Health and wellness

Artificial Intelligence could help diagnose cancer in future, say Hong Kong doctors

  • Hospital Authority launches investigation after medical staff fail to spot disease in three patients
  • Leading radiologist says AI may help one day, but the technology is far from being ready
PUBLISHED : Thursday, 22 November, 2018, 9:35pm
UPDATED : Friday, 23 November, 2018, 12:05pm

Artificial intelligence could one day prevent the failure to diagnose cancer in Hong Kong patients, doctors in the city said on Thursday – but it will not happen any time soon.

Dr Poon Wai-lun, a council member of the Hong Kong College of Radiologists, was speaking a day after it was revealed doctors had misread X-rays that showed lung cancer in three people.

In a case involving Prince of Wales Hospital in Sha Tin, doctors failed to follow up on a shadow on the left lung shown in a patient’s X-ray film, which taken in March last year.

The patient was later confirmed to have a metastatic lung tumour, after being admitted to the hospital this month suffering gallbladder inflammation.

Other cases at Queen Mary Hospital and Princess Margaret Hospital have also come to light this week, prompting the Hospital Authority to form a panel to explore ways to ensure health care workers spot abnormalities in chest X-rays sooner.

A report by the panel is expected to be completed and released in eight weeks.

Poon said AI had the potential to help doctors process X-rays, and said a public hospital in the city had been testing such a system.

“If the AI system notices something suspicious with an image, it would require radiologists to study and write report about the image earlier,” Poon said.

But, he said people should not overestimate the ability of AI to assist in image diagnosis.

“Do not deify AI system, as it is still in the stage of development in the field of radiology,” Poon said.

New cancer cases in Hong Kong projected to rise by up to 40 per cent by 2030

At Queen Mary Hospital, doctors failed to notice an abnormal mass in a patient, who had been having regular chest X-rays since 2016, three times. A report said the mass in a lung was believed to be malignant after the patient sought medical treatment at Ruttonjee Hospital last month.

In the last case, at Princess Margaret Hospital, a patient was confirmed to have lung cancer after being admitted after an accident. Doctors later revealed the patient’s chest X-ray films taken in February last year, and again in May this year, showed a shadow in the patient’s right lung had gone unnoticed.

Poon said public hospitals handled more than 1.8 million X-ray films each year, meaning it would not be possible for radiologists to go through every single film.

In general, clinical doctors who requested the X-ray would study the findings themselves. If they had concerns, they would then pass the image to a radiologist.

According to Poon, fewer than 15 per cent of those films in public hospitals are studied by radiologists annually.

Speaking on an RTHK radio programme on Thursday morning, Thomas Yau Chung-cheung, a medical oncologist from the University of Hong Kong, agreed that AI would be helpful in assisting doctors.

“An AI system could help reduce clinical mistakes,” Yau said. “It could identify even a small spot. If clinical doctors could not see them, they could at least be alerted.”

Poon said other possible ways to plug the oversight would be asking a more senior clinical doctor to help review an X-ray image, as well as hiring more radiologists.

Public hospitals suffered from a limited manpower shortage, including doctors specialising in radiology. Last year, turnover rate of public radiologists was 8 per cent, which was the second highest of all specialities.

As of March this year, there were 290 doctors working in radiology departments in the city’s public hospitals.

Patients’ rights advocate Tim Pang Hung-cheong said the authority should review the entire imaging diagnostic process.

“Be it the system, or the working attitude of staff, there should be a proper review,” Pang said.

Additional reporting by Karen Zhang