Scientist off Massachusetts Institute of Technology (MIT) created an A new learning model based on artificial intelligence that can predict breast cancer up to five years before it appears.
The model learned patterns in breast tissue which are precursors to Malignant tumors from a mammogram, which is explained in a release Regina Barzilay, responsible for the project.
Instead of manually identifying such patternsThe research team trained a deep learning model to derive them directly from data of more than 90,000 mammograms.
"Since the 1960s, radiologists have noticed it Women have unique and highly variable patterns of breast tissue visible on mammography"Constance Lehman said.
This new technology is placed with accuracy to 31% of all breast cancer patients in their highest risk category, compared to only 18% of traditional models.
This was translated as a much better result to predict risk than existing approaches, because The patterns are too subtle for the human eye to discover them.
The project also aims to make the risk assessment more accurate for racial minorities, Many models have been developed in white populations and have been less accurate for other breeds.
"This is particularly important given that Black women have been shown to be 42% more likely to die from breast cancer Due to a wide range of factors that may include differences in detection and access to healthcare, "the article reads.
According to the authors, this system can also one day allow doctors to use mammograms to see if patients are at increased risk cardiovascular diseases or other types of cancer.