Share this story!
- Capio St. Goran's Hospital uses AI to review mammograms.
- More cases of cancer are found thanks to AI.
- The new technology may decrease the burden on breast radiologists.
AI becomes a helping hand in healthcare
At Capio St. Goran's Hospital in Stockholm, Sweden, AI algorithms will soon start reviewing mammograms in breast cancer screening.
The decision was made after a scientific study showed that more cases of cancer are detected when an AI and a radiologist work in parallel, compared to two radiologists.
The study investigated whether an AI algorithm could replace one of the two radiologists in the double-review of mammograms in screening.
The study included 55,581 women who had been examined with mammography at the hospital's Breast Center. In addition to being reviewed as usual by two radiologists, an AI algorithm also reviewed the images and analyzed whether there were signs of a tumor in them.
The study shows that using the AI algorithm as a complement finds more cancer than before, without increasing the risk of unnecessary examinations.
The AI algorithm can detect patterns in the images that human radiologists can't, and can mark and warn about these. This provides more accurate screening and can enable detection of more cases of cancer.
First in the world
"We will now be the first breast center in the world to, after thorough and structured validation of the algorithm, replace one of the two radiologists with AI in breast cancer screening," says Karin Dembrower, responsible for the ScreenTrust CAD study.
The introduction of AI does not mean that the computer takes over. In practice, it will work so that all mammograms generated at the Breast Center are reviewed by a breast radiologist, and by the AI algorithm, independently of each other.
If any of them see signs of suspected cancer, they "flag" the images. The flagged images are then discussed by at least two human radiologists who look at why the images were flagged and decide if the woman in question should be recalled for additional imaging and ultrasound examination. The AI algorithm does not diagnose or make any decisions but is used as support in the so-called consensus discussion.
Can shorten queues
By replacing one of the two radiologists with AI, the hospital can also keep the queues for screening and clinical examinations shorter.
"There is generally a major shortage of breast radiologists in Sweden, and this could be a way to address that problem," says Karin Dembrower.