A team from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and Massachusetts General Hospital (MGH) has developed a new AI model that can predict the likelihood of a woman developing breast cancer, up to five years in advance.
Existing technology for detecting breast cancer is largely based on human knowledge and risk factors in individual women. On the other hand, this new model learned from 90,000 mammograms to derive signs common in breast cancer patients and foresee “subtle patterns in breast tissue that are precursors to malignant tumors”. As a result, the model was successful at identifying 31 percent of cancer patients as high risk, compared to 18 percent by other traditional models.
Breast cancer is the most common form of cancer in women, both in developed and developing countries and late detection of tumours remains a leading cause of death. Mammography is so far the most effective tool for breast cancer screening but, studies to date have not shown benefits from regular mammography in women under the age of 40. The new model is based on risk based rather than age based factors and will encourage screening based on “risk assessment at individual level”.
The researchers responsible for developing the model also aim to use it for more equitable detection. Unlike traditional models that were developed keeping only white populations in mind, this new model is equally accurate for both black and white women. In addition to that, the researchers think that their model will be able to detect risks for other health problems like cardiovascular disease or other cancers.
However, a considerable problem with mammography and other such technologies is that it is costly and leaves scope for developing alternate approaches. Where does this new system fit in terms of cost efficiency? How viable its scope to reach minorities and underdeveloped countries?