[ad_1]
OAK BROOK, Illinois – Massachusetts General Hospital (MGH) researchers have developed a deep learning model that identifies imaging biomarkers on screening mammograms to predict a patient’s risk of developing breast cancer more accurately than with traditional tools risk assessment. The results of the study are presented at the annual meeting of the Radiological Society of North America (RSNA).
“Traditional risk assessment models do not take advantage of the level of detail contained within a mammogram,” said Leslie Lamb, MD, M.Sc., breast radiologist at MGH. “Even the best existing traditional risk models can separate patient subgroups but are not as accurate on an individual level.”
Currently available risk assessment models incorporate only a small fraction of patient data such as family history, previous breast biopsies, and hormonal and reproductive history. Only one feature of the screening mammogram itself, breast density, is incorporated into traditional models.
“Why should we limit ourselves to breast density alone when such rich digital data is embedded in each woman’s mammogram?” said senior author Constance D. Lehman, MD, Ph.D., division chief of breast imaging at MGH. “Each woman’s mammogram is unique to her, just like her fingerprint. It contains imaging biomarkers that are highly predictive of future cancer risk, but until we had the deep learning tools, we weren’t able to to extract this information to improve patient care. “
Dr. Lamb and a team of researchers developed the new deep learning algorithm to predict breast cancer risk using data from five MGH breast cancer screening sites. The model was developed on a population that included women with a personal history of previous breast cancer, implants or biopsies.
The study included 245,753 consecutive 2D digital bilateral screening mammograms performed on 80,818 patients between 2009 and 2016. Of the total mammograms, 210,819 exams in 56,831 patients were used for training, 25,644 exams from 7,021 patients for testing and 9,290 examinations from 3,961 patients for validation.
Using statistical analysis, the researchers compared the accuracy of the deep learning image-only model with a commercially available risk assessment model (Tyrer-Cuzick version 8) to predict future breast cancer within five years of mammography. index. The deep learning model achieved a predictive rate of 0.71, significantly exceeding the traditional risk model, which achieved a rate of 0.61.
“Our deep learning model is capable of translating the full diversity of fine imaging biomarkers into mammography that can predict a woman’s future risk for breast cancer,” said Dr. Lamb.
Dr Lamb said the new deep learning model has been validated externally in Sweden and Taiwan, and further studies are planned for larger African American and minority populations.
In MGH, information on the risk of deep learning is available on the reporting software when the radiologist reads a patient’s screening mammogram.
“Traditional risk models can take a long time to acquire and rely on inconsistent or missing data,” said Dr. Lamb. “A risk model based only on deep learning images can provide greater access to more accurate and less expensive risk assessment and help deliver on the promise of precision medicine.”
###
Co-authors are Adam Yala, M.Eng., Peter Mikhael, BS, and Regina Barzilay, Ph.D.
For more information and images, visit RSNA.org/press20. Press the required account to view the embargoed materials.
RSNA is an association of radiologists, radiologists oncologists, medical physicists and related scientists that promotes excellence in patient care and healthcare delivery through education, research and technological innovation. The company is headquartered in Oak Brook, Illinois. (RSNA.org)
Editor’s Note:
The data in these publications may differ from those in the published abstract and from those actually presented at the meeting, as researchers continue to update their data until the meeting. To make sure you are using the most up-to-date information, call the RSNA Media Relations team in Newsroom at 1-630-590-7762.
For information on patient-friendly mammography, visit RadiologyInfo.org.
Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of press releases published on EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.
.
[ad_2]
Source link
