[ad_1]
Breast cancer is one of the top three causes of cancer death among women in the United States. While detection methods and technologies such as mammography have reduced the 39% mortality rate since 1989, more than 41,000 women will die in the United States. breast cancer this year, according to Cancer.net.
A new artificial intelligence model developed by researchers at MIT's Computer Science and Artificial Intelligence (CSAIL) labs can, however, analyze mammograms and predict breast cancer risks up to the end of the day. at five years in advance.
Risk models are imperfect
Since early detection is closely linked to the reduction of the mortality rate, field research focused on detecting symptoms as early as possible.
"Researchers have been creating risk models for breast cancer since the late 1980s. But the way scientists thought about it has not changed much until recently," says Adam Yala, PhD student. at MIT CSAIL and co-author of the study. was published in the medical journal Radiology.
Previous risk models were based on factors such as age, family history of breast cancer, breast density, and genetic factors. Although these models have helped to improve early detection, they lack a lot of important patient data and do not provide accurate results at the individual level.
"The problem with this approach is that you summarize the information that matters before you feed [it] in the model, which means that the models themselves have not been very accurate, "says Yala.
Scientists at MIT CSAIL, in partnership with Massachusetts General Hospital (MGH), developed an in-depth learning model that was trained on 90,000 high-resolution mammography exams in 60,000 patients over the course of several years. years, with different results.
The MIT's in-depth learning algorithm detected structures in the breast tissue that suggested a cancer risk but were too subtle to be detected. As a result, AI can detect signs of breast cancer several years earlier than human radiologists, which could reduce invasive treatments and reduce medical costs.
According to the study, the model accurately predicts 31% of cancer patients in the highest risk category. The accuracy of the existing models is about 18%.
Provide personalized care
One of the benefits of AI-based breast cancer detection is that physicians will be able to provide personalized analysis and prevention to the patients.
"Rather than taking a single approach, we can customize cancer risk screening for women," said Professor MIT, Regina Barzilay, lead author of the study and cancer survivor. breast. "For example, a doctor might recommend to a group of women to undergo a mammogram every two years, while another high-risk group might undergo additional MRI screening."
"If you give the right screening to the right person, you can both improve the experience and reduce the harms of mammography, but also detect cancers earlier, which has a huge impact on treatment decisions. because what you are doing for the early stages the cancers are very different, "said Yala.
Until now, the model has proved equally accurate for all groups of people of different races and ethnicities. This is one of the pain points of the other risk models, whose performance varies from one population to the other. According to Yala, risk models based on high level surface factors such as age and family history are not widespread well. For example, if they are created from data mainly from white women, their results are poor in non-white patients.
"Our model is based on the actual models of mammography, even though in our datasets, African-American women [comprise] 5% of the data set as a whole, the model still works equally well for both. This makes me understand that tissue information is more shared, while family history may not be, "says Yala.
Researchers are currently working with more hospitals to study and serve other groups and make the model even more equitable. They will also seek to expand their work to other types of cancer, particularly those with less effective risk models, such as pancreatic cancer.
The MIT CSAIL in-depth learning model is one of many projects aimed at applying artificial intelligence to the diagnosis and treatment of breast cancer. Major technology companies, including IBM, Google, and DeepMind, an affiliate of Alphabet, are fielding their efforts alongside universities such as the University of New York and Harvard Medical School.
"Our goal is to integrate these advances into the quality of care," Yala said. "By predicting who will develop cancer in the future, we can hope to save lives and stop cancer before the symptoms begin to manifest themselves."
[ad_2]
Source link