Watson is treating cancer cases

Doctors demonstrate Watson’s diagnosis of a breast cancer patient at Gachon University’s Gil Medical Center in Incheon. [GIL MEDICAL CENTER]

Cho Tae-hyun, 62, was diagnosed with stage 3 colorectal cancer in November 2016. A part of his colon was removed at Gacheon University’s Gil Medical Center, and for the following six months, he was treated with an anticancer drug called Folfox. Cho goes for check-ups every three months. The disease has not recurred.

The anticancer drug Cho takes was a recommendation by Watson, IBM’s artificial intelligence software. Doctors inputted Cho’s medical record and Watson made its recommendation after eight seconds. “It has a higher survival rate than other options,” Watson concluded in text.

Five doctors at the hospital had a 10-minute discussion of their own and came up with the same recommendation as the AI software.

It’s been a year since Gil Medical Center adopted Watson as its “AI doctor.” During that period, the system has helped treat 557 cancer patients. Having started with four types of cancer - colorectal, stomach, breast, lung - it now can recommend treatments for more: cancer of the uterus, ovaries, prostate gland and bladder.

Six other hospitals in Korea are using Watson.

On Tuesday, Gil Medical Center held a symposium to reveal Watson’s achievements over the year by analyzing the 557 cases it has worked on.

The most notable fact is that the Watson’s recommendations are matching those of human doctors at a greater rate. Watson is supposed to evolve or get smarter by itself analyzing patients’ medical records and referring to medical journals and text books. Its suggestions come in three types -recommended treatments, treatments that should be considered and non-recommended treatments.

In the case of colorectal cancer patients, 55.9 percent received the same suggestion for recommended primary treatments from doctors and Watson, according to Professor Baek Jeong-heum of Gil Medical Center’s surgery department.

“Watson’s ability has improved over the year and so has our medical team’s perspective and confidence in working with it,” said Baek.

However, there are still divergences between men and machine. Watson gives conservative treatment recommendations solely based on medical knowledge. For real doctors, the patient and family’s opinion and insurance status are factors that are also considered. Watson is not designed to understand insurance systems nor patient’s personal circumstances.

“Some patients ask for anticancer pills [instead of having to go to clinics to receive treatment], saying they have to work to make a living - even if they may end up dying,” said Baek. “This is the point where Watson and human doctors part ways.”

The level of satisfaction among patients is generally high: 48 of 51 patients who received treatment recommendations by Watson between October and December said they were “very satisfied” with their AI doctor.

But many remain skeptical as to whether Watson could outperform human doctors - or replace them in the near future.

“In the current stage, Watson is more of a tool to help doctors make their decisions,” said Professor Yoon-Young-ho of Seoul National University’s Cancer Hospital.

“Moreover, Watson is not reflecting characteristics particular to Korean patients. Its current role is to provide a second opinion - which in the past would have been sought from another hospital.”

“The level of satisfaction among patients has risen but this does not ensure medical quality,” he added. “We need to make a comparative analysis based on the five-year survival rate and the recurrence rate within one to two years of patients that Watson treated.”

BY JEONG JONG-HOON [song.kyoungson@joongang.co.kr]