About 260,000 women die of cervical cancer each year, with India accounting for as much as 25 per cent, meaning that 67,000 Indian women die from the cancer each year. Cervical cancer can now be prevented through early detection, so finding more effective screening methods can reduce the incidence of this cancer. Microsoft believes the introduction of artificial intelligence technology will help to analyze Pap smear samples more quickly. Last year, the company partnered with SRL Diagnostics to create an AI network for Pathology. SRL Diagnostics performs cell analysis on more than 100,000 cervical smear samples each year – 2% of which are abnormal and require further analysis. With AI technology, cytopathologists will be able to perform their work more efficiently.
Dr Arnab Roy, Head of New Program and Knowledge Management Technology at SRL Diagnostics, said: “We are looking for ways to ensure that our cell pathologists are able to find those 2% abnormal samples more quickly. “
Manish Gupta, principal application researcher at Microsoft Azure Global Engineering, said the AI algorithm could also help bridge differences between cell pathologists and allow “consensus in areas of assessment.” “Even if the overall diagnosis is the same, different cell pathologists will examine the different elements of the coating in a unique way,” says Dr Roy. This is a subjective factor throughout the process, and a lot of it is related to the experience of the experts”.
The algorithm was created based on thousands of cervical smear notes from cell pathologists in multiple laboratories. Microsoft said it found inconsistent annotations (i.e. if three team members observed them differently) and sent them to advanced cell pathologists for final analysis.
The technology has been so successful that it has now been validated in the lab for 3-6 months. During this period, more than half a million anonymous digital slides were evaluated and previewed at hospitals and other diagnostic centers. Cell pathologists can now reduce the area to 20 areas and verify positive cases by examining the entire area of the cytological image based on the glass liquid, increasing efficiency and speeding up the initial screening process.
Dr Roy said: “The API has the potential to increase the productivity of cell pathology slices by about four times. In the future of automatic preparation of slides with AI assistance, cytopathologists can complete their work in two hours, compared with eight hours previously”
In addition to cervical cancer, this AI technology can help diagnose many pathologies early, including oral, pancreatic and liver cancers. It will undoubtedly lead to tremendous advances in medicine, which could ultimately save hundreds of thousands of lives each year.