Artificial intelligence (AI) showed excellent results in predicting a five -year risk of developing breast cancer compared to a standard clinical risk model using self -report and other information about the patient. this is reported in the study published in the journal Radiology .
Researchers used data related to negative (not showing visible signs of cancer) screening mammograms made in Kaiser Permanente North California in 2016. Of the 324 009 women who passed the screening in 2016 and corresponding to the selection criteria, an accidental subclause of 13,628 women was chosen for analysis. In addition, all 4,584 patients from the collection pool were studied, in whom cancer was diagnosed for five years after the initial mammogram of 2016. All women were observed until 2021.
Using deep training, the researchers developed the AI algorithms that have extracted hundreds or thousands of additional mammographic signs from mammograms and used them to predict the risk of breast cancer. AI algorithms showed significantly better accuracy of predicting risk of breast cancer for a five -year period compared to the clinical risk model BCSC.
“This is the first study that shows the superiority of AI over the clinical risk models on such a large sample,” said the leading researcher Vignesh A. Aras, doctor of medicine, doctor of philosophy, researcher and practitioner radiologist at the Kaiser Permanente North California.
According to Dr. Aras, the AI algorithms can help individualize patient care and increase the effectiveness of forecasting. For example, they can help determine how often it is necessary to screening in women with a high risk of breast cancer or what additional examination methods can be useful.
“We hope that our AI algorithms will be integrated into clinical practice in the future,” he added.