Interval Cancer Detection Using a Neural Network and Breast Density in Women with Negative Screening Mammograms

https://pubs.rsna.org/doi/10.1148/radiol.210832 

Background:

Inclusion of mammographic breast density in breast cancer risk models improves accuracy, but accuracy remains modest. Interval cancer risk prediction may be improved by combining assessments of BD and an artificial intelligence cancer detection system.

Objectives:

To evaluate the performance of a neural network (NN)-based model that combines the assessments of breast density and Transpara in the prediction of risk of developing IC among women with negative screening mammography results.

Materials and Methods:

This retrospective nested case-control study performed with screening examinations included women who developed IC and women with normal follow-up findings (from January 2011 to January 2015). Transpara (version 1.6) analyzed all studies yielding a score of 1–10, representing increasing likelihood of malignancy. Breast density was automatically computed using publicly available software (Laboratory for Individualized Breast Radiodensity Assessment). An NN model was trained by combining the Transpara score and breast density using 10-fold cross-validation. Bootstrap analysis was used to calculate the area under the receiver operating characteristic curve (AUC), sensitivity at 90% specificity, and 95% CIs of Transpara, breast density, and NN models.

Results:

A total of 2222 women with interval cancer and 4661 women in the control group were included (mean age, 61 years; age range, 49–76 years). AUC of the NN model was 0.79, which was significantly higher than AUC of Transpara or the breast density measure alone (AUC, 0.73 and 0.69, respectively). At 90% specificity, the NN model had a sensitivity of 50.9% (339 of 666 women) to predict interval cancer, which was significantly higher than that of Transpara alone (37.5%; 250 of 666 women) or breast density percentage alone (22.4%; 149 of 666 women).

Conclusion:

The combined assessment of the Transpara score and breast density measurements enabled identification of a larger proportion of women who would develop interval cancer compared with either method alone.