Dense mammary tissue
Most women with a possible abnormality in the breast will have a mammogram: an X-ray of the breast. But a mammogram cannot always show whether there is an abnormality in the breast. This is more often the case in women with dense mammary gland tissue (dense breast tissue). These women have proportionally more mammary glands and less fat in their breasts.
If the mammogram does not give a clear result, a woman with dense breast tissue may benefit from an MRI scan. This has been shown by the Dutch DENSE study. This is a major study into the added value of the MRI scan for dense mammary gland tissue. If the MRI scan becomes a standard examination in the screening of women with dense mammary gland tissue, this will be about 40,000 women per year. All of these MRI images must be reviewed by a radiologist with breast cancer expertise. The DENSE study also confirmed that most MRI images of dense breasts show no abnormalities. ‘This research shows that it is possible with artificial intelligence to perform an automated breast cancer screening of dense breast tissue.’
A triage tool for the radiologist
To reduce the workload of the radiologists, researchers at UMC Utrecht developed an automated system based on artificial intelligence (AI). They used the MRI images from the DENSE study to train the system and learn to distinguish between breasts with and without abnormalities. Lead researcher Kenneth Gilhuijs explains: ‘The system can inspect MRI images quickly and accurately. It only sends the pictures with abnormalities to the radiologist and stops photos without abnormalities. This makes it a triage instrument for the radiologist.’
The system extracts almost half of the photos
The system is now ‘smart’ enough to extract 40% of MRI images without abnormalities. In addition, the system did not miss a single photo with a malignant breast tumor on it. All those pictures ended up at the radiologist.
Reduction of the workload of radiologists
Kenneth further explains: ‘In this study we show that it is possible with artificial intelligence to perform an automated breast cancer screening of dense breast tissue, without missing a malignant breast tumor. Safely rejecting almost half of the normal pictures would mean a major reduction in the workload for radiologists.’
Making system even smarter
The researchers are satisfied with these initial findings. ‘The results are better than we expected: 40% is a good start. Now we want to make the system even smarter’, says Kenneth. ‘The ultimate goal is for the system to become as smart as an experienced radiologist. To achieve that, more research is needed. We will continue to train the system with other MRI images and with the next rounds of screening of the DENSE trial.’
Artificial intelligence (AI) helps detect abnormalities in dense breast tissue
Source link Artificial intelligence (AI) helps detect abnormalities in dense breast tissue