2023-06-06 16:00:06
Artificial intelligence can read mammograms and intercept women destined to develop breast cancer “better than the classic 5-year risk prediction model”. And integrating the two systems could improve physician analysis. This is what emerges from a maxi study in which several AI algorithms were tested on thousands of mammograms. The results are published in the journal ‘Radiology’.
A woman’s breast cancer risk is typically calculated using clinical models such as that of the Breast Cancer Surveillance Consortium (BCSC), which uses self-reported information and other data about the patient – including age, family history of the disease, whether had children, if he has dense breasts – to define a score. “Clinical risk models depend on gathering information from many different sources, which is not always available or collected,” says principal investigator Vignesh A. Arasu, a radiologist trainee at Kaiser Permanente Northern California. ‘Ai provide us with the ability to extract hundreds to thousands of additional mammography features.”
In the retrospective study, Arasu used data associated with negative screening 2D mammograms (with no visible signs of cancer) performed at Kaiser Permanente Northern California in 2016. Of the 324,009 women screened in 2016 who met eligibility criteria, a random subcohort of 13,628 women were selected for analysis. And all 4,584 patients in the total pool who were diagnosed with cancer within 5 years of their original 2016 mammogram were also studied. All women were followed up through 2021.
Using mammograms, 5-year breast cancer risk scores were generated by 5 AI algorithms, including 2 academic algorithms used by the researchers and three commercially available. The risk scores were then compared with each other and with that obtained from the BCSC standard model. “All five AI algorithms performed better than the BCSC risk model for predicting 0-to-5-year breast cancer risk,” reports Arasu. “This strong predictive performance suggests that AI is identifying both missing tumors and breast tissue characteristics that help predict future cancer development. Something about mammograms allows us to monitor breast cancer risk. This is the ‘black box’ of artificial intelligence”.
Some of Ai’s algorithms excelled at predicting high-risk patient cases of an often aggressive type of cancer that may require a second reading of mammograms, additional screenings, or short-term follow-up imaging. For example, when women with the 10% highest risk were evaluated, AI predicted up to 28% of cancers, compared to 21% predicted by the classic method. Used in combination, the Ai and BCSC risk models further improved cancer prediction.
“We are looking for an accurate, efficient and extensible means of understanding women’s breast cancer risk. Mammography-based AI risk models offer practical advantages over traditional ones, because they use a single source of data: the mammography itself.” explains Arasu, highlighting that some institutions are already using artificial intelligence to help radiologists detect cancer in mammograms. A person’s future risk score, which takes seconds to generate from artificial intelligence, could be integrated into the radiology report shared with the patient and their doctor. “It’s a tool that could help us deliver personalized precision medicine nationwide,” concludes the researcher.
#predicts #breast #cancer #risk #classical #models #study