Combining the expert opinion of doctors with insights from artificial intelligence (AI) could provide the best option for skin cancer diagnosis, according to researchers from the University of Queensland.
UQ’s Professor Monika Handa was part of an international team that examined how collaboration between humans and machines worked in real life.
“This is important because AI decision support has slowly started to infiltrate healthcare settings, and yet few studies have tested its performance in real world settings or how clinicians interact with it,” Janda said.
The study was performed using a convolutional neural network (a type of algorithm inspired by biological vision processing pathways that can classify images faster than some other kinds of algorithms) that analysed pigmented skin lesions and compared the findings with human evaluations on three types of AI-based decision support.
The highest diagnostic accuracy was achieved when the crowd wisdom of doctors was combined with AI, suggesting either human-AI or crowd-AI collaborations are preferable to individual experts or AI operating alone.
However, the benefit for doctors evaluating potential skin cancer cases wasn’t equal across the board.
“Inexperienced evaluators gained the highest benefit from AI decision support and expert evaluators confident in skin cancer diagnosis achieved modest or no benefit.”
The study supports predictions from the likes of the Victorian Melanoma Service, who in the wake of AI’s growing diagnostic prowess, said that AI would be more likely to support doctors in diagnosis rather than replace them.
Janda said it also shows how AI-enabled clinical tools can be further improved.
“Our study found that good quality AI support was useful to clinicians but needed to be simple, concrete, and in accordance with a given task," she said.
“For clinicians of the future this means that AI-based screening and diagnosis might soon be available to support them on a daily basis.”
However, she warned that more work is needed to ensure the best outcomes for patients.
“Implementation of any AI software needs extensive testing to understand the impact it has on clinical decision making," she said.
Data from the Cancer Council shows GPs currently conduct more than 1 million patient consultations for skin cancer every year, with roughly two-thirds of Australians being diagnosed with skin cancer by the time they are 70.