three tertiary referral centers in China. Comparative survival analysis with propensity score matching (PSM) method was conducted to assess whether the WFO-recommended modality could benefit patients in survival when compared to MDT. Univariate and multivariate regression analyses were performed to identify factors associated with discordance. Results The overall concordance rate was 58.5% in all cases, and 53.7%, 61.4%, 47.3% and 61.7% for patients with BCLC stage 0, A, B and C, respectively. For BCLC stage 0, radiofrequency ablation (RFA) was the first recommended treatment by both MDT and WFO without significant difference (52.6% vs. 50.0%, P=0.867) while hepatectomy was for BCLC stage A (75.7% vs. 65.6%, P=0.066). For BCLC stage B/C, TACE was recommended more by WFO (100.0% vs. 6.3%, P Conclusions For HCC patients, MDT recommended more aggressive treatments than WFO. WFO-recommended treatments did not show survival superiority to MDT, in patients with BCLC stage B/C.