How much radiologist time can be saved by implementing AI in screen-reading mammograms?
[OBJECTIVE] There is a lack of breast radiologists in Norway and in Europe.
APA
Hovda T, Holen ÅS, Hofvind S (2026). How much radiologist time can be saved by implementing AI in screen-reading mammograms?. European radiology. https://doi.org/10.1007/s00330-025-12273-x
MLA
Hovda T, et al.. "How much radiologist time can be saved by implementing AI in screen-reading mammograms?." European radiology, 2026.
PMID
41498825
Abstract
[OBJECTIVE] There is a lack of breast radiologists in Norway and in Europe. Artificial intelligence (AI) offers an alternative to solely human readers and has demonstrated promising results in cancer detection in mammographic screening. We aimed to estimate the potential reduction in radiologists' workload by replacing one of the two radiologists with AI in screen-reading mammograms in BreastScreen Norway.
[MATERIALS AND METHODS] BreastScreen Norway targets about 680,000 women aged 50-69 who are invited biennially. The participation rates for each screening round are about 75%. All the screening mammograms are independently read by two radiologists. We collected information about the number of radiologist positions from all 16 breast centers in the country in 2024, while the number of screening examinations performed and the time spent on screen-reading and consensus were extracted from the screening database. We used 1 min for each screen-reading to estimate the screen-reading workload performed by the radiologists and calculated the time saved if one reader were replaced by AI.
[RESULTS] Screen-reading required a total of 6.5 man-years in BreastScreen Norway. Implementing AI as one of the two readers is thus able to reduce the screen-reading workload by 50%, from 6.5 to 3.3 man-years. The workload reduction corresponds to a reduction from 9% to 4.5% of the total workload for radiologists at the Norwegian breast centers.
[CONCLUSION] Implementation of AI in mammographic screening has the potential to reduce the screen-reading workload for breast radiologists. The reduced screening volume is of moderate influence on the overall workload for breast radiologists.
[KEY POINTS] Question How much radiology time is expected to be saved if AI were used as one of the two readers of screening mammograms? Findings Use of AI as one of the two readers, reducing the screen reading volume by 50%, was of moderate influence on the total workload for breast radiologists. Clinical relevance Implementing AI was shown to have limited potential in saving radiologists' time in screen-reading mammograms. The main benefit of implementing AI in screen-reading might thus be related to increased sensitivity of the screening test.
[MATERIALS AND METHODS] BreastScreen Norway targets about 680,000 women aged 50-69 who are invited biennially. The participation rates for each screening round are about 75%. All the screening mammograms are independently read by two radiologists. We collected information about the number of radiologist positions from all 16 breast centers in the country in 2024, while the number of screening examinations performed and the time spent on screen-reading and consensus were extracted from the screening database. We used 1 min for each screen-reading to estimate the screen-reading workload performed by the radiologists and calculated the time saved if one reader were replaced by AI.
[RESULTS] Screen-reading required a total of 6.5 man-years in BreastScreen Norway. Implementing AI as one of the two readers is thus able to reduce the screen-reading workload by 50%, from 6.5 to 3.3 man-years. The workload reduction corresponds to a reduction from 9% to 4.5% of the total workload for radiologists at the Norwegian breast centers.
[CONCLUSION] Implementation of AI in mammographic screening has the potential to reduce the screen-reading workload for breast radiologists. The reduced screening volume is of moderate influence on the overall workload for breast radiologists.
[KEY POINTS] Question How much radiology time is expected to be saved if AI were used as one of the two readers of screening mammograms? Findings Use of AI as one of the two readers, reducing the screen reading volume by 50%, was of moderate influence on the total workload for breast radiologists. Clinical relevance Implementing AI was shown to have limited potential in saving radiologists' time in screen-reading mammograms. The main benefit of implementing AI in screen-reading might thus be related to increased sensitivity of the screening test.