AI breast cancer detector
NEWSLINE PAPER,-
Breast oncologist in Bucks County, Pennsylvania, Dr. Monique Gary, DO says AI mammography, or X-rays that detect early breast cancer with the help of artificial intelligence, is still experimental and should not be considered a standard treatment.
Besides that the advanced mammogram is not insured, so it costs $40 to $100 more than a standard mammogram. AI mammography needs to be trained so that the algorithm learns more and better, says Dr. Gary, according to Pop Sugar, Monday.
"It's not a standard treatment, and it's still experimental," Dr. Gary said.
When trained, AI can help eliminate some of the mistakes that cause cancer cells to be often overlooked by standard mammograms.
As with women with dense breasts, imaging errors may be avoided by an AI mammogram, so early detection and treatment can be done.
"The reason why mammograms are generally difficult to do is because it's a density-based study. The classification is dense, white-looking; dense tumors, looking white; young women who have dense breast tissue, also looking white," Dr. Gary said. It's like staring at the clouds and looking for a certain cloud. There's no contrast and no blue sky in between to pick a particular cloud.
AI mammography can help provide a clearer image because it is trained or validated by looking at hundreds of thousands of mammograms to detect patterns that may be worrying.
Then the difficulty of early detection using mammograms can be overcome faster and the visible images are potentially more accurate. Plus, computers learn from every case.
In other words, AI should be trained on a diverse population: candidates of different ages (especially given the increased incidence of cancer among young adults); with different types of breasts (e.g. fatty breasts, dense breasts and breasts with scar tissue); and from different races and ethnic groups.
But right now, Dr. Gary feels AI mammography hasn't reached everything. Although he has seen joint efforts to recruit more black candidates in research and machine learning, however, he adds, most of them are still validated in white Americans and Europeans.
According to Yale Medicine, standard mammograms have an accuracy of about 85 to 90 percent, but in women with dense breasts, the figure could be down to 30 percent. Adding AI support to mammogram can help bridge the gap, by providing a sharper radiological view.
Those who have microclassifications - or calcium deposits that look like tiny white spots, but can signal cancer - can also benefit from an AI mammogram.
These resources can also make a difference to certain communities whose mammogram quality is not so good. These include less-served communities, such as the black community, as well as military or veteran affairs, where general radiologists read mammograms and not those who specialize in breast imaging.
Professor of radiology at the George Washington University Medical Center, Dr. Rachel Brem believes that AI support can benefit general radiologists to function at the same level as sub-specialized radiologist.
(Wahyu Fatih/Newsline Paper Teams)