AI and Pathology

31 May 2023

How AI is looking to the future

SA Pathology’s Associate Professor David Ross, a consultant haematologist at The Royal Adelaide Hospital and Flinders Medical Centre, is passionate about leveraging technology in healthcare and says the potential opportunities for artificial intelligence (AI) in pathology are endless.

In collaboration with the University of Adelaide’s Australian Institute for Machine Learning (AIML),  Associate Professor Johan Verjans and Dr Zhibin Liao, David is looking into how we can improve its accuracy for a variety of disease and blood biomarkers. 

He says pathology and AI make a fantastic partnership because “As humans, there’s a limit to what we can see and our ability to recognise patterns.

A lot of what we do as pathologists is looking at a sample and recognising a certain pattern to make a diagnosis. Machine learning and AI strategies are very good at pattern recognition - it seems to me to be a natural fit.”

 

Here and now 

SA Pathology is already using AI in some of its haematology work. CellaVision is a commercially developed system that uses artificial intelligence to interrogate images and perform routine blood counts for the team.

SA Pathology has invested in CellaVision to perform routine blood counts, and by the end of this year, aims to operate these instruments in most of its metropolitan and regional laboratories, and to network them across the state. 

David says, “SA Pathology has a broad network across the state. The fact that we can link up all these instruments across our sites is very positive. 

CellaVision takes photographs of the individual cells in a blood smear and uses an internal algorithm to identify, count and classify each cell it sees. 

The instrument scans at least 200 cells and therefore has a higher chance of detecting rare abnormal cells.” 

In addition to improving pathology service across the state, CellaVision has significantly improved patient care in the regions. In the past many regional laboratories sent a physical sample (a blood smear) to Adelaide for clinical review when a blood disease was suspected. Staff now lodge high-resolution images in CellaVision for immediate transfer and review in Adelaide, leading to regional clinicians and their patients receiving their results more quickly.

While David admits there is a limit to the range of abnormalities CellaVision can detect, it already performs basic functions and sometimes picks up rare abnormalities a person may miss. 

 

Partnership driving innovation 

David knows there’s opportunity to improve, and that’s what prompted his partnership with Johan, Zhibin and AIML. 

The collaboration has been running since 2021. The team is analysing existing images CellaVision has taken to identify features that correlate with certain rare diseases. This information is used to teach the computer to detect and discriminate disease states and build an AI solution that will suggest a disease diagnosis.  

Johan, who had long been interested in partnering with SA Pathology, says he and David met through their mutual passion for biomedical research. 

“David is fantastic and the relationship with SA Pathology has been very positive. SA Pathology is one of Australia’s largest pathology services and there’s huge potential for collaboration.”

David provides the data source of images, and Johan and Zhibin feed them to the machine so it can learn how to spot abnormal cells and disease.

 

Flagging disease

David compares the breakthrough of AI to the first computers. 

“With everything, there is a lead-in phase, and then a bit of an explosion. There was a time when computers were the size of a caravan and only used at universities and government facilities – now we all have one in our pocket. I think this is what we’re seeing with AI. It’s gone from that nerdy, high-tech concept to being more widely applicable.” 

While AI takes on much of the repetitive work, David maintains there will always be a need for pathologists. 

“Following our current trajectory, in the future I can see us using our own captured images for an initial analysis by CellaVision which then produces a flag telling us to look at case number X because it may have disease Y, which is where the pathologist gets involved.” 

 

Exciting times ahead

David says proposed developments in the AI space would mean quicker turnaround times for patients, and extra time for pathologists to investigate more complex cases. 

Johan agrees, “In the near future, we anticipate remarkable opportunities to automate processes with AI; ideally that may improve processes, to intervene earlier, keeping people healthier.”

David is excited that SA Pathology are getting involved on the ground with this technology.

“SA Pathology is a dynamic organisation and it’s looking to the future. 

AI is going to become more and more important, and it's going to accelerate, it’s exciting to be engaged in this work.”