Features linked to false-positive marks and false-negative cancers on AI-CAD automated breast ultrasound

Stand-alone AI-CAD on ABUS had 74.2% sensitivity and 57.7% specificity. False negatives were often non-mass, isoechoic, lacking echogenic rind. False positives linked to younger age, heterogeneous background. Use as an adjunct, not standalone.

• AI-CAD sensitivity 74.2% (72/97 cancers) and specificity 57.7% in stand-alone analysis of ABUS.
• False-negative cancers more often non-mass lesions, isoechoic masses, or without an echogenic rind.
• False positives associated with younger age, heterogeneous background echotexture, dense breasts, posterior location; retrospective, single-center design limits generalizability.

Automated summary

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