DBT radiomics-clinical nomogram for breast microcalcifications: validation AUC 0.790 versus clinical-only 0.881

In 181 breast microcalcification lesions, a digital breast tomosynthesis radiomics plus clinical nomogram had validation AUC 0.790, below the clinical-only 0.881, suggesting combined modeling needs optimization.

• Retrospective study of 181 microcalcification lesions from 172 patients, split 7:3 into training/validation sets; the intratumoral radiomics model alone had validation AUC 0.695.
• Independent clinical predictors were BI-RADS category (P<.001), diabetes history (P=.023), and personal breast cancer history (P=.034).
• Training nomogram AUC was 0.942; validation AUC was 0.790, with authors concluding combined modeling requires further optimization.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.