AI identifies up to 13% more cases of breast cancer – 11/22/2023 – Science

Artificial intelligence could dramatically reduce the number of undetected early-stage breast cancers, according to research showing the technology’s potential to improve and speed up medical diagnoses.

The analysis using artificial intelligence identified up to 13% more cases than doctors had identified — it is estimated that 20% or more of cancers are not detected with current screening tests without the use of AI.

The new study, published last Thursday (16) in Nature Medicine, highlights how technology can help combat this type of disease, identifying errors or signs that are difficult to interpret and that end up ignored by humans.

“Our study shows that the use of AI can act as an effective safety net, a tool to prevent more subtle signs of cancer from going unnoticed,” said Ben Glocker, study co-author and professor of machine learning for imaging at Imperial College London.

“Seeing firsthand that the use of AI could substantially reduce the rate of cancers missed in breast screening is a huge boost to our mission to transform cancer care with this capability.”

The study used an AI tool, known as Mia, developed by Kheiron Medical Technologies, a UK company specializing in medical diagnoses with AI. Researchers analyzed data from 25,000 women who underwent breast cancer screening exams in Hungary between 2021 and 2023.

The study had three phases, and in each of them the radiologists used AI differently. The groups showed improvements in cancer detection rates of 5%, 10% and 13%, compared to standard reading done by at least two radiologists.

Most of the additional cancers discovered were invasive, meaning they had the potential to spread to other parts of the body.

The work provides important evidence that AI can improve accuracy in identifying malignant tissue, as well as speeding up the process. Research from Sweden, published in late August, indicated that AI-enhanced mammogram analysis resulted in a cancer detection rate similar to standard human double reading.

Hungary’s latest research was “a promising example of how we can use AI to speed up diagnosis and treatment” in the NHS, said Katharine Halliday, president of the UK’s Royal College of Radiologists, who did not participate in the research.

“The results highlight the potential of AI to improve the accuracy of mammogram interpretations and support clinical decision making,” she explained. “The study emphasizes the complementary nature of AI and radiologists, envisioning a collaborative future that combines the strengths of both.”

The use of artificial intelligence also offers the possibility of speeding up analysis. The authors of the Hungarian paper said Mia could save up to 45% of the time spent reading breast cancer scans.

According to Kheiron, the tool has been tested in 16 hospitals in the United Kingdom and is being implemented in the United States.

The researchers highlighted the importance of continuing research to broaden and deepen how AI can be used to detect cancer. Areas to prioritize included getting results from more countries and using other AI systems, as well as monitoring the emergence of more cancer cases in their study group, they said.