🇪🇸 Leer este artículo en español

Title: Automated breast ultrasound (ABUS) and AI: better classification of BI-RADS 3–4 lesions without triggering unnecessary biopsies
Type: A (Medical imaging AI)
Tags: breast ultrasound, ABUS, artificial intelligence, BI-RADS, biopsy
PMID: 42572099
DOI: 10.1007/s40477-026-01212-1
Journal: Journal of Ultrasound (2026)
Category: AI in Medicine

---

⚖️ Transparency notice: this article was written with AI assistance and reviewed by the author, medical oncologist.

The problem nobody wants to have

BI-RADS 3 and 4 lesions on breast ultrasound are a gray zone: suspicious enough to cause anxiety, not suspicious enough to biopsy all of them. The result is a cascade of core needle biopsies (CNB), many unnecessary. An AI-based computer-aided diagnosis (CAD) tool applied to automated breast ultrasound (ABUS) images could help separate what truly matters [PMID: 42572099].

What they did

Italian study (University of Bari) with 54 patients undergoing ABUS (three sweeps per breast, axial/coronal/sagittal reconstruction). Two expert radiologists marked the region of interest (ROI) on 68 BI-RADS 3–4 lesions (32 type 3, 36 type 4), and a machine learning-based CAD system automatically assessed the level of suspicion [PMID: 42572099].

Critical reading

It is a small study (68 lesions, single center) without an external validation group or blind comparison against radiologists in real-world conditions. The ROI is drawn by the experts themselves, so the CAD does not start from scratch — it is already “guided”. It remains to be seen whether this scales to daily practice outside Bari [PMID: 42572099].

💡 Implications for clinical practice

ABUS + AI addresses a real problem: reducing unnecessary biopsies in the BI-RADS 3–4 zone. The concept fits with MARTE (our AI project on imaging reports). But until there is multicenter and blind validation, it is a promise, not a practice.

Reference: Moschetta M, et al. J Ultrasound. 2026. doi:10.1007/s40477-026-01212-1. PMID: 42572099.

— This analysis was generated by ANGIE (Always Next to Guide, Inspire and Empower), an artificial intelligence system with SOUL profiles, designed by Dr. Javier Pumares Pérez.

Connect on LinkedIn · ppj.es