NZ radiology centres to trial AI diagnosis

We're all familiar with turning to Dr Google for a quick medical diagnosis, but a new era of automatic diagnosis using artificial intelligence has arrived.

A New Zealand medical scanning company plans to use algorithms to help detect prostate cancer.

The future of medicine is not a miraculous machine or a holographic doctor - it's about deciphering data. Far from a single X-ray, scanning machines now slice and dice each patient.

Australian company Maxwell Plus says it has developed artificial intelligence algorithms to ease the workload on doctors.

"We are looking at lung cancer, prostate cancer and breast cancer, as well as some work around neurodegenerative disease, so dementia and Alzheimer's," says Dr Elliot Smith.

New Zealand medical scanning company Mercy Radiology is planning to trial one of the algorithms to help radiologists diagnose prostate cancer.

Dr Lloyd McCann says this doesn't mean radiologists will suddenly be redundant.

"The radiologist will report the scan as usual, but the algorithm will be running in parallel, looking at the image data that is captured and assessing whether there is something that needs to be reported or not."

This is perhaps not the case in the United States, where AI software that detects eye disease without specialist input has been given the go-ahead.

It's a new era of automated decision making that's sparking ethical dilemmas.

The new Privacy Bill is currently before select committee. The Privacy Commissioner wants it to include a new principle - algorithmic transparency, the right to explanation, meaning when machines make automated decisions you can ask how they were made.

But will we understand the answers we get?

"If we allow people to have this right of explanation, sometimes it can be a bit tricky," says lawyer Benjamin Liu from University of Auckland.

"When scientists or doctors make a decision based on this recommendation from machine learning and AI, they just cannot explain it."

Mercy Radiology is planning its algorithmic trial within the next few months, then we might see who can be trusted - man or machine.