We know that AI is powerful and can do a lot for radiologists and for other areas of the world. During a recent presentation held at the European Congress of Radiology (ECR), it was shared that AI can not only help a radiologist with reducing their daily reading list, but decreasing their recall rates as well.
In this particular study, the researchers found that triaging mammography exams based on the score produced from an AI algorithm resulted in a lower recall rate, and reduced the initial screening workload reading by more than 70%. It was also found that with these improvements, the AI and radiologists still found cancers at an equal rate. The study also noted that the AI found five additional cancers among the most suspicious exams.
Although this study shows the wonderful results that AI can produce, it isn’t quite ready to be implemented into practices. In the tested strategy, the cases that were not suspicious were just not read. In the total exam cases (approximately 37,500) 73% were not referred to the radiologists to review. Currently, there is no AI available that the FDA has cleared to enable radiologists to not review the case.
MammoScreen®, a leading FDA-cleared breast cancer detection software, can be compared to the tool used for the study. MammoScreen provides one comprehensive score that incorporates the finding, breast, and overall case assessment with one score for radiologists to be able to triage the most important cases.
With MammoScreen’s unique scoring system, radiologists are able to quickly identify the most suspicious cases (the red scores of 7-10) and the cases that are not suspicious (green cases of 1-4). Anything in a 5 or 6 (yellow scores) are not as confident in a suspicion level and therefore require more attention than other cases.
Although all cases that MammoScreen scores still need to be reviewed by radiologists, we have still proven a reduction in reading time. When batch reading, there has been up to a 35% reduction in reading time, mostly on the green cases that represents about 75% of the cases, and echoes what this study identified.
In the future, we can hope that radiologists will have more time to treat patients and be able to use AI services to help with reading. For now, MammoScreen is available to help find more cancer, reduce your reading time, and improve accuracy.
View our full brochure or contact us.