health no1 AI can estimate the risk of breast cancer for 5 years before diagnosis
Breast cancer is the most common cancer in women and is responsible for approximately 500,000 deaths worldwide every year. There are now many effective remedies for breast cancer, but a successful outcome is still dependent on initial diagnosis.
Later diagnosis requires more aggressive treatment, which come with great side effects and often fails. Therefore, identifying patients at risk of developing breast cancer has been an important focus for researchers to reduce the number of deaths related to breast cancer.
Screening programs that use mammography are employed to enable early detection and treatment of breast cancer. However, for this kind of screening, the sign of abnormality requires investigation of each mammogram, which is highly labor-intensive for the large number of women to check.
Apart from this, because images are reviewed manually, therefore there is an element of the subject and the danger of human error remains. In order to speed up the review of the mammogram and to enable the objective assessment of the risk, researchers are working on developing computer models that can screen the mammogram images faster and stronger for the risk of breast cancer.
'Unique patterns of breast tissue'
Using the information from more than 90,000 mammograms taken at the Massachusetts General Hospital (MGH), a team of MIT's Computer Science and Artificial Intelligence Laboratories has developed a new deep-learning model that changes the tissue of the breast Detects the subtle pattern, which is unable to human eye. To find out.
Mammograms and more than 60,000 patients are programmed using known results, the model identifies the pioneers as deadly tumors and can predict with a mammogram whether the patient is likely to have breast cancer.
Contrary to the assessment of major risk factors, such as age, family history of breast cancer, hormonal conditions and breast density, the MIT intensive-learning model identifies patterns of breast cancer. Predictions are thus data-driven and may be up to 5 years before the development of cancer. Thus screening provides a personal risk assessment that can be used to optimize screening and prevention programs on the basis of patient-to-patient basis.
Constance Lehman, Head of Department of Breast Imaging in Radiology Professor of Radiology at Harvard Medical School and MGH, commented on this research:
Since the 1960s radiologists have seen that there are unique and widely varying patterns of breast tissue appearing on mammograms in women ... these patterns represent the effect of genetics, hormones, pregnancy, breastfeeding, diet, weight loss and weight gain. can do. Now we can take advantage of this detailed information to be more detailed in our assessment of our risk at the individual level. "
Between 2009 and 2012, about 89,000 continuous screening mammograms were used to identify women at high risk of breast cancer development. It correctly kept 31% of all those patients, who later developed breast cancer in the top-risk bile. The corresponding value received using the existing Tyrer-Cuzick model was only 18%.
Alison Kurien is an Associate Professor of Medicine and Health Research / Policy at the Stanford University School of Medicine.
It is particularly striking that the model performs equally for white and black people, which is not in the case of earlier equipment ... if made available for valid and comprehensive use, then it is actually In order to estimate the risk, we can improve our current strategies. "
The goal of the team is to make a model of its model standard of care. According to estimating which person will develop cancer in the future, according to the development of breast cancer and preventing life, management strategies can be done accordingly.