BACKGROUND
Artificial intelligence（AI）is developing quickly in the medical field, which gives convenience to both medical staff and patients. The AI CDSS Watson for Oncology (WFO) is an outstanding representative of AI in medical field, it can give prompt treatment recommendations to cancer patients, just as an excellent oncologist. It is more and more widely used in China, but there is no report about whether it suits Chinese patients. Here we report a retrospective study about the consistency of WFO's recommendations and commendations of the same patient given by our multi-disciplinary team board（MDT） at our center on lung cancer patients’ treatment.
OBJECTIVE
Our objective was to explore practicability of the WFO to the lung cancer cases in China and how to make it more suitable for Chinese patients with lung cancer.
METHODS
We selected all the lung cancer patients who were hospitalized and received antitumor treatment in the 2nd Xiangya hospital cancer center for the first time from September to December in 2017. WFO gave treatment recommendations to all qualified cases. If our actual therapeutic regimen (which was given by our MDT) was “recommended” or “for consideration” in WFO, we thought it was of consistency, if it was “not recommended” or WFO didn’t have this option, we thought it was of no consistency. Blind second round reviews were made by our MDT to reassess the incongruent cases.
RESULTS
WFO didn’t support 18% of all the cases (n=182). Of the 149 cases (82%) supported, 66% were consistent with our MDT recommendations. Subgroup analysis showed that pathological type and staging had a significant effect on consistency (p=0.004, p=0, respectively). Age, gender, whether there was gene mutation or not had no effect on consistency. In 81% of the inconsistent cases, our MDT gave two treatments with Chinese characteristics , which were different from recommendations given by WFO but were also with excellent effect. If WFO can bring the two alternative treatments into the recommended or considered range, the overall consistency can be elevated from 66% to 93%.
CONCLUSIONS
In China, the majority of the treatment recommendations of WFO are consistent with the recommendation of the expert group, but there is still a relative high proportion of cases which are not supported by WFO. WFO cannot substitute our oncologists currently. As doctors’ assistant, it can improve the efficiency of our work, and it needs to learn local characteristics of patients to become a better assistant.
CLINICALTRIAL
none