The application of AI and deep learning can improve the results of IVF treatment by up to 50 per cent, according to Life Whisper, an Adelaide-based startup which provides an image diagnostics platform to assist doctors identify viable embryos.
“It’s a huge accuracy uplift in picking the best embryos,” says Life Whisperer co-founder Dr Michelle Perugini, speaking to Which-50 about her startup and the role of data and analytics in medical care.
A critical part of IVF treatment includes analysing embryos to determine their suitability. Traditionally this step was completed by eye, with embryologists analysing images under a microscope to determine their viability – “a typically manual and imprecise process”, according to Life Whisperer.
“It’s an area where there’s a lot of subjectivity, at the moment, in selecting the best embryos. It’s very difficult for clinicians to do that,” Perugini told Which-50.
An IVF patient will usually produce between one and 10 embryos, but ultimately only one will be selected to transfer into the patient, Perugini explained. “Picking which one has the highest chance of creating a pregnancy is very difficult for them visually.”
Life Whisperer has applied artificial intelligence and machine learning to images of potential embryos in an attempt to assist clinicians in selecting the most viable embryos. The images are added to a web based system which can then run large scale analysis and matching to help identify embryos with higher chance of success.
Images are scanned for complex patterns and features common in more healthy embryos, ultimately providing clinicians with information on which embryo has the best chance of success.
“The AI essentially provides an extra set of eyes via the computer that can help them to make the best decision and pick the right embryos first time.”
Perugini says clinicians can identify the embryos on either end of the quality spectrum with relative ease but “the 90 per cent in the middle” are difficult to assess.
Adding AI and machine learning appears to have dramatically improved accuracy. Perugini cited two clinical studies which had shown adding the AI tool had improved success rates between 30 and 50 per cent.
Technology to support change
According to Perugini, the fertility sector and patients have been particularly receptive to the application of emerging technology like Life Whisperer. Due in large part, Perugini says, to the existing struggles and difficulty in identifying viable embryos.
“What [clinics are] doing is not working particularly well and it’s not the case where they’re getting [a] high level of accuracy when they’re selecting these embryos. So they understand that they need technology to be able to support them.”
For patients the potential is a low cost, non invasive process that can greatly improve accuracy.
IVF treatment costs can range from $5,000 to $20,000 in Australia. But more important than cost savings, Perugini says, is the potential to lessen the emotional strain of unsuccessful IVF.
“It’s hugely expensive but I think there’s other more emotional challenges for the patient [with IVF]. It’s extremely taxing emotionally with all the hormone treatments. It’s a very difficult procedure to got through. It’s very time consuming.”
“Not coming out with a baby at the end of that process is really emotionally taxing on the females undergoing the process.”
The Life Whisperer solution is a continually improving process – machine learning generally benefits from more data – and an example of where healthcare is headed, according to Perugni. She believes emerging technology like AI and data and analytics will play a significant role in assisting healthcare professionals whose job increasingly involves trawling through data.
However, the technology is unlikely to replace clinicians, Perugni said. Instead it will play and augmented role supporting decision making, as Life Whisperer does now.
For example, clinicians retain the final say on embryo selection, including overruling the algorithm’s suggested embryo despite Life Whisperer’s accuracy.
“The final decision rests with the clinician. And I think that’s really important until they get a higher level of confidence in the algorithm’s ability to predict … Over time as the technology improves itself in the clinic they will put more reliance on it,” Perugni said.
Founding and running Life Whisper meant Perugini putting a successful academic career on hold in a bid to offer solutions to more patients through the commercial application of research.
“I really loved being an academic and a research scientist. It really scratched my itch around researching really tough problems. The thing that it didn’t give me is to then take those [solutions] to the world in terms of a commercial product that could actually benefit patients.”
Life Whisperer is part of a larger image based medical diagnostic platform.
“What that does is automate the process of going from medical images to creating diagnostic tools that can actually be used in the clinic. We’ve kind of automated that whole process and built the capability ourselves.”
Developing the technology in house is important, Perugini says, because it closes the gap between the AI based tools provided by larger players, like IBM Watson, and specific applications.
Dr Perugini will be speaking about Life Whisperer and other emerging technology during the IAPA Advancing Analytics 2018 National Conference in Melbourne on the 18th of October.