What if we could design a computer application like Google Earth where scientists can zoom in to see not land or buildings, but microscopic human cells up close – and so spot disease?

This is exactly the type of technology UQ researchers have recently developed to boost the efficiency of pathologists diagnosing medical conditions – and it’s set to revolutionise the healthcare industry worldwide, according to Professor Brian Lovell, AI specialist in UQ’s School of Electrical Engineering and Computer Science.

“A change in pathology lifts the whole system up, and this is what our new digital pathology system will do.”

Working with Queensland-based Sullivan and Nicolaides Pathology (SNP), an intelligent system has been created that not only increases the accuracy of results and reduces the time taken to process specimens but is also improving the job satisfaction of lab employees.

Traditional pathology testing involves a lot of repetitive manual work by scientists and technicians – preparing slides, scanning them, classifying them, looking at them under the microscope, checking them, filing them, and then writing a report. And then there’s the issue of storing the slides and trying to find them again later, when the specimens themselves may have decomposed.

The new system automates many of the functions previously done by hand.

“We’ve got automated pipettors and we’ve got things that carry specimens around the laboratory – none of that had happened before; it was a manual process, just like it was 50 years ago,” Dr Harrison said.

“Applying good digital imaging and artificial intelligence has provided an aid to make life better and get a better outcome as well.”

And with more automation comes fewer sampling errors and faster turnarounds – in some cases, pathology can even be done in the operating theatre.

“In traditional microscopy, a person leans over a microscope for hours and hours looking at glass slides, then they scan them, diagnosing them and then archiving them – but no-one else usually gets to see that slide again,” Professor Lovell said.

“Whereas with our system, we digitise the slides as huge images, which can be up to a terabyte in size, and diagnose from the images so that anyone can view them anywhere, anytime – even 10 years later.

“Our system is intelligent: it actually understands what it's scanning and looks for the important elements and makes sure they're in focus. It can use UV (fluorescent) or natural brightfield light sources and configures the light as each slide is processed.

“It also finds the diagnostic cells and rescans them at very high magnification, if necessary.”

Artificial intelligence is used to analyse the barcoded specimen and automatically configure the microscope for whatever test is required. In particular, the number and types of bacteria on the slide are identified and counted while the slide is being scanned.

“So, by the time the pathologist gets the slide it says these bacteria were detected there, this is how many, and they just have to confirm this and sign off on the report. So, it makes the diagnosis by the human much faster and more accurate,” Professor Lovell said.

The research to create this award-winning system has been more than a decade in the making, receiving support from Sullivan Nicolaides Pathology, the Australian Research Council and a Queensland Government Advance Queensland Fellowship. But without human involvement, there would be no automation.

“We had to train an algorithm to recognise the things that we recognise down the microscope,” Dr Harrison said.

“So, we were converting human thoughts into a computer program. And that computer program we found could be trained to be very, very reliable in coming to the same decision that we were making. More reliable, in fact, than humans because it never tires and it applies the same rules every time.”

In traditional testing, a pathologist looks down a microscope in a dark room, and a second person may check their results. Now, a person sits in front of a computer screen looking at the digital pictures taken through the microscope, and then confirms that the algorithm has interpreted that digital pattern correctly.

“It's not as though we're automating everything and taking the person out of it. We're giving them an aid through the diagnostic process to make their life better, and to get a better outcome as well.”

The system is currently able to conduct 17 different tests, including Q fever (used in the abattoir industry) and immunology, and has been accredited by the National Association of Testing Authorities (NATA) ready for rollout globally.

The next milestone is the ability to process Gram stains, a microbiology test that checks for bacteria at suspected infection sites such as the throat, lungs, genitals or skin to determine which antibiotics will be the most effective treatment.

“And that's probably been the hardest test of all to implement because it needs x100 oil immersion photography, which is very difficult to get right,” Professor Lovell said.

But Dr Harrison has no doubts it will happen.

“I don't believe this has been done anywhere else in the world: it's very sophisticated,” he said.

Professor Lovell also hopes that images can be put on pathology slips in future, so that people can see their results instead of just reading about them.

But the best thing will be the much quicker pathology results that we will all receive in the future: ‘googling the human’ may well save humanity.

STOP PRESS: Since publishing this story, the Digital Pathology System by UQ/Sullivan and Nicolaides has won the national Australian Information Industry Association award, Business and Innovation category.

Each month, Research News uses the latest findings to help explain the issues facing your community.