Home Hematology and Oncology AI-Supported Mammography Screen Reading Seems Promising

AI-Supported Mammography Screen Reading Seems Promising

Cancer detection rate similar compared with standard double reading, with lower screen-reading workload

By Elana Gotkine HealthDay Reporter

MONDAY, Aug. 7, 2023 (HealthDay News) — Cancer detection is similar for artificial intelligence (AI)-supported mammography screen reading compared with standard double reading, according to a study published online Aug. 1 in The Lancet Oncology.

Kristina LÃ¥ng, Ph.D., from Lund University in Sweden, and colleagues conducted a randomized, controlled, population-based trial involving women aged 40 to 80 years who were eligible for mammography screening at four sites in Sweden. Participants were randomly assigned to AI-supported screening or standard double reading without AI. The AI system provided an examination-based malignancy risk score, which was used to triage screening examinations to single-reading or double-reading risk scores. This study reported the prespecified clinical safety analysis for 80,033 women randomly assigned to AI-supported screening (40,003 women) or double reading without AI (40,030 women).

The researchers found that cancer detection rates were 6.1 and 5.1 per 1,000 screened participants in the intervention and control groups, respectively; the detection rate in the intervention group was above the lowest acceptable limit for safety. Recall rates were 2.2 and 2.0 percent in the intervention and control groups, respectively. In both groups, the false-positive rate was 1.5 percent. The positive predictive value of recall was 28.3 and 24.8 percent in the intervention and control groups, respectively. In the intervention group, 75 and 25 percent of cancers detected were invasive and in situ, respectively, compared with 81 and 19 percent, respectively, in the control group. Using AI, the screen-reading workload was reduced by 44.3 percent.

“The greatest potential of AI right now is that it could allow radiologists to be less burdened by the excessive amount of reading,” LÃ¥ng said in a statement.

One author disclosed ties to the pharmaceutical and medical technology industries.

Editorial (subscription or payment may be required)

Copyright © 2023 HealthDay. All rights reserved.